@withgauge/cli 0.25.1 → 0.27.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +181 -68
- package/dist/batchWorker.js +1 -1
- package/dist/{chatApi-W6HW3HKD.js → chatApi-HF2R3SVR.js} +1 -1
- package/dist/{chunk-E24F53BZ.js → chunk-AZKXHO6C.js} +1 -1
- package/dist/chunk-PMX3NOGH.js +2 -0
- package/dist/chunk-PUNARKS2.js +64 -0
- package/dist/chunk-PVTU7NGL.js +1 -0
- package/dist/chunk-TAFB66SK.js +6 -0
- package/dist/chunk-Y4Y2AKDY.js +1 -0
- package/dist/index.js +315 -142
- package/dist/{orgDirectory-BZW2ZAD3.js → orgDirectory-QCOOVF65.js} +1 -1
- package/dist/watch/app.js +576 -0
- package/dist/watch/favicon.svg +11 -0
- package/dist/watch/index.html +21 -0
- package/dist/watch/logo-for-dark.png +0 -0
- package/dist/watch/logo-for-light.png +0 -0
- package/dist/watch/style.css +133 -0
- package/dist/watchServer.js +1 -0
- package/package.json +1 -1
- package/dist/chunk-RRLAI72I.js +0 -2
- package/dist/chunk-UANK5LHE.js +0 -69
- package/dist/chunk-WJTCHKG6.js +0 -1
package/dist/index.js
CHANGED
|
@@ -1,28 +1,28 @@
|
|
|
1
1
|
#!/usr/bin/env node
|
|
2
|
-
import{LOCAL_AGENTS,LOCAL_MCP_HINT,LOCAL_TARGETS,analysisNext,assertLocalCapabilities,
|
|
2
|
+
import{LOCAL_AGENTS,LOCAL_MCP_HINT,LOCAL_TARGETS,analysisNext,assertLocalCapabilities,commentSourceSchema,deploymentSourceSchema,detachLocal,freezeWebFixtures,inspectLocalCredential,inspectLocalGateway,inspectLocalRuntime,mergeWebFixtures,parseWebFixtureInput,planLocal,pullLocalGateway,pullLocalRuntime,readProfiles,readWebFixtureArchive,requireProfile,runLocal,setProfile,websiteSourceSchema,websiteUseSchema,writeProfiles}from"./chunk-TAFB66SK.js";import{setTelemetryEnabled,telemetryStatus}from"./chunk-WYYRMPRX.js";import{readJson}from"./chunk-Y4Y2AKDY.js";import{cancelLocal,cleanupLocal,localBatchStatus,localStatus,openLocalStore}from"./chunk-PUNARKS2.js";import"./chunk-QV7MOO3M.js";import{CHAT_BATCH_SIZE,ChatApiClient,OUTPUT_FORMATS,addPageOptions,categoryScope,chatPath,chatScope,chunks,createChatApiClient,csv,describeScopes,fetchPages,findWorkspace,gaugeState,header,hostedOnly,instructionMode,printItems,printJson,printPage,readGaugeState,requireHosted,resolveChatCategory,resolveOrg,resolveOutput,resolveTagIds,resolveTopicId,selectedOrg,style,truncate,withCategory}from"./chunk-PVTU7NGL.js";import{createdOrganization,credentialKind,fetchDirectory,findOrganization,isUuid,legacySelection,productsLabel,resolveOrgValue}from"./chunk-AZKXHO6C.js";import{ANALYSIS_SCHEMA_VERSION,ApiClient,CliError,DEFAULT_LIFECYCLE_TIMEOUT_SECONDS,MAX_LIFECYCLE_SOURCE_LENGTH,MAX_LIFECYCLE_TIMEOUT_SECONDS,agentConfigInputSchema,agentConfigSchema,agentSchema,analysisSubmissionSchema,assetRefListSchema,batteries_exports,brands_exports,clearToken,createClient,createdBySchema,credentialsPath,errorEnvelopeSchema,exitCodeFor,feedback_exports,keys_exports,lifecycleSchema,listResponseSchema,readCredentials,resolveAuthUrl,resolveBaseUrl,resolveToken,runConfig_exports,runCreditsChargedSchema,runEventSchema,runTriggeredBySchema,saveCredentials,saveToken,sharedTokenSchema,storedSharedCredentials,withCredentialLock}from"./chunk-PMX3NOGH.js";import{CLI_VERSION,CONFIG_KEYS,USER_AGENT,configPath,readConfig,updateConfig}from"./chunk-OIGVDK3F.js";import{__export}from"./chunk-PSBF4TEK.js";import{Command,CommanderError as CommanderError2,Option as Option6}from"commander";import{fileURLToPath as fileURLToPath2}from"url";var auth_exports={};__export(auth_exports,{browserLogin:()=>browserLogin,deviceLogin:()=>deviceLogin,logout:()=>logout,parseLoopbackCallback:()=>parseLoopbackCallback,readStdin:()=>readStdin,register:()=>register,useOnlyOrg:()=>useOnlyOrg});import{spawn}from"child_process";import{createHash,randomBytes,randomUUID}from"crypto";import{createServer}from"http";import{hostname}from"os";import{createInterface}from"readline";import{setTimeout as sleep}from"timers/promises";import{z}from"zod";var deviceAuthorizeRequestSchema=z.object({deviceName:z.string().trim().max(64).optional(),email:z.string().trim().email().max(320).optional()});var deviceAuthorizeResponseSchema=z.object({deviceCode:z.string(),userCode:z.string(),verificationUri:z.string(),verificationUriComplete:z.string(),interval:z.number().int().positive(),expiresIn:z.number().int().positive(),emailSent:z.boolean()});var deviceTokenRequestSchema=z.object({deviceCode:z.string().min(1)});var deviceTokenApprovedSchema=z.object({status:z.literal("approved"),accessToken:z.string(),tokenType:z.literal("bearer"),expiresAt:z.string().nullable()});var deviceTokenResultSchema=z.discriminatedUnion("status",[z.object({status:z.literal("pending")}),z.object({status:z.literal("slow_down"),interval:z.number().int()}),z.object({status:z.literal("denied")}),z.object({status:z.literal("expired")}),deviceTokenApprovedSchema]);async function credentialIdentity(){if(credentialKind()==="legacy"){const{user}=await createClient().get("/api/v1/me");return{user,apiKey:null,token:{kind:"legacy",scopes:null,expiresAt:null}}}return createChatApiClient().me()}async function principalIdentity(){const{user,apiKey,token}=await credentialIdentity();const orgs=token.kind==="legacy"?null:apiKey?.organization?[apiKey.organization]:await fetchDirectory();return{user,apiKey,orgs}}function printPrincipal(identity){const{user,apiKey,orgs}=identity;if(apiKey)console.log(`Org API key "${apiKey.name}" (${apiKey.prefix}\u2026)`);else if(user)console.log(`Signed in as ${user.email||user.id}${user.name?` (${user.name})`:""}`);if(orgs)console.log(`Organizations: ${orgs.map(org=>`${org.name} (${org.slug})`).join(", ")||"(none)"}`);else if(orgs===null)console.log("Use `gauge auth login` to list your Gauge organizations.")}async function fetchAllPages(client,path,query={}){const items=[];let cursor;do{const page=await client.get(path,{limit:200,...query,cursor});items.push(...page.items);cursor=page.nextCursor}while(cursor);return items}var LOGIN_TIMEOUT_MS=2*60*1e3;function tokenSource(){if(process.env.GAUGE_API_TOKEN?.trim())return"env";return resolveToken()?"file":"none"}async function fetchMe(client,shared=false){return shared?client.get("/api/cli/me"):client.get("/api/v1/me")}function sessionCredentials(raw){return{version:1,sessionId:randomUUID(),...sharedTokenSchema.parse(raw),authUrl:resolveAuthUrl(),apiUrl:resolveBaseUrl()}}async function verifyIssuedToken(token){const shared=token.startsWith("gauge_cli_");return fetchMe(new ApiClient({token,baseUrl:shared?resolveAuthUrl():resolveBaseUrl()}),shared)}function parseLoopbackCallback(rawUrl,expectedState){const url=new URL(rawUrl??"/","http://127.0.0.1");if(url.pathname!=="/callback")return null;if(url.searchParams.get("state")===expectedState&&url.searchParams.get("error")==="access_denied"){throw new CliError("unauthorized","Authorization was denied in the browser.")}const code=url.searchParams.get("code");if(url.searchParams.get("state")!==expectedState||!code)return null;return code}function printIdentity(me){const orgs=me.orgs.map(o=>o.slug).join(", ")||"(none)";console.log(`Logged in as ${me.user.email??me.user.id}`);console.log(`Orgs: ${orgs}`);console.log(`API URL: ${resolveBaseUrl()}`)}function useOnlyOrg(me,shared=false,log=console.log){const org=me.orgs.length===1?me.orgs[0]:void 0;const only=!org?void 0:shared&&"id"in org?org.id:"canonicalGaugeOrganizationId"in org?org.canonicalGaugeOrganizationId:void 0;if(!only||readConfig().orgId)return;updateConfig({orgId:only,defaultOrg:void 0});log(`Default organization set to ${org?.name??only}.`)}function openBrowser(url){const[cmd,args]=process.platform==="darwin"?["open",[url]]:process.platform==="win32"?["cmd",["/c","start","",url]]:["xdg-open",[url]];try{const child=spawn(cmd,args,{stdio:"ignore",detached:true});child.on("error",()=>{});child.unref()}catch{}}async function browserLogin(){const state=randomBytes(16).toString("base64url");const codeVerifier=randomBytes(32).toString("base64url");const codeChallenge=createHash("sha256").update(codeVerifier).digest("base64url");const server=createServer();const codePromise=new Promise((resolve5,reject)=>{server.on("request",(req,res)=>{const url=new URL(req.url??"/","http://127.0.0.1");if(url.pathname!=="/callback"){res.writeHead(404).end();return}if(req.method!=="GET"){res.writeHead(405,{allow:"GET"}).end();return}let code;try{code=parseLoopbackCallback(req.url,state)}catch(error){res.writeHead(200,{"content-type":"text/plain"}).end("Authorization denied. Return to your terminal.");reject(error);return}if(!code){res.writeHead(400,{"content-type":"text/plain; charset=utf-8"});res.end("Could not complete sign-in. Run gauge auth login again.");return}res.writeHead(200,{"content-type":"text/plain; charset=utf-8"});res.end("Authorized. Return to your terminal.");resolve5(code)});server.on("error",e=>reject(new CliError("network",`Could not receive browser sign-in on this computer: ${e.message}`,{hint:"Try `gauge auth login --device` to sign in using a browser on another device."})))});await new Promise(resolve5=>{server.listen(0,"127.0.0.1",resolve5)});const address=server.address();if(!address||typeof address==="string"){server.close();throw new CliError("internal","Could not start browser sign-in on this computer.",{hint:"Try `gauge auth login --device` instead."})}const base=resolveAuthUrl();const authorizeUrl=new URL(`${base}/cli/authorize`);authorizeUrl.searchParams.set("port",String(address.port));authorizeUrl.searchParams.set("state",state);authorizeUrl.searchParams.set("name",hostname()||"this device");authorizeUrl.searchParams.set("code_challenge",codeChallenge);authorizeUrl.searchParams.set("code_challenge_method","S256");console.log("Opening your browser to authorize the Gauge CLI\u2026");console.log(`If it doesn't open, visit:
|
|
3
3
|
|
|
4
4
|
${authorizeUrl}
|
|
5
5
|
`);console.log("Waiting for browser authorization\u2026");console.log("No local network access or using SSH? Re-run with gauge auth login --device.");openBrowser(authorizeUrl.toString());let timer;try{const code=await Promise.race([codePromise,new Promise((_,reject)=>{timer=setTimeout(()=>{reject(new CliError("network","Timed out waiting for browser authorization (2 minutes). Retry, or use gauge auth login --device. New to Gauge? Sign-up needs a work email."))},LOGIN_TIMEOUT_MS)})]);const raw=await bootstrapPost("/api/cli/token",{code,codeVerifier});const parsed=sharedTokenSchema.safeParse(raw);if(!parsed.success){throw new CliError("internal","Gauge returned a sign-in response the CLI could not read. Try `gauge auth login` again.")}const exchange=parsed.data;const token=exchange.accessToken;const me=await verifyIssuedToken(token);await withCredentialLock(async()=>{saveCredentials(sessionCredentials(raw))});console.log(`Token saved to ${credentialsPath()}`);printIdentity(me);useOnlyOrg(me,token.startsWith("gauge_cli_"))}finally{if(timer)clearTimeout(timer);server.close();server.closeAllConnections()}}async function bootstrapPost(path,body){const base=resolveAuthUrl();let res;try{res=await fetch(`${base}${path}`,{method:"POST",redirect:"error",signal:AbortSignal.timeout(15e3),headers:{"Content-Type":"application/json",Accept:"application/json","User-Agent":USER_AGENT},body:JSON.stringify(body)})}catch(e){throw new CliError("network",`Request to ${base} failed: ${e instanceof Error?e.message:String(e)}`)}const text=await res.text();if(!res.ok){try{const parsed=errorEnvelopeSchema.safeParse(JSON.parse(text));if(parsed.success){throw new CliError(parsed.data.error.code,parsed.data.error.message)}}catch(e){if(e instanceof CliError)throw e}throw new CliError("internal",`Authentication request failed: HTTP ${res.status} ${res.statusText}`.trim())}return JSON.parse(text)}async function deviceLogin(opts){const start=await bootstrapPost("/api/cli/device/authorize",{deviceName:hostname()||"this device",email:opts.email});console.log("To sign in, open this URL in a browser:\n");console.log(` ${start.verificationUriComplete}
|
|
6
6
|
`);console.log(`and confirm this code: ${start.userCode}
|
|
7
|
-
`);if(start.emailSent&&opts.email){console.log(`We also emailed a sign-in link to ${opts.email}.`)}console.log("Waiting for you to approve in the browser\u2026");openBrowser(start.verificationUriComplete);const deadline=Date.now()+start.expiresIn*1e3;const MAX_TRANSIENT=5;let interval=start.interval;let transient=0;while(Date.now()<deadline){await sleep(interval*1e3);let raw;try{raw=await bootstrapPost("/api/cli/device/token",{deviceCode:start.deviceCode})}catch(e){if(e instanceof CliError&&(e.code==="network"||e.code==="internal")&&++transient<=MAX_TRANSIENT){continue}throw e}transient=0;const parsed=deviceTokenResultSchema.safeParse(raw);if(!parsed.success){throw new CliError("internal","Gauge returned a sign-in response the CLI could not read. Try `gauge auth login --device` again.")}const result=parsed.data;switch(result.status){case"pending":continue;case"slow_down":interval=result.interval;continue;case"denied":throw new CliError("unauthorized","Authorization was denied in the browser.");case"expired":throw new CliError("unauthorized","The sign-in request expired before it was approved. Re-run it to try again.");case"approved":{const token=result.accessToken;try{const credentials=sessionCredentials(raw);const me=await verifyIssuedToken(token);return{token,me,credentials}}catch(e){throw new CliError("network",`You approved sign-in, but the CLI could not confirm access to your account (${e instanceof Error?e.message:String(e)}). Run \`gauge auth login --device\` again. Your saved login has not changed.`)}}default:throw new CliError("internal","Gauge could not complete sign-in. Try `gauge auth login --device` again.")}}throw new CliError("network","Timed out waiting for browser approval. Re-run to try again. New to Gauge? Sign-up needs a work email.")}async function readStdin(input=process.stdin){if(input.isTTY){const lines=createInterface({input,crlfDelay:Infinity});for await(const line of lines){lines.close();return line}return""}let data="";for await(const chunk of input)data+=chunk;return data}async function withTokenLogin(){if(process.stdin.isTTY){console.error("Paste your API token, then press Enter:")}const token=(await readStdin()).trim();if(!token)throw new CliError("usage","Expected a token on stdin");if(credentialKind(token)==="api_key"){const me2=await new ChatApiClient({token,baseUrl:resolveAuthUrl()}).me();await withCredentialLock(async()=>{saveToken(token)});console.log(`Token saved to ${credentialsPath()}`);printPrincipal({...me2,orgs:me2.apiKey?.organization?[me2.apiKey.organization]:void 0});return}const me=await verifyIssuedToken(token);await withCredentialLock(async()=>{if(token.startsWith("gauge_cli_"))saveCredentials({version:1,sessionId:randomUUID(),accessToken:token,authUrl:resolveAuthUrl(),apiUrl:resolveBaseUrl()});else saveToken(token)});console.log(`Token saved to ${credentialsPath()}`);printIdentity(me);useOnlyOrg(me,token.startsWith("gauge_cli_"))}var TOKENS_PATH="/api/v1/tokens";async function fetchAllTokens(client){return fetchAllPages(client,TOKENS_PATH)}function tokenStatus(t){if(t.revokedAt)return"revoked";if(t.expiresAt&&Date.parse(t.expiresAt)<=Date.now())return"expired";return"active"}function tokenRow(t){return{id:t.id,name:t.name,token:`\u2026${t.last4}`,status:tokenStatus(t),created:t.createdAt.slice(0,10),lastUsed:t.lastUsedAt?t.lastUsedAt.slice(0,10):"never",expires:t.expiresAt?t.expiresAt.slice(0,10):"never"}}function parseExpiresIn(raw){if(raw==="never")return void 0;const match=raw.match(/^(\d+)([dh])$/);if(!match){throw new CliError("usage",`Invalid --expires-in "${raw}" (expected e.g. 90d, 12h, or never)`)}const n=Number(match[1]);const ms=match[2]==="d"?n*24*60*60*1e3:n*60*60*1e3;if(ms<=0){throw new CliError("usage","--expires-in must be a positive duration")}return new Date(Date.now()+ms).toISOString()}async function whoami(output){const identity=await principalIdentity();if(output==="json")return printJson(identity);printPrincipal(identity)}async function logout(){await withCredentialLock(async()=>{const stored=readCredentials();try{if(stored&&typeof stored!=="string"&&stored.refreshToken){const response=await fetch(`${stored.authUrl}/api/cli/revoke`,{method:"POST",redirect:"error",signal:AbortSignal.timeout(15e3),headers:{"Content-Type":"application/json"},body:JSON.stringify({refreshToken:stored.refreshToken})});if(!response.ok)throw new Error(`HTTP ${response.status}`)}}catch(error){console.error(`Warning: your saved login was removed from this computer, but Gauge could not end it on the server (${error instanceof Error?error.message:String(error)}). That login may still allow access. Contact Gauge support if you need to confirm it has been ended.`)}finally{clearToken()}})}function
|
|
7
|
+
`);if(start.emailSent&&opts.email){console.log(`We also emailed a sign-in link to ${opts.email}.`)}console.log("Waiting for you to approve in the browser\u2026");openBrowser(start.verificationUriComplete);const deadline=Date.now()+start.expiresIn*1e3;const MAX_TRANSIENT=5;let interval=start.interval;let transient=0;while(Date.now()<deadline){await sleep(interval*1e3);let raw;try{raw=await bootstrapPost("/api/cli/device/token",{deviceCode:start.deviceCode})}catch(e){if(e instanceof CliError&&(e.code==="network"||e.code==="internal")&&++transient<=MAX_TRANSIENT){continue}throw e}transient=0;const parsed=deviceTokenResultSchema.safeParse(raw);if(!parsed.success){throw new CliError("internal","Gauge returned a sign-in response the CLI could not read. Try `gauge auth login --device` again.")}const result=parsed.data;switch(result.status){case"pending":continue;case"slow_down":interval=result.interval;continue;case"denied":throw new CliError("unauthorized","Authorization was denied in the browser.");case"expired":throw new CliError("unauthorized","The sign-in request expired before it was approved. Re-run it to try again.");case"approved":{const token=result.accessToken;try{const credentials=sessionCredentials(raw);const me=await verifyIssuedToken(token);return{token,me,credentials}}catch(e){throw new CliError("network",`You approved sign-in, but the CLI could not confirm access to your account (${e instanceof Error?e.message:String(e)}). Run \`gauge auth login --device\` again. Your saved login has not changed.`)}}default:throw new CliError("internal","Gauge could not complete sign-in. Try `gauge auth login --device` again.")}}throw new CliError("network","Timed out waiting for browser approval. Re-run to try again. New to Gauge? Sign-up needs a work email.")}async function readStdin(input=process.stdin){if(input.isTTY){const lines=createInterface({input,crlfDelay:Infinity});for await(const line of lines){lines.close();return line}return""}let data="";for await(const chunk of input)data+=chunk;return data}async function withTokenLogin(){if(process.stdin.isTTY){console.error("Paste your API token, then press Enter:")}const token=(await readStdin()).trim();if(!token)throw new CliError("usage","Expected a token on stdin");if(credentialKind(token)==="api_key"){const me2=await new ChatApiClient({token,baseUrl:resolveAuthUrl()}).me();await withCredentialLock(async()=>{saveToken(token)});console.log(`Token saved to ${credentialsPath()}`);printPrincipal({...me2,orgs:me2.apiKey?.organization?[me2.apiKey.organization]:void 0});return}const me=await verifyIssuedToken(token);await withCredentialLock(async()=>{if(token.startsWith("gauge_cli_"))saveCredentials({version:1,sessionId:randomUUID(),accessToken:token,authUrl:resolveAuthUrl(),apiUrl:resolveBaseUrl()});else saveToken(token)});console.log(`Token saved to ${credentialsPath()}`);printIdentity(me);useOnlyOrg(me,token.startsWith("gauge_cli_"))}var TOKENS_PATH="/api/v1/tokens";async function fetchAllTokens(client){return fetchAllPages(client,TOKENS_PATH)}function tokenStatus(t){if(t.revokedAt)return"revoked";if(t.expiresAt&&Date.parse(t.expiresAt)<=Date.now())return"expired";return"active"}function tokenRow(t){return{id:t.id,name:t.name,token:`\u2026${t.last4}`,status:tokenStatus(t),created:t.createdAt.slice(0,10),lastUsed:t.lastUsedAt?t.lastUsedAt.slice(0,10):"never",expires:t.expiresAt?t.expiresAt.slice(0,10):"never"}}function parseExpiresIn(raw){if(raw==="never")return void 0;const match=raw.match(/^(\d+)([dh])$/);if(!match){throw new CliError("usage",`Invalid --expires-in "${raw}" (expected e.g. 90d, 12h, or never)`)}const n=Number(match[1]);const ms=match[2]==="d"?n*24*60*60*1e3:n*60*60*1e3;if(ms<=0){throw new CliError("usage","--expires-in must be a positive duration")}return new Date(Date.now()+ms).toISOString()}async function whoami(output){const identity=await principalIdentity();if(output==="json")return printJson(identity);printPrincipal(identity)}async function logout(){await withCredentialLock(async()=>{const stored=readCredentials();try{if(stored&&typeof stored!=="string"&&stored.refreshToken){const response=await fetch(`${stored.authUrl}/api/cli/revoke`,{method:"POST",redirect:"error",signal:AbortSignal.timeout(15e3),headers:{"Content-Type":"application/json"},body:JSON.stringify({refreshToken:stored.refreshToken})});if(!response.ok)throw new Error(`HTTP ${response.status}`)}}catch(error){console.error(`Warning: your saved login was removed from this computer, but Gauge could not end it on the server (${error instanceof Error?error.message:String(error)}). That login may still allow access. Contact Gauge support if you need to confirm it has been ended.`)}finally{clearToken()}})}function register(program2){const auth=hostedOnly(program2.command("auth"),{account:true}).description("Authenticate the CLI");auth.command("login").description("Sign in to Gauge using your browser").option("--with-token","read a token from stdin (headless machines)").option("--device","use a device code instead (no local network access required)").option("--email <email>","with --device: also email a sign-in link to this address").action(async opts=>{if(opts.withToken&&opts.device){throw new CliError("usage","Choose only one of --with-token or --device.")}if(opts.email&&!opts.device){throw new CliError("usage","--email requires --device.")}if(opts.device){const{token,me,credentials}=await deviceLogin({email:opts.email});await withCredentialLock(async()=>{saveCredentials(credentials)});console.log(`Token saved to ${credentialsPath()}`);printIdentity(me);useOnlyOrg(me,token.startsWith("gauge_cli_"))}else if(opts.withToken){await withTokenLogin()}else{await browserLogin()}});auth.command("logout").description("Sign out of this CLI installation").action(async()=>{await logout();console.log(`Removed ${credentialsPath()}`);if(process.env.GAUGE_API_TOKEN?.trim()){console.error("warning: GAUGE_API_TOKEN is still set. Commands will continue using it; unset it to fully sign out.")}});auth.command("status").description("Check your login or API key, permissions, and expiration").action(async(_opts,command)=>{const output=resolveOutput(command);const source=tokenSource();if(source==="none"){throw new CliError("unauthorized","Not logged in. Run `gauge auth login` or set GAUGE_API_TOKEN.")}let me;try{me=await credentialIdentity()}catch(e){if(e instanceof CliError&&e.code==="unauthorized"){throw new CliError("unauthorized",source==="env"?"The API key or token in GAUGE_API_TOKEN was rejected. It may have expired or been revoked.":"Your saved login or API key was rejected. It may have expired or been revoked.",{hint:source==="env"?"Replace GAUGE_API_TOKEN with a valid key, or unset it and run `gauge auth login`.":"Run `gauge auth login` to sign in again, or import a replacement API key with `gauge auth login --with-token`."})}throw e}if(output==="json"){printJson({authenticated:true,source,apiUrl:resolveBaseUrl(),authUrl:storedSharedCredentials()?.authUrl??resolveAuthUrl(),credential:me.token});return}const label=me.token.kind==="session"?"Login":me.token.kind==="api_key"?"API key":"Agents token";console.log(`${label}: valid`);console.log(`Permissions: ${me.token.scopes?describeScopes(me.token.scopes):"not available for this token"}`);console.log(`Expires: ${me.token.expiresAt??(me.token.kind==="legacy"?"not available for this token":"never")}`);console.log(`Auth URL: ${storedSharedCredentials()?.authUrl??resolveAuthUrl()}`);console.log(`Agents API URL: ${resolveBaseUrl()}`);console.log(`Using: ${source==="env"?"GAUGE_API_TOKEN environment variable":credentialsPath()}`)});const tokens=auth.command("tokens").description("Manage tokens created by older versions of Gauge Agents");tokens.command("list").description("List your API tokens").action(async(_opts,command)=>{const output=resolveOutput(command);const items=await fetchAllTokens(createClient());if(output==="json"){printJson(items);return}printItems(items.map(tokenRow),"table")});tokens.command("create <name>").description("Create an API token (the secret is shown exactly once)").option("--expires-in <duration>","e.g. 90d, 12h, or never","never").action(async(name,opts,command)=>{const output=resolveOutput(command);if(resolveToken()?.startsWith("gauge_cli_"))throw new CliError("forbidden","You can't create this older type of Agents token with your current login.",{hint:"For scripts and CI, create an organization API key in Settings \u2192 API Keys and set GAUGE_API_TOKEN."});const expiresAt=parseExpiresIn(opts.expiresIn);const created=await createClient().post(TOKENS_PATH,{name,...expiresAt?{expiresAt}:{}});if(output==="json"){printJson(created);return}console.log(`Created token "${created.name}" (id ${created.id}, expires ${created.expiresAt??"never"})`);console.log(`
|
|
8
8
|
${created.token}
|
|
9
|
-
`);console.log("Store it now \u2014 it won't be shown again.")});tokens.command("revoke <id>").description("Revoke an API token").action(async(id,_opts,command)=>{await createClient().delete(`${TOKENS_PATH}/${encodeURIComponent(id)}`);if(resolveOutput(command)!=="json")console.log(`Revoked ${id}`)});program2.command("whoami").description("Show the user or organization API key you are acting as").action(async(_opts,command)=>{await whoami(resolveOutput(command))})}var billing_exports={};__export(billing_exports,{fmtOverage:()=>fmtOverage,register:()=>
|
|
9
|
+
`);console.log("Store it now \u2014 it won't be shown again.")});tokens.command("revoke <id>").description("Revoke an API token").action(async(id,_opts,command)=>{await createClient().delete(`${TOKENS_PATH}/${encodeURIComponent(id)}`);if(resolveOutput(command)!=="json")console.log(`Revoked ${id}`)});program2.command("whoami").description("Show the user or organization API key you are acting as").action(async(_opts,command)=>{await whoami(resolveOutput(command))})}var billing_exports={};__export(billing_exports,{fmtOverage:()=>fmtOverage,register:()=>register2});function fmtDate(iso){return iso?iso.slice(0,10):"\u2014"}function fmtOverage(s){const line=s.creditLine;if(!line)return s.paymentType==="card"?"n/a (card)":"\u2014";if(line.limitUsd<=0)return"off (credit line limit $0)";return`$${line.usedUsd.toFixed(2)} of $${line.limitUsd.toFixed(2)} credit line`}function statusLines(s){return[["Plan",s.plan??"free"],["Subscription",s.subscriptionStatus??"\u2014"],["Payment",s.paymentType??"\u2014"],["Balance",`${s.balance} credits`],["Allowance / cycle",s.allowance===null?"\u2014":`${s.allowance} credits`],["Period",s.currentPeriodStart||s.currentPeriodEnd?`${fmtDate(s.currentPeriodStart)} \u2192 ${fmtDate(s.currentPeriodEnd)}`:"\u2014"],["Used this period",`${s.creditsUsedThisPeriod} credits`],["Runs this period",String(s.runsThisPeriod)],["Overage",fmtOverage(s)],["Can run eval",s.canRunByWorkload.eval?"yes":"no"],["Can run preference",s.canRunByWorkload.visibility?"yes":"no"],["Can run generic",s.canRunByWorkload.generic?"yes":"no"],["Billing enforced",s.enforced?"yes":"no"]]}function register2(program2){const group=hostedOnly(program2.command("billing")).description("Billing & credits");group.command("status").description("Show credit balance, plan, payment, and overage").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const status=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/billing`);if(resolveOutput(cmd)==="json"){printJson(status);return}const lines=statusLines(status);const width=Math.max(...lines.map(([label])=>label.length));for(const[label,value]of lines){console.log(`${label.padEnd(width)} ${value}`)}})}var brands_exports2={};__export(brands_exports2,{register:()=>register3});import{Option}from"commander";var KINDS=brands_exports.BRAND_KINDS.map(k=>k.toLowerCase());var ECOSYSTEMS=brands_exports.BRAND_PACKAGE_ECOSYSTEMS;function brandsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/brands`}function collect(value,previous){return previous?[...previous,value]:[value]}function splitList(raw){return raw.split(",").map(v=>v.trim()).filter(Boolean)}function parsePackage(raw){const sep3=raw.lastIndexOf(":");if(sep3===-1)return{name:raw.trim(),ecosystem:null};const name=raw.slice(0,sep3).trim();const ecosystem=raw.slice(sep3+1).trim();if(!name)throw new CliError("usage",`Invalid package "${raw}"`);if(!ECOSYSTEMS.includes(ecosystem)){throw new CliError("usage",`Invalid ecosystem "${ecosystem}" (expected ${ECOSYSTEMS.join("|")})`)}return{name,ecosystem}}function toRow(brand){return{id:brand.id,name:brand.name,kind:brand.kind.toLowerCase(),aliases:brand.aliases.join(", "),websites:brand.websites.join(", "),packages:brand.packages.map(p=>p.name).join(", ")}}function register3(program2){const group=hostedOnly(program2.command("brands")).description("Manage org brands");group.command("list").description("List brands").addOption(new Option("--kind <kind>","filter by kind").choices(KINDS)).action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const items=await fetchAllPages(createClient(),brandsPath(org),{kind:opts.kind?.toUpperCase()});if(resolveOutput(cmd)==="json"){printJson(items);return}printItems(items.map(toRow),"table")});group.command("add <name>").description("Add a brand").addOption(new Option("--kind <kind>","brand kind").choices(KINDS).makeOptionMandatory()).option("--aliases <list>","comma-separated alias spellings").option("--websites <list>","comma-separated docs/marketing URLs").option("--package <name[:ecosystem]>","package identifier, repeatable (ecosystem: npm|pypi|cargo|go|rubygems|other)",collect).action(async(name,opts,cmd)=>{const org=await resolveOrg(cmd);const row2=await createClient().post(brandsPath(org),{name,kind:opts.kind.toUpperCase(),aliases:opts.aliases===void 0?[]:splitList(opts.aliases),websites:opts.websites===void 0?[]:splitList(opts.websites),packages:(opts.package??[]).map(parsePackage)});if(resolveOutput(cmd)==="json")printJson(row2);else console.log(`Created brand ${row2.id} (${row2.name})`)});group.command("edit <id>").description("Update a brand (only the provided fields change)").option("--name <name>","new canonical name").addOption(new Option("--kind <kind>","new kind").choices(KINDS)).option("--aliases <list>","replace aliases (comma-separated; pass '' to clear)").option("--websites <list>","replace websites (comma-separated; pass '' to clear)").option("--package <name[:ecosystem]>","replace packages, repeatable",collect).option("--clear-packages","remove all packages").action(async(id,opts,cmd)=>{const patch={};if(opts.name!==void 0)patch.name=opts.name;if(opts.kind!==void 0)patch.kind=opts.kind.toUpperCase();if(opts.aliases!==void 0)patch.aliases=splitList(opts.aliases);if(opts.websites!==void 0)patch.websites=splitList(opts.websites);if(opts.clearPackages)patch.packages=[];else if(opts.package!==void 0)patch.packages=opts.package.map(parsePackage);if(Object.keys(patch).length===0){throw new CliError("usage","Nothing to update \u2014 pass at least one of --name/--kind/--aliases/--websites/--package/--clear-packages.")}const org=await resolveOrg(cmd);const row2=await createClient().patch(`${brandsPath(org)}/${encodeURIComponent(id)}`,patch);if(resolveOutput(cmd)==="json")printJson(row2);else console.log(`Updated brand ${row2.id} (${row2.name})`)});group.command("rm <id>").description("Delete a brand").action(async(id,_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(`${brandsPath(org)}/${encodeURIComponent(id)}`);if(resolveOutput(cmd)!=="json")console.log(`Deleted brand ${id}`)})}var chat_exports={};__export(chat_exports,{register:()=>register5});import{Help}from"commander";var MAX_ANSWER_LIMIT=100;var list=raw=>raw?csv(raw):void 0;function rowFilters(command){return command.option("--since <date>","YYYY-MM-DD (default 7 days ago; ranges \u2264 92 days)").option("--until <date>","YYYY-MM-DD (default today)").option("--model <models>","comma-separated models").option("--prompt <ids>","comma-separated prm_ ids").option("--topic <ids>","comma-separated top_ ids").option("--tag <ids>","comma-separated tag_ ids").option("--tracking-type <type>","visibility (default) | branded_own | all")}function rowQuery(opts){return{since:opts.since,until:opts.until,model:list(opts.model),prompt:list(opts.prompt),topic:list(opts.topic),tag:list(opts.tag),trackingType:opts.trackingType}}function registerData(chat){const answers=chat.command("answers").description("Sampled AI answers in a category");addPageOptions(rowFilters(withCategory(answers.command("list",{isDefault:true})).description("List answers, newest first").option("--brand <ids>","only answers mentioning any of these brd_ ids").option("--brand-absent <ids>","only answers mentioning none of these brd_ ids")),MAX_ANSWER_LIMIT).action(async(opts,cmd)=>{const scope=await categoryScope(cmd);const page=await fetchPages(scope.client,chatPath(...scope.base,"answers"),{...rowQuery(opts),brand:list(opts.brand),brandAbsent:list(opts.brandAbsent)},opts,MAX_ANSWER_LIMIT);printPage(page,resolveOutput(cmd),a=>({id:a.id,day:a.day,model:a.model,prompt:a.promptId,brands:a.brandIds.length,text:truncate(a.text,60)}))});withCategory(answers.command("get <answer>")).description("Show an answer with its mentions, citations and search queries").action(async(id,_opts,cmd)=>{const scope=await categoryScope(cmd);const answer=await scope.client.get(chatPath(...scope.base,"answers",id));if(resolveOutput(cmd)==="json")return printJson(answer);console.log(`${answer.id} ${answer.model} ${answer.day}`);console.log(`Prompt (${answer.promptId}): ${answer.promptText}
|
|
10
10
|
`);console.log(`${answer.text}
|
|
11
11
|
`);printItems(answer.mentions.map(m=>({brand:m.brandId,name:m.name,count:m.count,types:m.mentionTypes.join(",")})),"table");console.log("");printItems(answer.citations.map(c=>({page:c.pageId,domain:c.domain,url:c.url})),"table");if(answer.searchQueries.length)console.log(`
|
|
12
12
|
Search queries: ${answer.searchQueries.join(" | ")}`)});const citations=chat.command("citations").description("Pages cited by AI answers in a category");addPageOptions(rowFilters(withCategory(citations.command("list",{isDefault:true})).description("List citations").option("--domain <domains>","comma-separated cited domains").option("--page <ids>","comma-separated pag_ ids"))).action(async(opts,cmd)=>{const scope=await categoryScope(cmd);const page=await fetchPages(scope.client,chatPath(...scope.base,"citations"),{...rowQuery(opts),domain:list(opts.domain),page:list(opts.page)},opts);printPage(page,resolveOutput(cmd),c=>({answer:c.answerId,day:c.day,model:c.model,domain:c.domain,url:c.url}))})}import{readFileSync}from"fs";var TRACKING_TYPES=["visibility","branded_own"];var collect2=(value,prev=[])=>[...prev,...csv(value)];function readSource(file){return readFileSync(file==="-"?0:file,"utf8")}function fromJson(value,where){if(typeof value==="string")return{text:value};if(!value||typeof value!=="object"||Array.isArray(value))throw new CliError("usage",`${where}: expected a prompt string or an object with "text"`);const item=value;if(typeof item.text!=="string")throw new CliError("usage",`${where}: "text" must be a string`);const topic=item.topicId??item.topic;const tags=item.tagIds??item.tags;if(topic!==void 0&&typeof topic!=="string")throw new CliError("usage",`${where}: "topicId" must be a string`);if(tags!==void 0&&!(Array.isArray(tags)&&tags.every(t=>typeof t==="string")))throw new CliError("usage",`${where}: "tagIds" must be a string array`);if(item.trackingType!==void 0&&typeof item.trackingType!=="string")throw new CliError("usage",`${where}: "trackingType" must be a string`);return{text:item.text,topic,trackingType:item.trackingType,tags}}function parsePromptInput(content){const trimmed=content.trim();if(trimmed.startsWith("[")){let parsed;try{parsed=JSON.parse(trimmed)}catch(error){throw new CliError("usage",`prompt file is not valid JSON: ${error.message}`)}return parsed.map((item,i)=>fromJson(item,`item ${i}`))}return content.split(/\r?\n/).map((line,i)=>({line:line.trim(),number:i+1})).filter(({line})=>line).map(({line,number})=>{if(!line.startsWith("{"))return{text:line};try{return fromJson(JSON.parse(line),`line ${number}`)}catch(error){if(error instanceof CliError)throw error;throw new CliError("usage",`line ${number} is not valid JSON: ${error.message}`)}})}function trackingType(value){if(value===void 0)return void 0;if(TRACKING_TYPES.includes(value))return value;throw new CliError("usage",`trackingType must be one of ${TRACKING_TYPES.join(", ")} (got "${value}")`)}async function resolvePrompts(scope,raw,defaults){const topics=new Map;for(const[i,item]of raw.entries()){const topic=item.topic??defaults.topic;if(!topic)throw new CliError("usage",`prompt ${i} has no topic. Pass --topic <id|slug|title> or set "topicId" per item.`);if(!topics.has(topic))topics.set(topic,await resolveTopicId(scope,topic))}const tagNames=[...new Set([...defaults.tags,...raw.flatMap(item=>item.tags??[])])];const tagIds=new Map;const resolved=await resolveTagIds(scope,tagNames);tagNames.forEach((name,i)=>tagIds.set(name,resolved[i]));return raw.map(item=>{const tags=[...new Set([...defaults.tags,...item.tags??[]])];const type=trackingType(item.trackingType??defaults.trackingType);return{text:item.text,topicId:topics.get(item.topic??defaults.topic),...type?{trackingType:type}:{},...tags.length?{tagIds:tags.map(t=>tagIds.get(t))}:{}}})}var normalized=text=>text.trim().toLowerCase().replace(/\s+/g," ");function outcome(prompts,responses,dryRun){const results=responses.flatMap((response,chunk)=>response.results.map(r=>{const index=chunk*CHAT_BATCH_SIZE+r.index;return{index,status:r.status,id:r.prompt?.id??null,text:prompts[index].text,prompt:r.prompt??null}}));const summary={created:0,already_exists:0,would_create:0};for(const r of results)summary[r.status]++;return{results,dryRun,remaining:responses.at(-1)?.remaining??null,summary}}function printOutcome(result,cmd){if(resolveOutput(cmd)==="json")return printJson(result);printItems(result.results.map(r=>({index:r.index,status:r.status,id:r.id,text:truncate(r.text,70)})),"table");const{created,already_exists,would_create}=result.summary;console.error(`${result.dryRun?"dry run: ":""}${result.dryRun?`${would_create} would be created`:`${created} created`}, ${already_exists} already existed${result.remaining===null?"":`; plan allows ${result.remaining} more`}.`)}async function addPrompts(scope,prompts,dryRun,onPartial){const path=chatPath(...scope.base,"prompts");const batches=chunks(prompts);const send=(batch,dry)=>scope.client.post(path,{prompts:batch},{dryRun:dry||void 0});if(!dryRun&&batches.length>1){const preview=[];for(const batch of batches)preview.push(await send(batch,true));const fresh=new Set;preview.forEach((response,chunk)=>{for(const r of response.results)if(r.status==="would_create")fresh.add(normalized(prompts[chunk*CHAT_BATCH_SIZE+r.index].text))});const first=preview[0];const remaining=first?.remaining===null||first?.remaining===void 0?null:first.remaining+first.results.filter(r=>r.status==="would_create").length;if(remaining!==null&&fresh.size>remaining)throw new CliError("insufficient_credits",`limit_exceeded: ${fresh.size} new prompts exceed the ${remaining} your plan allows; nothing was created.`,{apiCode:"limit_exceeded",details:{remaining,wouldCreate:fresh.size}})}const responses=[];for(const batch of batches){try{responses.push(await send(batch,dryRun))}catch(error){if(responses.length&&onPartial)onPartial(outcome(prompts,responses,dryRun));if(responses.length&&error instanceof CliError)throw new CliError(error.code,`${error.message} \u2014 the first ${responses.length*CHAT_BATCH_SIZE} prompts were already submitted; rerunning is safe (existing prompts report already_exists).`,{apiCode:error.apiCode,details:error.details,hint:error.hint});throw error}}return outcome(prompts,responses,dryRun)}function readIds(args,file){const ids=[...args,...file?readSource(file).split(/[\s,]+/).filter(Boolean):[]];if(!ids.length)throw new CliError("usage","pass prompt ids as arguments or with --file");return[...new Set(ids)]}function registerPrompts(chat){const prompts=chat.command("prompts").description("Tracked prompts in a category");addPageOptions(withCategory(prompts.command("list",{isDefault:true})).description("List prompts").option("--topic <topic>","only this topic (id, slug, or title)").option("--tag <tags>","only prompts with any of these tags (ids or names)",collect2).option("--tracking-type <type>","visibility | branded_own | all (default all)")).action(async(opts,cmd)=>{const scope=await categoryScope(cmd);const page=await fetchPages(scope.client,chatPath(...scope.base,"prompts"),{topic:opts.topic?await resolveTopicId(scope,opts.topic):void 0,tag:opts.tag?.length?await resolveTagIds(scope,opts.tag):void 0,trackingType:opts.trackingType},opts);printPage(page,resolveOutput(cmd),p=>({id:p.id,topic:p.topicId,type:p.trackingType,tags:p.tagIds.join(","),firstAnswer:p.firstAnswerDay,text:truncate(p.text,70)}))});withCategory(prompts.command("add [texts...]")).description("Add prompts from arguments or --file (one per line, JSON Lines, or a JSON array); existing texts report already_exists").option("--topic <topic>","topic for items without their own (id, slug, or title)").option("-f, --file <path>","read prompts from a file ('-' for stdin)").option("--tag <tags>","tags for every prompt (ids or names; repeatable)",collect2).option("--tracking-type <type>","visibility (default) | branded_own, for items without their own").option("--dry-run","report what would be created without creating").addHelpText("after",`
|
|
13
13
|
Items in a file may be strings or objects:
|
|
14
14
|
{"text": "...", "topicId": "top_1\u2026", "trackingType": "visibility", "tagIds": ["tag_1\u2026"]}
|
|
15
15
|
Inputs over ${CHAT_BATCH_SIZE} prompts are sent in chunks after a dry run of every chunk.
|
|
16
|
-
Each item reports created, already_exists, or would_create (with --dry-run).`).action(async(texts,opts,cmd)=>{const raw=[...texts.map(text=>({text})),...opts.file?parsePromptInput(readSource(opts.file)):[]];if(!raw.length)throw new CliError("usage","no prompts given; pass them as arguments or with --file");const scope=await categoryScope(cmd);const resolved=await resolvePrompts(scope,raw,{topic:opts.topic,trackingType:opts.trackingType,tags:opts.tag??[]});const result=await addPrompts(scope,resolved,opts.dryRun??false,partial=>printOutcome(partial,cmd));printOutcome(result,cmd)});withCategory(prompts.command("move [prompts...]")).description("Move prompts to another topic (ids as arguments or --file), in batches of 500").requiredOption("--topic <topic>","destination topic (id, slug, or title)").option("-f, --file <path>","read prompt ids from a file ('-' for stdin)").action(async(args,opts,cmd)=>{const ids=readIds(args,opts.file);const scope=await categoryScope(cmd);const topicId=await resolveTopicId(scope,opts.topic);const updated=[];const notFound=[];for(const batch of chunks(ids)){const res=await scope.client.patch(chatPath(...scope.base,"prompts"),{ids:batch,topicId});updated.push(...res.updated);notFound.push(...res.notFound)}if(resolveOutput(cmd)==="json")return printJson({updated,notFound,topicId});console.log(`Moved ${updated.length} prompt(s) to ${topicId}${notFound.length?`; not found: ${notFound.join(", ")}`:""}`)});for(const[name,method,verb]of[["tag","post","Tagged"],["untag","delete","Untagged"]])withCategory(prompts.command(`${name} [prompts...]`)).description(`${name==="tag"?"Apply":"Remove"} tags ${name==="tag"?"to":"from"} prompts (ids as arguments or --file)`).requiredOption("--tag <tags>","tag ids or names (comma-separated or repeatable)",collect2).option("-f, --file <path>","read prompt ids from a file ('-' for stdin)").action(async(args,opts,cmd)=>{const ids=readIds(args,opts.file);const scope=await categoryScope(cmd);const tagIds=await resolveTagIds(scope,opts.tag);const path=chatPath(...scope.base,"prompts","tags");const promptIds=[];for(const batch of chunks(ids)){const body={promptIds:batch,tagIds};const res=method==="post"?await scope.client.post(path,body):await scope.client.delete(path,void 0,body);promptIds.push(...res.promptIds)}if(resolveOutput(cmd)==="json")return printJson({promptIds,tagIds});console.log(`${verb} ${promptIds.length} prompt(s) with ${tagIds.join(", ")}`)})}import{Option as Option2}from"commander";var query_exports={};__export(query_exports,{noteFreshness:()=>noteFreshness,parseOrderBy:()=>parseOrderBy,pct:()=>pct,readStdin:()=>readStdin2,register:()=>register5});import{join as join4}from"path";var dimension=(sql,description,type="string")=>({sql,description,type});var metric=(sql,description,type="number")=>({sql,description,type});var common={run:dimension("s.id","Session ID"),batch:dimension("s.batch_id","Launch batch ID"),eval:dimension("CASE WHEN s.measurement_kind = 'eval' THEN s.measurement_id END","Saved eval ID"),preference:dimension("CASE WHEN s.measurement_kind = 'preference' THEN s.measurement_id END","Saved preference prompt ID"),measurement_revision:dimension("s.measurement_revision","Frozen measurement revision","number"),agent:dimension("s.agent","Coding agent"),model:dimension("s.model","Agent model"),date:dimension("date(s.created_at)","Session creation date (UTC)","date")};var catalogs=[{name:"runs",grain:"One local session (current attempt)",from:"sessions s",dimensions:{...common,case:dimension("s.case_path","File-based case path, when applicable"),outcome:dimension("s.outcome","Execution outcome; this is not an eval verdict"),analyzed:dimension("s.current_analysis_id IS NOT NULL","Whether this attempt has published analysis","boolean"),analysis_id:dimension("s.current_analysis_id","Current published analysis ID")},metrics:{run_count:metric("COUNT(*)","Number of sessions"),analyzed_runs:metric("COALESCE(SUM(CASE WHEN s.current_analysis_id IS NOT NULL THEN 1 ELSE 0 END),0)","Sessions with published analysis"),avg_turns:metric("AVG(s.turns)","Average recorded turns; unknown values excluded"),avg_duration_ms:metric("AVG(s.duration_ms)","Average recorded duration in milliseconds; unknown values excluded"),input_tokens:metric("SUM(s.input_tokens)","Recorded input tokens; null when none have been recorded"),output_tokens:metric("SUM(s.output_tokens)","Recorded output tokens; null when none have been recorded")}},{name:"findings",grain:"One finding in a session's current published analysis",from:"sessions s JOIN session_analyses a ON a.id = s.current_analysis_id JOIN session_findings f ON f.analysis_id = a.id",dimensions:{...common,analysis_id:dimension("a.id","Current published analysis ID"),kind:dimension("f.kind","FRICTION, OBJECTION, or OPPORTUNITY"),subject:dimension("f.subject","Product or surface discussed"),label:dimension("f.label","Finding title; matching labels do not imply a shared theme"),basis:dimension("f.basis","OBSERVED, AGENT_CLAIM, or INFERRED"),effect:dimension("f.effect","Effect on the session"),cause:dimension("f.cause","Attributed cause")},metrics:{finding_count:metric("COUNT(*)","Findings in current published analyses"),affected_runs:metric("COUNT(DISTINCT s.id)","Distinct sessions with matching findings")}}];function localQueryMetadata(){return{datasets:catalogs.map(catalog=>({name:catalog.name,grain:catalog.grain,dimensions:Object.entries(catalog.dimensions).map(([name,field])=>({name,type:field.type,description:field.description})),metrics:Object.entries(catalog.metrics).map(([name,field])=>({name,type:field.type,description:field.description}))}))}}function queryLocal(gaugeDir,input){const compiled=compileLocalQuery(input);const store=openLocalStore(gaugeDir);store.db.exec("BEGIN");try{if(compiled.catalog.name==="findings"){const versions=store.db.prepare(compiled.findingVersionsSql).all(...compiled.coverageParams);const unsupported=versions.find(row2=>row2.schema_version!==ANALYSIS_SCHEMA_VERSION);if(unsupported)throw new CliError("usage",`A selected session's analysis uses schema version ${unsupported.schema_version}; this CLI supports ${ANALYSIS_SCHEMA_VERSION}. Upgrade Gauge before querying these findings. Your data is intact.`)}const rows=store.db.prepare(compiled.sql).all(...compiled.params);for(const row2 of rows)for(const name of compiled.dimensions){if(compiled.catalog.dimensions[name].type==="boolean"&&row2[name]!==null)row2[name]=Boolean(row2[name])}const coverage=store.db.prepare(compiled.coverageSql).get(...compiled.coverageParams);const response={columns:[...compiled.dimensions.map(key=>({key,kind:"dimension"})),...compiled.metrics.map(key=>({key,kind:"metric"}))],rows,...input.explain?{sql:compiled.sql}:{},freshness:{available:true,pending:0,watermark:coverage.watermark},analysisCoverage:{sessions:Number(coverage.total),analyzedSessions:Number(coverage.analyzed),unanalyzedSessions:Number(coverage.total)-Number(coverage.analyzed)}};store.db.exec("COMMIT");return response}catch(error){store.db.exec("ROLLBACK");throw error}finally{store.close()}}function compileLocalQuery(input){if(!input||typeof input!=="object"||Array.isArray(input))throw new CliError("usage","A query must be a JSON object.");const topKeys=new Set(["dataset","dimensions","metrics","filters","where","dateRange","orderBy","limit","granularity","explain"]);for(const key of Object.keys(input))if(!topKeys.has(key))throw new CliError("usage",`Unknown local query option "${key}". See gauge query fields --local --dataset runs.`);const name=input.dataset??"runs";const catalog=catalogs.find(item=>item.name===name);if(!catalog)throw new CliError("usage",`Dataset "${name}" is not available for local runs. Available: runs, findings.`,{hint:"Run `gauge query datasets --local` to see available datasets, or use --hosted to query runs on Gauge."});const dimensions=input.dimensions??[];const metrics=input.metrics;if(!Array.isArray(dimensions)||dimensions.some(key=>typeof key!=="string"))throw new CliError("usage","Query dimensions must be an array of field names.");if(!Array.isArray(metrics)||metrics.length===0||metrics.some(key=>typeof key!=="string"))throw new CliError("usage","At least one metric is required.");if(new Set([...dimensions,...metrics]).size!==dimensions.length+metrics.length)throw new CliError("usage","Select each query field only once.");const unsupported=(kind,field)=>{throw new CliError("usage",`${kind} "${field}" is not supported by the local ${name} dataset.`,{hint:`Run \`gauge query fields --dataset ${name} --local\` to see which fields you can query for local runs.`})};const granularity=input.granularity??"day";if(!["day","week","month"].includes(granularity))throw new CliError("usage","Granularity must be day, week, or month.");const dateSql=granularity==="month"?"date(s.created_at,'start of month')":granularity==="week"?"date(s.created_at,'-' || ((CAST(strftime('%w',s.created_at) AS INTEGER) + 6) % 7) || ' days')":"date(s.created_at)";const expr=key=>key==="date"?dateSql:Object.hasOwn(catalog.dimensions,key)?catalog.dimensions[key].sql:unsupported("Dimension",key);for(const key of dimensions)expr(key);for(const key of metrics)if(!Object.hasOwn(catalog.metrics,key))unsupported("Metric",key);const limit=input.limit??50;if(!Number.isInteger(limit)||limit<1||limit>1e3)throw new CliError("usage","Query limit must be an integer from 1 to 1000.");const params=[];const predicates=[];const coveragePredicates=[];const coverageParams=[];const bind=value=>{if(typeof value!=="string"&&typeof value!=="boolean"&&(typeof value!=="number"||!Number.isFinite(value)))throw new CliError("usage","Query filter values must be strings, booleans, or finite numbers.");params.push(typeof value==="boolean"?Number(value):value);return"?"};const where=input.where??[];if(!Array.isArray(where))throw new CliError("usage","Query where must be an array of predicates.");for(const clause of where){const paramStart=params.length;const predicateStart=predicates.length;if(!clause||typeof clause!=="object"||Object.keys(clause).some(key=>!["dimension","op","value"].includes(key)))throw new CliError("usage","Each where predicate requires dimension, op, and value.");const sql2=expr(clause.dimension);if(clause.op==="in"||clause.op==="notIn"){if(!Array.isArray(clause.value)||clause.value.length===0)throw new CliError("usage",`${clause.op} requires a nonempty value array.`);predicates.push(`${sql2} ${clause.op==="in"?"IN":"NOT IN"} (${clause.value.map(bind).join(",")})`)}else{if(Array.isArray(clause.value))throw new CliError("usage",`${clause.op} requires a scalar value.`);if(clause.op==="contains"){if(typeof clause.value!=="string")throw new CliError("usage","contains requires a string value.");predicates.push(`instr(lower(CAST(${sql2} AS TEXT)),lower(${bind(clause.value)})) > 0`)}else{const operators={eq:"=",neq:"!=",gt:">",gte:">=",lt:"<",lte:"<="};const op=operators[clause.op];if(!op)throw new CliError("usage",`Unsupported filter operator "${clause.op}".`);let value=clause.value;if(catalog.dimensions[clause.dimension].type==="number"&&typeof value==="string"&&value.trim()&&Number.isFinite(Number(value)))value=Number(value);if(catalog.dimensions[clause.dimension].type==="boolean"&&typeof value==="string"&&["true","false"].includes(value))value=value==="true";predicates.push(`${sql2} ${op} ${bind(value)}`)}}if(Object.hasOwn(common,clause.dimension)||name==="runs"){coveragePredicates.push(...predicates.slice(predicateStart));coverageParams.push(...params.slice(paramStart))}}const filterParamStart=params.length;const filterPredicateStart=predicates.length;const filters=input.filters??{};if(!filters||typeof filters!=="object"||Array.isArray(filters))throw new CliError("usage","Query filters must be an object.");const supportedFilters={agent:"s.agent",model:"s.model",prompt:"s.measurement_id",eval:"CASE WHEN s.measurement_kind = 'eval' THEN s.measurement_id END",visibilityPrompt:"CASE WHEN s.measurement_kind = 'preference' THEN s.measurement_id END"};for(const[key,value]of Object.entries(filters)){if(value===void 0)continue;if(!Object.hasOwn(supportedFilters,key))unsupported("Filter",key);const values=key==="prompt"?[value]:value;if(!Array.isArray(values)||values.length===0||values.some(item=>typeof item!=="string"))throw new CliError("usage",`Filter ${key} requires ${key==="prompt"?"a string":"a nonempty array of strings"}.`);predicates.push(`${supportedFilters[key]} IN (${values.map(bind).join(",")})`)}const range=input.dateRange;if(range!==void 0){if(!range||typeof range!=="object"||Array.isArray(range))throw new CliError("usage","Query dateRange must be an object.");for(const key of Object.keys(range))if(!["since","until","lastDays"].includes(key))throw new CliError("usage",`Unknown dateRange field ${key}.`);if(range.since!==void 0&&range.lastDays!==void 0)throw new CliError("usage","Use dateRange.since or dateRange.lastDays, not both.");if(range.lastDays!==void 0&&(!Number.isInteger(range.lastDays)||range.lastDays<=0))throw new CliError("usage","lastDays must be a positive integer.");const since=range.since!==void 0?dateValue(range.since):range.lastDays!==void 0?new Date(Date.now()-range.lastDays*864e5).toISOString():null;const until=range.until!==void 0?dateValue(range.until):null;if(since&&until&&since>until)throw new CliError("usage","Query since must be before until.");if(since)predicates.push(`julianday(s.created_at) >= julianday(${bind(since)})`);if(until)predicates.push(`julianday(s.created_at) <= julianday(${bind(until)})`)}coveragePredicates.push(...predicates.slice(filterPredicateStart));coverageParams.push(...params.slice(filterParamStart));const selected=new Set([...dimensions,...metrics]);const order=input.orderBy??[];if(!Array.isArray(order))throw new CliError("usage","Query orderBy must be an array.");for(const item of order){if(!item||!selected.has(item.key))throw new CliError("usage",`Order field "${item?.key}" must be one of the selected dimensions or metrics.`);if(item.dir!==void 0&&item.dir!=="asc"&&item.dir!=="desc")throw new CliError("usage","Order direction must be asc or desc.")}const columns=[...dimensions.map(key=>`${expr(key)} AS "${key}"`),...metrics.map(key=>`${catalog.metrics[key].sql} AS "${key}"`)];const sql=`SELECT ${columns.join(", ")} FROM ${catalog.from}${predicates.length?` WHERE ${predicates.map(predicate=>`(${predicate})`).join(" AND ")}`:""}${dimensions.length?` GROUP BY ${dimensions.map(expr).join(", ")}`:""}${order.length?` ORDER BY ${order.map(item=>`"${item.key}" ${(item.dir??"desc").toUpperCase()}`).join(", ")}`:""} LIMIT ${limit}`;const coverageWhere=coveragePredicates.length?` WHERE ${coveragePredicates.map(predicate=>`(${predicate})`).join(" AND ")}`:"";const coverageSql=`SELECT COUNT(*) AS total,COUNT(current_analysis_id) AS analyzed,MAX(COALESCE(ended_at,created_at)) AS watermark FROM sessions s${coverageWhere}`;const findingVersionsSql=`SELECT DISTINCT a.schema_version FROM sessions s JOIN session_analyses a ON a.id = s.current_analysis_id${coverageWhere}`;return{sql,params,coverageSql,coverageParams,findingVersionsSql,catalog,dimensions,metrics}}function dateValue(value){if(typeof value!=="string")throw new CliError("usage","Date bounds must be ISO dates or relative windows such as 24h or 7d.");const match=/^(\d+)([mhd])$/.exec(value.trim());const date=match?new Date(Date.now()-Number(match[1])*({m:6e4,h:36e5,d:864e5}[match[2]]??0)):new Date(value);if(!Number.isFinite(date.getTime()))throw new CliError("usage",`Invalid date bound "${value}".`);return date.toISOString()}import{createReadStream,existsSync,realpathSync}from"fs";import{isAbsolute,join,relative}from"path";import{createInterface as createInterface2}from"readline";async function searchLocalTranscripts(gaugeDir,text,options={}){if(!text.trim())throw new CliError("usage","Transcript search requires nonempty text.");if(options.eval&&options.preference)throw new CliError("usage","Select either --eval or --preference for transcript search.");const limit=options.limit??50;if(!Number.isInteger(limit)||limit<1||limit>1e3)throw new CliError("usage","Search limit must be an integer from 1 to 1000.");const store=openLocalStore(gaugeDir);const sessions=(()=>{try{return store.listSessions({measurementId:options.eval??options.preference,measurementKind:options.eval?"eval":options.preference?"preference":void 0,batchId:options.batch}).filter(session=>!options.since||Date.parse(session.createdAt)>=options.since.getTime())}finally{store.close()}})();const result={matches:[],selectedSessions:sessions.length,scannedSessions:0,missingArtifacts:[],truncated:false};const needle=text.toLowerCase();for(const session of sessions){if(!/^[A-Za-z0-9_-]+$/.test(session.id))throw new CliError("usage",`Invalid local run ID: ${session.id}`);const attemptDir=join(gaugeDir,"runs",session.id,"attempts",String(session.attempt));const path=join(attemptDir,"session","events.ndjson");if(!existsSync(path)){result.missingArtifacts.push(session.id);continue}const rel=relative(realpathSync(attemptDir),realpathSync(path));if(isAbsolute(rel)||rel===".."||rel.startsWith("../")||rel.startsWith("..\\"))throw new CliError("usage",`Run ${session.id} transcript resolves outside its attempt.`);result.scannedSessions++;const stream=createReadStream(path,{encoding:"utf8"});const lines=createInterface2({input:stream,crlfDelay:Infinity});let lineNumber=0;try{for await(const line of lines){lineNumber++;let event;try{event=JSON.parse(line)}catch{continue}if(!event||typeof event!=="object")continue;const content=typeof event.payload==="string"?event.payload:JSON.stringify(event.payload??event);const index=content.toLowerCase().indexOf(needle);if(index<0)continue;if(result.matches.length===limit){result.truncated=true;return result}const start=Math.max(0,index-100);const end=Math.min(content.length,Math.max(index+needle.length+200,start+400));result.matches.push({run:session.id,attempt:session.attempt,turn:typeof event.turn==="number"?event.turn:null,artifact:"session/events.ndjson",line:lineNumber,snippet:`${start?"\u2026":""}${content.slice(start,end)}${end<content.length?"\u2026":""}`})}}finally{lines.close();stream.destroy()}}return result}import{createHash as createHash2,randomUUID as randomUUID2}from"crypto";import{existsSync as existsSync2,readFileSync as readFileSync2,realpathSync as realpathSync2,statSync}from"fs";import{isAbsolute as isAbsolute2,join as join2,relative as relative2,resolve}from"path";function saveLocalAnalysis(gaugeDir,runId,value){if(typeof value==="object"&&value!==null&&"schemaVersion"in value&&value.schemaVersion!==ANALYSIS_SCHEMA_VERSION){throw new CliError("usage",`Analysis schema version ${String(value.schemaVersion)} is not supported. This CLI accepts version ${ANALYSIS_SCHEMA_VERSION}; upgrade Gauge for newer analyses. No data was changed.`)}const parsed=analysisSubmissionSchema.safeParse(value);if(!parsed.success)throw new CliError("usage",`Invalid analysis: ${parsed.error.issues.map(issue=>`${issue.path.join(".")||"analysis"}: ${issue.message}`).join("; ")}`,{hint:"Use gauge run analysis schema for the analysis submission contract."});const submission=parsed.data;const store=openLocalStore(gaugeDir);try{const session=store.getSession(runId);if(!session)throw new CliError("usage",`Local run ${runId} was not found.`,{hint:"Use gauge run list --local, or repair session summaries if its artifacts exist."});if(!["completed","failed","canceled","interrupted"].includes(session.outcome))throw new CliError("usage",`Run ${runId} is ${session.outcome}. Wait until it finishes before publishing analysis.`);if(session.attempt!==submission.attempt)throw new CliError("usage",`Run ${runId} is on attempt ${session.attempt}; the submitted analysis targets attempt ${submission.attempt}.`);validateCriteria(session,submission);const attemptDirectory=safeAttemptDirectory(gaugeDir,runId,submission.attempt);const files=new Map;const refs=[...submission.findings.flatMap(finding=>finding.evidence),...submission.products.flatMap(product=>Object.values(product.evidence).flat()),...submission.criteria.flatMap(criterion=>criterion.evidence)];for(const ref of refs)validateEvidence(attemptDirectory,submission.attempt,ref,files);const hash=createHash2("sha256");for(const[path,contents]of[...files].sort(([a],[b])=>a.localeCompare(b)))hash.update(path).update("\0").update(contents).update("\0");const evidenceDigest=`sha256:${hash.digest("hex")}`;return store.transaction(()=>{const currentSession=store.getSession(runId);if(currentSession.attempt!==submission.attempt)throw new CliError("usage","The run advanced to another attempt while validating analysis. Read its new evidence before publishing.");const latest=store.db.prepare("SELECT MAX(revision) AS revision FROM session_analyses WHERE run_id = ?").get(runId);const analysis={...submission,id:`a_${randomUUID2()}`,runId,revision:Number(latest?.revision??0)+1,createdAt:new Date().toISOString(),evidenceDigest,findings:submission.findings.map(finding=>({...finding,id:`f_${randomUUID2()}`}))};store.db.prepare("INSERT INTO session_analyses VALUES (?,?,?,?,?,?,?,?,?)").run(analysis.id,runId,analysis.attempt,analysis.revision,analysis.schemaVersion,analysis.instructionsVersion,evidenceDigest,JSON.stringify(analysis),analysis.createdAt);const findingStatement=store.db.prepare("INSERT INTO session_findings VALUES (?,?,?,?,?,?,?,?,?,?,?)");for(const finding of analysis.findings)findingStatement.run(finding.id,analysis.id,finding.kind,finding.subject,finding.label,finding.detail,finding.basis,finding.effect,finding.cause,JSON.stringify(finding.tags),JSON.stringify(finding.evidence));const productStatement=store.db.prepare("INSERT INTO session_products VALUES (?,?,?,?)");analysis.products.forEach((product,position)=>productStatement.run(analysis.id,position,product.name,JSON.stringify(product)));const criterionStatement=store.db.prepare("INSERT INTO criterion_results VALUES (?,?,?,?,?)");for(const criterion of analysis.criteria)criterionStatement.run(analysis.id,criterion.criterionId,criterion.outcome,criterion.label,JSON.stringify(criterion.evidence));store.db.prepare("UPDATE sessions SET current_analysis_id = ? WHERE id = ?").run(analysis.id,runId);return analysis})}finally{store.close()}}function getLocalAnalysis(gaugeDir,runId){const store=openLocalStore(gaugeDir);try{return readCurrentAnalysis(store,runId)}finally{store.close()}}function readCurrentAnalysis(store,runId){const row2=store.db.prepare("SELECT a.schema_version,a.payload FROM session_analyses a JOIN sessions s ON s.current_analysis_id = a.id WHERE s.id = ?").get(runId);if(!row2)return null;if(row2.schema_version!==ANALYSIS_SCHEMA_VERSION)throw new CliError("usage",`Stored analysis uses schema version ${row2.schema_version}; upgrade Gauge to read it. Your data is intact.`);return JSON.parse(String(row2.payload))}function validateCriteria(session,submission){const configuration=session.configuration??{};const item=configuration.case;const criteria=item?.criteria??configuration.criteria??[];if(!Array.isArray(criteria))throw new CliError("usage","This run's frozen criteria are invalid; repair its session summary before analysis.");const expected=criteria.map(criterion=>{if(!criterion||typeof criterion.id!=="string"||!criterion.id)throw new CliError("usage","This run has a frozen criterion without an ID. Criterion IDs must be assigned before execution.");return criterion.id});const supplied=new Set(submission.criteria.map(criterion=>criterion.criterionId));const missing=expected.filter(id=>!supplied.has(id));const extra=[...supplied].filter(id=>!expected.includes(id));if(missing.length||extra.length)throw new CliError("usage",`Analysis must cover exactly the run's frozen criteria.${missing.length?` Missing: ${missing.join(", ")}.`:""}${extra.length?` Unknown: ${extra.join(", ")}.`:""} Use UNASSESSABLE with an explanation when evidence is insufficient.`)}function safeAttemptDirectory(gaugeDir,runId,attempt){if(!/^[A-Za-z0-9_-]+$/.test(runId))throw new CliError("usage","Invalid local run ID.");const directory=join2(gaugeDir,"runs",runId,"attempts",String(attempt));if(!existsSync2(directory))throw new CliError("usage",`Run evidence is missing: ${directory}`);return realpathSync2(directory)}function readEvidence(directory,path,files){const fullPath=resolve(directory,path);if(!isInside(directory,fullPath))throw new CliError("usage",`Evidence path ${path} leaves the run's attempt directory.`);if(!existsSync2(fullPath)||!statSync(fullPath).isFile())throw new CliError("usage",`Evidence artifact ${path} does not exist in this attempt.`);if(!isInside(directory,realpathSync2(fullPath)))throw new CliError("usage",`Evidence artifact ${path} resolves outside this attempt.`);const contents=files.get(path)??readFileSync2(fullPath,"utf8");files.set(path,contents);return contents}function isInside(directory,path){const rel=relative2(directory,path);return rel!==".."&&!rel.startsWith(`..${process.platform==="win32"?"\\":"/"}`)&&!isAbsolute2(rel)}function validateEvidence(directory,attempt,ref,files){if(ref.kind==="diff"){const diff2=readEvidence(directory,"session/diff.patch",files);const paths=new Set(diff2.split("\n").flatMap(line=>{const match=/^(?:\+\+\+|---) [ab]\/(.*)$/.exec(line);return match?[match[1].split(" ")[0]]:[]}));if(!paths.has(ref.path))throw new CliError("usage",`Diff evidence path ${ref.path} does not appear in this attempt's diff.patch.`);return}if(ref.kind==="artifact"&&ref.attempt!==void 0&&ref.attempt!==attempt)throw new CliError("usage",`Evidence references attempt ${ref.attempt}, but this analysis targets attempt ${attempt}.`);const artifact=ref.kind==="trace"?"session/events.ndjson":ref.artifact;const contents=readEvidence(directory,artifact,files);const lines=contents.split(/\r?\n/);if(lines.at(-1)==="")lines.pop();if(ref.kind==="artifact"&&ref.line!==void 0&&(ref.line>lines.length||(ref.endLine??ref.line)>lines.length))throw new CliError("usage",`Evidence ${artifact}:${ref.line}${ref.endLine?`-${ref.endLine}`:""} exceeds the artifact's ${lines.length} lines.`);if(ref.seq!==void 0){const hasSequence=lines.some(line=>{try{return JSON.parse(line).seq===ref.seq}catch{return false}});if(!hasSequence)throw new CliError("usage",`Evidence sequence ${ref.seq} does not exist in ${artifact}. Copy an event's recorded seq; do not use its array index or turn.`,{hint:`For evidence without seq, cite an original artifact line, for example ${JSON.stringify({kind:"artifact",artifact,attempt,line:1})}. Line numbers start at 1; choose the line containing your evidence.`})}}function renderLocalAnalysisMarkdown(analysis){const lines=[`# Session analysis: ${analysis.runId}`,"",`Analysis ${analysis.id}, revision ${analysis.revision}; attempt ${analysis.attempt}.`,`Assessment: ${analysis.assessability}.`,""];if(analysis.limitation)lines.push(`Limitation: ${analysis.limitation}`,"");for(const finding of analysis.findings)lines.push(`## ${finding.kind}: ${finding.label}`,"",`${finding.subject} \xB7 ${finding.effect} \xB7 ${finding.basis}`,"",finding.detail,"",`Evidence: ${finding.evidence.map(formatEvidence).join(", ")}`,"");if(!analysis.findings.length)lines.push("No findings were reported in this analysis.","");if(analysis.criteria.length){lines.push("## Criteria","");for(const criterion of analysis.criteria)lines.push(`- ${criterion.criterionId}: ${criterion.outcome} \u2014 ${criterion.label}`);lines.push("")}if(analysis.products.length){lines.push("## Products","");for(const product of analysis.products)lines.push(`- ${product.name}: ${product.chosen?"chosen":product.rejected?"rejected":product.considered?"considered":"mentioned"} \u2014 ${product.reason}`);lines.push("")}return lines.join("\n")}function formatEvidence(ref){if(ref.kind==="trace")return`trace ${ref.seq}`;if(ref.kind==="diff")return`diff ${ref.path}`;return`${ref.artifact}${ref.line!==void 0?`:${ref.line}${ref.endLine!==void 0?`-${ref.endLine}`:""}`:` seq ${ref.seq}`}`}import{join as join3}from"path";function listLocalRuns(gaugeDir,filter={}){const store=openLocalStore(gaugeDir);try{return store.listSessions().map(session=>({...session,evalId:session.measurementKind==="eval"?session.measurementId??null:null,preferenceId:session.measurementKind==="preference"?session.measurementId??null:null,endedBy:session.endedBy??"",endedAt:session.endedAt??"",attemptDir:join3(gaugeDir,"runs",session.id,"attempts",String(session.attempt)),resultsDir:join3(gaugeDir,"runs",session.id,"results")})).filter(run=>matches(run,filter)).map(run=>{const analysis=readCurrentAnalysis(store,run.id);if(!analysis)return run;const session=store.getSession(run.id);const item=session.configuration?.case;const names=new Map((item?.criteria??[]).map(criterion=>[criterion.id,criterion.name]));return{...run,criterionResults:analysis.criteria.map(criterion=>({id:criterion.criterionId,name:names.get(criterion.criterionId)??criterion.criterionId,outcome:criterion.outcome}))}})}finally{store.close()}}function matches(run,filter){return(!filter.ids?.length||filter.ids.includes(run.id))&&(!filter.batch||run.batchId===filter.batch)&&(!filter.evalId||run.evalId===filter.evalId)&&(!filter.preferenceId||run.preferenceId===filter.preferenceId)&&(filter.measurementRevision===void 0||run.measurementRevision===filter.measurementRevision)&&(!filter.agent||run.agent===filter.agent)&&(!filter.model||run.model===filter.model)&&(!filter.outcome||filter.outcome.split(",").includes(run.outcome))&&(!filter.since||Date.parse(run.createdAt??run.endedAt)>=filter.since.getTime())&&(!filter.until||Date.parse(run.createdAt??run.endedAt)<=filter.until.getTime())&&(!filter.case||run.case===filter.case)&&(filter.analyzed===void 0||run.analyzed===filter.analyzed)}var UNIT_MS={m:6e4,h:36e5,d:864e5};function parseSince(window,now=Date.now(),flag="--since"){const relative7=/^(\d+)([mhd])$/.exec(window.trim());if(relative7)return new Date(now-Number(relative7[1])*UNIT_MS[relative7[2]]);const at=Date.parse(window);if(Number.isNaN(at))throw new CliError("usage",`invalid ${flag} "${window}" (expected 24h, 7d, or an ISO date)`);return new Date(at)}var collect3=(value,prev=[])=>[...prev,value];var csv2=raw=>raw.split(",").map(s=>s.trim()).filter(Boolean);function parseWhere(expr){const inMatch=expr.match(/^\s*(\w+)\s+(in|contains)\s+(.*)$/i);if(inMatch){const[,dimension3,opRaw,rest2]=inMatch;const op=opRaw.toLowerCase();if(op==="in"){return{dimension:dimension3,op:"in",value:csv2(rest2)}}return{dimension:dimension3,op:"contains",value:rest2.trim()}}const opMatch=expr.match(/^\s*(\w+)\s*(!=|>=|<=|=|>|<)\s*(.*)$/);if(!opMatch){throw new CliError("usage",`Could not parse --where "${expr}". Use e.g. "action=install", "ecosystem in npm,pypi", or "name contains supa".`)}const[,dimension2,sym,rest]=opMatch;const opFor={"=":"eq","!=":"neq",">":"gt",">=":"gte","<":"lt","<=":"lte"};return{dimension:dimension2,op:opFor[sym],value:rest.trim()}}function parseOrderBy(raw){const[key,dir]=raw.split(":");if(!key.trim()||dir!==void 0&&!["asc","desc"].includes(dir.trim())||raw.split(":").length>2)throw new CliError("usage",`Invalid --order-by "${raw}". Use a selected field followed by :asc or :desc.`);return{key:key.trim(),dir:dir?.trim()==="asc"?"asc":"desc"}}async function readStdin2(){const chunks2=[];for await(const chunk of process.stdin)chunks2.push(chunk);return Buffer.concat(chunks2).toString("utf8")}async function buildSpec(opts){let spec={metrics:[]};if(opts.json){const raw=opts.json==="-"?await readStdin2():opts.json;try{spec=JSON.parse(raw)}catch(e){throw new CliError("usage",`--json is not valid JSON: ${e.message}`)}}if(!spec||typeof spec!=="object"||Array.isArray(spec))throw new CliError("usage","--json must contain a query object.");if(opts.dataset)spec.dataset=opts.dataset;if(opts.dimensions!==void 0)spec.dimensions=csv2(opts.dimensions);if(opts.metrics!==void 0)spec.metrics=csv2(opts.metrics);if(opts.where?.length)spec.where=opts.where.map(parseWhere);if(opts.orderBy?.length)spec.orderBy=opts.orderBy.map(parseOrderBy);if(opts.limit!==void 0)spec.limit=Number(opts.limit);if(opts.granularity)spec.granularity=opts.granularity;if(opts.explain)spec.explain=true;const date={...spec.dateRange??{}};if(opts.since){date.since=opts.since;delete date.lastDays}else if(opts.lastDays){date.lastDays=Number(opts.lastDays);delete date.since}if(opts.until)date.until=opts.until;if(Object.keys(date).length)spec.dateRange=date;if(opts.branded&&opts.unbranded){throw new CliError("usage","Pass at most one of --branded / --unbranded.")}const filters={...spec.filters??{}};if(opts.prompt)filters.prompt=opts.prompt;if(opts.eval)filters.eval=csv2(opts.eval);if(opts.preference)filters.visibilityPrompt=csv2(opts.preference);if(opts.tag)filters.tag=csv2(opts.tag);if(opts.repo)filters.repo=csv2(opts.repo);if(opts.language)filters.language=csv2(opts.language);if(opts.framework)filters.framework=csv2(opts.framework);if(opts.size)filters.size=csv2(opts.size);if(opts.agent)filters.agent=csv2(opts.agent);if(opts.model)filters.model=csv2(opts.model);if(opts.scenario)filters.scenario=csv2(opts.scenario);if(opts.experiment)filters.experiment=csv2(opts.experiment);if(opts.branded)filters.branded=true;if(opts.unbranded)filters.branded=false;if(Object.keys(filters).length)spec.filters=filters;if(!spec.metrics||spec.metrics.length===0){throw new CliError("usage","At least one metric is required. Pass --metrics <a,b> (or a `metrics` array in --json). See `gauge query fields --dataset <name>`.")}return spec}var pct=n=>`${(n*100).toFixed(1)}%`;function noteFreshness(f){if(!f.available){console.error("warning: analytics store is temporarily unavailable; results may be incomplete.")}else if(f.pending>0){console.error(`note: ${f.pending} recent run(s) still ingesting; results may not reflect them yet.`)}}async function runQuery(command,opts){const output=resolveOutput(command);const spec=await buildSpec(opts);const state=gaugeState(command);const res=state.mode==="local"?queryLocal(localDirectory(command),spec):await createClient().post(`/api/v1/orgs/${encodeURIComponent(await resolveOrg(command))}/query`,spec);if(output==="json"){printJson(res);return}if(res.sql)console.error(`
|
|
16
|
+
Each item reports created, already_exists, or would_create (with --dry-run).`).action(async(texts,opts,cmd)=>{const raw=[...texts.map(text=>({text})),...opts.file?parsePromptInput(readSource(opts.file)):[]];if(!raw.length)throw new CliError("usage","no prompts given; pass them as arguments or with --file");const scope=await categoryScope(cmd);const resolved=await resolvePrompts(scope,raw,{topic:opts.topic,trackingType:opts.trackingType,tags:opts.tag??[]});const result=await addPrompts(scope,resolved,opts.dryRun??false,partial=>printOutcome(partial,cmd));printOutcome(result,cmd)});withCategory(prompts.command("move [prompts...]")).description("Move prompts to another topic (ids as arguments or --file), in batches of 500").requiredOption("--topic <topic>","destination topic (id, slug, or title)").option("-f, --file <path>","read prompt ids from a file ('-' for stdin)").action(async(args,opts,cmd)=>{const ids=readIds(args,opts.file);const scope=await categoryScope(cmd);const topicId=await resolveTopicId(scope,opts.topic);const updated=[];const notFound=[];for(const batch of chunks(ids)){const res=await scope.client.patch(chatPath(...scope.base,"prompts"),{ids:batch,topicId});updated.push(...res.updated);notFound.push(...res.notFound)}if(resolveOutput(cmd)==="json")return printJson({updated,notFound,topicId});console.log(`Moved ${updated.length} prompt(s) to ${topicId}${notFound.length?`; not found: ${notFound.join(", ")}`:""}`)});for(const[name,method,verb]of[["tag","post","Tagged"],["untag","delete","Untagged"]])withCategory(prompts.command(`${name} [prompts...]`)).description(`${name==="tag"?"Apply":"Remove"} tags ${name==="tag"?"to":"from"} prompts (ids as arguments or --file)`).requiredOption("--tag <tags>","tag ids or names (comma-separated or repeatable)",collect2).option("-f, --file <path>","read prompt ids from a file ('-' for stdin)").action(async(args,opts,cmd)=>{const ids=readIds(args,opts.file);const scope=await categoryScope(cmd);const tagIds=await resolveTagIds(scope,opts.tag);const path=chatPath(...scope.base,"prompts","tags");const promptIds=[];for(const batch of chunks(ids)){const body={promptIds:batch,tagIds};const res=method==="post"?await scope.client.post(path,body):await scope.client.delete(path,void 0,body);promptIds.push(...res.promptIds)}if(resolveOutput(cmd)==="json")return printJson({promptIds,tagIds});console.log(`${verb} ${promptIds.length} prompt(s) with ${tagIds.join(", ")}`)})}import{Option as Option2}from"commander";var query_exports={};__export(query_exports,{noteFreshness:()=>noteFreshness,parseOrderBy:()=>parseOrderBy,pct:()=>pct,readStdin:()=>readStdin2,register:()=>register4});import{join as join4}from"path";var dimension=(sql,description,type="string")=>({sql,description,type});var metric=(sql,description,type="number")=>({sql,description,type});var common={run:dimension("s.id","Session ID"),batch:dimension("s.batch_id","Launch batch ID"),eval:dimension("CASE WHEN s.measurement_kind = 'eval' THEN s.measurement_id END","Saved eval ID"),preference:dimension("CASE WHEN s.measurement_kind = 'preference' THEN s.measurement_id END","Saved preference prompt ID"),measurement_revision:dimension("s.measurement_revision","Frozen measurement revision","number"),agent:dimension("s.agent","Coding agent"),model:dimension("s.model","Agent model"),date:dimension("date(s.created_at)","Session creation date (UTC)","date")};var catalogs=[{name:"runs",grain:"One local session (current attempt)",from:"sessions s",dimensions:{...common,case:dimension("s.case_path","File-based case path, when applicable"),outcome:dimension("s.outcome","Execution outcome; this is not an eval verdict"),analyzed:dimension("s.current_analysis_id IS NOT NULL","Whether this attempt has published analysis","boolean"),analysis_id:dimension("s.current_analysis_id","Current published analysis ID")},metrics:{run_count:metric("COUNT(*)","Number of sessions"),analyzed_runs:metric("COALESCE(SUM(CASE WHEN s.current_analysis_id IS NOT NULL THEN 1 ELSE 0 END),0)","Sessions with published analysis"),avg_turns:metric("AVG(s.turns)","Average recorded turns; unknown values excluded"),avg_duration_ms:metric("AVG(s.duration_ms)","Average recorded duration in milliseconds; unknown values excluded"),input_tokens:metric("SUM(s.input_tokens)","Recorded input tokens; null when none have been recorded"),output_tokens:metric("SUM(s.output_tokens)","Recorded output tokens; null when none have been recorded")}},{name:"findings",grain:"One finding in a session's current published analysis",from:"sessions s JOIN session_analyses a ON a.id = s.current_analysis_id JOIN session_findings f ON f.analysis_id = a.id",dimensions:{...common,analysis_id:dimension("a.id","Current published analysis ID"),kind:dimension("f.kind","FRICTION, OBJECTION, or OPPORTUNITY"),subject:dimension("f.subject","Product or surface discussed"),label:dimension("f.label","Finding title; matching labels do not imply a shared theme"),basis:dimension("f.basis","OBSERVED, AGENT_CLAIM, or INFERRED"),effect:dimension("f.effect","Effect on the session"),cause:dimension("f.cause","Attributed cause")},metrics:{finding_count:metric("COUNT(*)","Findings in current published analyses"),affected_runs:metric("COUNT(DISTINCT s.id)","Distinct sessions with matching findings")}}];function localQueryMetadata(){return{datasets:catalogs.map(catalog=>({name:catalog.name,grain:catalog.grain,dimensions:Object.entries(catalog.dimensions).map(([name,field])=>({name,type:field.type,description:field.description})),metrics:Object.entries(catalog.metrics).map(([name,field])=>({name,type:field.type,description:field.description}))}))}}function queryLocal(gaugeDir,input){const compiled=compileLocalQuery(input);const store=openLocalStore(gaugeDir);store.db.exec("BEGIN");try{if(compiled.catalog.name==="findings"){const versions=store.db.prepare(compiled.findingVersionsSql).all(...compiled.coverageParams);const unsupported=versions.find(row2=>row2.schema_version!==ANALYSIS_SCHEMA_VERSION);if(unsupported)throw new CliError("usage",`A selected session's analysis uses schema version ${unsupported.schema_version}; this CLI supports ${ANALYSIS_SCHEMA_VERSION}. Upgrade Gauge before querying these findings. Your data is intact.`)}const rows=store.db.prepare(compiled.sql).all(...compiled.params);for(const row2 of rows)for(const name of compiled.dimensions){if(compiled.catalog.dimensions[name].type==="boolean"&&row2[name]!==null)row2[name]=Boolean(row2[name])}const coverage=store.db.prepare(compiled.coverageSql).get(...compiled.coverageParams);const response={columns:[...compiled.dimensions.map(key=>({key,kind:"dimension"})),...compiled.metrics.map(key=>({key,kind:"metric"}))],rows,...input.explain?{sql:compiled.sql}:{},freshness:{available:true,pending:0,watermark:coverage.watermark},analysisCoverage:{sessions:Number(coverage.total),analyzedSessions:Number(coverage.analyzed),unanalyzedSessions:Number(coverage.total)-Number(coverage.analyzed)}};store.db.exec("COMMIT");return response}catch(error){store.db.exec("ROLLBACK");throw error}finally{store.close()}}function compileLocalQuery(input){if(!input||typeof input!=="object"||Array.isArray(input))throw new CliError("usage","A query must be a JSON object.");const topKeys=new Set(["dataset","dimensions","metrics","filters","where","dateRange","orderBy","limit","granularity","explain"]);for(const key of Object.keys(input))if(!topKeys.has(key))throw new CliError("usage",`Unknown local query option "${key}". See gauge query fields --local --dataset runs.`);const name=input.dataset??"runs";const catalog=catalogs.find(item=>item.name===name);if(!catalog)throw new CliError("usage",`Dataset "${name}" is not available for local runs. Available: runs, findings.`,{hint:"Run `gauge query datasets --local` to see available datasets, or use --hosted to query runs on Gauge."});const dimensions=input.dimensions??[];const metrics=input.metrics;if(!Array.isArray(dimensions)||dimensions.some(key=>typeof key!=="string"))throw new CliError("usage","Query dimensions must be an array of field names.");if(!Array.isArray(metrics)||metrics.length===0||metrics.some(key=>typeof key!=="string"))throw new CliError("usage","At least one metric is required.");if(new Set([...dimensions,...metrics]).size!==dimensions.length+metrics.length)throw new CliError("usage","Select each query field only once.");const unsupported=(kind,field)=>{throw new CliError("usage",`${kind} "${field}" is not supported by the local ${name} dataset.`,{hint:`Run \`gauge query fields --dataset ${name} --local\` to see which fields you can query for local runs.`})};const granularity=input.granularity??"day";if(!["day","week","month"].includes(granularity))throw new CliError("usage","Granularity must be day, week, or month.");const dateSql=granularity==="month"?"date(s.created_at,'start of month')":granularity==="week"?"date(s.created_at,'-' || ((CAST(strftime('%w',s.created_at) AS INTEGER) + 6) % 7) || ' days')":"date(s.created_at)";const expr=key=>key==="date"?dateSql:Object.hasOwn(catalog.dimensions,key)?catalog.dimensions[key].sql:unsupported("Dimension",key);for(const key of dimensions)expr(key);for(const key of metrics)if(!Object.hasOwn(catalog.metrics,key))unsupported("Metric",key);const limit=input.limit??50;if(!Number.isInteger(limit)||limit<1||limit>1e3)throw new CliError("usage","Query limit must be an integer from 1 to 1000.");const params=[];const predicates=[];const coveragePredicates=[];const coverageParams=[];const bind=value=>{if(typeof value!=="string"&&typeof value!=="boolean"&&(typeof value!=="number"||!Number.isFinite(value)))throw new CliError("usage","Query filter values must be strings, booleans, or finite numbers.");params.push(typeof value==="boolean"?Number(value):value);return"?"};const where=input.where??[];if(!Array.isArray(where))throw new CliError("usage","Query where must be an array of predicates.");for(const clause of where){const paramStart=params.length;const predicateStart=predicates.length;if(!clause||typeof clause!=="object"||Object.keys(clause).some(key=>!["dimension","op","value"].includes(key)))throw new CliError("usage","Each where predicate requires dimension, op, and value.");const sql2=expr(clause.dimension);if(clause.op==="in"||clause.op==="notIn"){if(!Array.isArray(clause.value)||clause.value.length===0)throw new CliError("usage",`${clause.op} requires a nonempty value array.`);predicates.push(`${sql2} ${clause.op==="in"?"IN":"NOT IN"} (${clause.value.map(bind).join(",")})`)}else{if(Array.isArray(clause.value))throw new CliError("usage",`${clause.op} requires a scalar value.`);if(clause.op==="contains"){if(typeof clause.value!=="string")throw new CliError("usage","contains requires a string value.");predicates.push(`instr(lower(CAST(${sql2} AS TEXT)),lower(${bind(clause.value)})) > 0`)}else{const operators={eq:"=",neq:"!=",gt:">",gte:">=",lt:"<",lte:"<="};const op=operators[clause.op];if(!op)throw new CliError("usage",`Unsupported filter operator "${clause.op}".`);let value=clause.value;if(catalog.dimensions[clause.dimension].type==="number"&&typeof value==="string"&&value.trim()&&Number.isFinite(Number(value)))value=Number(value);if(catalog.dimensions[clause.dimension].type==="boolean"&&typeof value==="string"&&["true","false"].includes(value))value=value==="true";predicates.push(`${sql2} ${op} ${bind(value)}`)}}if(Object.hasOwn(common,clause.dimension)||name==="runs"){coveragePredicates.push(...predicates.slice(predicateStart));coverageParams.push(...params.slice(paramStart))}}const filterParamStart=params.length;const filterPredicateStart=predicates.length;const filters=input.filters??{};if(!filters||typeof filters!=="object"||Array.isArray(filters))throw new CliError("usage","Query filters must be an object.");const supportedFilters={agent:"s.agent",model:"s.model",prompt:"s.measurement_id",eval:"CASE WHEN s.measurement_kind = 'eval' THEN s.measurement_id END",visibilityPrompt:"CASE WHEN s.measurement_kind = 'preference' THEN s.measurement_id END"};for(const[key,value]of Object.entries(filters)){if(value===void 0)continue;if(!Object.hasOwn(supportedFilters,key))unsupported("Filter",key);const values=key==="prompt"?[value]:value;if(!Array.isArray(values)||values.length===0||values.some(item=>typeof item!=="string"))throw new CliError("usage",`Filter ${key} requires ${key==="prompt"?"a string":"a nonempty array of strings"}.`);predicates.push(`${supportedFilters[key]} IN (${values.map(bind).join(",")})`)}const range=input.dateRange;if(range!==void 0){if(!range||typeof range!=="object"||Array.isArray(range))throw new CliError("usage","Query dateRange must be an object.");for(const key of Object.keys(range))if(!["since","until","lastDays"].includes(key))throw new CliError("usage",`Unknown dateRange field ${key}.`);if(range.since!==void 0&&range.lastDays!==void 0)throw new CliError("usage","Use dateRange.since or dateRange.lastDays, not both.");if(range.lastDays!==void 0&&(!Number.isInteger(range.lastDays)||range.lastDays<=0))throw new CliError("usage","lastDays must be a positive integer.");const since=range.since!==void 0?dateValue(range.since):range.lastDays!==void 0?new Date(Date.now()-range.lastDays*864e5).toISOString():null;const until=range.until!==void 0?dateValue(range.until):null;if(since&&until&&since>until)throw new CliError("usage","Query since must be before until.");if(since)predicates.push(`julianday(s.created_at) >= julianday(${bind(since)})`);if(until)predicates.push(`julianday(s.created_at) <= julianday(${bind(until)})`)}coveragePredicates.push(...predicates.slice(filterPredicateStart));coverageParams.push(...params.slice(filterParamStart));const selected=new Set([...dimensions,...metrics]);const order=input.orderBy??[];if(!Array.isArray(order))throw new CliError("usage","Query orderBy must be an array.");for(const item of order){if(!item||!selected.has(item.key))throw new CliError("usage",`Order field "${item?.key}" must be one of the selected dimensions or metrics.`);if(item.dir!==void 0&&item.dir!=="asc"&&item.dir!=="desc")throw new CliError("usage","Order direction must be asc or desc.")}const columns=[...dimensions.map(key=>`${expr(key)} AS "${key}"`),...metrics.map(key=>`${catalog.metrics[key].sql} AS "${key}"`)];const sql=`SELECT ${columns.join(", ")} FROM ${catalog.from}${predicates.length?` WHERE ${predicates.map(predicate=>`(${predicate})`).join(" AND ")}`:""}${dimensions.length?` GROUP BY ${dimensions.map(expr).join(", ")}`:""}${order.length?` ORDER BY ${order.map(item=>`"${item.key}" ${(item.dir??"desc").toUpperCase()}`).join(", ")}`:""} LIMIT ${limit}`;const coverageWhere=coveragePredicates.length?` WHERE ${coveragePredicates.map(predicate=>`(${predicate})`).join(" AND ")}`:"";const coverageSql=`SELECT COUNT(*) AS total,COUNT(current_analysis_id) AS analyzed,MAX(COALESCE(ended_at,created_at)) AS watermark FROM sessions s${coverageWhere}`;const findingVersionsSql=`SELECT DISTINCT a.schema_version FROM sessions s JOIN session_analyses a ON a.id = s.current_analysis_id${coverageWhere}`;return{sql,params,coverageSql,coverageParams,findingVersionsSql,catalog,dimensions,metrics}}function dateValue(value){if(typeof value!=="string")throw new CliError("usage","Date bounds must be ISO dates or relative windows such as 24h or 7d.");const match=/^(\d+)([mhd])$/.exec(value.trim());const date=match?new Date(Date.now()-Number(match[1])*({m:6e4,h:36e5,d:864e5}[match[2]]??0)):new Date(value);if(!Number.isFinite(date.getTime()))throw new CliError("usage",`Invalid date bound "${value}".`);return date.toISOString()}import{createReadStream,existsSync,realpathSync}from"fs";import{isAbsolute,join,relative}from"path";import{createInterface as createInterface2}from"readline";async function searchLocalTranscripts(gaugeDir,text,options={}){if(!text.trim())throw new CliError("usage","Transcript search requires nonempty text.");if(options.eval&&options.preference)throw new CliError("usage","Select either --eval or --preference for transcript search.");const limit=options.limit??50;if(!Number.isInteger(limit)||limit<1||limit>1e3)throw new CliError("usage","Search limit must be an integer from 1 to 1000.");const store=openLocalStore(gaugeDir);const sessions=(()=>{try{return store.listSessions({measurementId:options.eval??options.preference,measurementKind:options.eval?"eval":options.preference?"preference":void 0,batchId:options.batch}).filter(session=>!options.since||Date.parse(session.createdAt)>=options.since.getTime())}finally{store.close()}})();const result={matches:[],selectedSessions:sessions.length,scannedSessions:0,missingArtifacts:[],truncated:false};const needle=text.toLowerCase();for(const session of sessions){if(!/^[A-Za-z0-9_-]+$/.test(session.id))throw new CliError("usage",`Invalid local run ID: ${session.id}`);const attemptDir=join(gaugeDir,"runs",session.id,"attempts",String(session.attempt));const path=join(attemptDir,"session","events.ndjson");if(!existsSync(path)){result.missingArtifacts.push(session.id);continue}const rel=relative(realpathSync(attemptDir),realpathSync(path));if(isAbsolute(rel)||rel===".."||rel.startsWith("../")||rel.startsWith("..\\"))throw new CliError("usage",`Run ${session.id} transcript resolves outside its attempt.`);result.scannedSessions++;const stream=createReadStream(path,{encoding:"utf8"});const lines=createInterface2({input:stream,crlfDelay:Infinity});let lineNumber=0;try{for await(const line of lines){lineNumber++;let event;try{event=JSON.parse(line)}catch{continue}if(!event||typeof event!=="object")continue;const content=typeof event.payload==="string"?event.payload:JSON.stringify(event.payload??event);const index=content.toLowerCase().indexOf(needle);if(index<0)continue;if(result.matches.length===limit){result.truncated=true;return result}const start=Math.max(0,index-100);const end=Math.min(content.length,Math.max(index+needle.length+200,start+400));result.matches.push({run:session.id,attempt:session.attempt,turn:typeof event.turn==="number"?event.turn:null,artifact:"session/events.ndjson",line:lineNumber,snippet:`${start?"\u2026":""}${content.slice(start,end)}${end<content.length?"\u2026":""}`})}}finally{lines.close();stream.destroy()}}return result}import{createHash as createHash2,randomUUID as randomUUID2}from"crypto";import{existsSync as existsSync2,readFileSync as readFileSync2,realpathSync as realpathSync2,statSync}from"fs";import{isAbsolute as isAbsolute2,join as join2,relative as relative2,resolve}from"path";function saveLocalAnalysis(gaugeDir,runId,value){if(typeof value==="object"&&value!==null&&"schemaVersion"in value&&value.schemaVersion!==ANALYSIS_SCHEMA_VERSION){throw new CliError("usage",`Analysis schema version ${String(value.schemaVersion)} is not supported. This CLI accepts version ${ANALYSIS_SCHEMA_VERSION}; upgrade Gauge for newer analyses. No data was changed.`)}const parsed=analysisSubmissionSchema.safeParse(value);if(!parsed.success)throw new CliError("usage",`Invalid analysis: ${parsed.error.issues.map(issue=>`${issue.path.join(".")||"analysis"}: ${issue.message}`).join("; ")}`,{hint:"Use gauge run analysis schema for the analysis submission contract."});const submission=parsed.data;const store=openLocalStore(gaugeDir);try{const session=store.getSession(runId);if(!session)throw new CliError("usage",`Local run ${runId} was not found.`,{hint:"Use gauge run list --local, or repair session summaries if its artifacts exist."});if(!["completed","failed","canceled","interrupted"].includes(session.outcome))throw new CliError("usage",`Run ${runId} is ${session.outcome}. Wait until it finishes before publishing analysis.`);if(session.attempt!==submission.attempt)throw new CliError("usage",`Run ${runId} is on attempt ${session.attempt}; the submitted analysis targets attempt ${submission.attempt}.`);validateCriteria(session,submission);const attemptDirectory=safeAttemptDirectory(gaugeDir,runId,submission.attempt);const files=new Map;const refs=[...submission.findings.flatMap(finding=>finding.evidence),...submission.products.flatMap(product=>Object.values(product.evidence).flat()),...submission.criteria.flatMap(criterion=>criterion.evidence)];for(const ref of refs)validateEvidence(attemptDirectory,submission.attempt,ref,files);const hash=createHash2("sha256");for(const[path,contents]of[...files].sort(([a],[b])=>a.localeCompare(b)))hash.update(path).update("\0").update(contents).update("\0");const evidenceDigest=`sha256:${hash.digest("hex")}`;return store.transaction(()=>{const currentSession=store.getSession(runId);if(currentSession.attempt!==submission.attempt)throw new CliError("usage","The run advanced to another attempt while validating analysis. Read its new evidence before publishing.");const latest=store.db.prepare("SELECT MAX(revision) AS revision FROM session_analyses WHERE run_id = ?").get(runId);const analysis={...submission,id:`a_${randomUUID2()}`,runId,revision:Number(latest?.revision??0)+1,createdAt:new Date().toISOString(),evidenceDigest,findings:submission.findings.map(finding=>({...finding,id:`f_${randomUUID2()}`}))};store.db.prepare("INSERT INTO session_analyses VALUES (?,?,?,?,?,?,?,?,?)").run(analysis.id,runId,analysis.attempt,analysis.revision,analysis.schemaVersion,analysis.instructionsVersion,evidenceDigest,JSON.stringify(analysis),analysis.createdAt);const findingStatement=store.db.prepare("INSERT INTO session_findings VALUES (?,?,?,?,?,?,?,?,?,?,?)");for(const finding of analysis.findings)findingStatement.run(finding.id,analysis.id,finding.kind,finding.subject,finding.label,finding.detail,finding.basis,finding.effect,finding.cause,JSON.stringify(finding.tags),JSON.stringify(finding.evidence));const productStatement=store.db.prepare("INSERT INTO session_products VALUES (?,?,?,?)");analysis.products.forEach((product,position)=>productStatement.run(analysis.id,position,product.name,JSON.stringify(product)));const criterionStatement=store.db.prepare("INSERT INTO criterion_results VALUES (?,?,?,?,?)");for(const criterion of analysis.criteria)criterionStatement.run(analysis.id,criterion.criterionId,criterion.outcome,criterion.label,JSON.stringify(criterion.evidence));store.db.prepare("UPDATE sessions SET current_analysis_id = ? WHERE id = ?").run(analysis.id,runId);return analysis})}finally{store.close()}}function getLocalAnalysis(gaugeDir,runId){const store=openLocalStore(gaugeDir);try{return readCurrentAnalysis(store,runId)}finally{store.close()}}function readCurrentAnalysis(store,runId){const row2=store.db.prepare("SELECT a.schema_version,a.payload FROM session_analyses a JOIN sessions s ON s.current_analysis_id = a.id WHERE s.id = ?").get(runId);if(!row2)return null;if(row2.schema_version!==ANALYSIS_SCHEMA_VERSION)throw new CliError("usage",`Stored analysis uses schema version ${row2.schema_version}; upgrade Gauge to read it. Your data is intact.`);return JSON.parse(String(row2.payload))}function validateCriteria(session,submission){const configuration=session.configuration??{};const item=configuration.case;const criteria=item?.criteria??configuration.criteria??[];if(!Array.isArray(criteria))throw new CliError("usage","This run's frozen criteria are invalid; repair its session summary before analysis.");const expected=criteria.map(criterion=>{if(!criterion||typeof criterion.id!=="string"||!criterion.id)throw new CliError("usage","This run has a frozen criterion without an ID. Criterion IDs must be assigned before execution.");return criterion.id});const supplied=new Set(submission.criteria.map(criterion=>criterion.criterionId));const missing=expected.filter(id=>!supplied.has(id));const extra=[...supplied].filter(id=>!expected.includes(id));if(missing.length||extra.length)throw new CliError("usage",`Analysis must cover exactly the run's frozen criteria.${missing.length?` Missing: ${missing.join(", ")}.`:""}${extra.length?` Unknown: ${extra.join(", ")}.`:""} Use UNASSESSABLE with an explanation when evidence is insufficient.`)}function safeAttemptDirectory(gaugeDir,runId,attempt){if(!/^[A-Za-z0-9_-]+$/.test(runId))throw new CliError("usage","Invalid local run ID.");const directory=join2(gaugeDir,"runs",runId,"attempts",String(attempt));if(!existsSync2(directory))throw new CliError("usage",`Run evidence is missing: ${directory}`);return realpathSync2(directory)}function readEvidence(directory,path,files){const fullPath=resolve(directory,path);if(!isInside(directory,fullPath))throw new CliError("usage",`Evidence path ${path} leaves the run's attempt directory.`);if(!existsSync2(fullPath)||!statSync(fullPath).isFile())throw new CliError("usage",`Evidence artifact ${path} does not exist in this attempt.`);if(!isInside(directory,realpathSync2(fullPath)))throw new CliError("usage",`Evidence artifact ${path} resolves outside this attempt.`);const contents=files.get(path)??readFileSync2(fullPath,"utf8");files.set(path,contents);return contents}function isInside(directory,path){const rel=relative2(directory,path);return rel!==".."&&!rel.startsWith(`..${process.platform==="win32"?"\\":"/"}`)&&!isAbsolute2(rel)}function validateEvidence(directory,attempt,ref,files){if(ref.kind==="diff"){const diff2=readEvidence(directory,"session/diff.patch",files);const paths=new Set(diff2.split("\n").flatMap(line=>{const match=/^(?:\+\+\+|---) [ab]\/(.*)$/.exec(line);return match?[match[1].split(" ")[0]]:[]}));if(!paths.has(ref.path))throw new CliError("usage",`Diff evidence path ${ref.path} does not appear in this attempt's diff.patch.`);return}if(ref.kind==="artifact"&&ref.attempt!==void 0&&ref.attempt!==attempt)throw new CliError("usage",`Evidence references attempt ${ref.attempt}, but this analysis targets attempt ${attempt}.`);const artifact=ref.kind==="trace"?"session/events.ndjson":ref.artifact;const contents=readEvidence(directory,artifact,files);const lines=contents.split(/\r?\n/);if(lines.at(-1)==="")lines.pop();if(ref.kind==="artifact"&&ref.line!==void 0&&(ref.line>lines.length||(ref.endLine??ref.line)>lines.length))throw new CliError("usage",`Evidence ${artifact}:${ref.line}${ref.endLine?`-${ref.endLine}`:""} exceeds the artifact's ${lines.length} lines.`);if(ref.seq!==void 0){const hasSequence=lines.some(line=>{try{return JSON.parse(line).seq===ref.seq}catch{return false}});if(!hasSequence)throw new CliError("usage",`Evidence sequence ${ref.seq} does not exist in ${artifact}. Copy an event's recorded seq; do not use its array index or turn.`,{hint:`For evidence without seq, cite an original artifact line, for example ${JSON.stringify({kind:"artifact",artifact,attempt,line:1})}. Line numbers start at 1; choose the line containing your evidence.`})}}function renderLocalAnalysisMarkdown(analysis){const lines=[`# Session analysis: ${analysis.runId}`,"",`Analysis ${analysis.id}, revision ${analysis.revision}; attempt ${analysis.attempt}.`,`Assessment: ${analysis.assessability}.`,""];if(analysis.limitation)lines.push(`Limitation: ${analysis.limitation}`,"");for(const finding of analysis.findings)lines.push(`## ${finding.kind}: ${finding.label}`,"",`${finding.subject} \xB7 ${finding.effect} \xB7 ${finding.basis}`,"",finding.detail,"",`Evidence: ${finding.evidence.map(formatEvidence).join(", ")}`,"");if(!analysis.findings.length)lines.push("No findings were reported in this analysis.","");if(analysis.criteria.length){lines.push("## Criteria","");for(const criterion of analysis.criteria)lines.push(`- ${criterion.criterionId}: ${criterion.outcome} \u2014 ${criterion.label}`);lines.push("")}if(analysis.products.length){lines.push("## Products","");for(const product of analysis.products)lines.push(`- ${product.name}: ${product.chosen?"chosen":product.rejected?"rejected":product.considered?"considered":"mentioned"} \u2014 ${product.reason}`);lines.push("")}return lines.join("\n")}function formatEvidence(ref){if(ref.kind==="trace")return`trace ${ref.seq}`;if(ref.kind==="diff")return`diff ${ref.path}`;return`${ref.artifact}${ref.line!==void 0?`:${ref.line}${ref.endLine!==void 0?`-${ref.endLine}`:""}`:` seq ${ref.seq}`}`}import{join as join3}from"path";function listLocalRuns(gaugeDir,filter={}){const store=openLocalStore(gaugeDir);try{return store.listSessions().map(session=>({...session,evalId:session.measurementKind==="eval"?session.measurementId??null:null,preferenceId:session.measurementKind==="preference"?session.measurementId??null:null,endedBy:session.endedBy??"",endedAt:session.endedAt??"",attemptDir:join3(gaugeDir,"runs",session.id,"attempts",String(session.attempt)),resultsDir:join3(gaugeDir,"runs",session.id,"results")})).filter(run=>matches(run,filter)).map(run=>{const analysis=readCurrentAnalysis(store,run.id);if(!analysis)return run;const session=store.getSession(run.id);const item=session.configuration?.case;const names=new Map((item?.criteria??[]).map(criterion=>[criterion.id,criterion.name]));return{...run,criterionResults:analysis.criteria.map(criterion=>({id:criterion.criterionId,name:names.get(criterion.criterionId)??criterion.criterionId,outcome:criterion.outcome}))}})}finally{store.close()}}function matches(run,filter){return(!filter.ids?.length||filter.ids.includes(run.id))&&(!filter.batch||run.batchId===filter.batch)&&(!filter.evalId||run.evalId===filter.evalId)&&(!filter.preferenceId||run.preferenceId===filter.preferenceId)&&(filter.measurementRevision===void 0||run.measurementRevision===filter.measurementRevision)&&(!filter.agent||run.agent===filter.agent)&&(!filter.model||run.model===filter.model)&&(!filter.outcome||filter.outcome.split(",").includes(run.outcome))&&(!filter.since||Date.parse(run.createdAt??run.endedAt)>=filter.since.getTime())&&(!filter.until||Date.parse(run.createdAt??run.endedAt)<=filter.until.getTime())&&(!filter.case||run.case===filter.case)&&(filter.analyzed===void 0||run.analyzed===filter.analyzed)}var UNIT_MS={m:6e4,h:36e5,d:864e5};function parseSince(window,now=Date.now(),flag="--since"){const relative6=/^(\d+)([mhd])$/.exec(window.trim());if(relative6)return new Date(now-Number(relative6[1])*UNIT_MS[relative6[2]]);const at=Date.parse(window);if(Number.isNaN(at))throw new CliError("usage",`invalid ${flag} "${window}" (expected 24h, 7d, or an ISO date)`);return new Date(at)}var collect3=(value,prev=[])=>[...prev,value];var csv2=raw=>raw.split(",").map(s=>s.trim()).filter(Boolean);function parseWhere(expr){const inMatch=expr.match(/^\s*(\w+)\s+(in|contains)\s+(.*)$/i);if(inMatch){const[,dimension3,opRaw,rest2]=inMatch;const op=opRaw.toLowerCase();if(op==="in"){return{dimension:dimension3,op:"in",value:csv2(rest2)}}return{dimension:dimension3,op:"contains",value:rest2.trim()}}const opMatch=expr.match(/^\s*(\w+)\s*(!=|>=|<=|=|>|<)\s*(.*)$/);if(!opMatch){throw new CliError("usage",`Could not parse --where "${expr}". Use e.g. "action=install", "ecosystem in npm,pypi", or "name contains supa".`)}const[,dimension2,sym,rest]=opMatch;const opFor={"=":"eq","!=":"neq",">":"gt",">=":"gte","<":"lt","<=":"lte"};return{dimension:dimension2,op:opFor[sym],value:rest.trim()}}function parseOrderBy(raw){const[key,dir]=raw.split(":");if(!key.trim()||dir!==void 0&&!["asc","desc"].includes(dir.trim())||raw.split(":").length>2)throw new CliError("usage",`Invalid --order-by "${raw}". Use a selected field followed by :asc or :desc.`);return{key:key.trim(),dir:dir?.trim()==="asc"?"asc":"desc"}}async function readStdin2(){const chunks2=[];for await(const chunk of process.stdin)chunks2.push(chunk);return Buffer.concat(chunks2).toString("utf8")}async function buildSpec(opts){let spec={metrics:[]};if(opts.json){const raw=opts.json==="-"?await readStdin2():opts.json;try{spec=JSON.parse(raw)}catch(e){throw new CliError("usage",`--json is not valid JSON: ${e.message}`)}}if(!spec||typeof spec!=="object"||Array.isArray(spec))throw new CliError("usage","--json must contain a query object.");if(opts.dataset)spec.dataset=opts.dataset;if(opts.dimensions!==void 0)spec.dimensions=csv2(opts.dimensions);if(opts.metrics!==void 0)spec.metrics=csv2(opts.metrics);if(opts.where?.length)spec.where=opts.where.map(parseWhere);if(opts.orderBy?.length)spec.orderBy=opts.orderBy.map(parseOrderBy);if(opts.limit!==void 0)spec.limit=Number(opts.limit);if(opts.granularity)spec.granularity=opts.granularity;if(opts.explain)spec.explain=true;const date={...spec.dateRange??{}};if(opts.since){date.since=opts.since;delete date.lastDays}else if(opts.lastDays){date.lastDays=Number(opts.lastDays);delete date.since}if(opts.until)date.until=opts.until;if(Object.keys(date).length)spec.dateRange=date;if(opts.branded&&opts.unbranded){throw new CliError("usage","Pass at most one of --branded / --unbranded.")}const filters={...spec.filters??{}};if(opts.prompt)filters.prompt=opts.prompt;if(opts.eval)filters.eval=csv2(opts.eval);if(opts.preference)filters.visibilityPrompt=csv2(opts.preference);if(opts.tag)filters.tag=csv2(opts.tag);if(opts.repo)filters.repo=csv2(opts.repo);if(opts.language)filters.language=csv2(opts.language);if(opts.framework)filters.framework=csv2(opts.framework);if(opts.size)filters.size=csv2(opts.size);if(opts.agent)filters.agent=csv2(opts.agent);if(opts.model)filters.model=csv2(opts.model);if(opts.scenario)filters.scenario=csv2(opts.scenario);if(opts.experiment)filters.experiment=csv2(opts.experiment);if(opts.branded)filters.branded=true;if(opts.unbranded)filters.branded=false;if(Object.keys(filters).length)spec.filters=filters;if(!spec.metrics||spec.metrics.length===0){throw new CliError("usage","At least one metric is required. Pass --metrics <a,b> (or a `metrics` array in --json). See `gauge query fields --dataset <name>`.")}return spec}var pct=n=>`${(n*100).toFixed(1)}%`;function noteFreshness(f){if(!f.available){console.error("warning: analytics store is temporarily unavailable; results may be incomplete.")}else if(f.pending>0){console.error(`note: ${f.pending} recent run(s) still ingesting; results may not reflect them yet.`)}}async function runQuery(command,opts){const output=resolveOutput(command);const spec=await buildSpec(opts);const state=gaugeState(command);const res=state.mode==="local"?queryLocal(localDirectory(command),spec):await createClient().post(`/api/v1/orgs/${encodeURIComponent(await resolveOrg(command))}/query`,spec);if(output==="json"){printJson(res);return}if(res.sql)console.error(`
|
|
17
17
|
${res.sql}
|
|
18
|
-
`);const rateKeys=new Set(res.columns.filter(c=>c.kind==="metric"&&c.key.endsWith("_rate")).map(c=>c.key));const rows=res.rows.map(row2=>{const out={};for(const col of res.columns){const v=row2[col.key];out[col.key]=rateKeys.has(col.key)&&typeof v==="number"?pct(v):v}return out});printItems(rows,"table",res.columns.map(c=>({key:c.key})));noteFreshness(res.freshness);if("analysisCoverage"in res&&res.analysisCoverage){const coverage=res.analysisCoverage;console.error(`Analysis coverage: ${coverage.analyzedSessions}/${coverage.sessions} selected sessions analyzed; ${coverage.unanalyzedSessions} unanalyzed.`)}}async function showMetadata(command,mode,dataset){const output=resolveOutput(command);const meta=gaugeState(command).mode==="local"?localQueryMetadata():await createClient().get(`/api/v1/orgs/${encodeURIComponent(await resolveOrg(command))}/query/metadata`);if(mode==="datasets"){if(output==="json")return printJson(meta.datasets.map(d=>({name:d.name,grain:d.grain})));printItems(meta.datasets.map(d=>({dataset:d.name,grain:d.grain})),"table");return}const ds=meta.datasets.find(d=>d.name===dataset);if(!ds){throw new CliError("usage",`Unknown dataset "${dataset}". Valid: ${meta.datasets.map(d=>d.name).join(", ")}.`)}if(output==="json")return printJson(ds);const rows=[...ds.dimensions.map(f=>({field:f.name,kind:"dimension",type:f.type,about:f.description??""})),...ds.metrics.map(f=>({field:f.name,kind:"metric",type:f.type,about:f.description??""}))];console.error(`dataset: ${ds.name} \u2014 ${ds.grain}`);printItems(rows,"table")}function
|
|
18
|
+
`);const rateKeys=new Set(res.columns.filter(c=>c.kind==="metric"&&c.key.endsWith("_rate")).map(c=>c.key));const rows=res.rows.map(row2=>{const out={};for(const col of res.columns){const v=row2[col.key];out[col.key]=rateKeys.has(col.key)&&typeof v==="number"?pct(v):v}return out});printItems(rows,"table",res.columns.map(c=>({key:c.key})));noteFreshness(res.freshness);if("analysisCoverage"in res&&res.analysisCoverage){const coverage=res.analysisCoverage;console.error(`Analysis coverage: ${coverage.analyzedSessions}/${coverage.sessions} selected sessions analyzed; ${coverage.unanalyzedSessions} unanalyzed.`)}}async function showMetadata(command,mode,dataset){const output=resolveOutput(command);const meta=gaugeState(command).mode==="local"?localQueryMetadata():await createClient().get(`/api/v1/orgs/${encodeURIComponent(await resolveOrg(command))}/query/metadata`);if(mode==="datasets"){if(output==="json")return printJson(meta.datasets.map(d=>({name:d.name,grain:d.grain})));printItems(meta.datasets.map(d=>({dataset:d.name,grain:d.grain})),"table");return}const ds=meta.datasets.find(d=>d.name===dataset);if(!ds){throw new CliError("usage",`Unknown dataset "${dataset}". Valid: ${meta.datasets.map(d=>d.name).join(", ")}.`)}if(output==="json")return printJson(ds);const rows=[...ds.dimensions.map(f=>({field:f.name,kind:"dimension",type:f.type,about:f.description??""})),...ds.metrics.map(f=>({field:f.name,kind:"metric",type:f.type,about:f.description??""}))];console.error(`dataset: ${ds.name} \u2014 ${ds.grain}`);printItems(rows,"table")}function register4(program2){const group=program2.command("query").description("Query local runs and findings or hosted analytics; discover backend capabilities");group.command("run",{isDefault:true}).description("Run a report; use query fields --dataset <name> with --local or --hosted to discover supported fields").option("--json <spec>","full query spec as JSON (or '-' to read stdin); flags override its keys").option("--dataset <name>","dataset name (local: runs | findings; use query datasets for hosted capabilities)").option("-d, --dimensions <names>","comma-separated GROUP BY dimensions").option("-m, --metrics <names>","comma-separated metrics (at least one)").option("--where <expr>",'predicate, e.g. "action=install" (repeatable)',collect3).option("--since <val>","date range start \u2014 ISO date or window (30d, 24h)").option("--until <val>","date range end \u2014 ISO date or window").option("--last-days <n>","shorthand for --since <n> days ago").option("--order-by <key:dir>","sort by a selected key, e.g. installs:desc (repeatable)",collect3).option("--limit <n>","max rows (default 50, max 1000)").option("--granularity <g>","date bucket: day | week | month (default day)").option("--explain","print the compiled SQL to stderr").option("--prompt <id>","filter to a single prompt").option("--eval <ids>","comma-separated saved eval IDs").option("--preference <ids>","comma-separated saved preference prompt IDs").option("--tag <ids>","comma-separated tag ids (hosted only)").option("--repo <urls>","comma-separated base repo urls (hosted only)").option("--language <langs>","comma-separated repo languages (hosted only)").option("--framework <names>","comma-separated repo frameworks (hosted only)").option("--size <buckets>","comma-separated codebase-size buckets (hosted only)").option("--agent <names>","comma-separated agent names").option("--model <models>","comma-separated models").option("--scenario <types>","comma-separated scenario types (greenfield|migration; hosted only)").option("--experiment <labels>","comma-separated experiment labels (hosted only)").option("--branded","only branded prompts (hosted only)").option("--unbranded","only non-branded prompts (hosted only)").action((opts,command)=>runQuery(command,opts));group.command("datasets").description("List the datasets you can query").action((_opts,command)=>showMetadata(command,"datasets"));group.command("fields").description("List the dimensions and metrics for a dataset").requiredOption("--dataset <name>","dataset to describe").action((opts,command)=>showMetadata(command,"fields",opts.dataset));group.command("search <text>").description("Search local transcript evidence without building an index").option("--eval <id>","filter to a saved eval").option("--preference <id>","filter to a saved preference prompt").option("--batch <id>","filter to one launch batch").option("--since <value>","session creation date or relative window, such as 7d").option("--limit <n>","maximum matching passages (default 50, maximum 1000)").action(async(text,opts,command)=>{if(gaugeState(command).mode!=="local")throw new CliError("usage","Transcript search is not supported by the hosted query backend. Use gauge run get <id> --logs --hosted to read a hosted transcript, or query search --local for local sessions.");const result=await searchLocalTranscripts(localDirectory(command),text,{...opts,since:opts.since?parseSince(opts.since):void 0,limit:opts.limit===void 0?void 0:Number(opts.limit)});if(resolveOutput(command)==="json"){printJson(result);return}printItems(result.matches.map(match=>({run:match.run,attempt:match.attempt,turn:match.turn,evidence:`${match.artifact}:${match.line}`,snippet:match.snippet})),"table");console.error(`Searched ${result.scannedSessions}/${result.selectedSessions} selected sessions.${result.truncated?" More matches exist; narrow the cohort or increase --limit.":""}${result.missingArtifacts.length?` Missing transcripts: ${result.missingArtifacts.join(", ")}.`:""}`)})}function localDirectory(command){const state=gaugeState(command);if(!state.root)throw new CliError("usage","No Gauge workspace found. Run gauge init --local in this directory first.");return join4(state.root,".gauge")}var CHAT_DATASETS=["answers","citations","searchQueries","aiTraffic"];var ORG_LEVEL_DATASETS=new Set(["aiTraffic"]);function parseChatWhere(expr){const list2=expr.match(/^\s*(\w+)\s+(not\s+in|in|contains)\s+(.*)$/i);if(list2){const[,dimension3,rawOp,rest2]=list2;const op=rawOp.toLowerCase().replace(/\s+/," ");if(op==="contains")return{dimension:dimension3,op:"contains",value:rest2.trim()};return{dimension:dimension3,op:op==="in"?"in":"notIn",value:csv(rest2)}}const cmp=expr.match(/^\s*(\w+)\s*(!=|=)\s*(.*)$/);if(!cmp)throw new CliError("usage",`Could not parse --where "${expr}". Use e.g. "domain=example.com", "model in gpt-5,claude", "domain not in a.com,b.com", or "query contains pricing".`);const[,dimension2,sym,rest]=cmp;return{dimension:dimension2,op:sym==="="?"eq":"neq",value:rest.trim()}}var collect4=(value,prev=[])=>[...prev,value];async function buildChatSpec(opts){let spec={};if(opts.json){const raw=opts.json==="-"?await readStdin2():opts.json;try{spec=JSON.parse(raw)}catch(e){throw new CliError("usage",`--json is not valid JSON: ${e.message}`)}if(!spec||typeof spec!=="object"||Array.isArray(spec))throw new CliError("usage","--json must contain a query object.")}if(opts.dataset)spec.dataset=opts.dataset;if(opts.dimensions!==void 0)spec.dimensions=csv(opts.dimensions);if(opts.metrics!==void 0)spec.metrics=csv(opts.metrics);if(opts.where?.length)spec.where=opts.where.map(parseChatWhere);if(opts.orderBy?.length)spec.orderBy=opts.orderBy.map(parseOrderBy);if(opts.limit!==void 0)spec.limit=Number(opts.limit);if(opts.granularity)spec.granularity=opts.granularity;if(opts.explain)spec.explain=true;const date={...spec.dateRange??{}};if(opts.since){date.since=opts.since;delete date.lastDays}else if(opts.lastDays){date.lastDays=Number(opts.lastDays);delete date.since}if(opts.until)date.until=opts.until;if(Object.keys(date).length)spec.dateRange=date;const filters={...spec.filters??{}};for(const key of["brand","model","topic","tag","prompt","domain","page"])if(opts[key])filters[key]=csv(opts[key]);if(opts.tagMode)filters.tagMode=opts.tagMode;if(opts.trackingType)filters.trackingType=opts.trackingType;if(Object.keys(filters).length)spec.filters=filters;if(!spec.dataset)throw new CliError("usage",`--dataset is required (${CHAT_DATASETS.join(", ")}). See \`gauge chat query datasets\`.`);if(!spec.metrics?.length)throw new CliError("usage",`At least one metric is required. Pass --metrics <a,b> (or a \`metrics\` array in --json). See \`gauge chat query fields --dataset ${spec.dataset}\`.`);return spec}function labelFor(labels,type,value){if(!type.startsWith("id:")||typeof value!=="string")return value;const label=labels[type.slice(3)]?.[value];const text=label?.name??label?.title??label?.text??label?.url;return typeof text==="string"?truncate(text,60):value}async function runQuery2(command,opts){const spec=await buildChatSpec(opts);const{client,org,chatOrg}=await chatScope(command);const{category}=command.optsWithGlobals();if(category)spec.filters={...spec.filters??{},category};else if(!ORG_LEVEL_DATASETS.has(spec.dataset)&&!spec.filters?.category)spec.filters={...spec.filters??{},category:await resolveChatCategory(command,client,chatOrg)};const res=await client.post(chatPath("orgs",org,"query"),spec);if(resolveOutput(command)==="json")return printJson(res);if(res.sql)console.error(`
|
|
19
19
|
${res.sql}
|
|
20
20
|
`);printItems(res.rows.map(row2=>{const out={};for(const col of res.columns){const v=row2[col.key];out[col.key]=col.kind==="dimension"?labelFor(res.labels,col.type,v):col.type==="rate"&&typeof v==="number"?pct(v):v}return out}),"table",res.columns.map(c=>({key:c.key})));console.error(`dateRange: ${res.dateRange.since} to ${res.dateRange.until}`);if(res.truncated)console.error("note: more rows exist; raise --limit or narrow the query.");noteFreshness(res.freshness)}async function metadata(command,dataset){const{client,org,chatOrg}=await chatScope(command);return client.get(chatPath("orgs",org,"query","metadata"),{dataset})}function registerQuery(chat){const group=chat.command("query").description("Query AI visibility data (answers, citations, searchQueries, aiTraffic)");withCategory(group.command("run",{isDefault:true})).description("Run a query; discover fields with `gauge chat query fields --dataset <name>`").option("--json <spec>","full query spec as JSON (or '-' to read stdin); flags override its keys").option("--dataset <name>",CHAT_DATASETS.join(" | ")).option("-d, --dimensions <names>","comma-separated GROUP BY dimensions").option("-m, --metrics <names>","comma-separated metrics (at least one)").option("--where <expr>",'predicate, e.g. "domain=example.com" or "model not in a,b" (repeatable)',collect4).option("--since <val>","date range start \u2014 YYYY-MM-DD or window (30d)").option("--until <val>","date range end \u2014 YYYY-MM-DD (inclusive)").option("--last-days <n>","shorthand for the last n days").option("--order-by <key:dir>","sort by a selected key, e.g. visibility:desc (repeatable)",collect4).option("--limit <n>","max rows (default 50, max 1000)").option("--granularity <g>","date bucket: hour (aiTraffic only, \u2264 14 days) | day | week | month (default day)").addOption(new Option2("--explain","print the compiled SQL to stderr").hideHelp()).option("--brand <ids>","comma-separated brd_ ids").option("--model <models>","comma-separated models").option("--topic <ids>","comma-separated top_ ids").option("--tag <ids>","comma-separated tag_ ids").option("--tag-mode <mode>","any (default) | all").option("--prompt <ids>","comma-separated prm_ ids").option("--tracking-type <type>","visibility (default) | branded_own | all").option("--domain <domains>","comma-separated cited domains").option("--page <ids>","comma-separated pag_ ids").addHelpText("after",`
|
|
21
21
|
Answer datasets (answers, citations, searchQueries) are scoped to a category:
|
|
22
22
|
--category, else the chatCategory default, else the org's only category.
|
|
23
23
|
aiTraffic is organization-level; with --granularity hour its date column holds
|
|
24
24
|
UTC timestamps (YYYY-MM-DDTHH:MM:SSZ) instead of days. Rate metrics are 0-1
|
|
25
|
-
fractions (shown as percentages in tables).`).action((opts,command)=>runQuery2(command,opts));group.command("datasets").description("List the datasets you can query").action(async(_opts,command)=>{const meta=await metadata(command);const rows=meta.datasets.map(d=>({name:d.name,grain:d.grain,requiresCategory:d.requiresCategory,description:d.description}));if(resolveOutput(command)==="json")return printJson(rows);printItems(rows.map(r=>({...r,description:truncate(r.description,70)})),"table")});group.command("fields").description("List the dimensions, metrics, filters and where columns for a dataset").requiredOption("--dataset <name>","dataset to describe").action(async(opts,command)=>{const meta=await metadata(command,opts.dataset);const ds=meta.datasets.find(d=>d.name===opts.dataset);if(!ds)throw new CliError("usage",`Unknown dataset "${opts.dataset}". Valid: ${CHAT_DATASETS.join(", ")}.`);if(resolveOutput(command)==="json")return printJson(ds);console.error(`dataset: ${ds.name} \u2014 ${ds.grain}. ${ds.description}`);printItems([...ds.dimensions.map(f=>({field:f.name,kind:"dimension",type:f.type,about:f.description??""})),...ds.metrics.map(f=>({field:f.name,kind:"metric",type:`${f.type}${f.unit?` (${f.unit})`:""}`,about:`${f.description??""}${f.requires?.length?` Needs ${f.requires.join(" or ")}.`:""}`})),...ds.where.map(w=>({field:w.name,kind:"where",type:w.ops.join("|"),about:w.description}))],"table");console.error(`filters: ${ds.filters.join(", ")||"(none)"}; granularities: ${ds.granularities.join(", ")}`)});group.command("metadata").description("Print the full query catalog with limits and this org's valid values (JSON)").option("--dataset <name>","only this dataset").action(async(opts,command)=>{printJson(await metadata(command,opts.dataset))})}function done(cmd,value,text){if(resolveOutput(cmd)==="json")printJson(value);else console.log(text)}function registerCategories(chat){const categories=chat.command("categories").description("Categories (markets) your organization tracks");categories.command("list",{isDefault:true}).description("List categories (* marks the chatCategory default)").action(async(_opts,cmd)=>{const{client,org,chatOrg}=await chatScope(cmd);const{data}=await client.get(chatPath("orgs",org,"categories"));const{chatCategory,chatCategoryOrg}=readConfig();const applies=chatCategoryOrg?chatCategoryOrg===org:chatOrg.source==="config";const isDefault=c=>applies&&(chatCategory===c.slug||chatCategory===c.id);if(resolveOutput(cmd)==="json")return printJson({data:data.map(c=>({...c,isDefault:isDefault(c)}))});printItems(data.map(c=>({default:isDefault(c)?"*":"",id:c.id,slug:c.slug,name:c.name})),"table")});categories.command("use <category>").description("Set the default category (slug or id) for the selected organization").action(async(category,_opts,cmd)=>{const{client,org,chatOrg}=await chatScope(cmd);const found=await client.get(chatPath("orgs",org,"categories",category));updateConfig({chatCategory:found.slug,chatCategoryOrg:org});done(cmd,{org,chatCategory:found.slug},`Default Chat category: ${found.slug} (org ${chatOrg.name??org})`)})}function registerTopics(chat){const topics=chat.command("topics").description("Topics group a category's prompts");addPageOptions(withCategory(topics.command("list",{isDefault:true}).description("List topics"))).action(async(opts,cmd)=>{const scope=await categoryScope(cmd);const page=await fetchPages(scope.client,chatPath(...scope.base,"topics"),{},opts);printPage(page,resolveOutput(cmd),t=>({id:t.id,title:t.title,slug:t.slug,prompts:t.promptCount}))});withCategory(topics.command("create <title>")).description("Create a topic; exits 0 with the existing topic when the title already exists (--if-exists)").option("--if-exists","return the existing topic instead of failing").action(async(title,opts,cmd)=>{const scope=await categoryScope(cmd);try{const topic=await scope.client.post(chatPath(...scope.base,"topics"),{title});done(cmd,topic,`Created topic ${topic.id} (${topic.title})`)}catch(error){const id=error instanceof CliError&&error.apiCode==="already_exists"?error.details?.id:void 0;if(!opts.ifExists||typeof id!=="string")throw error;const topic=await scope.client.get(chatPath(...scope.base,"topics",id));done(cmd,topic,`Topic already exists: ${topic.id} (${topic.title})`)}});withCategory(topics.command("rename <topic> <title>")).description("Rename a topic (id, slug, or title); its slug follows").action(async(topic,title,_opts,cmd)=>{const scope=await categoryScope(cmd);const id=await resolveTopicId(scope,topic);const updated=await scope.client.patch(chatPath(...scope.base,"topics",id),{title});done(cmd,updated,`Renamed topic ${updated.id} to ${updated.title}`)})}function registerTags(chat){const tags=chat.command("tags").description("Organization prompt tags");tags.command("list",{isDefault:true}).description("List tags").action(async(_opts,cmd)=>{const{client,org}=await chatScope(cmd);const page=await client.get(chatPath("orgs",org,"tags"));printPage(page,resolveOutput(cmd),t=>({id:t.id,name:t.name,color:t.color,prompts:t.promptCount}))});tags.command("create <name>").alias("add").description("Create a tag").option("--color <color>","tag color").action(async(name,opts,cmd)=>{const{client,org}=await chatScope(cmd);const tag=await client.post(chatPath("orgs",org,"tags"),{name,...opts.color?{color:opts.color}:{}});done(cmd,tag,`Created tag ${tag.id} (${tag.name})`)});tags.command("rename <tag> <name>").description("Rename a tag (id or current name)").option("--color <color>","also change the color").action(async(tag,name,opts,cmd)=>{const scope=await chatScope(cmd);const[id]=await resolveTagIds(scope,[tag]);const updated=await scope.client.patch(chatPath("orgs",scope.org,"tags",id),{name,...opts.color?{color:opts.color}:{}});done(cmd,updated,`Renamed tag ${updated.id} to ${updated.name}`)});tags.command("delete <tag>").alias("rm").description("Delete a tag (id or name); prompts are kept").action(async(tag,_opts,cmd)=>{const scope=await chatScope(cmd);const[id]=await resolveTagIds(scope,[tag]);await scope.client.delete(chatPath("orgs",scope.org,"tags",id));done(cmd,{id,deleted:true},`Deleted tag ${id}`)})}function registerBrands(chat){withCategory(chat.command("brands")).description("Owned and competitor brands tracked in the category").action(async(_opts,cmd)=>{const scope=await categoryScope(cmd);const page=await scope.client.get(chatPath(...scope.base,"brands"));printPage(page,resolveOutput(cmd),b=>({id:b.id,role:b.role,name:b.name,website:b.website}))})}var CHAT_ORG_HELP="organization slug or id (default: gauge.json orgId, then your saved default, else your only Chat org; an org API key implies its org)";function scopeGlobalOptions(program2,chat){const isAgentsOnly=option=>option.long==="--local"||option.long==="--hosted";chat.configureHelp({...program2.configureHelp(),visibleGlobalOptions(cmd){return Help.prototype.visibleGlobalOptions.call(this,cmd).filter(option=>!isAgentsOnly(option))},optionDescription(option){return option.long==="--org"&&program2.options.includes(option)?CHAT_ORG_HELP:Help.prototype.optionDescription.call(this,option)}});chat.hook("preAction",()=>{const flag=program2.options.find(option=>isAgentsOnly(option)&&program2.getOptionValueSource(option.attributeName())==="cli");if(flag)throw new CliError("usage",`${flag.long} applies only to Agents commands; gauge chat always uses the hosted Chat API. Remove ${flag.long}.`)})}function
|
|
25
|
+
fractions (shown as percentages in tables).`).action((opts,command)=>runQuery2(command,opts));group.command("datasets").description("List the datasets you can query").action(async(_opts,command)=>{const meta=await metadata(command);const rows=meta.datasets.map(d=>({name:d.name,grain:d.grain,requiresCategory:d.requiresCategory,description:d.description}));if(resolveOutput(command)==="json")return printJson(rows);printItems(rows.map(r=>({...r,description:truncate(r.description,70)})),"table")});group.command("fields").description("List the dimensions, metrics, filters and where columns for a dataset").requiredOption("--dataset <name>","dataset to describe").action(async(opts,command)=>{const meta=await metadata(command,opts.dataset);const ds=meta.datasets.find(d=>d.name===opts.dataset);if(!ds)throw new CliError("usage",`Unknown dataset "${opts.dataset}". Valid: ${CHAT_DATASETS.join(", ")}.`);if(resolveOutput(command)==="json")return printJson(ds);console.error(`dataset: ${ds.name} \u2014 ${ds.grain}. ${ds.description}`);printItems([...ds.dimensions.map(f=>({field:f.name,kind:"dimension",type:f.type,about:f.description??""})),...ds.metrics.map(f=>({field:f.name,kind:"metric",type:`${f.type}${f.unit?` (${f.unit})`:""}`,about:`${f.description??""}${f.requires?.length?` Needs ${f.requires.join(" or ")}.`:""}`})),...ds.where.map(w=>({field:w.name,kind:"where",type:w.ops.join("|"),about:w.description}))],"table");console.error(`filters: ${ds.filters.join(", ")||"(none)"}; granularities: ${ds.granularities.join(", ")}`)});group.command("metadata").description("Print the full query catalog with limits and this org's valid values (JSON)").option("--dataset <name>","only this dataset").action(async(opts,command)=>{printJson(await metadata(command,opts.dataset))})}function done(cmd,value,text){if(resolveOutput(cmd)==="json")printJson(value);else console.log(text)}function registerCategories(chat){const categories=chat.command("categories").description("Categories (markets) your organization tracks");categories.command("list",{isDefault:true}).description("List categories (* marks the chatCategory default)").action(async(_opts,cmd)=>{const{client,org,chatOrg}=await chatScope(cmd);const{data}=await client.get(chatPath("orgs",org,"categories"));const{chatCategory,chatCategoryOrg}=readConfig();const applies=chatCategoryOrg?chatCategoryOrg===org:chatOrg.source==="config";const isDefault=c=>applies&&(chatCategory===c.slug||chatCategory===c.id);if(resolveOutput(cmd)==="json")return printJson({data:data.map(c=>({...c,isDefault:isDefault(c)}))});printItems(data.map(c=>({default:isDefault(c)?"*":"",id:c.id,slug:c.slug,name:c.name})),"table")});categories.command("use <category>").description("Set the default category (slug or id) for the selected organization").action(async(category,_opts,cmd)=>{const{client,org,chatOrg}=await chatScope(cmd);const found=await client.get(chatPath("orgs",org,"categories",category));updateConfig({chatCategory:found.slug,chatCategoryOrg:org});done(cmd,{org,chatCategory:found.slug},`Default Chat category: ${found.slug} (org ${chatOrg.name??org})`)})}function registerTopics(chat){const topics=chat.command("topics").description("Topics group a category's prompts");addPageOptions(withCategory(topics.command("list",{isDefault:true}).description("List topics"))).action(async(opts,cmd)=>{const scope=await categoryScope(cmd);const page=await fetchPages(scope.client,chatPath(...scope.base,"topics"),{},opts);printPage(page,resolveOutput(cmd),t=>({id:t.id,title:t.title,slug:t.slug,prompts:t.promptCount}))});withCategory(topics.command("create <title>")).description("Create a topic; exits 0 with the existing topic when the title already exists (--if-exists)").option("--if-exists","return the existing topic instead of failing").action(async(title,opts,cmd)=>{const scope=await categoryScope(cmd);try{const topic=await scope.client.post(chatPath(...scope.base,"topics"),{title});done(cmd,topic,`Created topic ${topic.id} (${topic.title})`)}catch(error){const id=error instanceof CliError&&error.apiCode==="already_exists"?error.details?.id:void 0;if(!opts.ifExists||typeof id!=="string")throw error;const topic=await scope.client.get(chatPath(...scope.base,"topics",id));done(cmd,topic,`Topic already exists: ${topic.id} (${topic.title})`)}});withCategory(topics.command("rename <topic> <title>")).description("Rename a topic (id, slug, or title); its slug follows").action(async(topic,title,_opts,cmd)=>{const scope=await categoryScope(cmd);const id=await resolveTopicId(scope,topic);const updated=await scope.client.patch(chatPath(...scope.base,"topics",id),{title});done(cmd,updated,`Renamed topic ${updated.id} to ${updated.title}`)})}function registerTags(chat){const tags=chat.command("tags").description("Organization prompt tags");tags.command("list",{isDefault:true}).description("List tags").action(async(_opts,cmd)=>{const{client,org}=await chatScope(cmd);const page=await client.get(chatPath("orgs",org,"tags"));printPage(page,resolveOutput(cmd),t=>({id:t.id,name:t.name,color:t.color,prompts:t.promptCount}))});tags.command("create <name>").alias("add").description("Create a tag").option("--color <color>","tag color").action(async(name,opts,cmd)=>{const{client,org}=await chatScope(cmd);const tag=await client.post(chatPath("orgs",org,"tags"),{name,...opts.color?{color:opts.color}:{}});done(cmd,tag,`Created tag ${tag.id} (${tag.name})`)});tags.command("rename <tag> <name>").description("Rename a tag (id or current name)").option("--color <color>","also change the color").action(async(tag,name,opts,cmd)=>{const scope=await chatScope(cmd);const[id]=await resolveTagIds(scope,[tag]);const updated=await scope.client.patch(chatPath("orgs",scope.org,"tags",id),{name,...opts.color?{color:opts.color}:{}});done(cmd,updated,`Renamed tag ${updated.id} to ${updated.name}`)});tags.command("delete <tag>").alias("rm").description("Delete a tag (id or name); prompts are kept").action(async(tag,_opts,cmd)=>{const scope=await chatScope(cmd);const[id]=await resolveTagIds(scope,[tag]);await scope.client.delete(chatPath("orgs",scope.org,"tags",id));done(cmd,{id,deleted:true},`Deleted tag ${id}`)})}function registerBrands(chat){withCategory(chat.command("brands")).description("Owned and competitor brands tracked in the category").action(async(_opts,cmd)=>{const scope=await categoryScope(cmd);const page=await scope.client.get(chatPath(...scope.base,"brands"));printPage(page,resolveOutput(cmd),b=>({id:b.id,role:b.role,name:b.name,website:b.website}))})}var CHAT_ORG_HELP="organization slug or id (default: gauge.json orgId, then your saved default, else your only Chat org; an org API key implies its org)";function scopeGlobalOptions(program2,chat){const isAgentsOnly=option=>option.long==="--local"||option.long==="--hosted";chat.configureHelp({...program2.configureHelp(),visibleGlobalOptions(cmd){return Help.prototype.visibleGlobalOptions.call(this,cmd).filter(option=>!isAgentsOnly(option))},optionDescription(option){return option.long==="--org"&&program2.options.includes(option)?CHAT_ORG_HELP:Help.prototype.optionDescription.call(this,option)}});chat.hook("preAction",()=>{const flag=program2.options.find(option=>isAgentsOnly(option)&&program2.getOptionValueSource(option.attributeName())==="cli");if(flag)throw new CliError("usage",`${flag.long} applies only to Agents commands; gauge chat always uses the hosted Chat API. Remove ${flag.long}.`)})}function register5(program2){const chat=program2.command("chat").description("Gauge Chat AI visibility: prompts, topics, tags, answers, citations, queries").addHelpText("after",`
|
|
26
26
|
Category-scoped commands take --category (slug or cat_ id; default: chatCategory
|
|
27
27
|
config, else the org's only category). Authenticate with \`gauge auth login\`,
|
|
28
28
|
or for scripts set GAUGE_API_TOKEN to an org API key (gauge_sk_\u2026, created in
|
|
@@ -33,10 +33,39 @@ an org API key implies its own org, and --org may pick one of an agency key's
|
|
|
33
33
|
clients. Pass -o json for
|
|
34
34
|
machine-readable output; errors carry the API error code. Exit codes: 0 ok,
|
|
35
35
|
1 error, 2 invalid request, 3 plan limit.
|
|
36
|
-
Rate limits per token and organization: 600 requests/min, 60 queries/min.`);scopeGlobalOptions(program2,chat);registerCategories(chat);registerTopics(chat);registerPrompts(chat);registerTags(chat);registerBrands(chat);registerQuery(chat);registerData(chat)}var cleanup_exports={};__export(cleanup_exports,{register:()=>register7});import{join as join5}from"path";function register7(program2){program2.command("cleanup").description("Clean interrupted Gauge sessions in this repository (local only); preserve results and images").action(async(_opts,command)=>{const state=gaugeState(command);if(state.mode!=="local"||!state.root)throw new CliError("usage","Use --local inside the batch's Git repository");const results=await cleanupLocal(join5(state.root,".gauge"));if(results.some(r=>r.status==="error"))process.exitCode=1;if(resolveOutput(command)==="json")printJson({sessions:results});else{if(!results.length)console.log("No interrupted sessions need cleanup.");for(const result of results)console.log(`${result.id}: ${result.status}${result.detail?` (${result.detail})`:""}`)}})}var config_exports={};__export(config_exports,{register:()=>register8});function assertKey(key){if(CONFIG_KEYS.includes(key)){return key}throw new CliError("usage",`Unknown config key '${key}'. Valid keys: ${CONFIG_KEYS.join(", ")}`)}function register8(program2){const config=program2.command("config").description("Local client config (~/.config/gauge/config.json)");config.command("get [key]").description("Print the whole config, or a single key's value").action(key=>{const current=readConfig();if(key===void 0){printJson(current);return}console.log(current[assertKey(key)]??"")});config.command("set <key> <value>").description(`Set a config key (${CONFIG_KEYS.join(", ")})`).action((key,value)=>{const configKey=assertKey(key);if(configKey==="defaultOrg"||configKey==="orgId"&&!isUuid(value))throw new CliError("usage","Select an organization with `gauge orgs use <slug>`.");if(configKey==="backend"&&value!=="local"&&value!=="hosted")throw new CliError("usage","backend must be local or hosted");updateConfig({[configKey]:value});console.log(`${configKey} = ${value} (${configPath()})`)});config.command("unset <key>").description("Remove a config key").action(key=>{const configKey=assertKey(key);updateConfig({[configKey]:void 0});console.log(`${configKey} unset (${configPath()})`)})}var connections_exports={};__export(connections_exports,{register:()=>register9});function connectionsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/connections`}function register9(program2){const group=hostedOnly(program2.command("connections")).description("Inspect the org's stored credentials (no secrets)");group.command("list").description("List credentials with the ids --connection accepts").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(connectionsPath(org));if(resolveOutput(cmd)==="json"){printJson(res.items);return}printItems(res.items.map(c=>({id:c.id,name:c.handle,vendor:c.profileLabel,placeholder:[c.credentialKey,c.basicPasswordKey].filter(Boolean).join(", "),domains:c.allowedHosts.join(", "),kind:c.kind,value:c.last4?`\xB7\xB7\xB7\xB7${c.last4}`:""})),"table")})}var dashboards_exports={};__export(dashboards_exports,{register:()=>register10});import{readFileSync as readFileSync3}from"fs";function dashboardsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/dashboards`}function register10(program2){const group=hostedOnly(program2.command("dashboards")).description("Saved dashboards (CRUD + view filters; layout is UI-only)");group.command("list").description("List your dashboard library").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(dashboardsPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(d=>({key:d.key,name:d.name,source:d.source,default:d.isUserDefault?"you":d.isOrgDefault?"org":""})),"table")});group.command("get <key>").description("Show one resolved dashboard (widgets + saved filters)").action(async(key,_opts,cmd)=>{const org=await resolveOrg(cmd);const dashboard=await createClient().get(`${dashboardsPath(org)}/${encodeURIComponent(key)}`);if(resolveOutput(cmd)==="json"){printJson(dashboard);return}console.log(`key: ${dashboard.key}`);console.log(`name: ${dashboard.name}`);console.log(`filters: ${JSON.stringify(dashboard.defaultFilters)}`);console.log("widgets:");for(const w of dashboard.widgets)console.log(` [${w.id}] ${w.title} (${w.viz})`)});group.command("create <name>").description("Create an empty personal dashboard").action(async(name,_opts,cmd)=>{const org=await resolveOrg(cmd);const created=await createClient().post(dashboardsPath(org),{name});if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created dashboard ${created.key}`)});group.command("rename <key> <name>").description("Rename a dashboard (in your overlay)").action(async(key,name,_opts,cmd)=>{const org=await resolveOrg(cmd);const updated=await createClient().patch(`${dashboardsPath(org)}/${encodeURIComponent(key)}`,{name});if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Renamed to ${updated.name}`)});group.command("rm <key>").description("Delete a dashboard (built-in templates refuse)").action(async(key,_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(`${dashboardsPath(org)}/${encodeURIComponent(key)}`);if(resolveOutput(cmd)!=="json")console.log(`Deleted dashboard ${key}`)});group.command("filters <key>").description("Replace a dashboard's saved view-level filters").option("--json <json>","the filters object inline").option("--file <path>",'the filters object from a file ("-" for stdin)').option("--clear","remove all saved filters").action(async(key,opts,cmd)=>{const org=await resolveOrg(cmd);const provided=[opts.json,opts.file,opts.clear].filter(v=>v!==void 0&&v!==false).length;if(provided!==1)throw new CliError("usage","pass exactly one of --json, --file, or --clear");let filters={};if(!opts.clear){const raw=opts.json??readFileSync3(opts.file==="-"?0:opts.file,"utf8");try{filters=JSON.parse(raw)}catch(e){throw new CliError("usage",`filters are not valid JSON: ${e instanceof Error?e.message:e}`)}}const updated=await createClient().put(`${dashboardsPath(org)}/${encodeURIComponent(key)}/filters`,filters);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Saved filters: ${JSON.stringify(updated.defaultFilters)}`)})}var eval_exports={};__export(eval_exports,{register:()=>register11});import{randomUUID as randomUUID3}from"crypto";import{join as join6}from"path";import{z as z3}from"zod";import{z as z2}from"zod";var sampleCountSchema=z2.number().int().min(1).max(50);var nextRunAtInputSchema=z2.string().regex(/^\d{4}-\d{2}-\d{2}/,"Next run must be a calendar date").nullable();var ownedCadenceSchema=z2.enum(["NONE","DAILY","WEEKLY","MONTHLY"]);var ownedConfigurationInputShape={repoUrl:z2.string().nullish(),repoRef:z2.string().nullish(),profileId:z2.string().nullish(),agents:z2.array(agentConfigInputSchema).min(1,"Select at least one agent"),skillRefs:assetRefListSchema.optional().default([]),mcpRefs:assetRefListSchema.optional().default([]),connectionIds:z2.array(z2.string().min(1)).max(32).optional().default([]),cadence:ownedCadenceSchema.optional().default("NONE"),nextRunAt:nextRunAtInputSchema.optional(),sampleCount:sampleCountSchema.optional().default(1)};var ownedConfigurationInputSchema=z2.object(ownedConfigurationInputShape).strict();var ownedConfigurationPatchShape={repoUrl:z2.string().nullish(),repoRef:z2.string().nullish(),profileId:z2.string().nullish(),agents:z2.array(agentConfigInputSchema).min(1,"Select at least one agent").optional(),skillRefs:assetRefListSchema.optional(),mcpRefs:assetRefListSchema.optional(),connectionIds:z2.array(z2.string().min(1)).max(32).optional(),cadence:ownedCadenceSchema.optional(),nextRunAt:nextRunAtInputSchema.optional(),sampleCount:sampleCountSchema.optional()};var ownedConfigurationSchema=z2.object({repoUrl:z2.string().nullable(),repoRef:z2.string().nullable(),profile:z2.object({id:z2.string(),name:z2.string()}).nullable(),agents:z2.array(agentConfigSchema),skillRefs:z2.array(z2.string()),mcpRefs:z2.array(z2.string()),connectionIds:z2.array(z2.string()),cadence:ownedCadenceSchema,sampleCount:sampleCountSchema,lastScheduledAt:z2.string().nullable(),nextRunAt:z2.string().nullable()});var visibilityScenarioSchema=z2.object({id:z2.string(),repoUrl:z2.string().nullable(),repoRef:z2.string().nullable(),profile:z2.object({id:z2.string(),name:z2.string()}).nullable()});var visibilityScenarioInputSchema=z2.object({repoUrl:z2.string().nullish(),repoRef:z2.string().nullish(),profileId:z2.string().nullish()}).strict();var visibilityPromptSettingsInputShape={agents:ownedConfigurationInputShape.agents,skillRefs:ownedConfigurationInputShape.skillRefs,mcpRefs:ownedConfigurationInputShape.mcpRefs,connectionIds:ownedConfigurationInputShape.connectionIds,cadence:ownedConfigurationInputShape.cadence,nextRunAt:ownedConfigurationInputShape.nextRunAt,sampleCount:ownedConfigurationInputShape.sampleCount};var visibilityPromptSettingsPatchShape={agents:ownedConfigurationPatchShape.agents,skillRefs:ownedConfigurationPatchShape.skillRefs,mcpRefs:ownedConfigurationPatchShape.mcpRefs,connectionIds:ownedConfigurationPatchShape.connectionIds,cadence:ownedConfigurationPatchShape.cadence,nextRunAt:ownedConfigurationPatchShape.nextRunAt,sampleCount:ownedConfigurationPatchShape.sampleCount};var visibilityPromptSettingsSchema=z2.object({agents:z2.array(agentConfigSchema),skillRefs:z2.array(z2.string()),mcpRefs:z2.array(z2.string()),connectionIds:z2.array(z2.string()),cadence:ownedCadenceSchema,sampleCount:sampleCountSchema,lastScheduledAt:z2.string().nullable(),nextRunAt:z2.string().nullable()});var runAgentsOverrideSchema=z2.array(agentConfigInputSchema).min(1,"Select at least one agent");var runSampleCountOverrideSchema=sampleCountSchema;var runEvalBodySchema=z2.object({agents:runAgentsOverrideSchema.optional(),sampleCount:runSampleCountOverrideSchema.optional()}).strict();var runVisibilityBodySchema=z2.object({agents:runAgentsOverrideSchema.optional(),sampleCount:runSampleCountOverrideSchema.optional(),visibilityScenarioIds:z2.array(z2.string().min(1)).min(1).optional()}).strict();var criterionSchema=z3.object({id:z3.string(),name:z3.string(),rubric:z3.string(),position:z3.number().int()});var criterionInputSchema=z3.object({id:z3.string().optional(),name:z3.string().min(1).optional(),rubric:z3.string().min(1)});var evalSetSchema=z3.object({id:z3.string(),suiteId:z3.string().nullable(),name:z3.string(),promptText:z3.string(),createdAt:z3.string(),createdBy:createdBySchema,criteria:z3.array(criterionSchema),tags:z3.array(z3.string())}).extend(ownedConfigurationSchema.shape);var agentPassRateSchema=z3.object({agent:agentSchema,model:z3.string(),runId:z3.string(),finishedAt:z3.string().nullable(),passed:z3.number().int(),total:z3.number().int()});var evalSetStatSchema=z3.object({passRates:z3.array(agentPassRateSchema),runCount:z3.number().int(),judgedRuns:z3.number().int(),passedRuns:z3.number().int()});var csv3=z3.string().transform(s=>s.split(",").map(x=>x.trim()).filter(Boolean));var listEvalSetsQuerySchema=z3.object({agents:csv3.pipe(z3.array(agentSchema)).optional(),models:csv3.optional(),since:z3.coerce.date().optional()});var evalSetListItemSchema=evalSetSchema.extend({stats:evalSetStatSchema});var listEvalSetsResponseSchema=z3.object({items:z3.array(evalSetListItemSchema)});var createEvalSetBodySchema=z3.object({name:z3.string().min(1),promptText:z3.string().min(1),criteria:z3.array(criterionInputSchema).min(1),tags:z3.array(z3.string()).optional()}).extend(ownedConfigurationInputSchema.shape).strict();var updateEvalSetBodySchema=z3.object({name:z3.string().min(1).optional(),promptText:z3.string().min(1).optional(),criteria:z3.array(criterionInputSchema).min(1).optional(),tags:z3.array(z3.string()).optional()}).extend(ownedConfigurationPatchShape).strict();var runEvalSetResponseSchema=z3.object({batchIds:z3.array(z3.string()),runIds:z3.array(z3.string()),runRequestId:z3.string().optional()});var evalSetRunSchema=z3.object({id:z3.string(),status:z3.string(),agent:agentSchema,model:z3.string().nullable(),resolvedModel:z3.string().nullable(),createdAt:z3.string(),startedAt:z3.string().nullable(),finishedAt:z3.string().nullable(),failureReason:z3.string().nullable(),creditsCharged:runCreditsChargedSchema.optional(),triggeredBy:runTriggeredBySchema.optional(),evalRevision:z3.string().optional(),judge:z3.object({status:z3.string(),error:z3.string().nullable()}).nullable(),criterionResults:z3.array(z3.object({criterionId:z3.string().nullable(),name:z3.string(),passed:z3.boolean().nullable(),reasoning:z3.string().nullable()}))});var listEvalSetRunsResponseSchema=listResponseSchema(evalSetRunSchema);import{z as z4}from"zod";var VISIBILITY_PROMPT_KINDS=["HEAD_TO_HEAD","RECOMMEND"];var visibilityPromptKindSchema=z4.enum(VISIBILITY_PROMPT_KINDS);var visibilityPromptSchema=visibilityPromptSettingsSchema.extend({id:z4.string(),name:z4.string(),kind:visibilityPromptKindSchema,promptText:z4.string(),branded:z4.boolean(),brandA:z4.string().nullable(),brandB:z4.string().nullable(),market:z4.object({id:z4.string(),slug:z4.string(),name:z4.string()}).nullable(),position:z4.number().int(),createdAt:z4.string(),createdBy:createdBySchema,tags:z4.array(z4.string()),runCount:z4.number().int(),lastRunAt:z4.string().nullable(),scenarios:z4.array(visibilityScenarioSchema)});var listVisibilityPromptsResponseSchema=z4.object({items:z4.array(visibilityPromptSchema)});var createVisibilityPromptBodySchema=z4.object({name:z4.string().min(1),promptText:z4.string().min(1),kind:visibilityPromptKindSchema,branded:z4.boolean().optional(),brandA:z4.string().nullish(),brandB:z4.string().nullish(),marketId:z4.string().nullish(),tags:z4.array(z4.string()).optional(),agents:visibilityPromptSettingsInputShape.agents.optional(),skillRefs:visibilityPromptSettingsInputShape.skillRefs,mcpRefs:visibilityPromptSettingsInputShape.mcpRefs,connectionIds:visibilityPromptSettingsInputShape.connectionIds,cadence:visibilityPromptSettingsInputShape.cadence,sampleCount:visibilityPromptSettingsInputShape.sampleCount,scenarios:z4.array(visibilityScenarioInputSchema).optional().default([])}).strict();var updateVisibilityPromptBodySchema=z4.object({name:z4.string().min(1).optional(),promptText:z4.string().min(1).optional(),branded:z4.boolean().optional(),brandA:z4.string().nullish(),brandB:z4.string().nullish(),marketId:z4.string().nullish(),tags:z4.array(z4.string()).optional(),...visibilityPromptSettingsPatchShape,scenarios:z4.array(visibilityScenarioInputSchema).optional()}).strict();var runVisibilityPromptResponseSchema=z4.object({batchIds:z4.array(z4.string()),runIds:z4.array(z4.string())});var visibilityPromptRunSchema=z4.object({id:z4.string(),status:z4.string(),agent:agentSchema,model:z4.string().nullable(),resolvedModel:z4.string().nullable(),createdAt:z4.string(),startedAt:z4.string().nullable(),finishedAt:z4.string().nullable(),failureReason:z4.string().nullable(),creditsCharged:runCreditsChargedSchema.optional(),triggeredBy:runTriggeredBySchema.optional(),verdict:z4.object({status:z4.string(),error:z4.string().nullable(),payload:z4.unknown().nullable()}).nullable()});var listVisibilityPromptRunsResponseSchema=listResponseSchema(visibilityPromptRunSchema);var orgMarketSchema=z4.object({marketId:z4.string(),slug:z4.string(),name:z4.string(),capability:z4.string(),source:z4.string()});var addableMarketSchema=orgMarketSchema.omit({source:true});var listMarketsResponseSchema=z4.object({tracked:z4.array(orgMarketSchema),addable:z4.array(addableMarketSchema)});var addMarketBodySchema=z4.object({marketId:z4.string().min(1)});import{readFileSync as readFileSync4}from"fs";function parseLimit(raw){const limit=Number(raw);if(!Number.isInteger(limit)||limit<1||limit>200){throw new CliError("usage","--limit must be an integer between 1 and 200")}return limit}function collect5(value,previous){return[...previous,value]}function once(flag){return(value,previous)=>{if(previous!==void 0)throw new CliError("usage",`${flag} takes one id; start one run per id`);return value}}var AGENT_ALIASES={"claude-code":"CLAUDE_CODE",claude_code:"CLAUDE_CODE",claude:"CLAUDE_CODE",codex:"CODEX_CLI","codex-cli":"CODEX_CLI",codex_cli:"CODEX_CLI",cursor:"CURSOR",pi:"PI",opencode:"OPENCODE","open-code":"OPENCODE",open_code:"OPENCODE",antigravity:"ANTIGRAVITY"};function parseAgents(raw){const out=[];for(const part of raw.split(",")){const name=part.trim();if(!name)continue;const agent=AGENT_ALIASES[name.toLowerCase()];if(!agent){throw new CliError("usage",`unknown agent "${name}" (expected: claude-code, codex, cursor, pi, opencode, antigravity)`)}out.push(agent)}if(out.length===0)throw new CliError("usage","no agents given");return[...new Set(out)]}function parseAgentSpec(raw){const colon=raw.indexOf(":");const name=(colon<0?raw:raw.slice(0,colon)).trim();const agent=AGENT_ALIASES[name.toLowerCase()];if(!agent){throw new CliError("usage",`unknown agent "${name}" in --agent "${raw}" (expected: claude-code, codex, pi, opencode, or antigravity, optionally ":model1,model2")`)}if(agent==="CURSOR"){throw new CliError("usage","cursor is not schedulable \u2014 presets may roster claude-code, codex, pi, opencode, and hosted antigravity")}const models=colon<0?[]:raw.slice(colon+1).split(",").map(m=>m.trim()).filter(Boolean);if(colon>=0&&models.length===0){throw new CliError("usage",`--agent "${raw}" has a ":" but no models \u2014 drop the colon for the agent default`)}return{agent,models:[...new Set(models)]}}function buildAgentRoster(specs){const byAgent=new Map;for(const raw of specs){const{agent,models}=parseAgentSpec(raw);const set=byAgent.get(agent)??new Set;for(const m of models)set.add(m);byAgent.set(agent,set)}return[...byAgent.entries()].map(([agent,models])=>({agent,models:[...models]}))}function validateAssetRefs(kind,refs){const parsed=runConfig_exports.assetRefListSchema.safeParse(refs);if(!parsed.success){const first=parsed.error.issues[0];throw new CliError("usage",`invalid --${kind} reference: ${first?.message??"expected name@label"}`)}return refs}function readPromptText(opts){if(opts.text===void 0===(opts.file===void 0)){throw new CliError("usage","pass exactly one of --text or --file")}if(opts.text!==void 0)return opts.text;const text=readFileSync4(opts.file==="-"?0:opts.file,"utf8").trim();if(!text)throw new CliError("usage","prompt file is empty");return text}function parseCadence(raw){const cadence=(raw??"NONE").trim().toUpperCase();if(!["NONE","DAILY","WEEKLY","MONTHLY"].includes(cadence))throw new CliError("usage",`invalid cadence "${raw}" (expected none, daily, weekly, or monthly)`);return cadence}function parseSampleCount(raw){const value=Number(raw??"1");if(!Number.isSafeInteger(value)||value<1||value>50)throw new CliError("usage","--samples must be an integer from 1 to 50");return value}function ownedConfigFromOpts(opts){if(opts.ref!==void 0&&opts.repo===void 0)throw new CliError("usage","--ref requires --repo");if(opts.agent.length===0)throw new CliError("usage","at least one --agent is required");return{repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:opts.persona??null,agents:buildAgentRoster(opts.agent),skillRefs:validateAssetRefs("skill",opts.skill),mcpRefs:validateAssetRefs("mcp",opts.mcp),connectionIds:opts.connection,cadence:parseCadence(opts.cadence),sampleCount:parseSampleCount(opts.samples)}}function ownedConfigPatchFromOpts(opts){const patch={};if((opts.repo!==void 0||opts.ref!==void 0)&&opts.clearRepo)throw new CliError("usage","--repo/--ref and --clear-repo conflict");if(opts.ref!==void 0&&opts.clearRef)throw new CliError("usage","--ref and --clear-ref conflict");if(opts.clearRepo){patch.repoUrl=null;patch.repoRef=null}else{if(opts.repo!==void 0)patch.repoUrl=opts.repo;if(opts.clearRef)patch.repoRef=null;else if(opts.ref!==void 0)patch.repoRef=opts.ref}if(opts.persona!==void 0&&opts.clearPersona)throw new CliError("usage","--persona and --clear-persona conflict");if(opts.clearPersona)patch.profileId=null;else if(opts.persona!==void 0)patch.profileId=opts.persona;if(opts.agent.length>0)patch.agents=buildAgentRoster(opts.agent);if(opts.skill.length>0&&opts.clearSkills)throw new CliError("usage","--skill and --clear-skills conflict");if(opts.clearSkills)patch.skillRefs=[];else if(opts.skill.length>0)patch.skillRefs=validateAssetRefs("skill",opts.skill);if(opts.mcp.length>0&&opts.clearMcp)throw new CliError("usage","--mcp and --clear-mcp conflict");if(opts.clearMcp)patch.mcpRefs=[];else if(opts.mcp.length>0)patch.mcpRefs=validateAssetRefs("mcp",opts.mcp);if(opts.connection.length>0&&opts.clearConnections)throw new CliError("usage","--connection and --clear-connections conflict");if(opts.clearConnections)patch.connectionIds=[];else if(opts.connection.length>0)patch.connectionIds=opts.connection;if(opts.cadence!==void 0)patch.cadence=parseCadence(opts.cadence);if(opts.samples!==void 0)patch.sampleCount=parseSampleCount(opts.samples);return patch}function addRunSettingsCreateOptions(command){return command.option("--agent <agent[:m1,m2]>","agent and optional model pins (repeatable)",collect5,[]).option("--skill <name@label>","staged skill (repeatable)",collect5,[]).option("--mcp <name@label>","MCP server (repeatable)",collect5,[]).option("--connection <id>","connection id (repeatable)",collect5,[]).option("--cadence <cadence>","none | daily | weekly | monthly","none").option("--samples <n>","sample count from 1 to 50","1")}function addRunSettingsEditOptions(command){return command.option("--agent <agent[:m1,m2]>","replace agents and model pins (repeatable)",collect5,[]).option("--skill <name@label>","replace skills (repeatable)",collect5,[]).option("--clear-skills","remove all skills").option("--mcp <name@label>","replace MCP servers (repeatable)",collect5,[]).option("--clear-mcp","remove all MCP servers").option("--connection <id>","replace connections (repeatable)",collect5,[]).option("--clear-connections","remove all connections").option("--cadence <cadence>","none | daily | weekly | monthly").option("--samples <n>","sample count from 1 to 50")}function hasRunSettingsFlags(opts){return opts.agent.length>0||opts.skill.length>0||opts.mcp.length>0||opts.connection.length>0||opts.cadence.toUpperCase()!=="NONE"||opts.samples!=="1"}function runSettingsFromOpts(opts){return{...opts.agent.length>0?{agents:buildAgentRoster(opts.agent)}:{},skillRefs:validateAssetRefs("skill",opts.skill),mcpRefs:validateAssetRefs("mcp",opts.mcp),connectionIds:opts.connection,cadence:parseCadence(opts.cadence),sampleCount:parseSampleCount(opts.samples)}}function runSettingsPatchFromOpts(opts){return ownedConfigPatchFromOpts(opts)}function runOverridesFromOpts(opts){return{...opts.agent.length>0?{agents:buildAgentRoster(opts.agent)}:{},...opts.samples!==void 0?{sampleCount:parseSampleCount(opts.samples)}:{}}}function cadenceLabel(cadence){return cadence==="NONE"?"manual":cadence.toLowerCase()}function localDirectory2(command){const state=gaugeState(command);if(!state.root)throw new CliError("usage","No Gauge workspace found. Run `gauge init` in this directory first.");if(state.config===null)throw new CliError("usage","gauge.json is not valid JSON");return join6(state.root,".gauge")}function isLocal(command){return gaugeState(command).mode==="local"}function withLocalStore(command,fn){const store=openLocalStore(localDirectory2(command));try{return fn(store)}finally{store.close()}}function normalize(kind,raw,previous){const{inputs:_inputs,...definition}=raw;const scenarioInputs=Array.isArray(definition.scenarios)?definition.scenarios:void 0;if(scenarioInputs)definition.scenarios=scenarioInputs.map(({id:_id,...scenario})=>scenario);const parsed=(kind==="eval"?createEvalSetBodySchema:createVisibilityPromptBodySchema).safeParse(definition);if(!parsed.success)throw new CliError("usage",`Invalid ${kind}: ${parsed.error.issues.map(i=>`${i.path.join(".")}: ${i.message}`).join("; ")}`);const value=parsed.data;value.name=value.name.trim();value.promptText=value.promptText.trim();if(!value.name||!value.promptText)throw new CliError("usage","A measurement needs a nonempty name and prompt.");if(value.cadence&&value.cadence!=="NONE")throw new CliError("usage","Local measurements do not support managed schedules. Use --cadence none or --hosted.");if(value.nextRunAt)throw new CliError("usage","Local measurements do not support scheduled start times. Use --hosted.");assertLocalCapabilities(value);value.tags=[...new Set((value.tags??[]).map(tag=>tag.trim()).filter(Boolean))];if(kind==="eval"){if(value.repoRef&&!value.repoUrl)throw new CliError("usage","A repository ref requires a repository URL.");const ids=new Set;const existing=new Set(previous?.criteria?.map(criterion=>criterion.id));value.criteria=value.criteria.map(c=>{const rubric=c.rubric.trim();if(!rubric)throw new CliError("usage","Every criterion needs a nonempty rubric.");if(c.id&&previous&&!existing.has(c.id))throw new CliError("usage",`Unknown criterion ID: ${c.id}. Omit id to add a new criterion.`);const id=c.id??`cr_${randomUUID3()}`;if(!id||ids.has(id))throw new CliError("usage",`Duplicate or empty criterion ID: ${id}`);ids.add(id);return{id,name:c.name?.trim()||deriveCriterionName(rubric),rubric}})}else{value.brandA=typeof value.brandA==="string"?value.brandA.trim()||null:null;value.brandB=typeof value.brandB==="string"?value.brandB.trim()||null:null;if(value.kind==="HEAD_TO_HEAD"){if(!value.brandA||!value.brandB||String(value.brandA).toLowerCase()===String(value.brandB).toLowerCase())throw new CliError("usage","Head-to-head preference needs two different --brand-a and --brand-b values");[value.brandA,value.brandB]=[String(value.brandA),String(value.brandB)].sort((a,b)=>a.toLowerCase().localeCompare(b.toLowerCase()))}value.branded=value.kind==="HEAD_TO_HEAD"||Boolean(value.branded);value.agents??=[{agent:"CLAUDE_CODE",models:[]},{agent:"CODEX_CLI",models:[]}];const scenarios=value.scenarios;if(!previous&&scenarios.length===0)scenarios.push({repoUrl:null,repoRef:null,profileId:null});if(scenarios.some(s=>s.profileId))throw new CliError("usage","Local scenarios do not support hosted personas");const seen=new Set;const ids=new Set;const existing=new Set(previous?.scenarios?.map(scenario=>scenario.id));value.scenarios=scenarios.map((scenario,index)=>{const repoUrl=typeof scenario.repoUrl==="string"?scenario.repoUrl.trim()||null:null;const repoRef=typeof scenario.repoRef==="string"?scenario.repoRef.trim()||null:null;if(repoRef&&!repoUrl)throw new CliError("usage","A scenario ref requires a repository URL.");const key=JSON.stringify([repoUrl,repoRef]);if(seen.has(key))throw new CliError("usage","This prompt already has the same scenario.");seen.add(key);const suppliedId=scenarioInputs?.[index]?.id;if(suppliedId!==void 0&&(typeof suppliedId!=="string"||!previous||!existing.has(suppliedId)))throw new CliError("usage",`Unknown scenario ID: ${suppliedId}. Omit id to add a scenario.`);const id=typeof suppliedId==="string"?suppliedId:`sc_${randomUUID3()}`;if(ids.has(id))throw new CliError("usage",`Duplicate scenario ID: ${id}`);ids.add(id);return{id,repoUrl,repoRef,profileId:null}})}return value}function editable(row2){const{id,revision,measurementKind,createdAt,updatedAt,...definition}=row2;return definition}function createLocalMeasurement(command,kind,definition){return withLocalStore(command,store=>store.createMeasurement(kind,normalize(kind,definition)))}function editLocalMeasurement(command,kind,id,patch){return withLocalStore(command,store=>{const current=store.requireMeasurement(id,kind);const value=normalize(kind,{...editable(current),...patch},current);if(!("criteria"in patch)&¤t.criteria)value.criteria=current.criteria;if(!("scenarios"in patch)&¤t.scenarios)value.scenarios=current.scenarios;return store.updateMeasurement(id,value)})}function getLocalMeasurement(command,kind,id){return withLocalStore(command,store=>measurementWithStats(store,store.requireMeasurement(id,kind)))}function removeLocalMeasurement(command,kind,id){withLocalStore(command,store=>{store.requireMeasurement(id,kind);store.deleteMeasurement(id)})}function localMeasurementView(row2){return{profile:null,suiteId:null,market:null,position:0,runCount:0,lastRunAt:null,cadence:"NONE",sampleCount:1,repoUrl:null,repoRef:null,agents:[],skillRefs:[],mcpRefs:[],connectionIds:[],stats:{passRates:[],runCount:0,judgedRuns:0,passedRuns:0},...row2,scenarios:Array.isArray(row2.scenarios)?row2.scenarios.map(s=>({...s,profile:null})):[]}}function listLocalMeasurements(command,kind){return withLocalStore(command,store=>store.listMeasurements(kind).map(row2=>localMeasurementView(measurementWithStats(store,row2))))}function measurementWithStats(store,row2){const sessions=store.listSessions({measurementId:row2.id});const stats={runCount:sessions.length,judgedRuns:0,passedRuns:0,passRates:[]};for(const session of sessions){const analysis=readCurrentAnalysis(store,session.id);if(!analysis)continue;const outcomes=analysis.criteria.map(criterion=>criterion.outcome);const verdict=outcomes.includes("FAIL")?false:!outcomes.length||outcomes.includes("UNASSESSABLE")?null:true;if(verdict===null)continue;stats.judgedRuns++;if(verdict)stats.passedRuns++;if(!stats.passRates.some(rate=>rate.agent===session.agent))stats.passRates.push({agent:session.agent,model:session.model,runId:session.id,finishedAt:session.endedAt,passed:outcomes.filter(outcome2=>outcome2==="PASS").length,total:outcomes.length})}return{...row2,runCount:sessions.length,lastRunAt:sessions.map(session=>session.createdAt).sort().at(-1)??null,stats}}function deriveCriterionName(rubric){const trimmed=rubric.replace(/^\s*(?:[-*•]|\(?\d+[.)])\s+/,"").trim();const first=trimmed.split(/(?<=[.!?])\s|\n/)[0]?.trim()??"";return(new RegExp("\\p{L}","u").test(first)?first:trimmed.replace(/\s+/g," ")).replace(/[.!?]+$/,"").trim()}function localMeasurementSpec(command,kind,id,opts){const row2=getLocalMeasurement(command,kind,id);const stored=row2.inputs;if(stored?.length&&!opts.inputs.length)console.error(`note: saved inputs are no longer attached (${stored.join(", ")}). Pass --input <name> at launch.`);const settings={...row2,...runOverridesFromOpts(opts)};const agents=settings.agents;if(!agents?.length)throw new CliError("usage","Choose at least one agent with --agent before launching this local measurement");let scenarios=kind==="preference"?settings.scenarios:[{repoUrl:settings.repoUrl??null,repoRef:settings.repoRef??null}];if(!scenarios.length)throw new CliError("usage","This preference prompt has no scenarios. Add one with gauge preference scenarios add <id> before running it.");if(opts.scenario?.length){for(const selected of opts.scenario)if(!scenarios.some(s=>s.id===selected))throw new CliError("usage",`Unknown local scenario: ${selected}`);scenarios=scenarios.filter(s=>s.id!==void 0&&opts.scenario.includes(s.id))}const cases=scenarios.map((scenario,index)=>({key:`${kind}-${index+1}`,prompt:String(settings.promptText),...kind==="eval"?{criteria:settings.criteria.map(c=>({name:c.name,rubric:c.rubric}))}:{preference:settings.kind==="HEAD_TO_HEAD"?{kind:"head-to-head",brandA:String(settings.brandA),brandB:String(settings.brandB)}:{kind:"open-ended",branded:Boolean(settings.branded)}},config:{agents,sampleCount:Number(settings.sampleCount??1),repoUrl:scenario.repoUrl,repoRef:scenario.repoRef,inputs:opts.inputs}}));return{spec:{version:2,cases},measurement:{id,kind,revision:row2.revision,...kind==="eval"?{criterionIds:row2.criteria.map(c=>c.id)}:{}}}}import{readFileSync as readFileSync5}from"fs";import{createInterface as createInterface3}from"readline";async function confirmOrAbort(question,yes){if(yes)return;if(!process.stdin.isTTY){throw new CliError("usage","refusing to proceed without confirmation on a non-interactive stdin \u2014 pass --yes")}const rl=createInterface3({input:process.stdin,output:process.stderr});const answer=await new Promise(resolve5=>rl.question(`${question} [y/N] `,resolve5));rl.close();if(!/^y(es)?$/i.test(answer.trim())){throw new CliError("usage","aborted")}}function creatorLabel(createdBy){return createdBy?.name||createdBy?.email||"--"}function triggeredByLabel(t){if(!t)return"";return t.email?`${t.source} (${t.email})`:t.source}async function filterByCreator(items,createdBy){if(createdBy===void 0)return items;const wanted=createdBy.trim().toLowerCase();if(wanted==="me"){const{user}=await createClient().get("/api/v1/me");const email=user.email?.toLowerCase();return items.filter(item=>item.createdBy?.userId===user.id||email!==void 0&&item.createdBy?.email?.toLowerCase()===email)}return items.filter(item=>item.createdBy?.email?.toLowerCase()===wanted)}function evalSetsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/eval-sets`}function evalSetPath(org,id){return`${evalSetsPath(org)}/${encodeURIComponent(id)}`}function parseCriterion(raw){const idx=raw.indexOf(":");if(idx<=0||!raw.slice(idx+1).trim()){throw new CliError("usage",`invalid --criterion "${raw}" (expected "Name: rubric text")`)}return{name:raw.slice(0,idx).trim(),rubric:raw.slice(idx+1).trim()}}function readCriteriaFile(path){let parsed;try{parsed=JSON.parse(readFileSync5(path==="-"?0:path,"utf8"))}catch(e){throw new CliError("usage",`could not read --criteria-file: ${e instanceof Error?e.message:e}`)}if(!Array.isArray(parsed)||parsed.length===0)throw new CliError("usage","--criteria-file must be a non-empty JSON array of {id?, name, rubric}");return parsed}function criteriaFromOpts(opts){if(opts.criterion.length>0&&opts.criteriaFile!==void 0)throw new CliError("usage","--criterion and --criteria-file conflict");if(opts.criteriaFile!==void 0)return readCriteriaFile(opts.criteriaFile);if(opts.criterion.length>0)return opts.criterion.map(parseCriterion);return void 0}function repoLabel(e){if(!e.repoUrl)return"(default scaffold)";return e.repoRef?`${e.repoUrl}@${e.repoRef}`:e.repoUrl}function passRateLabel(stats){if(stats.judgedRuns===0)return"-";return`${Math.round(stats.passedRuns/stats.judgedRuns*100)}% (${stats.passedRuns}/${stats.judgedRuns})`}function evalSetRow(e){return{id:e.id,name:e.name,repo:repoLabel(e),criteria:e.criteria.length,tags:e.tags.join(", "),"pass rate":passRateLabel(e.stats),runs:e.stats.runCount,"created by":creatorLabel(e.createdBy)}}function printDetail(e){console.log(`id: ${e.id}`);console.log(`name: ${e.name}`);console.log(`repo: ${repoLabel(e)}`);console.log(`persona: ${e.profile?e.profile.name:"(default persona)"}`);console.log(`tags: ${e.tags.join(", ")||"(none)"}`);console.log(`cadence: ${cadenceLabel(e.cadence)}`);console.log(`samples: ${e.sampleCount}`);console.log(`agents: ${e.agents.map(agent=>`${agent.agent}${agent.models.length>0?`:${agent.models.join(",")}`:""}`).join("; ")}`);console.log(`skills: ${e.skillRefs.join(", ")||"(none)"}`);console.log(`mcp: ${e.mcpRefs.join(", ")||"(none)"}`);console.log(`conns: ${e.connectionIds.join(", ")||"(none)"}`);console.log(`creator: ${creatorLabel(e.createdBy)}`);console.log(`prompt: ${e.promptText}`);console.log("criteria:");for(const c of e.criteria)console.log(` [${c.id}] ${c.name}: ${c.rubric}`)}async function listEvalSets(cmd,opts){if(isLocal(cmd)){if(opts.agents||opts.models||opts.since)throw new CliError("usage","Local eval list does not support stats filters yet; use gauge query --local.");return listLocalMeasurements(cmd,"eval")}const res=await createClient().get(evalSetsPath(await resolveOrg(cmd)),{agents:opts.agents===void 0?void 0:parseAgents(opts.agents).join(","),models:opts.models,since:opts.since});return res.items}async function getEvalSet(cmd,id){const evalSet=isLocal(cmd)?localMeasurementView(getLocalMeasurement(cmd,"eval",id)):await createClient().get(evalSetPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson(evalSet);else printDetail(evalSet)}async function createEvalSet(cmd,opts){const criteria=criteriaFromOpts(opts);if(!criteria)throw new CliError("usage","at least one --criterion (or --criteria-file) is required");const body={name:opts.name,promptText:readPromptText(opts),...ownedConfigFromOpts(opts),criteria,tags:opts.tag.length>0?opts.tag:void 0};if(!isLocal(cmd)&&opts.repo===void 0)console.error("note: no --repo \u2014 this eval set will run against the default Gauge scaffold, NOT your repository. Use `gauge eval edit <id> --repo <url>` to retarget.");const created=isLocal(cmd)?createLocalMeasurement(cmd,"eval",body):await createClient().post(evalSetsPath(await resolveOrg(cmd)),body);if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created eval set ${created.id}`)}async function editEvalSet(cmd,id,opts){const body=ownedConfigPatchFromOpts(opts);if(opts.name!==void 0)body.name=opts.name;if(opts.text!==void 0||opts.file!==void 0)body.promptText=readPromptText(opts);const criteria=criteriaFromOpts(opts);if(criteria)body.criteria=criteria;if(opts.tag.length>0&&opts.clearTags)throw new CliError("usage","--tag and --clear-tags conflict");if(opts.clearTags)body.tags=[];else if(opts.tag.length>0)body.tags=opts.tag;if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");const updated=isLocal(cmd)?editLocalMeasurement(cmd,"eval",id,body):await createClient().patch(evalSetPath(await resolveOrg(cmd),id),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated eval set ${updated.id}`)}async function removeEvalSet(cmd,id){if(isLocal(cmd))removeLocalMeasurement(cmd,"eval",id);else await createClient().delete(evalSetPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)!=="json")console.log(`Deleted eval set ${id}`)}async function runEvalSet(cmd,id,opts){const org=await resolveOrg(cmd);const overrides=runOverridesFromOpts(opts);await confirmOrAbort(Object.keys(overrides).length>0?"Run this eval now with these overrides? The stored configuration stays unchanged.":"Run this eval's stored configuration now?",opts.yes);const result=await createClient().post(`${evalSetPath(org,id)}/run`,overrides);if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`Started ${result.runIds.length} run${result.runIds.length===1?"":"s"} in batch ${result.batchIds[0]??""}`);for(const batch of result.batchIds)console.log(`Runs: gauge run list --batch ${batch}`);if(result.runRequestId)console.log(`Wait: gauge --org ${org} run wait ${result.runRequestId}`)}}function register11(program2){const evals=program2.command("eval").description("Manage eval sets: saved prompts with judged criteria. Launch one with `gauge run --eval <id>`");evals.command("list").description("List eval sets").option("--agents <agents>","stats for these agents only (csv)").option("--models <models>","stats for these model slugs only (csv)").option("--since <date>","stats for runs since this ISO date only").option("--created-by <email|me>","only evals created by this user").action(async(opts,cmd)=>{const items=await filterByCreator(await listEvalSets(cmd,opts),opts.createdBy);if(resolveOutput(cmd)==="json")printJson(items);else printItems(items.map(evalSetRow),"table")});evals.command("get <id>").description("Show one eval set").action(async(id,_opts,cmd)=>{await getEvalSet(cmd,id)});addRunSettingsCreateOptions(evals.command("create").description("Create an eval set").requiredOption("--name <name>","name").option("--text <text>","prompt text inline").option("--file <path>",'prompt text from a file ("-" for stdin)').option("--repo <url>","consumer repository (default: empty locally, Gauge scaffold hosted)").option("--ref <ref>","branch, tag, or commit (with --repo)").option("--persona <profileId>","persona profile id")).option("--criterion <spec>",'judged criterion as "Name: rubric text" (repeatable)',collect5,[]).option("--criteria-file <path>",'JSON array of {name, rubric} ("-" for stdin)').option("--tag <name>","user tag name, create-on-type (repeatable)",collect5,[]).action(async(opts,cmd)=>{await createEvalSet(cmd,opts)});addRunSettingsEditOptions(evals.command("edit <id>").description("Update an eval set. Omitted flags keep their value. A criteria list REPLACES the whole set (use --criteria-file with ids to edit in place).").option("--name <name>","rename it").option("--text <text>","replace the prompt text inline").option("--file <path>",'replace the prompt text from a file ("-" stdin)').option("--repo <url>","set the repository (keeps a ref pin)").option("--ref <ref>","set the branch, tag, or commit pin").option("--clear-ref","drop the ref pin").option("--clear-repo","use an empty workspace locally or the default scaffold hosted").option("--persona <profileId>","set the persona profile id").option("--clear-persona","use the default persona")).option("--criterion <spec>",'criterion as "Name: rubric", replaces the whole set (repeatable)',collect5,[]).option("--criteria-file <path>","JSON array of {id?, name, rubric}; entries with id update in place").option("--tag <name>","replace the user tag set (repeatable)",collect5,[]).option("--clear-tags","remove all user tags").action(async(id,opts,cmd)=>{await editEvalSet(cmd,id,opts)});evals.command("rm <id>").description("Delete an eval set (run history stays)").action(async(id,_opts,cmd)=>{await removeEvalSet(cmd,id)})}var feedback_exports2={};__export(feedback_exports2,{register:()=>register12});import{Option as Option3}from"commander";function register12(program2){program2.command("feedback").description("Send feedback about this CLI to the Gauge team: friction you hit, missing commands or flags, bugs").requiredOption("-m, --message <text>","what happened").addOption(new Option3("--kind <kind>","feedback type").choices([...feedback_exports.FEEDBACK_KINDS]).default("other")).option("--command <command>","the gauge command the feedback is about").action(async(opts,cmd)=>{const{org}=cmd.optsWithGlobals();const client=createClient({authOptional:true});const body={message:opts.message,kind:opts.kind,command:opts.command,org:client.signedIn?org?(await resolveOrgValue(org)).id:readConfig().orgId:void 0};const res=await client.post("/api/v1/feedback",body);if(resolveOutput(cmd)==="json")printJson(res);else console.log(`Feedback sent (${res.id}${client.signedIn?"":", anonymous"}). Thank you.`)})}var instructions_exports={};__export(instructions_exports,{register:()=>register13});var EVAL=`---
|
|
36
|
+
Rate limits per token and organization: 600 requests/min, 60 queries/min.`);scopeGlobalOptions(program2,chat);registerCategories(chat);registerTopics(chat);registerPrompts(chat);registerTags(chat);registerBrands(chat);registerQuery(chat);registerData(chat)}var cleanup_exports={};__export(cleanup_exports,{register:()=>register6});import{join as join5}from"path";function register6(program2){program2.command("cleanup").description("Clean interrupted Gauge sessions in this repository (local only); preserve results and images").action(async(_opts,command)=>{const state=gaugeState(command);if(state.mode!=="local"||!state.root)throw new CliError("usage","Use --local inside the batch's Git repository");const results=await cleanupLocal(join5(state.root,".gauge"));if(results.some(r=>r.status==="error"))process.exitCode=1;if(resolveOutput(command)==="json")printJson({sessions:results});else{if(!results.length)console.log("No interrupted sessions need cleanup.");for(const result of results)console.log(`${result.id}: ${result.status}${result.detail?` (${result.detail})`:""}`)}})}var config_exports={};__export(config_exports,{register:()=>register7});function assertKey(key){if(CONFIG_KEYS.includes(key)){return key}throw new CliError("usage",`Unknown config key '${key}'. Valid keys: ${CONFIG_KEYS.join(", ")}`)}function register7(program2){const config=program2.command("config").description("Local client config (~/.config/gauge/config.json)");config.command("get [key]").description("Print the whole config, or a single key's value").action(key=>{const current=readConfig();if(key===void 0){printJson(current);return}console.log(current[assertKey(key)]??"")});config.command("set <key> <value>").description(`Set a config key (${CONFIG_KEYS.join(", ")})`).action((key,value)=>{const configKey=assertKey(key);if(configKey==="defaultOrg"||configKey==="orgId"&&!isUuid(value))throw new CliError("usage","Select an organization with `gauge orgs use <slug>`.");if(configKey==="backend"&&value!=="local"&&value!=="hosted")throw new CliError("usage","backend must be local or hosted");updateConfig({[configKey]:value});console.log(`${configKey} = ${value} (${configPath()})`)});config.command("unset <key>").description("Remove a config key").action(key=>{const configKey=assertKey(key);updateConfig({[configKey]:void 0});console.log(`${configKey} unset (${configPath()})`)})}var connections_exports={};__export(connections_exports,{register:()=>register8});function connectionsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/connections`}function register8(program2){const group=hostedOnly(program2.command("connections")).description("Inspect the org's stored credentials (no secrets)");group.command("list").description("List credentials with the ids --connection accepts").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(connectionsPath(org));if(resolveOutput(cmd)==="json"){printJson(res.items);return}printItems(res.items.map(c=>({id:c.id,name:c.handle,vendor:c.profileLabel,placeholder:[c.credentialKey,c.basicPasswordKey].filter(Boolean).join(", "),domains:c.allowedHosts.join(", "),kind:c.kind,value:c.last4?`\xB7\xB7\xB7\xB7${c.last4}`:""})),"table")})}var dashboards_exports={};__export(dashboards_exports,{register:()=>register9});import{readFileSync as readFileSync3}from"fs";function dashboardsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/dashboards`}function register9(program2){const group=hostedOnly(program2.command("dashboards")).description("Saved dashboards (CRUD + view filters; layout is UI-only)");group.command("list").description("List your dashboard library").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(dashboardsPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(d=>({key:d.key,name:d.name,source:d.source,default:d.isUserDefault?"you":d.isOrgDefault?"org":""})),"table")});group.command("get <key>").description("Show one resolved dashboard (widgets + saved filters)").action(async(key,_opts,cmd)=>{const org=await resolveOrg(cmd);const dashboard=await createClient().get(`${dashboardsPath(org)}/${encodeURIComponent(key)}`);if(resolveOutput(cmd)==="json"){printJson(dashboard);return}console.log(`key: ${dashboard.key}`);console.log(`name: ${dashboard.name}`);console.log(`filters: ${JSON.stringify(dashboard.defaultFilters)}`);console.log("widgets:");for(const w of dashboard.widgets)console.log(` [${w.id}] ${w.title} (${w.viz})`)});group.command("create <name>").description("Create an empty personal dashboard").action(async(name,_opts,cmd)=>{const org=await resolveOrg(cmd);const created=await createClient().post(dashboardsPath(org),{name});if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created dashboard ${created.key}`)});group.command("rename <key> <name>").description("Rename a dashboard (in your overlay)").action(async(key,name,_opts,cmd)=>{const org=await resolveOrg(cmd);const updated=await createClient().patch(`${dashboardsPath(org)}/${encodeURIComponent(key)}`,{name});if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Renamed to ${updated.name}`)});group.command("rm <key>").description("Delete a dashboard (built-in templates refuse)").action(async(key,_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(`${dashboardsPath(org)}/${encodeURIComponent(key)}`);if(resolveOutput(cmd)!=="json")console.log(`Deleted dashboard ${key}`)});group.command("filters <key>").description("Replace a dashboard's saved view-level filters").option("--json <json>","the filters object inline").option("--file <path>",'the filters object from a file ("-" for stdin)').option("--clear","remove all saved filters").action(async(key,opts,cmd)=>{const org=await resolveOrg(cmd);const provided=[opts.json,opts.file,opts.clear].filter(v=>v!==void 0&&v!==false).length;if(provided!==1)throw new CliError("usage","pass exactly one of --json, --file, or --clear");let filters={};if(!opts.clear){const raw=opts.json??readFileSync3(opts.file==="-"?0:opts.file,"utf8");try{filters=JSON.parse(raw)}catch(e){throw new CliError("usage",`filters are not valid JSON: ${e instanceof Error?e.message:e}`)}}const updated=await createClient().put(`${dashboardsPath(org)}/${encodeURIComponent(key)}/filters`,filters);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Saved filters: ${JSON.stringify(updated.defaultFilters)}`)})}var eval_exports={};__export(eval_exports,{register:()=>register10});import{randomUUID as randomUUID3}from"crypto";import{join as join6}from"path";import{z as z3}from"zod";import{z as z2}from"zod";var sampleCountSchema=z2.number().int().min(1).max(50);var nextRunAtInputSchema=z2.string().regex(/^\d{4}-\d{2}-\d{2}/,"Next run must be a calendar date").nullable();var ownedCadenceSchema=z2.enum(["NONE","DAILY","WEEKLY","MONTHLY"]);var ownedConfigurationInputShape={repoUrl:z2.string().nullish(),repoRef:z2.string().nullish(),profileId:z2.string().nullish(),agents:z2.array(agentConfigInputSchema).min(1,"Select at least one agent"),skillRefs:assetRefListSchema.optional().default([]),mcpRefs:assetRefListSchema.optional().default([]),connectionIds:z2.array(z2.string().min(1)).max(32).optional().default([]),lifecycle:lifecycleSchema.nullish(),cadence:ownedCadenceSchema.optional().default("NONE"),nextRunAt:nextRunAtInputSchema.optional(),sampleCount:sampleCountSchema.optional().default(1)};var ownedConfigurationInputSchema=z2.object(ownedConfigurationInputShape).strict();var ownedConfigurationPatchShape={repoUrl:z2.string().nullish(),repoRef:z2.string().nullish(),profileId:z2.string().nullish(),agents:z2.array(agentConfigInputSchema).min(1,"Select at least one agent").optional(),skillRefs:assetRefListSchema.optional(),mcpRefs:assetRefListSchema.optional(),connectionIds:z2.array(z2.string().min(1)).max(32).optional(),lifecycle:lifecycleSchema.nullish(),cadence:ownedCadenceSchema.optional(),nextRunAt:nextRunAtInputSchema.optional(),sampleCount:sampleCountSchema.optional()};var ownedConfigurationSchema=z2.object({repoUrl:z2.string().nullable(),repoRef:z2.string().nullable(),profile:z2.object({id:z2.string(),name:z2.string()}).nullable(),agents:z2.array(agentConfigSchema),skillRefs:z2.array(z2.string()),mcpRefs:z2.array(z2.string()),connectionIds:z2.array(z2.string()),lifecycle:lifecycleSchema.nullish(),cadence:ownedCadenceSchema,sampleCount:sampleCountSchema,lastScheduledAt:z2.string().nullable(),nextRunAt:z2.string().nullable()});var visibilityScenarioSchema=z2.object({id:z2.string(),repoUrl:z2.string().nullable(),repoRef:z2.string().nullable(),profile:z2.object({id:z2.string(),name:z2.string()}).nullable()});var visibilityScenarioInputSchema=z2.object({repoUrl:z2.string().nullish(),repoRef:z2.string().nullish(),profileId:z2.string().nullish()}).strict();var visibilityPromptSettingsInputShape={agents:ownedConfigurationInputShape.agents,skillRefs:ownedConfigurationInputShape.skillRefs,mcpRefs:ownedConfigurationInputShape.mcpRefs,connectionIds:ownedConfigurationInputShape.connectionIds,lifecycle:ownedConfigurationInputShape.lifecycle,cadence:ownedConfigurationInputShape.cadence,nextRunAt:ownedConfigurationInputShape.nextRunAt,sampleCount:ownedConfigurationInputShape.sampleCount};var visibilityPromptSettingsPatchShape={agents:ownedConfigurationPatchShape.agents,skillRefs:ownedConfigurationPatchShape.skillRefs,mcpRefs:ownedConfigurationPatchShape.mcpRefs,connectionIds:ownedConfigurationPatchShape.connectionIds,lifecycle:ownedConfigurationPatchShape.lifecycle,cadence:ownedConfigurationPatchShape.cadence,nextRunAt:ownedConfigurationPatchShape.nextRunAt,sampleCount:ownedConfigurationPatchShape.sampleCount};var visibilityPromptSettingsSchema=z2.object({agents:z2.array(agentConfigSchema),skillRefs:z2.array(z2.string()),mcpRefs:z2.array(z2.string()),connectionIds:z2.array(z2.string()),lifecycle:lifecycleSchema.nullish(),cadence:ownedCadenceSchema,sampleCount:sampleCountSchema,lastScheduledAt:z2.string().nullable(),nextRunAt:z2.string().nullable()});var runAgentsOverrideSchema=z2.array(agentConfigInputSchema).min(1,"Select at least one agent");var runSampleCountOverrideSchema=sampleCountSchema;var runEvalBodySchema=z2.object({agents:runAgentsOverrideSchema.optional(),sampleCount:runSampleCountOverrideSchema.optional(),lifecycle:lifecycleSchema.nullish()}).strict();var runVisibilityBodySchema=z2.object({agents:runAgentsOverrideSchema.optional(),sampleCount:runSampleCountOverrideSchema.optional(),lifecycle:lifecycleSchema.nullish(),visibilityScenarioIds:z2.array(z2.string().min(1)).min(1).optional()}).strict();var criterionSchema=z3.object({id:z3.string(),name:z3.string(),rubric:z3.string(),position:z3.number().int()});var criterionInputSchema=z3.object({id:z3.string().optional(),name:z3.string().min(1).optional(),rubric:z3.string().min(1)});var evalSetSchema=z3.object({id:z3.string(),suiteId:z3.string().nullable(),name:z3.string(),promptText:z3.string(),createdAt:z3.string(),createdBy:createdBySchema,criteria:z3.array(criterionSchema),tags:z3.array(z3.string())}).extend(ownedConfigurationSchema.shape);var agentPassRateSchema=z3.object({agent:agentSchema,model:z3.string(),runId:z3.string(),finishedAt:z3.string().nullable(),passed:z3.number().int(),total:z3.number().int()});var evalSetStatSchema=z3.object({passRates:z3.array(agentPassRateSchema),runCount:z3.number().int(),judgedRuns:z3.number().int(),passedRuns:z3.number().int()});var csv3=z3.string().transform(s=>s.split(",").map(x=>x.trim()).filter(Boolean));var listEvalSetsQuerySchema=z3.object({agents:csv3.pipe(z3.array(agentSchema)).optional(),models:csv3.optional(),since:z3.coerce.date().optional()});var evalSetListItemSchema=evalSetSchema.extend({stats:evalSetStatSchema});var listEvalSetsResponseSchema=z3.object({items:z3.array(evalSetListItemSchema)});var createEvalSetBodySchema=z3.object({name:z3.string().min(1),promptText:z3.string().min(1),criteria:z3.array(criterionInputSchema).min(1),tags:z3.array(z3.string()).optional()}).extend(ownedConfigurationInputSchema.shape).strict();var updateEvalSetBodySchema=z3.object({name:z3.string().min(1).optional(),promptText:z3.string().min(1).optional(),criteria:z3.array(criterionInputSchema).min(1).optional(),tags:z3.array(z3.string()).optional()}).extend(ownedConfigurationPatchShape).strict();var runEvalSetResponseSchema=z3.object({batchIds:z3.array(z3.string()),runIds:z3.array(z3.string()),runRequestId:z3.string().optional()});var evalSetRunSchema=z3.object({id:z3.string(),status:z3.string(),agent:agentSchema,model:z3.string().nullable(),resolvedModel:z3.string().nullable(),createdAt:z3.string(),startedAt:z3.string().nullable(),finishedAt:z3.string().nullable(),failureReason:z3.string().nullable(),creditsCharged:runCreditsChargedSchema.optional(),triggeredBy:runTriggeredBySchema.optional(),evalRevision:z3.string().optional(),judge:z3.object({status:z3.string(),error:z3.string().nullable()}).nullable(),criterionResults:z3.array(z3.object({criterionId:z3.string().nullable(),name:z3.string(),passed:z3.boolean().nullable(),reasoning:z3.string().nullable()}))});var listEvalSetRunsResponseSchema=listResponseSchema(evalSetRunSchema);import{z as z4}from"zod";var VISIBILITY_PROMPT_KINDS=["HEAD_TO_HEAD","RECOMMEND"];var visibilityPromptKindSchema=z4.enum(VISIBILITY_PROMPT_KINDS);var visibilityPromptSchema=visibilityPromptSettingsSchema.extend({id:z4.string(),name:z4.string(),kind:visibilityPromptKindSchema,promptText:z4.string(),branded:z4.boolean(),brandA:z4.string().nullable(),brandB:z4.string().nullable(),market:z4.object({id:z4.string(),slug:z4.string(),name:z4.string()}).nullable(),position:z4.number().int(),createdAt:z4.string(),createdBy:createdBySchema,tags:z4.array(z4.string()),runCount:z4.number().int(),lastRunAt:z4.string().nullable(),scenarios:z4.array(visibilityScenarioSchema)});var listVisibilityPromptsResponseSchema=z4.object({items:z4.array(visibilityPromptSchema)});var createVisibilityPromptBodySchema=z4.object({name:z4.string().min(1),promptText:z4.string().min(1),kind:visibilityPromptKindSchema,branded:z4.boolean().optional(),brandA:z4.string().nullish(),brandB:z4.string().nullish(),marketId:z4.string().nullish(),tags:z4.array(z4.string()).optional(),agents:visibilityPromptSettingsInputShape.agents.optional(),skillRefs:visibilityPromptSettingsInputShape.skillRefs,mcpRefs:visibilityPromptSettingsInputShape.mcpRefs,connectionIds:visibilityPromptSettingsInputShape.connectionIds,lifecycle:visibilityPromptSettingsInputShape.lifecycle,cadence:visibilityPromptSettingsInputShape.cadence,sampleCount:visibilityPromptSettingsInputShape.sampleCount,scenarios:z4.array(visibilityScenarioInputSchema).optional().default([])}).strict();var updateVisibilityPromptBodySchema=z4.object({name:z4.string().min(1).optional(),promptText:z4.string().min(1).optional(),branded:z4.boolean().optional(),brandA:z4.string().nullish(),brandB:z4.string().nullish(),marketId:z4.string().nullish(),tags:z4.array(z4.string()).optional(),...visibilityPromptSettingsPatchShape,scenarios:z4.array(visibilityScenarioInputSchema).optional()}).strict();var runVisibilityPromptResponseSchema=z4.object({batchIds:z4.array(z4.string()),runIds:z4.array(z4.string())});var visibilityPromptRunSchema=z4.object({id:z4.string(),status:z4.string(),agent:agentSchema,model:z4.string().nullable(),resolvedModel:z4.string().nullable(),createdAt:z4.string(),startedAt:z4.string().nullable(),finishedAt:z4.string().nullable(),failureReason:z4.string().nullable(),creditsCharged:runCreditsChargedSchema.optional(),triggeredBy:runTriggeredBySchema.optional(),verdict:z4.object({status:z4.string(),error:z4.string().nullable(),payload:z4.unknown().nullable()}).nullable()});var listVisibilityPromptRunsResponseSchema=listResponseSchema(visibilityPromptRunSchema);var orgMarketSchema=z4.object({marketId:z4.string(),slug:z4.string(),name:z4.string(),capability:z4.string(),source:z4.string()});var addableMarketSchema=orgMarketSchema.omit({source:true});var listMarketsResponseSchema=z4.object({tracked:z4.array(orgMarketSchema),addable:z4.array(addableMarketSchema)});var addMarketBodySchema=z4.object({marketId:z4.string().min(1)});import{readFileSync as readFileSync4,statSync as statSync2}from"fs";function hasLifecycleFlags(opts){return opts.setupFile!==void 0||opts.teardownFile!==void 0||opts.setupTimeout!==void 0||opts.teardownTimeout!==void 0||opts.lifecycleParams!==void 0||Boolean(opts.clearSetup||opts.clearTeardown||opts.clearLifecycle)}function addLifecycleOptions(command){return command.option("--setup-file <path>","upload a UTF-8 setup script for hosted sessions (JavaScript/TypeScript)").option("--teardown-file <path>","upload a UTF-8 teardown script for hosted sessions (JavaScript/TypeScript)").option("--setup-timeout <seconds>","setup deadline, 1\u2013600 seconds (new scripts default to 120)").option("--teardown-timeout <seconds>","teardown deadline, 1\u2013600 seconds (new scripts default to 120)").option("--lifecycle-params <json>","replace the scripts' JSON parameter object (maximum 64 KiB)")}function addLifecycleEditOptions(command){return addLifecycleOptions(command).option("--clear-setup","remove setup, preserving teardown and parameters").option("--clear-teardown","remove teardown, preserving setup and parameters").option("--clear-lifecycle","remove both scripts and their parameters")}function readScript(path,phase){const flag=`--${phase}-file`;if(path==="-")throw new CliError("usage",`${flag} requires a file path; stdin is not supported`);let bytes;try{const stat2=statSync2(path);if(!stat2.isFile())throw new Error("expected a regular file");if(stat2.size>MAX_LIFECYCLE_SOURCE_LENGTH)throw new Error("script must fit within 128 KiB");bytes=readFileSync4(path)}catch(error){throw new CliError("usage",`could not read ${flag} ${JSON.stringify(path)}: ${error instanceof Error?error.message:String(error)}`)}if(bytes.byteLength>MAX_LIFECYCLE_SOURCE_LENGTH)throw new CliError("usage",`${flag} script must fit within 128 KiB`);let source;try{source=new TextDecoder("utf-8",{fatal:true,ignoreBOM:true}).decode(bytes)}catch{throw new CliError("usage",`${flag} must contain valid UTF-8`)}if(!source.trim()||source.includes("\0"))throw new CliError("usage",`${flag} must contain nonempty script text without NUL bytes`);return source}function timeout(raw,phase){if(raw===void 0)return void 0;const value=Number(raw);if(!Number.isInteger(value)||value<1||value>MAX_LIFECYCLE_TIMEOUT_SECONDS)throw new CliError("usage",`--${phase}-timeout must be an integer from 1 to ${MAX_LIFECYCLE_TIMEOUT_SECONDS}`);return value}function lifecyclePatchFromOpts(opts,existing){if(!hasLifecycleFlags(opts))return{};if(opts.clearLifecycle){if(hasLifecycleFlags({...opts,clearLifecycle:false}))throw new CliError("usage","--clear-lifecycle conflicts with other lifecycle flags");return{lifecycle:null}}const current=existing==null?void 0:lifecycleSchema.safeParse(existing);if(current&&!current.success)throw new CliError("usage","The saved lifecycle configuration is unsupported; update the CLI before editing it");const value=current?.success?{...current.data}:{version:1,params:{}};for(const phase of["setup","teardown"]){const path=phase==="setup"?opts.setupFile:opts.teardownFile;const rawTimeout=phase==="setup"?opts.setupTimeout:opts.teardownTimeout;const clear=phase==="setup"?opts.clearSetup:opts.clearTeardown;if(clear&&(path!==void 0||rawTimeout!==void 0))throw new CliError("usage",`--clear-${phase} conflicts with --${phase}-file/--${phase}-timeout`);if(clear){delete value[phase];continue}const seconds=timeout(rawTimeout,phase);const source=path===void 0?value[phase]?.source:readScript(path,phase);if(seconds!==void 0&&source===void 0)throw new CliError("usage",`--${phase}-timeout requires --${phase}-file or a saved ${phase} script`);if(source!==void 0)value[phase]={source,timeoutSeconds:seconds??value[phase]?.timeoutSeconds??DEFAULT_LIFECYCLE_TIMEOUT_SECONDS}}if(opts.lifecycleParams!==void 0){try{value.params=JSON.parse(opts.lifecycleParams)}catch{throw new CliError("usage","--lifecycle-params must be a JSON object")}}if(!value.setup&&!value.teardown){if(opts.lifecycleParams!==void 0)throw new CliError("usage","--lifecycle-params requires a setup or teardown script");return{lifecycle:null}}const parsed=lifecycleSchema.safeParse(value);if(!parsed.success)throw new CliError("usage",parsed.error.issues[0]?.message??"Invalid lifecycle configuration");return{lifecycle:parsed.data}}import{readFileSync as readFileSync5}from"fs";function parseLimit(raw){const limit=Number(raw);if(!Number.isInteger(limit)||limit<1||limit>200){throw new CliError("usage","--limit must be an integer between 1 and 200")}return limit}function collect5(value,previous){return[...previous,value]}function once(flag){return(value,previous)=>{if(previous!==void 0)throw new CliError("usage",`${flag} takes one id; start one run per id`);return value}}var AGENT_ALIASES={"claude-code":"CLAUDE_CODE",claude_code:"CLAUDE_CODE",claude:"CLAUDE_CODE",codex:"CODEX_CLI","codex-cli":"CODEX_CLI",codex_cli:"CODEX_CLI",cursor:"CURSOR",pi:"PI",opencode:"OPENCODE","open-code":"OPENCODE",open_code:"OPENCODE",antigravity:"ANTIGRAVITY"};function parseAgents(raw){const out=[];for(const part of raw.split(",")){const name=part.trim();if(!name)continue;const agent=AGENT_ALIASES[name.toLowerCase()];if(!agent){throw new CliError("usage",`unknown agent "${name}" (expected: claude-code, codex, cursor, pi, opencode, antigravity)`)}out.push(agent)}if(out.length===0)throw new CliError("usage","no agents given");return[...new Set(out)]}function parseAgentSpec(raw){const colon=raw.indexOf(":");const name=(colon<0?raw:raw.slice(0,colon)).trim();const agent=AGENT_ALIASES[name.toLowerCase()];if(!agent){throw new CliError("usage",`unknown agent "${name}" in --agent "${raw}" (expected: claude-code, codex, pi, opencode, or antigravity, optionally ":model1,model2")`)}if(agent==="CURSOR"){throw new CliError("usage","cursor is not schedulable \u2014 presets may roster claude-code, codex, pi, opencode, and hosted antigravity")}const models=colon<0?[]:raw.slice(colon+1).split(",").map(m=>m.trim()).filter(Boolean);if(colon>=0&&models.length===0){throw new CliError("usage",`--agent "${raw}" has a ":" but no models \u2014 drop the colon for the agent default`)}return{agent,models:[...new Set(models)]}}function buildAgentRoster(specs){const byAgent=new Map;for(const raw of specs){const{agent,models}=parseAgentSpec(raw);const set=byAgent.get(agent)??new Set;for(const m of models)set.add(m);byAgent.set(agent,set)}return[...byAgent.entries()].map(([agent,models])=>({agent,models:[...models]}))}function validateAssetRefs(kind,refs){const parsed=runConfig_exports.assetRefListSchema.safeParse(refs);if(!parsed.success){const first=parsed.error.issues[0];throw new CliError("usage",`invalid --${kind} reference: ${first?.message??"expected name@label"}`)}return refs}function readPromptText(opts){if(opts.text===void 0===(opts.file===void 0)){throw new CliError("usage","pass exactly one of --text or --file")}if(opts.text!==void 0)return opts.text;const text=readFileSync5(opts.file==="-"?0:opts.file,"utf8").trim();if(!text)throw new CliError("usage","prompt file is empty");return text}function parseCadence(raw){const cadence=(raw??"NONE").trim().toUpperCase();if(!["NONE","DAILY","WEEKLY","MONTHLY"].includes(cadence))throw new CliError("usage",`invalid cadence "${raw}" (expected none, daily, weekly, or monthly)`);return cadence}function parseSampleCount(raw){const value=Number(raw??"1");if(!Number.isSafeInteger(value)||value<1||value>50)throw new CliError("usage","--samples must be an integer from 1 to 50");return value}function ownedConfigFromOpts(opts){if(opts.ref!==void 0&&opts.repo===void 0)throw new CliError("usage","--ref requires --repo");if(opts.agent.length===0)throw new CliError("usage","at least one --agent is required");return{...lifecyclePatchFromOpts(opts),repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:opts.persona??null,agents:buildAgentRoster(opts.agent),skillRefs:validateAssetRefs("skill",opts.skill),mcpRefs:validateAssetRefs("mcp",opts.mcp),connectionIds:opts.connection,cadence:parseCadence(opts.cadence),sampleCount:parseSampleCount(opts.samples)}}function ownedConfigPatchFromOpts(opts,lifecycle){const patch=lifecyclePatchFromOpts(opts,lifecycle);if((opts.repo!==void 0||opts.ref!==void 0)&&opts.clearRepo)throw new CliError("usage","--repo/--ref and --clear-repo conflict");if(opts.ref!==void 0&&opts.clearRef)throw new CliError("usage","--ref and --clear-ref conflict");if(opts.clearRepo){patch.repoUrl=null;patch.repoRef=null}else{if(opts.repo!==void 0)patch.repoUrl=opts.repo;if(opts.clearRef)patch.repoRef=null;else if(opts.ref!==void 0)patch.repoRef=opts.ref}if(opts.persona!==void 0&&opts.clearPersona)throw new CliError("usage","--persona and --clear-persona conflict");if(opts.clearPersona)patch.profileId=null;else if(opts.persona!==void 0)patch.profileId=opts.persona;if(opts.agent.length>0)patch.agents=buildAgentRoster(opts.agent);if(opts.skill.length>0&&opts.clearSkills)throw new CliError("usage","--skill and --clear-skills conflict");if(opts.clearSkills)patch.skillRefs=[];else if(opts.skill.length>0)patch.skillRefs=validateAssetRefs("skill",opts.skill);if(opts.mcp.length>0&&opts.clearMcp)throw new CliError("usage","--mcp and --clear-mcp conflict");if(opts.clearMcp)patch.mcpRefs=[];else if(opts.mcp.length>0)patch.mcpRefs=validateAssetRefs("mcp",opts.mcp);if(opts.connection.length>0&&opts.clearConnections)throw new CliError("usage","--connection and --clear-connections conflict");if(opts.clearConnections)patch.connectionIds=[];else if(opts.connection.length>0)patch.connectionIds=opts.connection;if(opts.cadence!==void 0)patch.cadence=parseCadence(opts.cadence);if(opts.samples!==void 0)patch.sampleCount=parseSampleCount(opts.samples);return patch}function addRunSettingsCreateOptions(command){return addLifecycleOptions(command).option("--agent <agent[:m1,m2]>","agent and optional model pins (repeatable)",collect5,[]).option("--skill <name@label>","staged skill (repeatable)",collect5,[]).option("--mcp <name@label>","MCP server (repeatable)",collect5,[]).option("--connection <id>","connection id (repeatable)",collect5,[]).option("--cadence <cadence>","none | daily | weekly | monthly","none").option("--samples <n>","sample count from 1 to 50","1")}function addRunSettingsEditOptions(command){return addLifecycleEditOptions(command).option("--agent <agent[:m1,m2]>","replace agents and model pins (repeatable)",collect5,[]).option("--skill <name@label>","replace skills (repeatable)",collect5,[]).option("--clear-skills","remove all skills").option("--mcp <name@label>","replace MCP servers (repeatable)",collect5,[]).option("--clear-mcp","remove all MCP servers").option("--connection <id>","replace connections (repeatable)",collect5,[]).option("--clear-connections","remove all connections").option("--cadence <cadence>","none | daily | weekly | monthly").option("--samples <n>","sample count from 1 to 50")}function hasRunSettingsFlags(opts){return hasLifecycleFlags(opts)||opts.agent.length>0||opts.skill.length>0||opts.mcp.length>0||opts.connection.length>0||opts.cadence.toUpperCase()!=="NONE"||opts.samples!=="1"}function runSettingsFromOpts(opts){return{...lifecyclePatchFromOpts(opts),...opts.agent.length>0?{agents:buildAgentRoster(opts.agent)}:{},skillRefs:validateAssetRefs("skill",opts.skill),mcpRefs:validateAssetRefs("mcp",opts.mcp),connectionIds:opts.connection,cadence:parseCadence(opts.cadence),sampleCount:parseSampleCount(opts.samples)}}function runSettingsPatchFromOpts(opts,lifecycle){return ownedConfigPatchFromOpts(opts,lifecycle)}function runOverridesFromOpts(opts){return{...opts.agent.length>0?{agents:buildAgentRoster(opts.agent)}:{},...opts.samples!==void 0?{sampleCount:parseSampleCount(opts.samples)}:{}}}function cadenceLabel(cadence){return cadence==="NONE"?"manual":cadence.toLowerCase()}function localDirectory2(command){const state=gaugeState(command);if(!state.root)throw new CliError("usage","No Gauge workspace found. Run `gauge init` in this directory first.");if(state.config===null)throw new CliError("usage","gauge.json is not valid JSON");return join6(state.root,".gauge")}function isLocal(command){return gaugeState(command).mode==="local"}function withLocalStore(command,fn){const store=openLocalStore(localDirectory2(command));try{return fn(store)}finally{store.close()}}function normalize(kind,raw,previous){const{inputs:_inputs,...definition}=raw;const scenarioInputs=Array.isArray(definition.scenarios)?definition.scenarios:void 0;if(scenarioInputs)definition.scenarios=scenarioInputs.map(({id:_id,...scenario})=>scenario);const parsed=(kind==="eval"?createEvalSetBodySchema:createVisibilityPromptBodySchema).safeParse(definition);if(!parsed.success)throw new CliError("usage",`Invalid ${kind}: ${parsed.error.issues.map(i=>`${i.path.join(".")}: ${i.message}`).join("; ")}`);const value=parsed.data;value.name=value.name.trim();value.promptText=value.promptText.trim();if(!value.name||!value.promptText)throw new CliError("usage","A measurement needs a nonempty name and prompt.");if(value.cadence&&value.cadence!=="NONE")throw new CliError("usage","Local measurements do not support managed schedules. Use --cadence none or --hosted.");if(value.nextRunAt)throw new CliError("usage","Local measurements do not support scheduled start times. Use --hosted.");assertLocalCapabilities(value);value.tags=[...new Set((value.tags??[]).map(tag=>tag.trim()).filter(Boolean))];if(kind==="eval"){if(value.repoRef&&!value.repoUrl)throw new CliError("usage","A repository ref requires a repository URL.");const ids=new Set;const existing=new Set(previous?.criteria?.map(criterion=>criterion.id));value.criteria=value.criteria.map(c=>{const rubric=c.rubric.trim();if(!rubric)throw new CliError("usage","Every criterion needs a nonempty rubric.");if(c.id&&previous&&!existing.has(c.id))throw new CliError("usage",`Unknown criterion ID: ${c.id}. Omit id to add a new criterion.`);const id=c.id??`cr_${randomUUID3()}`;if(!id||ids.has(id))throw new CliError("usage",`Duplicate or empty criterion ID: ${id}`);ids.add(id);return{id,name:c.name?.trim()||deriveCriterionName(rubric),rubric}})}else{value.brandA=typeof value.brandA==="string"?value.brandA.trim()||null:null;value.brandB=typeof value.brandB==="string"?value.brandB.trim()||null:null;if(value.kind==="HEAD_TO_HEAD"){if(!value.brandA||!value.brandB||String(value.brandA).toLowerCase()===String(value.brandB).toLowerCase())throw new CliError("usage","Head-to-head preference needs two different --brand-a and --brand-b values");[value.brandA,value.brandB]=[String(value.brandA),String(value.brandB)].sort((a,b)=>a.toLowerCase().localeCompare(b.toLowerCase()))}value.branded=value.kind==="HEAD_TO_HEAD"||Boolean(value.branded);value.agents??=[{agent:"CLAUDE_CODE",models:[]},{agent:"CODEX_CLI",models:[]}];const scenarios=value.scenarios;if(!previous&&scenarios.length===0)scenarios.push({repoUrl:null,repoRef:null,profileId:null});if(scenarios.some(s=>s.profileId))throw new CliError("usage","Local scenarios do not support hosted personas");const seen=new Set;const ids=new Set;const existing=new Set(previous?.scenarios?.map(scenario=>scenario.id));value.scenarios=scenarios.map((scenario,index)=>{const repoUrl=typeof scenario.repoUrl==="string"?scenario.repoUrl.trim()||null:null;const repoRef=typeof scenario.repoRef==="string"?scenario.repoRef.trim()||null:null;if(repoRef&&!repoUrl)throw new CliError("usage","A scenario ref requires a repository URL.");const key=JSON.stringify([repoUrl,repoRef]);if(seen.has(key))throw new CliError("usage","This prompt already has the same scenario.");seen.add(key);const suppliedId=scenarioInputs?.[index]?.id;if(suppliedId!==void 0&&(typeof suppliedId!=="string"||!previous||!existing.has(suppliedId)))throw new CliError("usage",`Unknown scenario ID: ${suppliedId}. Omit id to add a scenario.`);const id=typeof suppliedId==="string"?suppliedId:`sc_${randomUUID3()}`;if(ids.has(id))throw new CliError("usage",`Duplicate scenario ID: ${id}`);ids.add(id);return{id,repoUrl,repoRef,profileId:null}})}return value}function editable(row2){const{id,revision,measurementKind,createdAt,updatedAt,...definition}=row2;return definition}function createLocalMeasurement(command,kind,definition){return withLocalStore(command,store=>store.createMeasurement(kind,normalize(kind,definition)))}function editLocalMeasurement(command,kind,id,patch){return withLocalStore(command,store=>{const current=store.requireMeasurement(id,kind);const value=normalize(kind,{...editable(current),...patch},current);if(!("criteria"in patch)&¤t.criteria)value.criteria=current.criteria;if(!("scenarios"in patch)&¤t.scenarios)value.scenarios=current.scenarios;return store.updateMeasurement(id,value)})}function getLocalMeasurement(command,kind,id){return withLocalStore(command,store=>measurementWithStats(store,store.requireMeasurement(id,kind)))}function removeLocalMeasurement(command,kind,id){withLocalStore(command,store=>{store.requireMeasurement(id,kind);store.deleteMeasurement(id)})}function localMeasurementView(row2){return{profile:null,suiteId:null,market:null,position:0,runCount:0,lastRunAt:null,cadence:"NONE",sampleCount:1,repoUrl:null,repoRef:null,agents:[],skillRefs:[],mcpRefs:[],connectionIds:[],stats:{passRates:[],runCount:0,judgedRuns:0,passedRuns:0},...row2,scenarios:Array.isArray(row2.scenarios)?row2.scenarios.map(s=>({...s,profile:null})):[]}}function listLocalMeasurements(command,kind){return withLocalStore(command,store=>store.listMeasurements(kind).map(row2=>localMeasurementView(measurementWithStats(store,row2))))}function measurementWithStats(store,row2){const sessions=store.listSessions({measurementId:row2.id});const stats={runCount:sessions.length,judgedRuns:0,passedRuns:0,passRates:[]};for(const session of sessions){const analysis=readCurrentAnalysis(store,session.id);if(!analysis)continue;const outcomes=analysis.criteria.map(criterion=>criterion.outcome);const verdict=outcomes.includes("FAIL")?false:!outcomes.length||outcomes.includes("UNASSESSABLE")?null:true;if(verdict===null)continue;stats.judgedRuns++;if(verdict)stats.passedRuns++;if(!stats.passRates.some(rate=>rate.agent===session.agent))stats.passRates.push({agent:session.agent,model:session.model,runId:session.id,finishedAt:session.endedAt,passed:outcomes.filter(outcome2=>outcome2==="PASS").length,total:outcomes.length})}return{...row2,runCount:sessions.length,lastRunAt:sessions.map(session=>session.createdAt).sort().at(-1)??null,stats}}function deriveCriterionName(rubric){const trimmed=rubric.replace(/^\s*(?:[-*•]|\(?\d+[.)])\s+/,"").trim();const first=trimmed.split(/(?<=[.!?])\s|\n/)[0]?.trim()??"";return(new RegExp("\\p{L}","u").test(first)?first:trimmed.replace(/\s+/g," ")).replace(/[.!?]+$/,"").trim()}function localMeasurementSpec(command,kind,id,opts){const row2=getLocalMeasurement(command,kind,id);const stored=row2.inputs;if(stored?.length&&!opts.inputs.length)console.error(`note: saved inputs are no longer attached (${stored.join(", ")}). Pass --input <name> at launch.`);const settings={...row2,...runOverridesFromOpts(opts)};const agents=settings.agents;if(!agents?.length)throw new CliError("usage","Choose at least one agent with --agent before launching this local measurement");let scenarios=kind==="preference"?settings.scenarios:[{repoUrl:settings.repoUrl??null,repoRef:settings.repoRef??null}];if(!scenarios.length)throw new CliError("usage","This preference prompt has no scenarios. Add one with gauge preference scenarios add <id> before running it.");if(opts.scenario?.length){for(const selected of opts.scenario)if(!scenarios.some(s=>s.id===selected))throw new CliError("usage",`Unknown local scenario: ${selected}`);scenarios=scenarios.filter(s=>s.id!==void 0&&opts.scenario.includes(s.id))}const cases=scenarios.map((scenario,index)=>({key:`${kind}-${index+1}`,prompt:String(settings.promptText),...kind==="eval"?{criteria:settings.criteria.map(c=>({name:c.name,rubric:c.rubric}))}:{preference:settings.kind==="HEAD_TO_HEAD"?{kind:"head-to-head",brandA:String(settings.brandA),brandB:String(settings.brandB)}:{kind:"open-ended",branded:Boolean(settings.branded)}},config:{agents,sampleCount:Number(settings.sampleCount??1),repoUrl:scenario.repoUrl,repoRef:scenario.repoRef,inputs:opts.inputs}}));return{spec:{version:2,cases},measurement:{id,kind,revision:row2.revision,...kind==="eval"?{criterionIds:row2.criteria.map(c=>c.id)}:{}}}}import{readFileSync as readFileSync6}from"fs";import{createInterface as createInterface3}from"readline";async function confirmOrAbort(question,yes){if(yes)return;if(!process.stdin.isTTY){throw new CliError("usage","refusing to proceed without confirmation on a non-interactive stdin \u2014 pass --yes")}const rl=createInterface3({input:process.stdin,output:process.stderr});const answer=await new Promise(resolve5=>rl.question(`${question} [y/N] `,resolve5));rl.close();if(!/^y(es)?$/i.test(answer.trim())){throw new CliError("usage","aborted")}}function creatorLabel(createdBy){return createdBy?.name||createdBy?.email||"--"}function triggeredByLabel(t){if(!t)return"";return t.email?`${t.source} (${t.email})`:t.source}async function filterByCreator(items,createdBy){if(createdBy===void 0)return items;const wanted=createdBy.trim().toLowerCase();if(wanted==="me"){const{user}=await createClient().get("/api/v1/me");const email=user.email?.toLowerCase();return items.filter(item=>item.createdBy?.userId===user.id||email!==void 0&&item.createdBy?.email?.toLowerCase()===email)}return items.filter(item=>item.createdBy?.email?.toLowerCase()===wanted)}function evalSetsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/eval-sets`}function evalSetPath(org,id){return`${evalSetsPath(org)}/${encodeURIComponent(id)}`}function parseCriterion(raw){const idx=raw.indexOf(":");if(idx<=0||!raw.slice(idx+1).trim()){throw new CliError("usage",`invalid --criterion "${raw}" (expected "Name: rubric text")`)}return{name:raw.slice(0,idx).trim(),rubric:raw.slice(idx+1).trim()}}function readCriteriaFile(path){let parsed;try{parsed=JSON.parse(readFileSync6(path==="-"?0:path,"utf8"))}catch(e){throw new CliError("usage",`could not read --criteria-file: ${e instanceof Error?e.message:e}`)}if(!Array.isArray(parsed)||parsed.length===0)throw new CliError("usage","--criteria-file must be a non-empty JSON array of {id?, name, rubric}");return parsed}function criteriaFromOpts(opts){if(opts.criterion.length>0&&opts.criteriaFile!==void 0)throw new CliError("usage","--criterion and --criteria-file conflict");if(opts.criteriaFile!==void 0)return readCriteriaFile(opts.criteriaFile);if(opts.criterion.length>0)return opts.criterion.map(parseCriterion);return void 0}function repoLabel(e){if(!e.repoUrl)return"(default scaffold)";return e.repoRef?`${e.repoUrl}@${e.repoRef}`:e.repoUrl}function passRateLabel(stats){if(stats.judgedRuns===0)return"-";return`${Math.round(stats.passedRuns/stats.judgedRuns*100)}% (${stats.passedRuns}/${stats.judgedRuns})`}function evalSetRow(e){return{id:e.id,name:e.name,repo:repoLabel(e),criteria:e.criteria.length,tags:e.tags.join(", "),"pass rate":passRateLabel(e.stats),runs:e.stats.runCount,"created by":creatorLabel(e.createdBy)}}function printDetail(e){console.log(`id: ${e.id}`);console.log(`name: ${e.name}`);console.log(`repo: ${repoLabel(e)}`);console.log(`persona: ${e.profile?e.profile.name:"(default persona)"}`);console.log(`tags: ${e.tags.join(", ")||"(none)"}`);console.log(`cadence: ${cadenceLabel(e.cadence)}`);console.log(`samples: ${e.sampleCount}`);console.log(`agents: ${e.agents.map(agent=>`${agent.agent}${agent.models.length>0?`:${agent.models.join(",")}`:""}`).join("; ")}`);console.log(`skills: ${e.skillRefs.join(", ")||"(none)"}`);console.log(`mcp: ${e.mcpRefs.join(", ")||"(none)"}`);console.log(`conns: ${e.connectionIds.join(", ")||"(none)"}`);console.log(`creator: ${creatorLabel(e.createdBy)}`);console.log(`prompt: ${e.promptText}`);console.log("criteria:");for(const c of e.criteria)console.log(` [${c.id}] ${c.name}: ${c.rubric}`)}async function listEvalSets(cmd,opts){if(isLocal(cmd)){if(opts.agents||opts.models||opts.since)throw new CliError("usage","Local eval list does not support stats filters yet; use gauge query --local.");return listLocalMeasurements(cmd,"eval")}const res=await createClient().get(evalSetsPath(await resolveOrg(cmd)),{agents:opts.agents===void 0?void 0:parseAgents(opts.agents).join(","),models:opts.models,since:opts.since});return res.items}async function getEvalSet(cmd,id){const evalSet=isLocal(cmd)?localMeasurementView(getLocalMeasurement(cmd,"eval",id)):await createClient().get(evalSetPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson(evalSet);else printDetail(evalSet)}async function createEvalSet(cmd,opts){if(hasLifecycleFlags(opts)&&isLocal(cmd))throw new CliError("usage","Lifecycle scripts require hosted execution; pass --hosted");const criteria=criteriaFromOpts(opts);if(!criteria)throw new CliError("usage","at least one --criterion (or --criteria-file) is required");const body={name:opts.name,promptText:readPromptText(opts),...ownedConfigFromOpts(opts),criteria,tags:opts.tag.length>0?opts.tag:void 0};if(!isLocal(cmd)&&opts.repo===void 0)console.error("note: no --repo \u2014 this eval set will run against the default Gauge scaffold, NOT your repository. Use `gauge eval edit <id> --repo <url>` to retarget.");const created=isLocal(cmd)?createLocalMeasurement(cmd,"eval",body):await createClient().post(evalSetsPath(await resolveOrg(cmd)),body);if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created eval set ${created.id}`)}async function editEvalSet(cmd,id,opts){if(hasLifecycleFlags(opts)&&isLocal(cmd))throw new CliError("usage","Lifecycle scripts require hosted execution; pass --hosted");const lifecyclePath=hasLifecycleFlags(opts)&&!opts.clearLifecycle?evalSetPath(await resolveOrg(cmd),id):void 0;const existing=lifecyclePath?await createClient().get(lifecyclePath):void 0;const body=ownedConfigPatchFromOpts(opts,existing?.lifecycle);if(opts.name!==void 0)body.name=opts.name;if(opts.text!==void 0||opts.file!==void 0)body.promptText=readPromptText(opts);const criteria=criteriaFromOpts(opts);if(criteria)body.criteria=criteria;if(opts.tag.length>0&&opts.clearTags)throw new CliError("usage","--tag and --clear-tags conflict");if(opts.clearTags)body.tags=[];else if(opts.tag.length>0)body.tags=opts.tag;if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");const updated=isLocal(cmd)?editLocalMeasurement(cmd,"eval",id,body):await createClient().patch(lifecyclePath??evalSetPath(await resolveOrg(cmd),id),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated eval set ${updated.id}`)}async function removeEvalSet(cmd,id){if(isLocal(cmd))removeLocalMeasurement(cmd,"eval",id);else await createClient().delete(evalSetPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)!=="json")console.log(`Deleted eval set ${id}`)}async function runEvalSet(cmd,id,opts){const org=await resolveOrg(cmd);const overrides={...runOverridesFromOpts(opts),...hasLifecycleFlags(opts)?lifecyclePatchFromOpts(opts,(await createClient().get(evalSetPath(org,id))).lifecycle):{}};await confirmOrAbort(Object.keys(overrides).length>0?"Run this eval now with these overrides? The stored configuration stays unchanged.":"Run this eval's stored configuration now?",opts.yes);const result=await createClient().post(`${evalSetPath(org,id)}/run`,overrides);if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`Started ${result.runIds.length} run${result.runIds.length===1?"":"s"} in batch ${result.batchIds[0]??""}`);for(const batch of result.batchIds)console.log(`Runs: gauge run list --batch ${batch}`);if(result.runRequestId)console.log(`Wait: gauge --org ${org} run wait ${result.runRequestId}`)}return result}function register10(program2){const evals=program2.command("eval").description("Manage eval sets: saved prompts with judged criteria. Launch one with `gauge run --eval <id>`");evals.command("list").description("List eval sets").option("--agents <agents>","stats for these agents only (csv)").option("--models <models>","stats for these model slugs only (csv)").option("--since <date>","stats for runs since this ISO date only").option("--created-by <email|me>","only evals created by this user").action(async(opts,cmd)=>{const items=await filterByCreator(await listEvalSets(cmd,opts),opts.createdBy);if(resolveOutput(cmd)==="json")printJson(items);else printItems(items.map(evalSetRow),"table")});evals.command("get <id>").description("Show one eval set").action(async(id,_opts,cmd)=>{await getEvalSet(cmd,id)});addRunSettingsCreateOptions(evals.command("create").description("Create an eval set").requiredOption("--name <name>","name").option("--text <text>","prompt text inline").option("--file <path>",'prompt text from a file ("-" for stdin)').option("--repo <url>","consumer repository (default: empty locally, Gauge scaffold hosted)").option("--ref <ref>","branch, tag, or commit (with --repo)").option("--persona <profileId>","persona profile id")).option("--criterion <spec>",'judged criterion as "Name: rubric text" (repeatable)',collect5,[]).option("--criteria-file <path>",'JSON array of {name, rubric} ("-" for stdin)').option("--tag <name>","user tag name, create-on-type (repeatable)",collect5,[]).action(async(opts,cmd)=>{await createEvalSet(cmd,opts)});addRunSettingsEditOptions(evals.command("edit <id>").description("Update an eval set. Omitted flags keep their value. A criteria list REPLACES the whole set (use --criteria-file with ids to edit in place).").option("--name <name>","rename it").option("--text <text>","replace the prompt text inline").option("--file <path>",'replace the prompt text from a file ("-" stdin)').option("--repo <url>","set the repository (keeps a ref pin)").option("--ref <ref>","set the branch, tag, or commit pin").option("--clear-ref","drop the ref pin").option("--clear-repo","use an empty workspace locally or the default scaffold hosted").option("--persona <profileId>","set the persona profile id").option("--clear-persona","use the default persona")).option("--criterion <spec>",'criterion as "Name: rubric", replaces the whole set (repeatable)',collect5,[]).option("--criteria-file <path>","JSON array of {id?, name, rubric}; entries with id update in place").option("--tag <name>","replace the user tag set (repeatable)",collect5,[]).option("--clear-tags","remove all user tags").action(async(id,opts,cmd)=>{await editEvalSet(cmd,id,opts)});evals.command("rm <id>").description("Delete an eval set (run history stays)").action(async(id,_opts,cmd)=>{await removeEvalSet(cmd,id)})}var feedback_exports2={};__export(feedback_exports2,{register:()=>register11});import{Option as Option3}from"commander";function register11(program2){program2.command("feedback").description("Send feedback about this CLI to the Gauge team: friction you hit, missing commands or flags, bugs").requiredOption("-m, --message <text>","what happened").addOption(new Option3("--kind <kind>","feedback type").choices([...feedback_exports.FEEDBACK_KINDS]).default("other")).option("--command <command>","the gauge command the feedback is about").action(async(opts,cmd)=>{const{org}=cmd.optsWithGlobals();const client=createClient({authOptional:true});const body={message:opts.message,kind:opts.kind,command:opts.command,org:client.signedIn?org?(await resolveOrgValue(org)).id:readConfig().orgId:void 0};const res=await client.post("/api/v1/feedback",body);if(resolveOutput(cmd)==="json")printJson(res);else console.log(`Feedback sent (${res.id}${client.signedIn?"":", anonymous"}). Thank you.`)})}var instructions_exports={};__export(instructions_exports,{register:()=>register12});var AUTHORING=`- Measure one meaningful developer outcome. Record the learning goal, reproducible starting state, observable finish line, and execution budget; a coherent migration can span browser and server workflows.
|
|
37
|
+
- Write the prompt as a real user's request, in their language. Describe their goal, relevant context, and intended outcome. Preserve constraints they actually supplied; do not assume technical expertise or invent features, UI actions, business rules, commands, filenames, or verification recipes. Leave discovery and implementation choices to the agent.
|
|
38
|
+
- Align a small set of independently assessable criteria with the requested outcome and existing application. Accept equivalent supported approaches and execution evidence; remove hidden requirements when revising the prompt.`;var MIGRATION_EXAMPLE=`## Migration prompt and criteria
|
|
39
|
+
|
|
40
|
+
This example tests migration of an existing feature-flag application across browser and server workflows. Its starting repository contains the application, while attached connections supply destination credentials. The source account's targeting configuration is unavailable.
|
|
41
|
+
|
|
42
|
+
### Keep the substantial task
|
|
43
|
+
|
|
44
|
+
A bad version of this migration prompt would add specific features, such as a flag selector, an account switcher, or an activity log. It might also specify exact UI layouts, clicks to perform, SDK calls, or files to edit. Those directions assume the user already knows the implementation and add requirements beyond their goal.
|
|
45
|
+
|
|
46
|
+
Use the kind of language a user would naturally use to ask for a change. For example, "Let me pass a feature to the CLI and get instructions for it" describes the desired behavior without prescribing command registration or rendering code. Keep details the user actually requests; do not add exact specifications or UI/UX actions just to make grading easier.
|
|
47
|
+
|
|
48
|
+
> Migrate this app from its current feature-flag provider to the replacement provider while preserving its existing functionality. Get it running locally, verify the migration works, and explain any limitations.
|
|
49
|
+
|
|
50
|
+
If the user also asks for a complete replacement, carry that decision into the rubric. Do not add backward compatibility or a gradual production rollout solely because a migration guide recommends them.
|
|
51
|
+
|
|
52
|
+
### Define observable success
|
|
53
|
+
|
|
54
|
+
Inspect the starting application and relevant migration guidance before drafting criteria. For this application, success means existing browser and server workflows continue to work through the destination provider, live evaluations drive the running UI, and active dependencies on the source provider are removed when full replacement is requested. The rubric can assess behavior present in the source without dictating SDK methods, edited files, or an exact test sequence.
|
|
55
|
+
|
|
56
|
+
Suitable criteria for this fixture:
|
|
57
|
+
|
|
58
|
+
- **Complete migration:** Required flags and code references are accounted for. The active integration uses the destination provider, and the app no longer needs the source provider at runtime when full replacement is requested. A separate demo or hardcoded replacement is insufficient.
|
|
59
|
+
- **Preserved behavior:** Existing browser/server workflows, context changes, and server-provided initial values continue to work. Available configuration is preserved or differences explained. Exact historical targeting parity cannot be established without source rules.
|
|
60
|
+
- **Verified working application:** Execution evidence connects live flag evaluation from the destination provider to the running app's behavior across its browser and server workflows. A build or standalone SDK probe alone does not demonstrate the migration.
|
|
61
|
+
- **Appropriate product surfaces:** Supported SDKs/providers fit the browser and server runtimes; supported management interfaces handle configuration; credentials fit each surface. Equivalent approaches are acceptable, and using every tool is unnecessary.
|
|
62
|
+
|
|
63
|
+
Keep environment identifiers, credential mappings, repository revision, and resource isolation in supported connection or fixture mechanisms where possible. Authoring notes and source limitations must not become an extra implementation checklist.
|
|
64
|
+
|
|
65
|
+
Missing source rules limit the parity claim. Record that uncertainty and assess observable application behavior; do not replace it with an invented targeting contract. If account-configuration migration is the requested outcome, prepare the required source data before calling the fixture ready.`;var EVAL=`---
|
|
37
66
|
name: gauge-eval
|
|
38
|
-
description:
|
|
39
|
-
before creating or changing a saved eval set.
|
|
67
|
+
description: Write, save, edit, and schedule eval sets that gauge run launches. Read
|
|
68
|
+
before creating or changing a saved eval set; includes a migration example.
|
|
40
69
|
---
|
|
41
70
|
|
|
42
71
|
# Eval sets
|
|
@@ -54,6 +83,10 @@ eval set.
|
|
|
54
83
|
gauge eval get <id> -o json
|
|
55
84
|
gauge run list --eval <id> -o json # its runs; gauge run get <run-id> shows verdicts
|
|
56
85
|
|
|
86
|
+
## Write the prompt and criteria
|
|
87
|
+
|
|
88
|
+
${AUTHORING}
|
|
89
|
+
|
|
57
90
|
## Create and edit
|
|
58
91
|
|
|
59
92
|
gauge eval create --name <name> --file prompt.md --criteria-file criteria.json --repo <url> --agent claude-code
|
|
@@ -69,6 +102,7 @@ eval set.
|
|
|
69
102
|
- --agent belongs to the eval set. Hosted --skill, --mcp, and --connection
|
|
70
103
|
references require hosted execution. Read
|
|
71
104
|
\`gauge instructions run\` to choose models.
|
|
105
|
+
- Hosted --setup-file/--teardown-file upload UTF-8 JS/TS files; import bindings from "@withgauge/session". For one launch only, pass the same flags to gauge run --hosted --eval <id>; saved configuration stays unchanged. Edits keep the other hook/params; --clear-setup/--clear-teardown/--clear-lifecycle remove hooks (see --help).
|
|
72
106
|
- Hosted --cadence daily|weekly|monthly and --samples schedule repeat runs that spend
|
|
73
107
|
credits. Confirm the schedule with the user.
|
|
74
108
|
- An edit to the prompt, criteria, repo, or persona starts a new eval revision
|
|
@@ -83,7 +117,47 @@ After an input change, rebuild/repack and rerun the same eval ID. Local launches
|
|
|
83
117
|
snapshot current input files into a new batch. Keep prompts, criteria, and model
|
|
84
118
|
pins fixed; analyze and compare baseline/candidate batches before claiming improvement.
|
|
85
119
|
Creating or editing a measurement does not authorize additional paid launches.
|
|
86
|
-
|
|
120
|
+
|
|
121
|
+
${MIGRATION_EXAMPLE}
|
|
122
|
+
`;var LOGO=String.raw`
|
|
123
|
+
|
|
124
|
+
⣠⣶⣿⡿⠿⣿⣿⣦⡀ ⣼⣷⣄
|
|
125
|
+
⣰⣿⡿⠁ ⠙⠿⠿⠄ ⢀⣤⣤⣤⣤⡀ ⢠⣤⡄ ⢠⣤⡄ ⢀⣤⣤⣤⡀⣤⣤ ⣀⣤⣤⣤⣄⡀ ⢀⣼⣿⣿⣿⣀
|
|
126
|
+
⣿⣿⡇ ⣶⣶⣶⣶⣶⡆⠰⠿⠟⠉⠉⢿⣿⡆⢸⣿⡇ ⢸⣿⡇⢰⣿⡿⠋⠉⠻⣿⣿ ⣼⣿⠟⠉⠙⢿⣷⡄ ⢠⣾⣿⣿⣿⣿⠿⢿⣷⡄
|
|
127
|
+
⢿⣿⣇ ⠉⠉⠉⣿⣿⡇⢠⣶⣾⠿⠿⢿⣿⡇⢸⣿⡇ ⢸⣿⡇⢸⣿⡇ ⣿⣿ ⣿⣿⣿⣿⣿⣿⣿⡇ ⣿⣄⢀⣴⣿⣿⣿⣿⣿⣷⣶⣶⣿⣿⣶⣶⣶⡆
|
|
128
|
+
⠈⢿⣿⣦⣤⣀⣤⣾⢿⣿⡇⢿⣿⣇⡀⣀⣾⣿⡇⢸⣿⣷⣀⣀⣾⣿⡇⠘⣿⣿⣄⣀⣴⣿⣿ ⢻⣿⣦⣀⣀⣶⣶⠆ ⠙⠛⠋⠈⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠇
|
|
129
|
+
⠉⠛⠛⠛⠛⠁⠘⠛⠃⠈⠙⠛⠛⠋⠘⠛⠃ ⠙⠛⠛⠋⠘⠛⠃⢀⣈⣙⠛⠋⠁⣿⣿ ⠉⠛⠛⠛⠛⠁ ⠹⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠿⠋
|
|
130
|
+
⠻⢿⣷⣶⣿⠿⠋ ⠈⢿⠿⠋⠉
|
|
131
|
+
`;var SMALL_LOGO=`\u2588\u2580\u2580 \u2584\u2580\u2588 \u2588 \u2588 \u2588\u2580\u2580 \u2588\u2580\u2580
|
|
132
|
+
\u2588\u2584\u2588 \u2588\u2580\u2588 \u2588\u2584\u2588 \u2588\u2584\u2588 \u2588\u2588\u2584`;var SMALL_DOG_LOGO=` \u{1FB4B}\u{1FB4F}
|
|
133
|
+
\u{1FB4A}\u{1FB41}\u2588\u{1FB39}\u{1FB51}\u{1FB3C}
|
|
134
|
+
\u{1FB48}\u{1FB44}\u2588\u2588\u2588\u{1FB39}\u{1FB3B}\u{1FB4C}\u{1FB39}\u{1FB39}\u{1FB39}
|
|
135
|
+
\u{1FB65}\u{1FB52}\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u{1FB5E}\u{1FB5A}
|
|
136
|
+
\u{1FB62}\u{1FB55}\u{1FB5E}\u{1FB5C}\u{1FB02}\u{1FB02}\u{1FB02}\u{1FB58}`;var BARK=String.raw`
|
|
137
|
+
⣀
|
|
138
|
+
⣠⣶⣿⡿⠿⣿⣿⣦⡀ ⢠⣿⣷⣄
|
|
139
|
+
⣰⣿⡿⠁ ⠙⠿⠿⠄ ⢀⣤⣤⣤⣤⡀ ⢠⣤⡄ ⢠⣤⡄ ⢀⣤⣤⣤⡀⣤⣤ ⣀⣤⣤⣤⣄⡀ ⢀⣾⣿⣿⣿⣤⣄⣀ ⣠⡾⠋
|
|
140
|
+
⣿⣿⡇ ⣶⣶⣶⣶⣶⡆⠰⠿⠟⠉⠉⢿⣿⡆⢸⣿⡇ ⢸⣿⡇⢰⣿⡿⠋⠉⠻⣿⣿ ⣼⣿⠟⠉⠙⢿⣷⡄ ⢀⣾⣿⣿⣿⣿⠛⠛⣿⣧⣀⣠⣤⣄ ⠨⠶⠶⠶
|
|
141
|
+
⢿⣿⣇ ⠉⠉⠉⣿⣿⡇⢠⣶⣾⠿⠿⢿⣿⡇⢸⣿⡇ ⢸⣿⡇⢸⣿⡇ ⣿⣿ ⣿⣿⣿⣿⣿⣿⣿⡇ ⢰⣄ ⣠⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠿⠿⠿⠿ ⠘⢷⣤⡀
|
|
142
|
+
⠈⢿⣿⣦⣤⣀⣤⣾⢿⣿⡇⢿⣿⣇⡀⣀⣾⣿⡇⢸⣿⣷⣀⣀⣾⣿⡇⠘⣿⣿⣄⣀⣴⣿⣿ ⢻⣿⣦⣀⣀⣶⣶⠆ ⠾⠿⠋⠻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷⡦ ⠈
|
|
143
|
+
⠉⠛⠛⠛⠛⠁⠘⠛⠃⠈⠙⠛⠛⠋⠘⠛⠃ ⠙⠛⠛⠋⠘⠛⠃⢀⣈⣙⠛⠋⠁⣿⣿ ⠉⠛⠛⠛⠛⠁ ⠈⠻⣿⣿⣿⣿⡿⠿⠛⠛⠛⠋
|
|
144
|
+
⠻⢿⣷⣶⣿⠿⠋ ⠈⠿⠟⠁
|
|
145
|
+
`;var LOGO_ART=LOGO.replace(/^\n+/,"").trimEnd();var LOCAL_SETUP_PRESENTATION=`For the first local setup explanation, help users understand how Gauge OSS works before seeking any needed launch approval. Users cannot see CLI output. Show this explanation once during setup, starting with the logo:
|
|
146
|
+
Open with the FULL Gauge ASCII art below, including the dog, copied verbatim in a fenced text block.
|
|
147
|
+
Preserve the Braille characters and spacing in one code block with no blank lines between artwork rows; do not redraw it.
|
|
148
|
+
Add one sentence naming the planned case count, agent/model, and local Docker.
|
|
149
|
+
Draw a simple overview titled "How Gauge OSS works", with three boxes joined by arrows:
|
|
150
|
+
choose tasks + product inputs \u2192 agents try your product in local Docker sessions \u2192 this coding agent reviews the recorded sessions.
|
|
151
|
+
Use one or two short, plain-language lines per box. Make clear the runs happen on this machine, separately from this chat, and the review explains what worked and where agents struggled.
|
|
152
|
+
Keep the graphic conceptual: no actual prompts, implementation steps, rubric details, or findings. Do not expand the Docker box into a technical walkthrough.
|
|
153
|
+
Use aligned boxes with whitespace; place them side by side if they fit, or stack them on narrow displays. Below the graphic show cases \xD7 targets \xD7 samples = total sessions, without implying concurrency.
|
|
154
|
+
Follow with a compact table of case names, tasks, and success criteria. Use plain text in cells, with semicolons for multiple criteria; no HTML tags or nested lists.
|
|
155
|
+
Summarize changed files and routine setup in at most one line.
|
|
156
|
+
End with a short launch question, or the next step if launch is already approved.
|
|
157
|
+
Include only what will run, what success means, and what happens next. Explain checks
|
|
158
|
+
only on problems; omit rejected inputs, selection narration, and recaps unless requested.
|
|
159
|
+
Use the diagram to explain Gauge OSS and the table for this run's tasks/criteria; avoid repeating either in prose.
|
|
160
|
+
Show planned work as planned; honor existing approval without adding another approval step. For each approved launch, use the compact live update in \`gauge instructions run\` instead of repeating this diagram or full logo.`;var INIT=`---
|
|
87
161
|
name: gauge-init
|
|
88
162
|
description: Initialize a Gauge workspace, choose local or hosted execution,
|
|
89
163
|
and optionally scaffold shared cases. Read before running gauge init.
|
|
@@ -108,17 +182,17 @@ preference. The selection is private to this checkout.
|
|
|
108
182
|
|
|
109
183
|
## What it writes
|
|
110
184
|
|
|
111
|
-
gauge.json shared manifest: { "version": 2, "inputs": {} }
|
|
112
185
|
.gauge/config.json private backend preference
|
|
113
186
|
.gauge/gauge.sqlite local measurements, launches, summaries, and analyses
|
|
114
187
|
.gauge/.gitignore ignores workspace state, including the database
|
|
115
|
-
AGENTS.md a Gauge block pointing to these instructions
|
|
116
188
|
|
|
117
|
-
Plain init
|
|
118
|
-
|
|
119
|
-
|
|
120
|
-
|
|
121
|
-
|
|
189
|
+
Plain init changes no tracked files. Add \`--cases\` when shared, fully specified
|
|
190
|
+
cases are useful; it also writes \`gauge.json\` ({ "version": 2, "inputs": {} }),
|
|
191
|
+
\`gauge-cases/\`, and an AGENTS.md block pointing to these instructions.
|
|
192
|
+
Create \`gauge.json\` yourself to declare named inputs, case paths, and GitHub
|
|
193
|
+
checks; it never chooses the execution backend. A configured org binds repository
|
|
194
|
+
automation without forcing interactive runs onto hosted Gauge. Commands find the
|
|
195
|
+
nearest parent \`gauge.json\` or \`.gauge/\`.
|
|
122
196
|
|
|
123
197
|
Repeat init safely: it preserves measurements and existing shared configuration,
|
|
124
198
|
and refreshes the Gauge instructions block. Commit shared configuration, optional
|
|
@@ -139,6 +213,113 @@ inputs and previews use case files/specs; read \`gauge instructions run\`.
|
|
|
139
213
|
|
|
140
214
|
Gauge recommends the gauge-agents skill for future sessions:
|
|
141
215
|
\`gauge install-skill\`. Coding agents install it for themselves in this project, without prompts.
|
|
216
|
+
`;var INIT_LOCAL=INIT.replace("## Next\n",`## Explain the first local setup
|
|
217
|
+
|
|
218
|
+
${LOCAL_SETUP_PRESENTATION}
|
|
219
|
+
|
|
220
|
+
\`\`\`text
|
|
221
|
+
${LOGO_ART}
|
|
222
|
+
\`\`\`
|
|
223
|
+
|
|
224
|
+
## Next
|
|
225
|
+
`);var ISSUES=`---
|
|
226
|
+
name: gauge-issues
|
|
227
|
+
description: Find the friction agents hit with your product and their
|
|
228
|
+
objections to it, ranked by impact. Read before choosing what to improve.
|
|
229
|
+
---
|
|
230
|
+
|
|
231
|
+
# Issues
|
|
232
|
+
|
|
233
|
+
An issue is the same finding across many sessions. Each issue is friction,
|
|
234
|
+
which agents hit in any session, or an objection, which agents raise when they
|
|
235
|
+
pass on your product in a preference session. Findings is an alternative term
|
|
236
|
+
for issues. Gauge groups each analyzed run's findings into issues per product.
|
|
237
|
+
|
|
238
|
+
## Find and read
|
|
239
|
+
|
|
240
|
+
gauge issues list -o json # every issue, most sessions first
|
|
241
|
+
gauge issues list --type friction -o json # or --type objection
|
|
242
|
+
gauge issues get <id> -o json # the issue's findings, runs, and agents
|
|
243
|
+
gauge run insight <run> -o json # one run's findings with their issue
|
|
244
|
+
|
|
245
|
+
Friction shows impact by effect: BLOCKED, EXTRA_WORK, RECOVERED, or UNKNOWN.
|
|
246
|
+
Objection shows where agents went instead. Some findings are not in an issue
|
|
247
|
+
yet; \`issues list\` counts them.
|
|
248
|
+
|
|
249
|
+
## Close and recheck
|
|
250
|
+
|
|
251
|
+
Each issue is open, fixed, or muted. \`issues list\` shows open issues;
|
|
252
|
+
\`--status fixed\`, \`--status muted\`, or \`--status all\` shows the others.
|
|
253
|
+
|
|
254
|
+
gauge issues fix <id> # done; it reopens if agents hit it again
|
|
255
|
+
gauge issues mute <id> # not relevant; it never reopens
|
|
256
|
+
gauge issues reopen <id> # back to open (also unmutes)
|
|
257
|
+
gauge issues recheck <id> # preview the past runs it would repeat
|
|
258
|
+
gauge issues recheck <id> --confirm # rerun them; get shows which still show it
|
|
259
|
+
|
|
260
|
+
- Fix an issue after you ship a change for it. Gauge reopens it when a run
|
|
261
|
+
that started after the fix shows it again, and marks it "came back".
|
|
262
|
+
- Mute an issue only when the user says it is irrelevant or by design. Muted
|
|
263
|
+
issues still collect findings, so they never come back as new issues.
|
|
264
|
+
- \`recheck\` (alias \`verify\`) never prompts. Without \`--confirm\` it only
|
|
265
|
+
previews: the past runs it would repeat, the latest per prompt, agent, and
|
|
266
|
+
model, with the default 3 checked. Show the user the preview and ask before
|
|
267
|
+
you add \`--confirm\`, because reruns spend credits. \`--run <runId>...\` picks
|
|
268
|
+
others, at most 10. The result shows in \`gauge issues get <id>\` when the
|
|
269
|
+
reruns are analyzed: "looks fixed" if none show the issue, "still seen in n
|
|
270
|
+
of m" if some do. A recheck never closes an issue: tell the user the result,
|
|
271
|
+
and run \`gauge issues fix <id>\` when they agree. A fixed issue that shows it
|
|
272
|
+
again reopens.
|
|
273
|
+
|
|
274
|
+
## Choose what to improve
|
|
275
|
+
|
|
276
|
+
- Rank by sessions and effect. BLOCKED and EXTRA_WORK cost agents the most.
|
|
277
|
+
- Read the findings and their evidence before you name a cause.
|
|
278
|
+
- Prefer issues on surfaces the org owns: its docs, skills, and CLI.
|
|
279
|
+
- Improve an issue with \`gauge optimizations create --issue <id>\`:
|
|
280
|
+
\`gauge instructions optimization\`.
|
|
281
|
+
`;var LIFECYCLE=`---
|
|
282
|
+
name: gauge-lifecycle
|
|
283
|
+
description: Author setup and cleanup scripts for hosted sessions using the Gauge APIs.
|
|
284
|
+
---
|
|
285
|
+
|
|
286
|
+
# Setup and cleanup scripts (hosted only)
|
|
287
|
+
|
|
288
|
+
Write ordinary JavaScript or TypeScript files with top-level await. No exported function is required.
|
|
289
|
+
The CLI uploads each file's contents, not its path or companion files. Gauge supplies Node.js and the session's installed workspace packages; do not assume arbitrary dependencies exist.
|
|
290
|
+
Setup runs after the repository, MCP servers, and Connections are ready, before the coding agent. Cleanup (called teardown in CLI flags) is attempted after execution, cancellation, or partial setup failure when the VM remains usable.
|
|
291
|
+
|
|
292
|
+
## Author, preview, launch
|
|
293
|
+
|
|
294
|
+
gauge run --hosted "<task>" --setup-file ./setup.ts --teardown-file ./cleanup.ts --dry-run
|
|
295
|
+
gauge run --hosted "<task>" --setup-file ./setup.ts --teardown-file ./cleanup.ts
|
|
296
|
+
|
|
297
|
+
Preview validates syntax and configuration, not imports, types, or remote behavior. It does not execute scripts or launch sessions. Hosted commands contact Gauge; previews are not offline checks. Launch only within the user's authorization; hosted runs consume credits.
|
|
298
|
+
For one launch of a saved eval: \`gauge run --hosted --eval <id> --setup-file ./setup.ts\`. Unspecified scripts/timeouts/parameters inherit saved values. Use \`--preference <id>\` for a preference prompt. Saved launches do not support \`--dry-run\`.
|
|
299
|
+
To save scripts for future runs: \`gauge eval edit <id> --hosted --setup-file ./setup.ts --teardown-file ./cleanup.ts\`. The same flags work with eval/preference create and edit. These edit saved configuration; run flags do not.
|
|
300
|
+
Either script is optional. Defaults: 120 seconds per new script; \`--setup-timeout\` / \`--teardown-timeout\` allow 1\u2013600 seconds. Files must be nonempty UTF-8 and at most 128 KiB.
|
|
301
|
+
\`--lifecycle-params '{"fixture":"demo"}'\` replaces the shared nonsecret JSON object (64 KiB max). Existing scripts, timeouts, and parameters are preserved when their flags are omitted.
|
|
302
|
+
\`--clear-setup\`, \`--clear-teardown\`, or \`--clear-lifecycle\` remove scripts on saved edits or saved launches; clearing the last script also clears parameters. Hooks are unsupported for local runs, case files, and \`--apply\`.
|
|
303
|
+
|
|
304
|
+
## Runtime API \u2014 no package installation required
|
|
305
|
+
|
|
306
|
+
import { run, params, mcp, state, agent, signal } from "@withgauge/session";
|
|
307
|
+
|
|
308
|
+
| Binding | Use |
|
|
309
|
+
|---|---|
|
|
310
|
+
| run | Immutable id, attempt, phase (setup/teardown), workspacePath, setupOutcome, executionOutcome. The script runs from workspacePath. |
|
|
311
|
+
| params | Immutable shared JSON configuration supplied by --lifecycle-params. |
|
|
312
|
+
| mcp.connect(name) | Connect to an attached server by its exact handle, without @label; configured authentication is provided. Returns a client with listTools(), callTool({name, arguments}), and close(). |
|
|
313
|
+
| state.set(key, json), state.get(key), state.delete(key) | Await all calls. Durable JSON within one session attempt, shared by setup and cleanup. get returns undefined when absent. |
|
|
314
|
+
| agent.setEnv(key, value) | Await to pass a nonsecret string to the coding agent. Setup only. |
|
|
315
|
+
| signal | AbortSignal for cancellation and the phase deadline; signal.throwIfAborted() checks it. |
|
|
316
|
+
|
|
317
|
+
Setup example: \`await state.set("resourceId", "created-id"); await agent.setEnv("TEST_RESOURCE_ID", "created-id");\`
|
|
318
|
+
Cleanup example: \`const id = await state.get("resourceId");\` \u2014 handle missing state, restore/remove the resource, then \`await state.delete("resourceId");\`.
|
|
319
|
+
For MCP: \`const client = await mcp.connect("your-server"); const { tools } = await client.listTools();\` Inspect actual tool schemas; do not invent names or arguments. Check \`result.isError\` after callTool and throw on failure.
|
|
320
|
+
Throw or exit nonzero to fail. console.log provides phase progress. Keep credentials out of source, params, state, agent environment exports, and logs; use configured Connections for authentication.
|
|
321
|
+
Scripts do not serialize sessions or isolate shared remote state. Prefer disposable per-session fixtures. Save cleanup identifiers early, tolerate partial setup, and make cleanup safe to repeat. Cleanup is best effort if the VM is lost.
|
|
322
|
+
Use \`gauge run watch <run-id>\` to follow execution and \`gauge run --help\` for launch flags. Drafting a script or reading these instructions does not start a session.
|
|
142
323
|
`;var PREFERENCE=`---
|
|
143
324
|
name: gauge-preference
|
|
144
325
|
description: Save, edit, and schedule preference prompts that gauge run
|
|
@@ -180,6 +361,8 @@ Read \`gauge instructions session-analysis\` to analyze the sessions.
|
|
|
180
361
|
- --agent belongs to the prompt. Hosted --skill, --mcp, and --connection
|
|
181
362
|
references require hosted execution.
|
|
182
363
|
Read \`gauge instructions run\` to choose models.
|
|
364
|
+
- Hosted --setup-file/--teardown-file upload UTF-8 JS/TS files; import bindings from "@withgauge/session". For one launch only, pass the same flags to gauge run --hosted --preference <id>; saved configuration stays unchanged.
|
|
365
|
+
Edits keep the other hook/params; --clear-setup/--clear-teardown/--clear-lifecycle remove hooks (see --help).
|
|
183
366
|
- Hosted --cadence daily|weekly|monthly and --samples schedule repeat runs that spend
|
|
184
367
|
credits. Confirm the schedule with the user.
|
|
185
368
|
- For one launch with other agents or samples, use
|
|
@@ -193,6 +376,13 @@ After an input change, rebuild/repack and rerun the same preference ID. Local la
|
|
|
193
376
|
snapshot current input files into a new batch. Keep the prompt and model pins fixed;
|
|
194
377
|
analyze and compare baseline/candidate batches before claiming a preference change.
|
|
195
378
|
Creating or editing a measurement does not authorize additional paid launches.
|
|
379
|
+
`;var RUN_LOGO_BLOCK=`\`\`\`text
|
|
380
|
+
${SMALL_DOG_LOGO}
|
|
381
|
+
\`\`\``;var RUN_LAYOUT_INSTRUCTIONS=`Show the exact dog artwork below with a clickable "Watch running sessions" link immediately to its right, aligned with the dog's middle row. Preserve the artwork's characters and spacing. Wrap each artwork row in inline code and separate rows with Markdown hard line breaks (two trailing spaces); put the Markdown link outside the middle row's inline code so it stays clickable. The fenced block is reference artwork, not the final side-by-side layout. Do not redraw, abbreviate, or substitute the dog. Show this small dog once per actual launch, including reruns.`;var RUN_PRESENTATION_REMINDER=`Agents: the user cannot see this CLI output. In your launch update, show the session summary first. ${RUN_LAYOUT_INSTRUCTIONS} Run \`gauge watch\` with the same backend and organization as the launch to get the exact URL. Report the actual launch state. If already shown for this launch, do not repeat the card. If you omitted it, include it in your next update while sessions are still running, without asking the user whether to read more instructions.
|
|
382
|
+
${RUN_LOGO_BLOCK}`;var RUN_PRESENTATION=`## Give the user a live view
|
|
383
|
+
Users cannot see CLI output. After any needed launch approval, run \`gauge watch\` with the same backend (\`--local\` or \`--hosted\`) and organization (\`--org\`, when selected) as the launch; it returns immediately. Show a compact live update, then launch. The first local setup explanation in \`gauge instructions init\` is separate; do not repeat its full logo or diagram at launch.
|
|
384
|
+
Start with one short sentence with the planned sessions, targets, and where they run. ${RUN_LAYOUT_INSTRUCTIONS} Use the returned watch URL exactly, including \`?liveruns=true\` for hosted. Keep the dog and link together with no run details between them. Distinguish planned, started, and completed work accurately. Show this once per launch, including reruns, rather than on every poll; previews and verification do not start runs. If you omitted the card, include it in your next update while sessions are still running, without asking the user whether to read more instructions.
|
|
385
|
+
${RUN_LOGO_BLOCK}
|
|
196
386
|
`;var SNAPSHOTS=`Preview JSON separates the source \`path\` from \`gaugeArchive\` hash/bytes in \`inputSnapshots\`, and lists the archived \`files\` (first 100) with \`fileCount\`; check them before launching. \`unattachedInputs\` names gauge.json inputs this launch leaves out. The Gauge delivery archive contains the selected files; for npm it contains the .tgz, so its hash/size differ from the raw .tgz. Launch takes a fresh snapshot; preview does not reserve bytes or prove an eval passed.`;var RUN=`---
|
|
197
387
|
name: gauge-run
|
|
198
388
|
description: Write Gauge eval cases, choose models and judging criteria, and
|
|
@@ -203,9 +393,9 @@ description: Write Gauge eval cases, choose models and judging criteria, and
|
|
|
203
393
|
|
|
204
394
|
- Antigravity is hosted-only and requires organization enrollment. Discover available model identifiers with \`gauge models list --hosted\`. Use \`--agent antigravity:<model>\` when creating or launching saved evals/preferences, or \`agent: ANTIGRAVITY\` with explicit \`models\` in case configuration. Preview case files or JSON case specs with \`--hosted --dry-run\`; launch with \`--hosted\`. The server checks access and model availability. Hosted launches use credits; previews contact Gauge and may upload inputs.
|
|
205
395
|
- Launch with \`gauge run\` (this repo's gauge-cases/ cases), \`gauge run --eval <id>\` or \`--preference <id>\` (saved: \`gauge instructions eval\` or \`gauge instructions preference\`), or \`gauge run --apply <file>\` (\`-\` for stdin) with JSON: a case spec (\`cases\`: Markdown cases as JSON, each with criteria or preference; \`gauge run --example\`) or a battery spec (\`prompts\`: saved prompts and schedules; \`gauge run --example --format battery\`). Follow a launch with \`gauge run wait <request-id>\` (exits 0 once settled; \`--check\` exits 1 unless PASSED or MEASURED), \`gauge run status <request-id>\` (one snapshot), \`gauge run watch <run-id>\`, or \`gauge run get <run-id> --logs\`. For one launch with other agents or sample counts, pass \`--agent\` and \`--samples\` with \`--eval\` or \`--preference\`; do not edit or copy the eval set or preference prompt for a smoke test.
|
|
206
|
-
-
|
|
207
|
-
|
|
208
|
-
-
|
|
396
|
+
- \`gauge run "<prompt>"\` runs one open-ended preference prompt and saves no measurement. It uses the calling agent if codex or opencode, else claude-code; \`--agent\` overrides.
|
|
397
|
+
${AUTHORING}
|
|
398
|
+
- For one-off setup/teardown, use \`gauge run --hosted "<prompt>" --setup-file ./setup.ts --teardown-file ./teardown.ts\`; add \`--dry-run\` to validate without executing hooks or launching sessions. No saved measurement is created. Files upload as inline UTF-8 source; \`--setup-timeout\`, \`--teardown-timeout\`, and \`--lifecycle-params\` configure this launch. These flags also override a saved \`--eval <id>\` or \`--preference <id>\` for one launch, inheriting unspecified hooks/timeouts/params; \`--clear-setup\`/\`--clear-teardown\`/\`--clear-lifecycle\` remove saved hooks only for that launch. Saved configuration is never edited. Hook flags do not work with local execution, case files, or \`--apply\`; saved measurement launches do not support \`--dry-run\`. Hooks use \`@withgauge/session\` bindings and do not isolate shared remote state between concurrent sessions.
|
|
209
399
|
- Keep repository revisions, credentials, connections, runtime setup, and budgets in supported configuration or fixtures where possible. Inspect \`gauge eval create --help\` and \`gauge run --help\` for current controls; do not assume proposed limits are enforced.
|
|
210
400
|
- For shared Markdown cases, preview with \`gauge run --dry-run\` before \`gauge run\`; \`gauge-cases/**/*.md\` cases are discovered automatically and \`--files\` overrides discovery with paths/globs. Root \`gauge.json\` can attach existing GitHub Actions artifacts and committed files; Gauge consumes them without building. Website inputs route a canonical HTTPS origin to an existing public preview; sources are GitHub Deployments or an existing provider PR comment with exact author, readiness pattern, and candidate SHA. Comments without commit identity and private or path-prefixed previews are unsupported.
|
|
211
401
|
- Native command inputs use \`type: "binary"\`, \`platform: "linux-x64"\`, selected \`paths\`, and a \`commands\` map from command names to explicit relative ELF executable files. CI must upload already unpacked files. Gauge restores executable permissions and exposes declared commands on PATH; compatible system libraries are required, nested archives and arbitrary installers are unsupported.
|
|
@@ -248,35 +438,9 @@ Background: [Open models as preference proxies](https://www.withgauge.com/blog/o
|
|
|
248
438
|
Replace \`criteria\` with \`preference: { kind: open-ended }\`, adding \`branded: true\` when the prompt names brands. For a comparison use \`preference: { kind: head-to-head, brandA: PostHog, brandB: Amplitude }\`. Specify exactly one of criteria/preference; the body remains the task and must name the comparison brands itself.
|
|
249
439
|
Existing per-case settings, inputs, and path filters apply. The app's preference analyst reports advisory choices; competitor selections never fail criteria. Preference-only checks finish neutral; mixed case sets retain experience gating. With \`--check\`, the wait command succeeds for MEASURED/PASSED and fails on operational errors. No base/head delta is inferred.
|
|
250
440
|
|
|
251
|
-
|
|
252
|
-
|
|
253
|
-
This example tests migration of an existing feature-flag application across browser and server workflows. Its starting repository contains the application, while attached connections supply destination credentials. The source account's targeting configuration is unavailable.
|
|
441
|
+
${MIGRATION_EXAMPLE}
|
|
254
442
|
|
|
255
|
-
|
|
256
|
-
|
|
257
|
-
A bad version of this migration prompt would add specific features, such as a flag selector, an account switcher, or an activity log. It might also specify exact UI layouts, clicks to perform, SDK calls, or files to edit. Those directions assume the user already knows the implementation and add requirements beyond their goal.
|
|
258
|
-
|
|
259
|
-
Use the kind of language a user would naturally use to ask for a change. For example, "Let me pass a feature to the CLI and get instructions for it" describes the desired behavior without prescribing command registration or rendering code. Keep details the user actually requests; do not add exact specifications or UI/UX actions just to make grading easier.
|
|
260
|
-
|
|
261
|
-
> Migrate this app from its current feature-flag provider to the replacement provider while preserving its existing functionality. Get it running locally, verify the migration works, and explain any limitations.
|
|
262
|
-
|
|
263
|
-
If the user also asks for a complete replacement, carry that decision into the rubric. Do not add backward compatibility or a gradual production rollout solely because a migration guide recommends them.
|
|
264
|
-
|
|
265
|
-
### Define observable success
|
|
266
|
-
|
|
267
|
-
Inspect the starting application and relevant migration guidance before drafting criteria. For this application, success means existing browser and server workflows continue to work through the destination provider, live evaluations drive the running UI, and active dependencies on the source provider are removed when full replacement is requested. The rubric can assess behavior present in the source without dictating SDK methods, edited files, or an exact test sequence.
|
|
268
|
-
|
|
269
|
-
Suitable criteria for this fixture:
|
|
270
|
-
|
|
271
|
-
- **Complete migration:** Required flags and code references are accounted for. The active integration uses the destination provider, and the app no longer needs the source provider at runtime when full replacement is requested. A separate demo or hardcoded replacement is insufficient.
|
|
272
|
-
- **Preserved behavior:** Existing browser/server workflows, context changes, and server-provided initial values continue to work. Available configuration is preserved or differences explained. Exact historical targeting parity cannot be established without source rules.
|
|
273
|
-
- **Verified working application:** Execution evidence connects live flag evaluation from the destination provider to the running app's behavior across its browser and server workflows. A build or standalone SDK probe alone does not demonstrate the migration.
|
|
274
|
-
- **Appropriate product surfaces:** Supported SDKs/providers fit the browser and server runtimes; supported management interfaces handle configuration; credentials fit each surface. Equivalent approaches are acceptable, and using every tool is unnecessary.
|
|
275
|
-
|
|
276
|
-
Keep environment identifiers, credential mappings, repository revision, and resource isolation in supported connection or fixture mechanisms where possible. Authoring notes and source limitations must not become an extra implementation checklist.
|
|
277
|
-
|
|
278
|
-
Missing source rules limit the parity claim. Record that uncertainty and assess observable application behavior; do not replace it with an invented targeting contract. If account-configuration migration is the requested outcome, prepare the required source data before calling the fixture ready.
|
|
279
|
-
`;var RUN_LOCAL=`---
|
|
443
|
+
${RUN_PRESENTATION}`;var RUN_LOCAL=`---
|
|
280
444
|
name: gauge-run
|
|
281
445
|
description: Write local Gauge eval and Agent Preference cases for this repo's
|
|
282
446
|
skills, CLI, or SDK and verify them. Read before creating or editing cases.
|
|
@@ -285,6 +449,7 @@ description: Write local Gauge eval and Agent Preference cases for this repo's
|
|
|
285
449
|
# Run (local)
|
|
286
450
|
|
|
287
451
|
For a product hillclimb and docs exposure limits, read \`gauge instructions hillclimb\`.
|
|
452
|
+
|
|
288
453
|
## Choose a definition source
|
|
289
454
|
|
|
290
455
|
Save an ongoing measurement with \`gauge eval\` or \`gauge preference\`, then
|
|
@@ -293,6 +458,7 @@ Read their instruction topics for create/edit flags. Saved measurements live in
|
|
|
293
458
|
\`.gauge/gauge.sqlite\` with stable identity and revision history. They require
|
|
294
459
|
no committed case files. The file-based examples below are for measurements
|
|
295
460
|
you intend to share or automate. Both sources work with local and hosted backends.
|
|
461
|
+
\`gauge run "<prompt>"\` runs one open-ended preference prompt and saves no measurement. It uses the calling agent if codex or opencode, else claude-code; \`--agent\` overrides.
|
|
296
462
|
|
|
297
463
|
## How a run works
|
|
298
464
|
|
|
@@ -354,6 +520,8 @@ Put each case in gauge-cases/<input>/<case>.md, one folder per surface. An eval
|
|
|
354
520
|
gauge run --apply cases.json # cases as JSON, no Markdown; each with criteria or preference ("-" for stdin; see --example)
|
|
355
521
|
|
|
356
522
|
This runs every case in local Docker and waits. It needs Docker, the public runtime (install with \`gauge local runtime pull\`; inspect with \`gauge local runtime inspect\`), and agent logins: \`codex login\` for Codex, Pi, and OpenCode; \`claude\` or \`ANTHROPIC_API_KEY\` for Claude Code. Then do its "Next" step. Ctrl+C cancels and cleans up; results and cached images remain. Use \`gauge run --detach --local\` for a background batch; it prints exact status/wait/stop commands. Inspect with \`gauge run status --batch <id> --local\`; block until it settles with \`gauge run wait --batch <id> --local\` (do not write a polling loop); stop with \`gauge run cancel --batch <id> --local\` (or \`gauge run cancel <run-id> --local\`). Clean up interrupted sessions with \`gauge cleanup --local\`; it preserves evidence and leaves active sessions alone. For stale or missing database session summaries, use \`gauge run repair --local\` to rebuild them from artifacts while preserving measurements and analyses. Keep .gauge/gauge.sqlite: it is durable data, not a disposable cache. Local credentials can use native logins or API keys. Configure references with \`gauge local profiles set work --agent CODEX_CLI --provider openai --auth api_key --env TEAM_OPENAI_KEY\`, then \`gauge local profiles use work\`. Pass the variable name, never a key value. Inspect with \`gauge local profiles inspect work -o json\`; update with \`set ... --replace\`, list with \`list\`, and remove references with \`remove <profile>\`. \`gauge run --local --profile work\` overrides defaults and must be compatible with every target. Dry-run reports each actual model\u2019s credential readiness; local checks do not verify provider acceptance. Detached batches retain the selected references even if defaults change. For additional system tools,derive an image from the standard runtime and pass \`--runtime-image <image>\` to \`gauge run --local\` (also works with \`--dry-run\` and \`--detach\`). Build or pull that image first. Dry runs show its reference and immutable ID without starting a container; actual runs check required tools, selected harnesses, permissions and command forwarding before accepting the batch. Every queued session uses the frozen image ID even if its tag changes. Preserve runtime contract label \`sh.gauge.contract="1"\` and the base harnesses; an entrypoint must execute the supplied command.
|
|
523
|
+
|
|
524
|
+
${RUN_PRESENTATION}
|
|
357
525
|
`;var SESSION_ANALYSIS=`---
|
|
358
526
|
name: gauge-session-analysis
|
|
359
527
|
description: Analyze hosted Gauge runs: why agents chose or rejected a product,
|
|
@@ -491,7 +659,7 @@ Open the final answer with a compact findings list: **Friction:** what slowed or
|
|
|
491
659
|
Include every supported category, even when an Opportunity addresses the same incident as a Friction finding; omit empty categories. Give evidence and a practical next step for each, with uncertainty that changes its interpretation alongside it.
|
|
492
660
|
Put run counts, setup, command recaps, and general limitations afterward. Include contradictory sessions.
|
|
493
661
|
Fix product issues within the user's authorized scope, rerun the same measurement, publish the new analyses, and compare before/after evidence.
|
|
494
|
-
`;function hillclimbSteps(hosted){return[{title:"Identify inputs and propose a pilot",command:"gauge instructions hillclimb",details:["Inspect this project and the user's goal. Suggest representative tasks, including a mistake or boundary case; explain what they measure.","Choose inputs and the execution backend deliberately. Web fixture JSON supports HTTPS page replacement and origin mapping locally or hosted. GitHub-resolved website previews need hosted execution; supplied docs files do not replace the canonical site."]},{title:"Prepare and run",command:"gauge instructions run",details:["Save ongoing measurements with eval/preference; use gauge-cases/ for shared or automated cases. Build/pack inputs, verify cases, and preview.",hosted?"
|
|
662
|
+
`;function hillclimbSteps(hosted){return[{title:"Identify inputs and propose a pilot",command:"gauge instructions hillclimb",details:["Inspect this project and the user's goal. Suggest representative tasks, including a mistake or boundary case; explain what they measure.","Choose inputs and the execution backend deliberately. Web fixture JSON supports HTTPS page replacement and origin mapping locally or hosted. GitHub-resolved website previews need hosted execution; supplied docs files do not replace the canonical site."]},{title:"Prepare and run",command:"gauge instructions run",details:["Save ongoing measurements with eval/preference; use gauge-cases/ for shared or automated cases. Build/pack inputs, verify cases, and preview.",hosted?"Show the user each case's task and criteria, the targets, the session count, and the credit budget, then offer to start the runs with --hosted; say they run on Gauge, separate from this chat.":"Check Docker, the runtime image, and agent credentials. Show the user each case's task and criteria, the targets, and the session count, then offer to start the runs; say they run in Docker containers on this machine, separate from this chat. Mention verify and preview only when they fail.",...hosted?[]:["Read gauge instructions init for how to present the plan."]]},{title:"Analyze evidence",command:`gauge instructions session-analysis --${hosted?"hosted":"local"}`,details:[hosted?"Inspect transcripts and hosted analysis. Separate product friction from environment failures and report findings, affected sessions, and coverage.":"Read transcripts, judge outcomes, and publish structured analysis. Report findings, affected sessions, and coverage; an empty findings list can be a successful analyzed session."]},{title:"Improve, rerun, and compare",details:["Implement a focused change, rebuild/repack, and rerun affected measurements. Keep tasks, criteria, and models fixed; disclose comparison limits.",hosted?"Use gauge run compare --hosted with baseline/candidate request or batch selectors, then inspect whether the original friction disappeared.":"Use gauge run compare --local with baseline/candidate batch selectors after publishing analysis; inspect whether the original friction disappeared.","Repeat within the agreed budget. Report before/after evidence, new friction, and anything still unvalidated."]}]}var LOCAL_STEPS=hillclimbSteps(false);var HOSTED_STEPS=hillclimbSteps(true);function stepsMarkdown(steps=LOCAL_STEPS){return steps.map((step,i)=>[`${i+1}. **${step.title}**${step.command?`: \`${step.command}\``:""}`,...step.details.map(detail=>` - ${detail}`)].join("\n")).join("\n")}var HILLCLIMB=`---
|
|
495
663
|
name: gauge-hillclimb
|
|
496
664
|
description: Plan and carry out an evidence-driven improvement loop for a product,
|
|
497
665
|
CLI, SDK, Skill, or documentation. Use for open-ended setup and usability testing.
|
|
@@ -512,9 +680,9 @@ Ask only for decisions that affect the test. For narrow requests, use just the r
|
|
|
512
680
|
requirements. Honor requested models and state cases \xD7 targets \xD7 samples.
|
|
513
681
|
3. **Prepare and run a small baseline.** Preserve supplied product/docs content during setup;
|
|
514
682
|
propose fixes separately. Save groups with eval/preference or shared cases for automation. Build/pack,
|
|
515
|
-
verify cases, then preview selected files and runtime readiness
|
|
516
|
-
|
|
517
|
-
|
|
683
|
+
verify cases, then preview selected files and runtime readiness (mention them only on failure).
|
|
684
|
+
Preview is not execution. A blocked prerequisite is a setup result, not a product failure.
|
|
685
|
+
Before the first launch, show the user each case's task, criteria, and targets, and the files you changed, then offer to start the runs. Say they run separately from this chat: in local Docker containers or on hosted Gauge.
|
|
518
686
|
4. **Extract evidence and publish analysis.** Read transcripts and artifacts, not only
|
|
519
687
|
grades. Distinguish product friction, agent mistakes, environment failures, and missing
|
|
520
688
|
evidence. Locally your coding agent publishes structured analysis; hosted runs provide
|
|
@@ -559,7 +727,7 @@ Reuse authorization for agreed setup; do not ask permission at every step. See \
|
|
|
559
727
|
| Project inputs and previews on saved measurement launches | yes | use case files/specs |
|
|
560
728
|
| Structured findings and queries | yes, supported fields | yes |
|
|
561
729
|
| Personas, recurring schedules, PR checks | no | yes |
|
|
562
|
-
| Optimizations (
|
|
730
|
+
| Optimizations (edit and rerun sessions) | no | yes |
|
|
563
731
|
| Stored credentials, stats, org settings | no | yes |
|
|
564
732
|
|
|
565
733
|
Local uses Docker and agent credentials. Hosted needs a Gauge account (\`gauge onboard\`) and uses credits.
|
|
@@ -575,8 +743,8 @@ description: Understand Gauge, use its CLI, and choose the module guidance
|
|
|
575
743
|
Gauge helps you understand and improve how coding agents discover and use your
|
|
576
744
|
product. It runs real agents in sandboxes under reproducible conditions, then
|
|
577
745
|
judges their work. Evals measure successful product use; Agent Preference
|
|
578
|
-
measures which tools agents choose and why. Optimizations
|
|
579
|
-
|
|
746
|
+
measures which tools agents choose and why. Optimizations edit the docs, CLI
|
|
747
|
+
rules, and skills agents used, then rerun those sessions to test the change.
|
|
580
748
|
|
|
581
749
|
## Use Gauge and the CLI
|
|
582
750
|
|
|
@@ -594,10 +762,14 @@ identify inputs, pilot representative tasks, analyze evidence, fix, rerun, compa
|
|
|
594
762
|
- \`gauge run\` starts and reads runs. \`gauge eval\` and \`gauge preference\`
|
|
595
763
|
save what it runs again: \`gauge instructions eval|preference\`.
|
|
596
764
|
- Pass \`-o json\` on reads and parse it. Tables are for people.
|
|
597
|
-
- Confirm before launches
|
|
765
|
+
- Confirm before launches. Use \`--yes\` for work already approved.
|
|
598
766
|
- Retry writes with their supported idempotency flag; blind retries can launch
|
|
599
767
|
duplicate paid sessions.
|
|
600
|
-
-
|
|
768
|
+
- The user does not see gauge output. Show them the files you wrote, each case's task, criteria,
|
|
769
|
+
and targets before launch, and results with counts and limits. After any needed approval, before each actual launch, show a compact live update with the session summary first. ${RUN_LAYOUT_INSTRUCTIONS} Get the URL from \`gauge watch\` with the same backend and org. If you omit it, include it in your next update while sessions are still running, without asking the user whether to read more instructions.
|
|
770
|
+
|
|
771
|
+
${RUN_LOGO_BLOCK}
|
|
772
|
+
|
|
601
773
|
- Run \`gauge update\` to install the latest published CLI.
|
|
602
774
|
- Exit codes: 0 ok, 1 error, 2 usage, 3 billing-blocked, 4 needs your input.
|
|
603
775
|
- When the CLI gets in your way (friction, a missing command or flag, a bug),
|
|
@@ -623,7 +795,7 @@ Hold each case's prompt, fixture, and rubric fixed across targets. Pilot before
|
|
|
623
795
|
scaling, confirm paid launches, and report per-target evidence, sample counts,
|
|
624
796
|
failures, and missing evidence.
|
|
625
797
|
|
|
626
|
-
Full guidance and migration example: \`gauge instructions run\`.
|
|
798
|
+
Full guidance and migration example: \`gauge instructions run\`. Setup/cleanup scripts and their runtime API: \`gauge instructions lifecycle\` (hosted only).
|
|
627
799
|
|
|
628
800
|
### Session analysis
|
|
629
801
|
|
|
@@ -642,15 +814,14 @@ Full guidance: \`gauge instructions session-analysis\`.
|
|
|
642
814
|
|
|
643
815
|
### Optimization
|
|
644
816
|
|
|
645
|
-
|
|
646
|
-
|
|
647
|
-
|
|
648
|
-
|
|
649
|
-
|
|
650
|
-
|
|
651
|
-
|
|
652
|
-
|
|
653
|
-
winner; use \`handoff\` for implementation in your source repository.
|
|
817
|
+
Optimizations can help hill climb your agent surfaces to improve Agent Preference
|
|
818
|
+
and Agent Experience. Start from an issue (\`gauge instructions issues\`), one
|
|
819
|
+
finding, or a goal. Create an optimization, then read \`authoring\` for the source sessions' findings and
|
|
820
|
+
the editing practices. Copy each page, skill, or CLI rule set you change into
|
|
821
|
+
the tree with \`file add\`, edit it locally, and save it with \`file write\`. Keep your
|
|
822
|
+
diagnosis and plan in \`notes\`. Ask before \`launch\`: it reruns the source sessions
|
|
823
|
+
with the tree. \`watch\` waits for the reruns and shows which
|
|
824
|
+
findings resolved, returned, or appeared. Record the outcome with \`finish\`.
|
|
654
825
|
|
|
655
826
|
Full guidance: \`gauge instructions optimization\`.
|
|
656
827
|
`;var ROOT_LOCAL=`---
|
|
@@ -684,6 +855,16 @@ ${stepsMarkdown()}
|
|
|
684
855
|
- \`gauge instructions <topic>\` is workflow guidance, not a page per command.
|
|
685
856
|
- Pass \`-o json\` on reads and parse it. Tables are for people.
|
|
686
857
|
- Exit codes: 0 ok, 1 error, 2 usage.
|
|
858
|
+
- The user does not see gauge output. Stop and show them:
|
|
859
|
+
- before a launch, each case's task, criteria, targets, and session count,
|
|
860
|
+
and the files you changed, then offer to start the runs, which run in
|
|
861
|
+
Docker containers separate from this chat (mention verify and preview only
|
|
862
|
+
on failure);
|
|
863
|
+
- after runs, per-case results, top friction, and \`gauge run trace|diff <id>\`.
|
|
864
|
+
- After any needed approval, before each actual launch, show a compact live update with the session summary first. ${RUN_LAYOUT_INSTRUCTIONS} Get the URL from \`gauge watch\` with the same backend and org. The full logo and diagram in \`gauge instructions init\` are for the first setup explanation. If you omit the live update, include it in your next update while sessions are still running, without asking the user whether to read more instructions.
|
|
865
|
+
|
|
866
|
+
${RUN_LOGO_BLOCK}
|
|
867
|
+
|
|
687
868
|
- Report CLI friction, gaps, or bugs, with no account:
|
|
688
869
|
\`gauge feedback -m "<what happened>" --kind friction|missing|bug\`.
|
|
689
870
|
|
|
@@ -692,56 +873,66 @@ ${MODES}
|
|
|
692
873
|
To switch this repo to hosted, run \`gauge onboard\`, then \`gauge init --hosted\`.
|
|
693
874
|
`;var OPTIMIZATIONS=`---
|
|
694
875
|
name: gauge-optimizations
|
|
695
|
-
description:
|
|
696
|
-
|
|
876
|
+
description: Hill climb agent surfaces. Edit the pages, skills, and CLI rules agent
|
|
877
|
+
sessions used, rerun those sessions, and compare their findings.
|
|
697
878
|
---
|
|
698
879
|
|
|
699
880
|
# Optimizations
|
|
700
881
|
|
|
701
|
-
|
|
882
|
+
An optimization is a tree of files plus the sessions it came from. You edit the
|
|
883
|
+
tree; Gauge reruns the source sessions with it injected. Nothing deploys. Use -o json.
|
|
884
|
+
Choose an issue, and fix or recheck it once the user ships: \`gauge instructions issues\`.
|
|
702
885
|
|
|
703
886
|
## The loop
|
|
704
887
|
|
|
705
|
-
gauge
|
|
706
|
-
gauge
|
|
707
|
-
gauge
|
|
708
|
-
gauge optimizations
|
|
709
|
-
|
|
710
|
-
|
|
711
|
-
|
|
712
|
-
|
|
713
|
-
|
|
714
|
-
|
|
715
|
-
|
|
716
|
-
|
|
717
|
-
|
|
718
|
-
gauge optimizations
|
|
719
|
-
|
|
720
|
-
|
|
721
|
-
|
|
722
|
-
|
|
723
|
-
|
|
724
|
-
|
|
725
|
-
|
|
726
|
-
|
|
727
|
-
|
|
728
|
-
|
|
729
|
-
|
|
730
|
-
|
|
731
|
-
|
|
732
|
-
|
|
733
|
-
|
|
734
|
-
|
|
735
|
-
|
|
736
|
-
|
|
737
|
-
|
|
738
|
-
|
|
739
|
-
|
|
888
|
+
gauge issues list # by impact; --type friction|objection
|
|
889
|
+
gauge issues get <id> # the issue's findings and runs
|
|
890
|
+
gauge optimizations create --issue <id> # or --finding <id> --run <run>, or --goal <text>
|
|
891
|
+
gauge optimizations authoring <id> # findings, tree, editing practices
|
|
892
|
+
gauge run trace <run> --seq <n> # events around a finding's evidence seq
|
|
893
|
+
gauge run web <run> # every URL the agent fetched
|
|
894
|
+
gauge run commands <run> --query <text> # shell output the agent saw
|
|
895
|
+
|
|
896
|
+
Before you edit, name the finding, the seq where the agent stalled, and what it
|
|
897
|
+
will read differently there. Read a page before you change it, keep its purpose,
|
|
898
|
+
and source every new fact. A surface the org does not own is a what-if the user
|
|
899
|
+
cannot ship. Tree keys are page:<url>, skill:<name>, and cli:<package>.
|
|
900
|
+
|
|
901
|
+
gauge optimizations file add <id> --page <url> # copy the original in; or --skill, --cli
|
|
902
|
+
gauge optimizations file read <id> page:<url> --text # or --query <text>: exact source
|
|
903
|
+
gauge optimizations file edit <id> --page <url> --edits edits.json # [{find, replace}]
|
|
904
|
+
gauge optimizations file write <id> --skill <name> --path SKILL.md --from SKILL.md
|
|
905
|
+
gauge optimizations file write <id> --cli <package> --from rules.json
|
|
906
|
+
gauge optimizations notes <id> --from notes.md # diagnosis, plan, each launch's result
|
|
907
|
+
|
|
908
|
+
Copy find text from --query; check how much of the page each edit kept. Use
|
|
909
|
+
file write --page only for a new page or a full rewrite. Skills need SKILL.md;
|
|
910
|
+
CLI rules are {schemaVersion: 3, target, rules, shippingIntent}. Files not in
|
|
911
|
+
the tree rerun as the source session had them.
|
|
912
|
+
|
|
913
|
+
## Rerun and compare
|
|
914
|
+
|
|
915
|
+
gauge optimizations launch <id> --run <run> --yes # ask first; restores auto
|
|
916
|
+
gauge optimizations launch <id> --restore-before <rerun>:<seq> --yes # or --fresh
|
|
917
|
+
gauge optimizations watch <id> # waits for reruns and analysis
|
|
918
|
+
|
|
919
|
+
Launch restores each rerun before its first read of a changed surface, or
|
|
920
|
+
starts fresh, and prints the checkpoint and seq. Rerun only the sessions your
|
|
921
|
+
edits address; check that launch lists every edit as changed. Findings come
|
|
922
|
+
back resolved, returned (also the same barrier in other words), or introduced.
|
|
923
|
+
Before you credit an edit, check with \`gauge run web <rerun>\` that the rerun
|
|
924
|
+
read it. Judge findings first, report criteria and failed reruns, and iterate.
|
|
925
|
+
|
|
926
|
+
gauge optimizations finish <id> --from result.json --yes
|
|
927
|
+
|
|
928
|
+
result.json is {summary, changed, resolved, returned, introduced}: a PR-style
|
|
929
|
+
summary, the changed tree keys, and {runId, label} findings.
|
|
930
|
+
`;var TOPICS={root:{name:"gauge",hosted:ROOT,local:ROOT_LOCAL},init:{name:"gauge-init",hosted:INIT,local:INIT_LOCAL,command:"init"},hillclimb:{name:"gauge-hillclimb",hosted:HILLCLIMB,shared:true},run:{name:"gauge-run",hosted:RUN,local:RUN_LOCAL,command:"run"},lifecycle:{name:"gauge-lifecycle",hosted:LIFECYCLE},eval:{name:"gauge-eval",hosted:EVAL,shared:true,command:"eval"},preference:{name:"gauge-preference",hosted:PREFERENCE,shared:true,command:"preference"},"session-analysis":{name:"gauge-session-analysis",hosted:SESSION_ANALYSIS,local:SESSION_ANALYSIS_LOCAL},issues:{name:"gauge-issues",hosted:ISSUES,command:"issues"},optimization:{name:"gauge-optimizations",hosted:OPTIMIZATIONS,command:"optimizations"}};var TOPIC_NAMES=Object.keys(TOPICS).filter(topic=>topic!=="root");var NO_STATE={mode:"hosted",source:"default",inRepo:true,config:{},signedIn:true};function instructions(topic,state=NO_STATE,{setup=true}={}){const key=topicKey(topic??"root");const found=TOPICS[key];const markdown=instructionMode(state).mode==="local"&&found.local?found.local:found.hosted;const lines=header(setup?state:{...state,config:{}},{key,local:found.local!==void 0,shared:found.shared}).filter(line=>setup||!line.startsWith("Next:"));const end=markdown.indexOf("\n---\n",4)+5;const quoted=lines.map(line=>`> ${line}`).join("\n");const footer=found.command?`
|
|
740
931
|
Run \`gauge ${found.command} -h\` for usage and flag options.
|
|
741
932
|
`:"";return`${markdown.slice(0,end)}
|
|
742
933
|
${quoted}
|
|
743
|
-
${markdown.slice(end)}${footer}`}function normalize2(topic){const normalized2=topic.trim().toLowerCase();return normalized2==="optimizations"?"optimization":normalized2}function isTopic(topic){return TOPICS[normalize2(topic)]!==void 0}function topicForCommand(command){return TOPIC_NAMES.find(key=>TOPICS[key].command===command)}function topicKey(topic){const key=normalize2(topic);if(!TOPICS[key])throw new CliError("usage",`unknown topic "${topic}" (expected ${TOPIC_NAMES.join(", ")})`,{hint:"Topics are workflow guidance. Run `gauge --help` to list commands, and `gauge <command> -h` for flags."});return key}function
|
|
744
|
-
${command.helpInformation()}`);return}process.stdout.write(instructions(topic,gaugeState(cmd)))})}var init_exports={};__export(init_exports,{register:()=>
|
|
934
|
+
${markdown.slice(end)}${footer}`}function normalize2(topic){const normalized2=topic.trim().toLowerCase();return normalized2==="optimizations"?"optimization":normalized2}function isTopic(topic){return TOPICS[normalize2(topic)]!==void 0}function topicForCommand(command){return TOPIC_NAMES.find(key=>TOPICS[key].command===command)}function topicKey(topic){const key=normalize2(topic);if(!TOPICS[key])throw new CliError("usage",`unknown topic "${topic}" (expected ${TOPIC_NAMES.join(", ")})`,{hint:"Topics are workflow guidance. Run `gauge --help` to list commands, and `gauge <command> -h` for flags."});return key}function register12(program2){program2.command("instructions [topic]").description(`Instructions for a coding agent using this CLI. Run before your first gauge command; a topic loads the full workflow guidance: ${TOPIC_NAMES.join(", ")}. Topics are workflows, not a page per command: run \`gauge <command> -h\` for flags`).option("--list","print the topic names only").action((requested,opts,cmd)=>{if(opts.list){for(const name2 of TOPIC_NAMES)console.log(name2);return}const name=requested?.trim().toLowerCase();const command=name&&!isTopic(name)?cmd.parent?.commands.find(c=>c.name()===name||c.aliases().includes(name)):void 0;const topic=(command&&topicForCommand(command.name()))??requested;if(command&&topic===requested){const help=`gauge ${command.name()} -h`;process.stdout.write(`"${command.name()}" is a command, not an instructions topic. Topics: ${TOPIC_NAMES.join(", ")}. Its help (\`${help}\`):
|
|
935
|
+
${command.helpInformation()}`);return}process.stdout.write(instructions(topic,gaugeState(cmd)))})}var init_exports={};__export(init_exports,{register:()=>register13,summarize:()=>summarize});import{setTimeout as delay}from"timers/promises";var SKILL_AGENTS={CLAUDECODE:"claude-code",CODEX_SANDBOX:"codex",CODEX_THREAD_ID:"codex",CODEX_CI:"codex",CURSOR_AGENT:"cursor",GEMINI_CLI:"gemini-cli",CLINE_ACTIVE:"cline",OPENCODE:"opencode"};var AGENT_MARKERS=[...Object.keys(SKILL_AGENTS),"AI_AGENT"];function detectAgent(env,stdoutIsTTY){if(AGENT_MARKERS.some(name=>env[name]))return"agent";return stdoutIsTTY?"human":"agent"}function detectSkillAgent(env){return Object.entries(SKILL_AGENTS).find(([name])=>env[name])?.[1]}var DOCS_URL="https://docs.withgauge.com/agents";function noCasesError(){return new CliError("not_found","No case files found in this workspace.",{hint:"Plain `gauge init` creates no measurements. Save an eval or preference with `gauge eval create` or `gauge preference create`, then use `gauge run --eval <id>` or `--preference <id>`. For shared case files, read `gauge instructions run`. Start with `gauge instructions hillclimb`."})}import{resolve as resolve2}from"path";import{existsSync as existsSync3,mkdirSync,readFileSync as readFileSync7,writeFileSync}from"fs";import{join as join7}from"path";var START="<!-- gauge:start -->";var END="<!-- gauge:end -->";var AGENTS_BLOCK=`${START}
|
|
745
936
|
Gauge project inputs and optional shared cases are configured in gauge.json.
|
|
746
937
|
Saved evals and preference prompts, sessions, and analyses live in private .gauge/
|
|
747
938
|
state locally or the hosted organization. Shared cases use gauge-cases/.
|
|
@@ -750,53 +941,35 @@ For product improvement or a first pilot, use \`gauge instructions hillclimb\`:
|
|
|
750
941
|
identify inputs, run representative sessions, analyze, improve, rerun, compare.
|
|
751
942
|
After local runs, follow \`gauge instructions session-analysis\` to publish
|
|
752
943
|
structured findings. Add \`-o json\` for machine-readable output.
|
|
753
|
-
${END}`;function scaffoldWorkspace(root,backend,cases=false){mkdirSync(root,{recursive:true});const manifest=join7(root,"gauge.json");const casesPath=join7(root,"gauge-cases");const gaugeDir=join7(root,".gauge");const result={root,backend,gaugeJson:existsSync3(manifest)?"kept":"created",casesDir:cases?existsSync3(casesPath)?"kept":"created":"not-requested",agentsMd:writeAgentsBlock(join7(root,"AGENTS.md"))};if(result.gaugeJson==="created")writeFileSync(manifest,JSON.stringify({version:2,inputs:{}},null,2)+"\n");mkdirSync(gaugeDir,{recursive:true,mode:448});writeFileSync(join7(gaugeDir,".gitignore"),"*\n");writeFileSync(join7(gaugeDir,"config.json"),JSON.stringify({version:1,backend},null,2)+"\n",{mode:384});const store=openLocalStore(gaugeDir);store.close();if(result.casesDir==="created"){mkdirSync(casesPath);writeFileSync(join7(casesPath,".gitkeep"),"")}return result}function writeAgentsBlock(path){if(!existsSync3(path)){writeFileSync(path,`${AGENTS_BLOCK}
|
|
754
|
-
`);return"created"}const text=
|
|
755
|
-
`);return"added"}const next=text.slice(0,start)+AGENTS_BLOCK+text.slice(end+END.length);if(next===text)return"unchanged";writeFileSync(path,next);return"updated"}var LOGO=String.raw`
|
|
756
|
-
|
|
757
|
-
⣠⣶⣿⡿⠿⣿⣿⣦⡀ ⣼⣷⣄
|
|
758
|
-
⣰⣿⡿⠁ ⠙⠿⠿⠄ ⢀⣤⣤⣤⣤⡀ ⢠⣤⡄ ⢠⣤⡄ ⢀⣤⣤⣤⡀⣤⣤ ⣀⣤⣤⣤⣄⡀ ⢀⣼⣿⣿⣿⣀
|
|
759
|
-
⣿⣿⡇ ⣶⣶⣶⣶⣶⡆⠰⠿⠟⠉⠉⢿⣿⡆⢸⣿⡇ ⢸⣿⡇⢰⣿⡿⠋⠉⠻⣿⣿ ⣼⣿⠟⠉⠙⢿⣷⡄ ⢠⣾⣿⣿⣿⣿⠿⢿⣷⡄
|
|
760
|
-
⢿⣿⣇ ⠉⠉⠉⣿⣿⡇⢠⣶⣾⠿⠿⢿⣿⡇⢸⣿⡇ ⢸⣿⡇⢸⣿⡇ ⣿⣿ ⣿⣿⣿⣿⣿⣿⣿⡇ ⣿⣄⢀⣴⣿⣿⣿⣿⣿⣷⣶⣶⣿⣿⣶⣶⣶⡆
|
|
761
|
-
⠈⢿⣿⣦⣤⣀⣤⣾⢿⣿⡇⢿⣿⣇⡀⣀⣾⣿⡇⢸⣿⣷⣀⣀⣾⣿⡇⠘⣿⣿⣄⣀⣴⣿⣿ ⢻⣿⣦⣀⣀⣶⣶⠆ ⠙⠛⠋⠈⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠇
|
|
762
|
-
⠉⠛⠛⠛⠛⠁⠘⠛⠃⠈⠙⠛⠛⠋⠘⠛⠃ ⠙⠛⠛⠋⠘⠛⠃⢀⣈⣙⠛⠋⠁⣿⣿ ⠉⠛⠛⠛⠛⠁ ⠹⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠿⠋
|
|
763
|
-
⠻⢿⣷⣶⣿⠿⠋ ⠈⢿⠿⠋⠉
|
|
764
|
-
`;var BARK=String.raw`
|
|
765
|
-
⣀
|
|
766
|
-
⣠⣶⣿⡿⠿⣿⣿⣦⡀ ⢠⣿⣷⣄
|
|
767
|
-
⣰⣿⡿⠁ ⠙⠿⠿⠄ ⢀⣤⣤⣤⣤⡀ ⢠⣤⡄ ⢠⣤⡄ ⢀⣤⣤⣤⡀⣤⣤ ⣀⣤⣤⣤⣄⡀ ⢀⣾⣿⣿⣿⣤⣄⣀ ⣠⡾⠋
|
|
768
|
-
⣿⣿⡇ ⣶⣶⣶⣶⣶⡆⠰⠿⠟⠉⠉⢿⣿⡆⢸⣿⡇ ⢸⣿⡇⢰⣿⡿⠋⠉⠻⣿⣿ ⣼⣿⠟⠉⠙⢿⣷⡄ ⢀⣾⣿⣿⣿⣿⠛⠛⣿⣧⣀⣠⣤⣄ ⠨⠶⠶⠶
|
|
769
|
-
⢿⣿⣇ ⠉⠉⠉⣿⣿⡇⢠⣶⣾⠿⠿⢿⣿⡇⢸⣿⡇ ⢸⣿⡇⢸⣿⡇ ⣿⣿ ⣿⣿⣿⣿⣿⣿⣿⡇ ⢰⣄ ⣠⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠿⠿⠿⠿ ⠘⢷⣤⡀
|
|
770
|
-
⠈⢿⣿⣦⣤⣀⣤⣾⢿⣿⡇⢿⣿⣇⡀⣀⣾⣿⡇⢸⣿⣷⣀⣀⣾⣿⡇⠘⣿⣿⣄⣀⣴⣿⣿ ⢻⣿⣦⣀⣀⣶⣶⠆ ⠾⠿⠋⠻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷⡦ ⠈
|
|
771
|
-
⠉⠛⠛⠛⠛⠁⠘⠛⠃⠈⠙⠛⠛⠋⠘⠛⠃ ⠙⠛⠛⠋⠘⠛⠃⢀⣈⣙⠛⠋⠁⣿⣿ ⠉⠛⠛⠛⠛⠁ ⠈⠻⣿⣿⣿⣿⡿⠿⠛⠛⠛⠋
|
|
772
|
-
⠻⢿⣷⣶⣿⠿⠋ ⠈⠿⠟⠁
|
|
773
|
-
`;var SMALL_LOGO=String.raw`█▀▀ ▄▀█ █ █ █▀▀ █▀▀
|
|
774
|
-
█▄█ █▀█ █▄█ █▄█ ██▄`;async function printLogo(){const out=process.stdout;if(detectAgent(process.env,out.isTTY)==="agent"){console.log(SMALL_LOGO);return}if(process.env.CI||process.env.TERM==="dumb"){console.log(LOGO.replace(/^\n+/,""));return}out.write("\x1B[?25l");try{out.write(LOGO);const up=`\x1B[${LOGO.split("\n").length-1}A`;for(const[frame,wait]of[[BARK,220],[LOGO,150],[BARK,220],[LOGO,150],[BARK,220],[LOGO,150]]){await delay(wait);out.write(up+frame.replace(/\n/g,"\x1B[K\n"))}out.write("\n")}finally{out.write("\x1B[?25h")}}function nextSteps(mode){const bold=style("1");const cyan=style("36");const dim=style("2");const lines=[];lines.push(` ${bold("1. Install the Gauge agent skill")}`);lines.push(` ${cyan("$ gauge install-skill")}`);lines.push(` ${dim("\xB7")} Recommended. It teaches your coding agent the Gauge workflows.`);if(detectAgent(process.env,process.stdout.isTTY)==="agent")lines.push(` ${dim("\xB7")} Ask the user whether to install it, unless already installed or declined.`);lines.push("");(mode==="hosted"?HOSTED_STEPS:LOCAL_STEPS).forEach((step,i)=>{lines.push(` ${bold(`${i+2}. ${step.title}`)}`);if(step.command)lines.push(` ${cyan(`$ ${step.command}`)}`);for(const detail of step.details)lines.push(` ${dim("\xB7")} ${detail}`);lines.push("")});lines.push(`Agents: add ${cyan("-o json")} to any gauge command for JSON output.`);lines.push(`Docs: ${cyan(DOCS_URL)}`);return lines.join("\n")}function summarize(result){return`${result.gaugeJson==="created"?"Created":"Kept"} gauge.json. Backend: ${result.backend} (.gauge/config.json). Local state: .gauge/gauge.sqlite.${result.casesDir==="not-requested"?"":` Shared cases: gauge-cases/ (${result.casesDir}).`}`}function register14(program2,{cwd=process.cwd()}={}){program2.command("init").description("Initialize a Gauge workspace: shared gauge.json and private .gauge/ state. No Git or account required. Backend defaults to local independently of login.").option("--cases","also scaffold shared gauge-cases/ files").action(async(opts,cmd)=>{const state=readGaugeState(cmd.optsWithGlobals(),cwd);if(state.config===null)throw new CliError("usage","gauge.json is not valid JSON. Fix it before initializing.");const result=scaffoldWorkspace(state.root??resolve2(cwd),state.mode,opts.cases);if(resolveOutput(cmd)==="json"){printJson(result);return}await printLogo();console.log(`${style("32")("\u2713")} gauge init complete`);console.log(summarize(result));console.log(nextSteps(state.mode))})}var installSkill_exports={};__export(installSkill_exports,{register:()=>register15,summarizeInstaller:()=>summarizeInstaller});import{spawnSync}from"child_process";var ANSI=/\x1b\[[0-9;?]*[A-Za-z]/g;function summarizeInstaller(output){const lines=output.replace(ANSI,"").split("\n").map(line=>line.replace(/^\s*│/,"").replace(/│\s*$/,"").trim());const paths=[...new Set(lines.flatMap(line=>line.match(/^→ (\S+)$/)?.[1]??[]))];const audits={};const row2=lines.findIndex(line=>/^gauge-agents\s{2,}/.test(line));if(row2>0){const names=lines[row2-1].split(/\s{2,}/);const values=lines[row2].split(/\s{2,}/).slice(1);names.forEach((name,i)=>{if(values[i])audits[name]=values[i]})}const details=lines.flatMap(line=>line.match(/^Details: (\S+)$/)?.[1]??[])[0];return{paths,audits,details}}function register15(program2,{run=spawnSync,platform=process.platform,env=process.env,isTTY=process.stdout.isTTY}={}){program2.command("install-skill").description("Install Gauge's gauge-agents skill from gauge-sh/gauge-skills on GitHub using npx. People pick agents and scope in the installer; coding agents install for themselves in this project without prompts. No Gauge account needed").option("-a, --agent <name>","target an agent (e.g. claude-code or codex); repeat for multiple agents. Coding agents default to themselves",(value,previous)=>[...previous,value],[]).option("-g, --global","install across projects instead of in this repo").option("-y, --yes","skip installer prompts (requires --agent; defaults to project scope). Always on for coding agents").option("--copy","copy skill files instead of creating symlinks").action((opts,cmd)=>{const quiet=detectAgent(env,isTTY)==="agent";const self=quiet?detectSkillAgent(env):void 0;const agents=opts.agent.length||!self?opts.agent:[self];if(quiet&&!agents.length)throw new CliError("usage","Cannot detect the coding agent, and the installer cannot prompt here.",{hint:"Pass the agent, e.g. gauge install-skill --agent claude-code"});if(opts.yes&&!agents.length)throw new CliError("usage","--yes requires --agent to avoid installing for unrelated agents.",{hint:"For Claude Code: gauge install-skill --agent claude-code --yes"});if(agents.some(agent=>!/^[a-z0-9-]+$/.test(agent)))throw new CliError("usage","Agent names must contain only lowercase letters, digits, and hyphens.");const args=["--yes","skills@1.5.0","add","gauge-sh/gauge-skills","--skill","gauge-agents"];for(const agent of agents)args.push("--agent",agent);if(opts.global)args.push("--global");if(opts.yes||quiet)args.push("--yes");if(opts.copy)args.push("--copy");const json=resolveOutput(cmd)==="json";console.error("Installing gauge-agents from gauge-sh/gauge-skills...");const result=run(platform==="win32"?"npx.cmd":"npx",args,{shell:platform==="win32",stdio:quiet?["ignore","pipe","pipe"]:["inherit",json?2:"inherit","inherit"]});const output=quiet?`${result.stdout??""}${result.stderr??""}`:"";if(result.error||result.status!==0){if(output)process.stderr.write(output.replace(ANSI,""));const reason=result.error?.message??(result.signal?`installer interrupted by ${result.signal}`:`installer exited with code ${result.status}`);throw new CliError("internal",`Skill installation failed: ${reason}.`,{hint:"Run gauge install-skill again. Node.js and network access are required."})}if(!quiet){if(json)printJson({completed:true,skill:"gauge-agents"});else console.log("Gauge skill installer completed.");return}const summary=summarizeInstaller(output);if(json){printJson({completed:true,skill:"gauge-agents",agents,...summary});return}console.log(`Installed gauge-agents for ${agents.join(", ")}${summary.paths.length?`: ${summary.paths.join(", ")}`:"."}`);const audits=Object.entries(summary.audits).map(([name,value])=>`${name} ${value}`);if(audits.length)console.log(`skills.sh audits: ${audits.join(", ")}.${summary.details?` Details: ${summary.details}`:""}`)})}var keys_exports2={};__export(keys_exports2,{register:()=>register16});function keysPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/provider-keys`}function providerArg(raw){const provider=raw.toLowerCase();if(!keys_exports.PROVIDER_TYPES.includes(provider)){throw new CliError("usage",`Unknown provider "${raw}" (expected ${keys_exports.PROVIDER_TYPES.join("|")})`)}return provider}async function readKeyInput(opts){if(!opts.stdin){const env=process.env.GAUGE_PROVIDER_KEY?.trim();if(env)return env}if(process.stdin.isTTY){process.stderr.write("Paste the API key, press Enter, then Ctrl-D:\n")}process.stdin.setEncoding("utf8");let data="";for await(const chunk of process.stdin)data+=chunk;const key=data.trim();if(!key){throw new CliError("usage","No key provided. Pipe it to stdin (e.g. `gauge keys set anthropic < key.txt`) or set GAUGE_PROVIDER_KEY.")}return key}function register16(program2){const group=hostedOnly(program2.command("keys")).description("Manage BYOK provider keys (write-only; reads show last4)");group.command("set <provider>").description("Set a provider key (anthropic|openai); reads it from stdin or GAUGE_PROVIDER_KEY").option("--stdin","read the key from stdin (ignores GAUGE_PROVIDER_KEY)").action(async(provider,opts,cmd)=>{const type=providerArg(provider);const org=await resolveOrg(cmd);const key=await readKeyInput(opts);const res=await createClient().put(`${keysPath(org)}/${type}`,{key});if(resolveOutput(cmd)==="json")printJson(res);else console.log(`Saved ${res.providerType} key (\u2026${res.last4})`)});group.command("list").description("List configured provider keys").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(keysPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(k=>({providerType:k.providerType,last4:`\u2026${k.last4}`,createdAt:k.createdAt,updatedAt:k.updatedAt})),"table")});group.command("rm <provider>").description("Remove a provider key").action(async(provider,_opts,cmd)=>{const type=providerArg(provider);const org=await resolveOrg(cmd);await createClient().delete(`${keysPath(org)}/${type}`);if(resolveOutput(cmd)!=="json")console.log(`Removed ${type} key`)});group.command("disable-byok").description("Switch back to Gauge-managed metered billing (remove provider keys first)").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(keysPath(org));if(resolveOutput(cmd)!=="json")console.log("BYOK disabled \u2014 runs are Gauge-managed and metered again")})}var localRuntime_exports={};__export(localRuntime_exports,{register:()=>register17});function print(result,command){if(resolveOutput(command)==="json")printJson(result);else{console.log(`Runtime: ${result.reference}`);if(result.imageId)console.log(`Image ID: ${result.imageId}`);if(result.repoDigest)console.log(`Registry digest: ${result.repoDigest}`);console.log(`Status: ${result.ready?"ready":result.problem?.code}`);if(result.platform)console.log(`Platform: ${result.platform}`);if(result.problem)console.log(`${result.problem.message}
|
|
944
|
+
${END}`;function scaffoldWorkspace(root,backend,cases=false){mkdirSync(root,{recursive:true});const manifest=join7(root,"gauge.json");const casesPath=join7(root,"gauge-cases");const gaugeDir=join7(root,".gauge");const shared=cases||existsSync3(casesPath);const result={root,backend,gaugeJson:existsSync3(manifest)?"kept":shared?"created":"not-requested",casesDir:cases?existsSync3(casesPath)?"kept":"created":"not-requested",agentsMd:shared?writeAgentsBlock(join7(root,"AGENTS.md")):"not-requested"};if(result.gaugeJson==="created")writeFileSync(manifest,JSON.stringify({version:2,inputs:{}},null,2)+"\n");mkdirSync(gaugeDir,{recursive:true,mode:448});writeFileSync(join7(gaugeDir,".gitignore"),"*\n");writeFileSync(join7(gaugeDir,"config.json"),JSON.stringify({version:1,backend},null,2)+"\n",{mode:384});const store=openLocalStore(gaugeDir);store.close();if(result.casesDir==="created"){mkdirSync(casesPath);writeFileSync(join7(casesPath,".gitkeep"),"")}return result}function writeAgentsBlock(path){if(!existsSync3(path)){writeFileSync(path,`${AGENTS_BLOCK}
|
|
945
|
+
`);return"created"}const text=readFileSync7(path,"utf8");const start=text.indexOf(START);const end=start===-1?-1:text.indexOf(END,start);if(end===-1){const separator=text===""?"":text.endsWith("\n")?"\n":"\n\n";writeFileSync(path,`${text}${separator}${AGENTS_BLOCK}
|
|
946
|
+
`);return"added"}const next=text.slice(0,start)+AGENTS_BLOCK+text.slice(end+END.length);if(next===text)return"unchanged";writeFileSync(path,next);return"updated"}async function printLogo(){const out=process.stdout;if(detectAgent(process.env,out.isTTY)==="agent"){console.log(SMALL_LOGO);return}if(process.env.CI||process.env.TERM==="dumb"){console.log(LOGO_ART);return}out.write("\x1B[?25l");try{out.write(LOGO);const up=`\x1B[${LOGO.split("\n").length-1}A`;for(const[frame,wait]of[[BARK,220],[LOGO,150],[BARK,220],[LOGO,150],[BARK,220],[LOGO,150]]){await delay(wait);out.write(up+frame.replace(/\n/g,"\x1B[K\n"))}out.write("\n")}finally{out.write("\x1B[?25h")}}function nextSteps(mode){const bold=style("1");const cyan=style("36");const dim=style("2");const lines=[];lines.push(` ${bold("1. Install the Gauge agent skill")}`);lines.push(` ${cyan("$ gauge install-skill")}`);lines.push(` ${dim("\xB7")} Recommended. It teaches your coding agent the Gauge workflows.`);if(detectAgent(process.env,process.stdout.isTTY)==="agent")lines.push(` ${dim("\xB7")} Ask the user whether to install it, unless already installed or declined.`);lines.push("");(mode==="hosted"?HOSTED_STEPS:LOCAL_STEPS).forEach((step,i)=>{lines.push(` ${bold(`${i+2}. ${step.title}`)}`);if(step.command)lines.push(` ${cyan(`$ ${step.command}`)}`);for(const detail of step.details)lines.push(` ${dim("\xB7")} ${detail}`);lines.push("")});lines.push(`Agents: add ${cyan("-o json")} to any gauge command for JSON output.`);lines.push("Agents: the user does not see gauge output. Show them the files you write, each case before launch, and the results.");if(mode==="local")lines.push(`Agents: read ${cyan("gauge instructions init")} before you show the user the plan.`);lines.push(`Docs: ${cyan(DOCS_URL)}`);return lines.join("\n")}function summarize(result){const manifest=result.gaugeJson==="not-requested"?"":`${result.gaugeJson==="created"?"Created":"Kept"} gauge.json. `;return`${manifest}Backend: ${result.backend} (.gauge/config.json). Local state: .gauge/gauge.sqlite.${result.casesDir==="not-requested"?"":` Shared cases: gauge-cases/ (${result.casesDir}).`}`}function register13(program2,{cwd=process.cwd()}={}){program2.command("init").description("Initialize a Gauge workspace: private .gauge/ state, plus shared gauge.json, gauge-cases/, and an AGENTS.md block with --cases. No Git or account required. Backend defaults to local independently of login.").option("--cases","also write shared gauge.json, gauge-cases/, and an AGENTS.md block").action(async(opts,cmd)=>{const state=readGaugeState(cmd.optsWithGlobals(),cwd);if(state.config===null)throw new CliError("usage","gauge.json is not valid JSON. Fix it before initializing.");const result=scaffoldWorkspace(state.root??resolve2(cwd),state.mode,opts.cases);if(resolveOutput(cmd)==="json"){printJson(result);return}await printLogo();console.log(`${style("32")("\u2713")} gauge init complete`);console.log(summarize(result));console.log(nextSteps(state.mode))})}var installSkill_exports={};__export(installSkill_exports,{register:()=>register14,summarizeInstaller:()=>summarizeInstaller});import{spawnSync}from"child_process";var ANSI=/\x1b\[[0-9;?]*[A-Za-z]/g;function summarizeInstaller(output){const lines=output.replace(ANSI,"").split("\n").map(line=>line.replace(/^\s*│/,"").replace(/│\s*$/,"").trim());const paths=[...new Set(lines.flatMap(line=>line.match(/^→ (\S+)$/)?.[1]??[]))];const audits={};const row2=lines.findIndex(line=>/^gauge-agents\s{2,}/.test(line));if(row2>0){const names=lines[row2-1].split(/\s{2,}/);const values=lines[row2].split(/\s{2,}/).slice(1);names.forEach((name,i)=>{if(values[i])audits[name]=values[i]})}const details=lines.flatMap(line=>line.match(/^Details: (\S+)$/)?.[1]??[])[0];return{paths,audits,details}}function register14(program2,{run=spawnSync,platform=process.platform,env=process.env,isTTY=process.stdout.isTTY}={}){program2.command("install-skill").description("Install Gauge's gauge-agents skill from gauge-sh/gauge-skills on GitHub using npx. People pick agents and scope in the installer; coding agents install for themselves in this project without prompts. No Gauge account needed").option("-a, --agent <name>","target an agent (e.g. claude-code or codex); repeat for multiple agents. Coding agents default to themselves",(value,previous)=>[...previous,value],[]).option("-g, --global","install across projects instead of in this repo").option("-y, --yes","skip installer prompts (requires --agent; defaults to project scope). Always on for coding agents").option("--copy","copy skill files instead of creating symlinks").action((opts,cmd)=>{const quiet=detectAgent(env,isTTY)==="agent";const self=quiet?detectSkillAgent(env):void 0;const agents=opts.agent.length||!self?opts.agent:[self];if(quiet&&!agents.length)throw new CliError("usage","Cannot detect the coding agent, and the installer cannot prompt here.",{hint:"Pass the agent, e.g. gauge install-skill --agent claude-code"});if(opts.yes&&!agents.length)throw new CliError("usage","--yes requires --agent to avoid installing for unrelated agents.",{hint:"For Claude Code: gauge install-skill --agent claude-code --yes"});if(agents.some(agent=>!/^[a-z0-9-]+$/.test(agent)))throw new CliError("usage","Agent names must contain only lowercase letters, digits, and hyphens.");const args=["--yes","skills@1.5.0","add","gauge-sh/gauge-skills","--skill","gauge-agents"];for(const agent of agents)args.push("--agent",agent);if(opts.global)args.push("--global");if(opts.yes||quiet)args.push("--yes");if(opts.copy)args.push("--copy");const json=resolveOutput(cmd)==="json";console.error("Installing gauge-agents from gauge-sh/gauge-skills...");const result=run(platform==="win32"?"npx.cmd":"npx",args,{shell:platform==="win32",stdio:quiet?["ignore","pipe","pipe"]:["inherit",json?2:"inherit","inherit"]});const output=quiet?`${result.stdout??""}${result.stderr??""}`:"";if(result.error||result.status!==0){if(output)process.stderr.write(output.replace(ANSI,""));const reason=result.error?.message??(result.signal?`installer interrupted by ${result.signal}`:`installer exited with code ${result.status}`);throw new CliError("internal",`Skill installation failed: ${reason}.`,{hint:"Run gauge install-skill again. Node.js and network access are required."})}if(!quiet){if(json)printJson({completed:true,skill:"gauge-agents"});else console.log("Gauge skill installer completed.");return}const summary=summarizeInstaller(output);if(json){printJson({completed:true,skill:"gauge-agents",agents,...summary});return}console.log(`Installed gauge-agents for ${agents.join(", ")}${summary.paths.length?`: ${summary.paths.join(", ")}`:"."}`);const audits=Object.entries(summary.audits).map(([name,value])=>`${name} ${value}`);if(audits.length)console.log(`skills.sh audits: ${audits.join(", ")}.${summary.details?` Details: ${summary.details}`:""}`)})}var issues_exports={};__export(issues_exports,{flattenBlocks:()=>flattenBlocks,impactLine:()=>impactLine,register:()=>register15,statusLine:()=>statusLine});import{Option as Option4}from"commander";var TYPES={friction:"problems",objection:"reasons"};var STATUSES=["open","fixed","muted","all"];function statusLine(issue){const{check:check3}=issue;if(check3?.status==="RUNNING")return`${issue.status.toLowerCase()}, recheck running`;if(issue.status==="FIXED")return`fixed${issue.statusBy?` by ${issue.statusBy}`:""}`;if(issue.status==="MUTED")return`muted${issue.statusBy?` by ${issue.statusBy}`:""}`;const fresh=check3&&(issue.statusAt==null||check3.createdAt>issue.statusAt)&&(issue.cameBackAt==null||check3.createdAt>issue.cameBackAt);if(fresh&&check3.status==="VERIFIED")return`open, looks fixed: 0 of ${check3.sessions??0} runs on the last recheck`;if(fresh&&check3.status==="STILL_SEEN")return`open, still seen in ${check3.seen??0} of ${check3.sessions??0} runs on the last recheck`;if(issue.cameBackAt)return"open, came back after a fix";return"open"}function flattenBlocks(type,blocks){return{issues:blocks.flatMap(block=>block.issues.map(issue=>({...issue,type,product:block.product}))),ungrouped:blocks.reduce((sum,block)=>sum+block.ungrouped,0)}}function issuesPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/issues`}function impactLine(issue){if(issue.impact)return issue.impact.map(i=>`${i.effect.toLowerCase().replace("_"," ")} ${i.sessions}`).join(", ");return issue.wentTo.map(w=>`${w.name} ${w.sessions}`).join(", ")}function register15(program2){const group=hostedOnly(program2.command("issues")).description("The same finding across many sessions, by impact. Also called findings");group.command("list").description("Friction and objections across your products, most sessions first").addOption(new Option4("--type <type>","only this type").choices(Object.keys(TYPES))).addOption(new Option4("--status <status>","only issues with this status").choices(STATUSES).default("open")).action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const types=opts.type?[opts.type]:Object.keys(TYPES);const results=await Promise.all(types.map(async type=>{const res=await createClient().get(issuesPath(org),{kind:TYPES[type],status:opts.status});return flattenBlocks(type,res.blocks)}));const issues=results.flatMap(r=>r.issues).sort((a,b)=>b.sessions-a.sessions);if(resolveOutput(cmd)==="json")return printJson(issues);if(!issues.length)console.log("No issues yet. Issues group the findings from analyzed eval and preference runs.");else printItems(issues.map(issue=>({id:issue.id,type:issue.type,product:issue.product,sessions:issue.sessions,title:issue.title,status:statusLine(issue.issue),impact:impactLine(issue)})),"table");const ungrouped=results.reduce((sum,r)=>sum+r.ungrouped,0);if(ungrouped)console.log(`${ungrouped} more finding${ungrouped===1?" is":"s are"} not in an issue yet`)});group.command("get <id>").description("One issue and its findings: run, finding id, effect, and agent").action(async(id,_opts,cmd)=>{const issue=await createClient().get(`${issuesPath(await resolveOrg(cmd))}/${encodeURIComponent(id)}`);if(resolveOutput(cmd)==="json")return printJson(issue);console.log(`${issue.title} (${issue.product}, ${issue.kind})`);console.log(issue.description);console.log(`Status: ${statusLine(issue)}`);const reruns=issue.check?.runs??[];if(reruns.length)console.log(`Last recheck: ${reruns.map(r=>`${r.runId} ${r.seen?"showed it":r.seen===false?"clean":r.status.toLowerCase()}`).join(", ")}`);printItems(issue.instances.map(f=>({run:f.runId,finding:f.id,effect:f.effect,agent:`${f.agent} ${f.model}`,label:f.label})),"table")});for(const[name,status,description]of[["fix","FIXED","Mark an issue fixed. It reopens if agents hit it again"],["mute","MUTED","Mute an issue. New findings still collect, but it never reopens"],["reopen","OPEN","Move a fixed or muted issue back to open"]])group.command(`${name} <id>`).description(description).action(async(id,_opts,cmd)=>{const res=await createClient().patch(`${issuesPath(await resolveOrg(cmd))}/${encodeURIComponent(id)}`,{status});if(resolveOutput(cmd)==="json")return printJson(res);console.log(`${res.id} is ${res.status.toLowerCase()}`)});group.command("recheck <id>").alias("verify").description("Preview, then with --confirm rerun, past runs that showed an issue (spends credits). `issues get` shows which still show it; a fixed issue that shows it reopens").option("--run <runId...>","the past runs to repeat, from the preview (default: the 3 checked ones, at most 10)").option("--confirm","launch the reruns; without it, only preview them").action(async(id,opts,cmd)=>{const path=`${issuesPath(await resolveOrg(cmd))}/${encodeURIComponent(id)}/recheck`;const json=resolveOutput(cmd)==="json";if(opts.run&&opts.run.length>10)throw new CliError("usage","--run takes at most 10 runs");if(!opts.confirm){const{runs}=await createClient().get(path);const known=new Set(runs.map(r=>r.sourceRunId));const unknown=opts.run?.filter(r=>!known.has(r))??[];if(unknown.length)throw new CliError("usage",`Not a past run for ${id}: ${unknown.join(", ")}`);const picked=new Set(opts.run??runs.filter(r=>r.recommended).map(r=>r.sourceRunId));const confirm=picked.size?`gauge issues recheck ${id} --run ${[...picked].join(" ")} --confirm`:null;if(json)return printJson({runs:runs.map(r=>({...r,picked:picked.has(r.sourceRunId)})),confirm});console.log(`Recheck would rerun ${picked.size} of ${runs.length} past runs (the latest per prompt, agent, and model):`);printItems(runs.map(r=>({picked:picked.has(r.sourceRunId)?"\u2713":"",run:r.sourceRunId,prompt:r.title,agent:`${r.agent} ${r.modelName??r.model}`,"seen in":r.sessions})),"table");if(confirm)console.log(`Run it: ${confirm}`);console.log(`Choose: gauge issues recheck ${id} --run <runId>... --confirm`);return}const res=await createClient().post(path,opts.run?{runIds:opts.run}:{});if(json)return printJson(res);if(!res.runIds.length){console.log(`No runs to recheck. Choose some: gauge issues recheck ${id} --run <runId>... --confirm`);return}console.log(`Started ${res.runIds.length} run${res.runIds.length===1?"":"s"}. When they are analyzed, see the result with: gauge issues get ${id}`)})}var keys_exports2={};__export(keys_exports2,{register:()=>register16});function keysPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/provider-keys`}function providerArg(raw){const provider=raw.toLowerCase();if(!keys_exports.PROVIDER_TYPES.includes(provider)){throw new CliError("usage",`Unknown provider "${raw}" (expected ${keys_exports.PROVIDER_TYPES.join("|")})`)}return provider}async function readKeyInput(opts){if(!opts.stdin){const env=process.env.GAUGE_PROVIDER_KEY?.trim();if(env)return env}if(process.stdin.isTTY){process.stderr.write("Paste the API key, press Enter, then Ctrl-D:\n")}process.stdin.setEncoding("utf8");let data="";for await(const chunk of process.stdin)data+=chunk;const key=data.trim();if(!key){throw new CliError("usage","No key provided. Pipe it to stdin (e.g. `gauge keys set anthropic < key.txt`) or set GAUGE_PROVIDER_KEY.")}return key}function register16(program2){const group=hostedOnly(program2.command("keys")).description("Manage BYOK provider keys (write-only; reads show last4)");group.command("set <provider>").description("Set a provider key (anthropic|openai); reads it from stdin or GAUGE_PROVIDER_KEY").option("--stdin","read the key from stdin (ignores GAUGE_PROVIDER_KEY)").action(async(provider,opts,cmd)=>{const type=providerArg(provider);const org=await resolveOrg(cmd);const key=await readKeyInput(opts);const res=await createClient().put(`${keysPath(org)}/${type}`,{key});if(resolveOutput(cmd)==="json")printJson(res);else console.log(`Saved ${res.providerType} key (\u2026${res.last4})`)});group.command("list").description("List configured provider keys").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(keysPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(k=>({providerType:k.providerType,last4:`\u2026${k.last4}`,createdAt:k.createdAt,updatedAt:k.updatedAt})),"table")});group.command("rm <provider>").description("Remove a provider key").action(async(provider,_opts,cmd)=>{const type=providerArg(provider);const org=await resolveOrg(cmd);await createClient().delete(`${keysPath(org)}/${type}`);if(resolveOutput(cmd)!=="json")console.log(`Removed ${type} key`)});group.command("disable-byok").description("Switch back to Gauge-managed metered billing (remove provider keys first)").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(keysPath(org));if(resolveOutput(cmd)!=="json")console.log("BYOK disabled \u2014 runs are Gauge-managed and metered again")})}var localRuntime_exports={};__export(localRuntime_exports,{register:()=>register17});function print(result,command){if(resolveOutput(command)==="json")printJson(result);else{console.log(`Runtime: ${result.reference}`);if(result.imageId)console.log(`Image ID: ${result.imageId}`);if(result.repoDigest)console.log(`Registry digest: ${result.repoDigest}`);console.log(`Status: ${result.ready?"ready":result.problem?.code}`);if(result.platform)console.log(`Platform: ${result.platform}`);if(result.problem)console.log(`${result.problem.message}
|
|
775
947
|
${result.problem.hint}`);console.log(`Gateway: ${result.gateway.reference} (${result.gateway.ready?"ready":result.gateway.problem?.code})`);if(result.gateway.problem)console.log(`${result.gateway.problem.message}
|
|
776
|
-
${result.gateway.problem.hint}`)}if(!result.ready||!result.gateway.ready)process.exitCode=1}function register17(program2){const local=program2.commands.find(command=>command.name()==="local")??program2.command("local").description("Configure local execution (no Gauge account needed)");const runtime=local.command("runtime").description("Inspect or download the supported Docker runtime");runtime.command("inspect").description("Report local runtime availability and compatibility without downloading").action((_options,command)=>print({...inspectLocalRuntime(),gateway:inspectLocalGateway()},command));runtime.command("pull").description("Download the CLI's supported public runtime image").action(async(_options,command)=>print({...await pullLocalRuntime(),gateway:await pullLocalGateway()},command))}var mcp_exports={};__export(mcp_exports,{register:()=>register18});import{readFileSync as readFileSync7}from"fs";import{z as z6}from"zod";import{z as z5}from"zod";var sha256Hex=z5.custom(value=>typeof value==="string"&&/^sha256:[0-9a-f]{64}$/.test(value),"expected a sha256:<64 lowercase hex> digest");var skillHandleSchema=z5.string().min(1).max(128).regex(/^[A-Za-z0-9][A-Za-z0-9._-]*$/,"invalid skill handle").refine(name=>name!=="."&&name!=="..","invalid skill handle");var skillSetEntrySchema=z5.object({skillId:z5.string().min(1),name:skillHandleSchema,label:z5.string().min(1),contentDigest:sha256Hex,blobDigest:sha256Hex,blobKey:z5.string().min(1),rawBytes:z5.number().int().positive().optional(),storedBytes:z5.number().int().positive().optional(),exposure:z5.record(z5.string(),z5.unknown()).optional()});var skillSetSchema=z5.array(skillSetEntrySchema);var skillSourceWireSchema=z5.object({repoFullName:z5.string(),ref:z5.string(),subpath:z5.string(),autoIngest:z5.boolean(),lastSeenCommit:z5.string().nullable(),lastCheckedAt:z5.string().nullable(),lastError:z5.string().nullable()});var skillVersionWireSchema=z5.object({id:z5.string(),label:z5.string(),contentDigest:z5.string(),ingestTrigger:z5.string().nullable(),createdAt:z5.string()});var skillListItemSchema=z5.object({id:z5.string(),name:skillHandleSchema,note:z5.string().nullable(),source:skillSourceWireSchema.nullable(),latestVersion:skillVersionWireSchema.nullable(),versionCount:z5.number().int(),createdAt:z5.string(),updatedAt:z5.string()});var listSkillsResponseSchema=z5.object({items:z5.array(skillListItemSchema)});var skillDetailSchema=skillListItemSchema.omit({latestVersion:true}).extend({versions:z5.array(skillVersionWireSchema)});var addSkillBodySchema=z5.object({reference:z5.string().min(1),handle:skillHandleSchema.optional(),label:z5.string().min(1).max(120).optional(),skill:z5.string().min(1).optional(),note:z5.string().max(2e3).optional(),trackSource:z5.boolean().optional()});var addSkillResponseSchema=z5.object({skillId:z5.string().nullable(),handle:z5.string(),label:z5.string(),unchanged:z5.boolean()});var updateSkillBodySchema=z5.object({name:skillHandleSchema.optional(),note:z5.string().max(2e3).nullish(),source:z5.object({repoFullName:z5.string().min(1),ref:z5.string().min(1),subpath:z5.string().optional(),autoIngest:z5.boolean().optional()}).nullish()});var reingestSkillResponseSchema=z5.object({label:z5.string().nullable(),unchanged:z5.boolean()});var addSkillVersionBodySchema=z5.object({reference:z5.string().min(1),label:z5.string().min(1).max(120).optional()});var addSkillVersionResponseSchema=z5.object({label:z5.string(),unchanged:z5.boolean()});var skillRunWireSchema=z5.object({runId:z5.string(),agent:z5.string(),turns:z5.number().int().nullable(),durationMs:z5.number().int().nullable(),costUsd:z5.number().nullable(),finishedAt:z5.string().nullable(),versionLabel:z5.string().nullable()});var listSkillRunsResponseSchema=z5.object({items:z5.array(skillRunWireSchema)});var mcpServerHandleSchema=skillHandleSchema;var MCP_SPEC_VERSION=1;var envVarNameSchema=z6.string().min(1).max(128).regex(/^[A-Za-z_][A-Za-z0-9_]*$/,"an environment variable name (letters, digits, underscore; not starting with a digit)");var headerNameSchema=z6.string().min(1).max(128).regex(/^[A-Za-z0-9!#$%&'*+\-.^_`|~]+$/,"an HTTP header name (a token \u2014 no colon, no spaces)");var mcpToolSchema=z6.object({name:z6.string().min(1).max(128),description:z6.string().max(4096).optional()});var REFUSED_VALUE_KEYS={env:"envNames",headers:"headerNames",http_headers:"headerNames",env_http_headers:"headerNames",bearer_token_env_var:"headerNames"};var mcpSpecSchema=z6.object({specVersion:z6.literal(MCP_SPEC_VERSION),transport:z6.enum(["stdio","http"]),command:z6.string().min(1).max(512).optional(),args:z6.array(z6.string().max(4096)).max(64).optional(),envNames:z6.array(envVarNameSchema).max(32).optional(),url:z6.string().url().max(2048).optional(),headerNames:z6.array(headerNameSchema).max(32).optional(),serverName:mcpServerHandleSchema,description:z6.string().max(4096).optional(),tools:z6.array(mcpToolSchema).max(256).optional()}).strict().superRefine((spec,ctx)=>{if(spec.transport==="stdio"){if(!spec.command){ctx.addIssue({code:"custom",path:["command"],message:"a stdio server needs a command"})}for(const field of["url","headerNames"]){if(spec[field]!==void 0){ctx.addIssue({code:"custom",path:[field],message:`${field} belongs to an http server \u2014 a stdio server is a command`})}}return}if(!spec.url){ctx.addIssue({code:"custom",path:["url"],message:"an http server needs a url"})}else if(!/^https?:\/\//i.test(spec.url)){ctx.addIssue({code:"custom",path:["url"],message:"an http server's url must be http(s)"})}for(const field of["command","args","envNames"]){if(spec[field]!==void 0){ctx.addIssue({code:"custom",path:[field],message:`${field} belongs to a stdio server \u2014 an http server is a url`})}}});function parseMcpSpec(input){if(input&&typeof input==="object"&&!Array.isArray(input)){for(const[key,replacement]of Object.entries(REFUSED_VALUE_KEYS)){if(key in input){throw new Error(`"${key}" is not stored: Gauge never keeps a credential VALUE for an MCP server. Declare ${replacement} instead \u2014 the names are what the agent's config references.`)}}}return mcpSpecSchema.parse(input)}var mcpExposureSchema=z6.object({serverName:z6.string().min(1),description:z6.string().optional(),tools:z6.array(mcpToolSchema).default([]),envNames:z6.array(z6.string()).default([]),headerNames:z6.array(z6.string()).default([])});var mcpWarningSchema=z6.object({code:z6.enum(["unpinned_package","credential_required","value_discarded","flag_ignored"]),message:z6.string()});var mcpSetEntrySchema=z6.object({serverId:z6.string().min(1),name:mcpServerHandleSchema,label:z6.string().min(1),specDigest:z6.string().min(1),spec:mcpSpecSchema,exposure:z6.record(z6.string(),z6.unknown()).optional(),connectionId:z6.string().min(1).optional()});var MAX_MCP_SET_ENTRIES=8;var mcpSetSchema=z6.array(mcpSetEntrySchema).max(MAX_MCP_SET_ENTRIES,`a run may prepare at most ${MAX_MCP_SET_ENTRIES} MCP servers`);var mcpRefSchema=z6.string().min(3).max(320).regex(/^[A-Za-z0-9][A-Za-z0-9._-]*@[^,]+$/,'an MCP server reference must be "name@label" (a bare name is not a reference)');var mcpRefsSchema=z6.array(mcpRefSchema).max(MAX_MCP_SET_ENTRIES);var mcpVersionWireSchema=z6.object({connectionId:z6.string().nullable().optional(),id:z6.string(),label:z6.string(),specDigest:z6.string(),spec:mcpSpecSchema,exposure:mcpExposureSchema,warnings:z6.array(mcpWarningSchema),createdAt:z6.string()});var mcpServerListItemSchema=z6.object({id:z6.string(),name:mcpServerHandleSchema,note:z6.string().nullable(),latestVersion:mcpVersionWireSchema.nullable(),versionCount:z6.number().int(),createdAt:z6.string(),updatedAt:z6.string()});var listMcpServersResponseSchema=z6.object({items:z6.array(mcpServerListItemSchema)});var mcpServerDetailSchema=mcpServerListItemSchema.omit({latestVersion:true}).extend({versions:z6.array(mcpVersionWireSchema)});var saveMcpServerBodySchema=z6.object({connectionId:z6.string().min(1).max(128).optional(),name:mcpServerHandleSchema,spec:z6.unknown(),label:z6.string().min(1).max(120).optional(),note:z6.string().max(2e3).optional()});var beginMcpOauthBodySchema=z6.object({name:mcpServerHandleSchema,url:z6.url().max(2048).refine(url=>url.startsWith("https://"),"OAuth requires an HTTPS MCP URL"),clientId:z6.string().min(1).max(2048).optional(),clientSecret:z6.string().min(1).max(8192).optional()}).strict();var saveMcpServerResponseSchema=z6.object({connectionId:z6.string().nullable().optional(),serverId:z6.string().nullable(),handle:z6.string(),label:z6.string(),specDigest:z6.string(),warnings:z6.array(mcpWarningSchema),unchanged:z6.boolean(),invocation:z6.string()});var updateMcpServerBodySchema=z6.object({name:mcpServerHandleSchema.optional(),note:z6.string().max(2e3).nullish()});import{spawn as spawn2}from"child_process";import{setTimeout as sleep2}from"timers/promises";function openBrowser2(url){const[command,args]=process.platform==="darwin"?["open",[url]]:process.platform==="win32"?["rundll32",["url.dll,FileProtocolHandler",url]]:["xdg-open",[url]];try{const child=spawn2(command,args,{stdio:"ignore",detached:true});child.on("error",()=>{});child.unref()}catch{}}async function connectMcpOauth(client,org,input,options,overrides={}){const deps={openBrowser:openBrowser2,sleep:sleep2,now:Date.now,report:console.error,...overrides};const path=`/api/v1/orgs/${encodeURIComponent(org)}/mcp-authorizations`;const result=await client.post(path,input);if(result.kind==="connected")return result.connectionId;if(result.kind==="problem")throw new CliError("invalid_request",result.problem);if(result.kind==="needs-client-id"){throw new CliError("invalid_request",result.reason,{hint:`Register an OAuth client with redirect URI ${result.redirectUri}, then retry with --oauth-client-id. If required, supply its secret using --oauth-client-secret-env.`})}let url;try{url=new URL(result.authorizationUrl)}catch{throw new CliError("internal","Gauge returned an invalid authorization URL.")}if(url.protocol!=="https:"||url.username||url.password)throw new CliError("internal","Gauge returned an unsafe authorization URL.");const expiresAt=Date.parse(result.expiresAt);if(!Number.isFinite(expiresAt))throw new CliError("internal","Gauge returned an invalid authorization expiry.");deps.report(`Authorize this MCP connection in your browser:
|
|
948
|
+
${result.gateway.problem.hint}`)}if(!result.ready||!result.gateway.ready)process.exitCode=1}function register17(program2){const local=program2.commands.find(command=>command.name()==="local")??program2.command("local").description("Configure local execution (no Gauge account needed)");const runtime=local.command("runtime").description("Inspect or download the supported Docker runtime");runtime.command("inspect").description("Report local runtime availability and compatibility without downloading").action((_options,command)=>print({...inspectLocalRuntime(),gateway:inspectLocalGateway()},command));runtime.command("pull").description("Download the CLI's supported public runtime image").action(async(_options,command)=>print({...await pullLocalRuntime(),gateway:await pullLocalGateway()},command))}var mcp_exports={};__export(mcp_exports,{register:()=>register18});import{readFileSync as readFileSync8}from"fs";import{z as z6}from"zod";import{z as z5}from"zod";var sha256Hex=z5.custom(value=>typeof value==="string"&&/^sha256:[0-9a-f]{64}$/.test(value),"expected a sha256:<64 lowercase hex> digest");var skillHandleSchema=z5.string().min(1).max(128).regex(/^[A-Za-z0-9][A-Za-z0-9._-]*$/,"invalid skill handle").refine(name=>name!=="."&&name!=="..","invalid skill handle");var skillSetEntrySchema=z5.object({skillId:z5.string().min(1),name:skillHandleSchema,label:z5.string().min(1),contentDigest:sha256Hex,blobDigest:sha256Hex,blobKey:z5.string().min(1),rawBytes:z5.number().int().positive().optional(),storedBytes:z5.number().int().positive().optional(),exposure:z5.record(z5.string(),z5.unknown()).optional()});var skillSetSchema=z5.array(skillSetEntrySchema);var skillSourceWireSchema=z5.object({repoFullName:z5.string(),ref:z5.string(),subpath:z5.string(),autoIngest:z5.boolean(),lastSeenCommit:z5.string().nullable(),lastCheckedAt:z5.string().nullable(),lastError:z5.string().nullable()});var skillVersionWireSchema=z5.object({id:z5.string(),label:z5.string(),contentDigest:z5.string(),ingestTrigger:z5.string().nullable(),createdAt:z5.string()});var skillListItemSchema=z5.object({id:z5.string(),name:skillHandleSchema,note:z5.string().nullable(),source:skillSourceWireSchema.nullable(),latestVersion:skillVersionWireSchema.nullable(),versionCount:z5.number().int(),createdAt:z5.string(),updatedAt:z5.string()});var listSkillsResponseSchema=z5.object({items:z5.array(skillListItemSchema)});var skillDetailSchema=skillListItemSchema.omit({latestVersion:true}).extend({versions:z5.array(skillVersionWireSchema)});var addSkillBodySchema=z5.object({reference:z5.string().min(1),handle:skillHandleSchema.optional(),label:z5.string().min(1).max(120).optional(),skill:z5.string().min(1).optional(),note:z5.string().max(2e3).optional(),trackSource:z5.boolean().optional()});var addSkillResponseSchema=z5.object({skillId:z5.string().nullable(),handle:z5.string(),label:z5.string(),unchanged:z5.boolean()});var updateSkillBodySchema=z5.object({name:skillHandleSchema.optional(),note:z5.string().max(2e3).nullish(),source:z5.object({repoFullName:z5.string().min(1),ref:z5.string().min(1),subpath:z5.string().optional(),autoIngest:z5.boolean().optional()}).nullish()});var reingestSkillResponseSchema=z5.object({label:z5.string().nullable(),unchanged:z5.boolean()});var addSkillVersionBodySchema=z5.object({reference:z5.string().min(1),label:z5.string().min(1).max(120).optional()});var addSkillVersionResponseSchema=z5.object({label:z5.string(),unchanged:z5.boolean()});var skillRunWireSchema=z5.object({runId:z5.string(),agent:z5.string(),turns:z5.number().int().nullable(),durationMs:z5.number().int().nullable(),costUsd:z5.number().nullable(),finishedAt:z5.string().nullable(),versionLabel:z5.string().nullable()});var listSkillRunsResponseSchema=z5.object({items:z5.array(skillRunWireSchema)});var mcpServerHandleSchema=skillHandleSchema;var MCP_SPEC_VERSION=1;var envVarNameSchema=z6.string().min(1).max(128).regex(/^[A-Za-z_][A-Za-z0-9_]*$/,"an environment variable name (letters, digits, underscore; not starting with a digit)");var headerNameSchema=z6.string().min(1).max(128).regex(/^[A-Za-z0-9!#$%&'*+\-.^_`|~]+$/,"an HTTP header name (a token \u2014 no colon, no spaces)");var mcpToolSchema=z6.object({name:z6.string().min(1).max(128),description:z6.string().max(4096).optional()});var REFUSED_VALUE_KEYS={env:"envNames",headers:"headerNames",http_headers:"headerNames",env_http_headers:"headerNames",bearer_token_env_var:"headerNames"};var mcpSpecSchema=z6.object({specVersion:z6.literal(MCP_SPEC_VERSION),transport:z6.enum(["stdio","http"]),command:z6.string().min(1).max(512).optional(),args:z6.array(z6.string().max(4096)).max(64).optional(),envNames:z6.array(envVarNameSchema).max(32).optional(),url:z6.string().url().max(2048).optional(),headerNames:z6.array(headerNameSchema).max(32).optional(),serverName:mcpServerHandleSchema,description:z6.string().max(4096).optional(),tools:z6.array(mcpToolSchema).max(256).optional()}).strict().superRefine((spec,ctx)=>{if(spec.transport==="stdio"){if(!spec.command){ctx.addIssue({code:"custom",path:["command"],message:"a stdio server needs a command"})}for(const field of["url","headerNames"]){if(spec[field]!==void 0){ctx.addIssue({code:"custom",path:[field],message:`${field} belongs to an http server \u2014 a stdio server is a command`})}}return}if(!spec.url){ctx.addIssue({code:"custom",path:["url"],message:"an http server needs a url"})}else if(!/^https?:\/\//i.test(spec.url)){ctx.addIssue({code:"custom",path:["url"],message:"an http server's url must be http(s)"})}for(const field of["command","args","envNames"]){if(spec[field]!==void 0){ctx.addIssue({code:"custom",path:[field],message:`${field} belongs to a stdio server \u2014 an http server is a url`})}}});function parseMcpSpec(input){if(input&&typeof input==="object"&&!Array.isArray(input)){for(const[key,replacement]of Object.entries(REFUSED_VALUE_KEYS)){if(key in input){throw new Error(`"${key}" is not stored: Gauge never keeps a credential VALUE for an MCP server. Declare ${replacement} instead \u2014 the names are what the agent's config references.`)}}}return mcpSpecSchema.parse(input)}var mcpExposureSchema=z6.object({serverName:z6.string().min(1),description:z6.string().optional(),tools:z6.array(mcpToolSchema).default([]),envNames:z6.array(z6.string()).default([]),headerNames:z6.array(z6.string()).default([])});var mcpWarningSchema=z6.object({code:z6.enum(["unpinned_package","credential_required","value_discarded","flag_ignored"]),message:z6.string()});var mcpSetEntrySchema=z6.object({serverId:z6.string().min(1),name:mcpServerHandleSchema,label:z6.string().min(1),specDigest:z6.string().min(1),spec:mcpSpecSchema,exposure:z6.record(z6.string(),z6.unknown()).optional(),connectionId:z6.string().min(1).optional()});var MAX_MCP_SET_ENTRIES=8;var mcpSetSchema=z6.array(mcpSetEntrySchema).max(MAX_MCP_SET_ENTRIES,`a run may prepare at most ${MAX_MCP_SET_ENTRIES} MCP servers`);var mcpRefSchema=z6.string().min(3).max(320).regex(/^[A-Za-z0-9][A-Za-z0-9._-]*@[^,]+$/,'an MCP server reference must be "name@label" (a bare name is not a reference)');var mcpRefsSchema=z6.array(mcpRefSchema).max(MAX_MCP_SET_ENTRIES);var mcpVersionWireSchema=z6.object({connectionId:z6.string().nullable().optional(),id:z6.string(),label:z6.string(),specDigest:z6.string(),spec:mcpSpecSchema,exposure:mcpExposureSchema,warnings:z6.array(mcpWarningSchema),createdAt:z6.string()});var mcpServerListItemSchema=z6.object({id:z6.string(),name:mcpServerHandleSchema,note:z6.string().nullable(),latestVersion:mcpVersionWireSchema.nullable(),versionCount:z6.number().int(),createdAt:z6.string(),updatedAt:z6.string()});var listMcpServersResponseSchema=z6.object({items:z6.array(mcpServerListItemSchema)});var mcpServerDetailSchema=mcpServerListItemSchema.omit({latestVersion:true}).extend({versions:z6.array(mcpVersionWireSchema)});var saveMcpServerBodySchema=z6.object({connectionId:z6.string().min(1).max(128).optional(),name:mcpServerHandleSchema,spec:z6.unknown(),label:z6.string().min(1).max(120).optional(),note:z6.string().max(2e3).optional()});var beginMcpOauthBodySchema=z6.object({name:mcpServerHandleSchema,url:z6.url().max(2048).refine(url=>url.startsWith("https://"),"OAuth requires an HTTPS MCP URL"),clientId:z6.string().min(1).max(2048).optional(),clientSecret:z6.string().min(1).max(8192).optional()}).strict();var saveMcpServerResponseSchema=z6.object({connectionId:z6.string().nullable().optional(),serverId:z6.string().nullable(),handle:z6.string(),label:z6.string(),specDigest:z6.string(),warnings:z6.array(mcpWarningSchema),unchanged:z6.boolean(),invocation:z6.string()});var updateMcpServerBodySchema=z6.object({name:mcpServerHandleSchema.optional(),note:z6.string().max(2e3).nullish()});import{spawn as spawn2}from"child_process";import{setTimeout as sleep2}from"timers/promises";function openBrowser2(url){const[command,args]=process.platform==="darwin"?["open",[url]]:process.platform==="win32"?["rundll32",["url.dll,FileProtocolHandler",url]]:["xdg-open",[url]];try{const child=spawn2(command,args,{stdio:"ignore",detached:true});child.on("error",()=>{});child.unref()}catch{}}async function connectMcpOauth(client,org,input,options,overrides={}){const deps={openBrowser:openBrowser2,sleep:sleep2,now:Date.now,report:console.error,...overrides};const path=`/api/v1/orgs/${encodeURIComponent(org)}/mcp-authorizations`;const result=await client.post(path,input);if(result.kind==="connected")return result.connectionId;if(result.kind==="problem")throw new CliError("invalid_request",result.problem);if(result.kind==="needs-client-id"){throw new CliError("invalid_request",result.reason,{hint:`Register an OAuth client with redirect URI ${result.redirectUri}, then retry with --oauth-client-id. If required, supply its secret using --oauth-client-secret-env.`})}let url;try{url=new URL(result.authorizationUrl)}catch{throw new CliError("internal","Gauge returned an invalid authorization URL.")}if(url.protocol!=="https:"||url.username||url.password)throw new CliError("internal","Gauge returned an unsafe authorization URL.");const expiresAt=Date.parse(result.expiresAt);if(!Number.isFinite(expiresAt))throw new CliError("internal","Gauge returned an invalid authorization expiry.");deps.report(`Authorize this MCP connection in your browser:
|
|
777
949
|
${url}
|
|
778
|
-
Waiting for consent\u2026`);if(options.browser)deps.openBrowser(url.toString());const deadline=Math.min(expiresAt,deps.now()+options.timeoutSeconds*1e3);while(deps.now()<deadline){const signal=AbortSignal.timeout(Math.max(1,deadline-deps.now()));let status;try{status=await client.get(`${path}/${encodeURIComponent(result.authorizationId)}`,void 0,signal)}catch(error){if(signal.aborted)break;throw error}if(status.status==="connected")return status.connectionId;if(status.status==="failed"||status.status==="expired")throw new CliError("invalid_request",status.message);await deps.sleep(Math.min(2e3,Math.max(0,deadline-deps.now())))}throw new CliError("invalid_request","Timed out waiting for MCP authorization.",{hint:"If consent succeeded, use `gauge connections list --hosted` and retry with --connection <id>. Otherwise retry with --oauth; incomplete saved grants can resume without new consent."})}function mcpPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/mcp-servers`}function serverPath(org,ref){return`${mcpPath(org)}/${encodeURIComponent(ref)}`}function specLabel(spec){return spec.transport==="stdio"?[spec.command,...spec.args??[]].join(" "):spec.url??""}function specFromOpts(opts){if(opts.specFile!==void 0){if(opts.command!==void 0||opts.url!==void 0)throw new CliError("usage","--spec-file conflicts with --command/--url");try{return JSON.parse(readFileSync7(opts.specFile==="-"?0:opts.specFile,"utf8"))}catch(e){throw new CliError("usage",`could not read --spec-file: ${e instanceof Error?e.message:e}`)}}if(opts.command!==void 0===(opts.url!==void 0))throw new CliError("usage","pass exactly one of --command (stdio) or --url (http), or --spec-file");const tools=opts.tool.map(line=>{const colon=line.indexOf(":");return colon>0?{name:line.slice(0,colon).trim(),description:line.slice(colon+1).trim()}:{name:line.trim()}});return{transport:opts.command!==void 0?"stdio":"http",...opts.command!==void 0?{command:opts.command,args:opts.arg,envNames:opts.env}:{url:opts.url,headerNames:opts.header},...opts.serverName?{serverName:opts.serverName}:{},...opts.description?{description:opts.description}:{},...tools.length?{tools}:{}}}function register18(program2){const group=hostedOnly(program2.command("mcp"),{hint:LOCAL_MCP_HINT}).description("Manage MCP servers used by hosted eval and preference runs");group.command("list").description("List the org's MCP servers with their newest version").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(mcpPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(s=>({name:s.name,latest:s.latestVersion?.label??"",versions:s.versionCount,transport:s.latestVersion?.spec.transport??"",invocation:s.latestVersion?specLabel(s.latestVersion.spec):"",note:s.note??""})),"table")});group.command("get <ref>").description("Show one MCP server: note and every version").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const server=await createClient().get(serverPath(org,ref));if(resolveOutput(cmd)==="json"){printJson(server);return}console.log(`name: ${server.name}`);console.log(`id: ${server.id}`);console.log(`note: ${server.note??"(none)"}`);console.log("versions:");for(const v of server.versions){console.log(` ${server.name}@${v.label} [${v.spec.transport}] ${specLabel(v.spec)} ${v.createdAt}`);if(v.connectionId)console.log(` connection: ${v.connectionId}`);for(const w of v.warnings)console.log(` warning: ${w.message}`)}});group.command("add").description("Add a server, or append a version to an existing --name (versions are immutable)").requiredOption("--name <handle>","the handle (the agent config key)").option("--command <cmd>","stdio launch command, e.g. npx").option("--arg <arg>","stdio argument (repeatable)",collect5,[]).option("--env <NAME>","required env-var NAME \u2014 never a value (repeatable)",collect5,[]).option("--url <url>","http endpoint").option("--header <NAME>","required header NAME \u2014 never a value (repeatable)",collect5,[]).option("--server-name <name>","agent-visible server name (default: handle)").option("--description <text>","agent-visible description").option("--tool <name[: description]>","declared tool (repeatable)",collect5,[]).option("--spec-file <path>",'whole spec as JSON ("-" for stdin)').option("--label <label>","version label (a typed label always mints; omitted dedups by spec digest)").option("--note <text>","note shown on the index").option("--connection <id>","bind an existing connection to this HTTP MCP server").option("--oauth","authorize this HTTP MCP in a browser, then save its connection binding").option("--oauth-client-id <id>","pre-registered OAuth client id, when the server requires one").option("--oauth-client-secret-env <NAME>","environment variable holding an optional OAuth client secret").option("--no-browser","print the OAuth URL without opening a browser").option("--oauth-timeout <seconds>","maximum time to wait for OAuth consent (1-600)","600").action(async(opts,cmd)=>{const org=await resolveOrg(cmd);if(opts.oauth&&opts.connection)throw new CliError("usage","--oauth conflicts with --connection; choose new consent or an existing connection.");if(!opts.oauth&&(opts.oauthClientId||opts.oauthClientSecretEnv||opts.browser===false||opts.oauthTimeout!=="600"))throw new CliError("usage","OAuth options require --oauth.");const timeoutSeconds=Number(opts.oauthTimeout);if(!Number.isInteger(timeoutSeconds)||timeoutSeconds<1||timeoutSeconds>600)throw new CliError("usage","--oauth-timeout must be an integer from 1 to 600 seconds.");let spec;let body;try{const raw=specFromOpts(opts);spec=parseMcpSpec(raw&&typeof raw==="object"&&!Array.isArray(raw)?{specVersion:1,serverName:opts.name,...raw}:raw);body=saveMcpServerBodySchema.parse({name:opts.name,spec,label:opts.label,note:opts.note,connectionId:opts.connection})}catch(error){if(error instanceof CliError)throw error;throw new CliError("usage",error instanceof Error?error.message:"Invalid MCP configuration.")}if((opts.oauth||opts.connection)&&spec.transport!=="http")throw new CliError("usage","--oauth and --connection require an HTTP MCP server.");const client=createClient();if(opts.oauth){let clientSecret;if(opts.oauthClientSecretEnv){if(!/^[A-Za-z_][A-Za-z0-9_]*$/.test(opts.oauthClientSecretEnv))throw new CliError("usage","--oauth-client-secret-env must name an environment variable.");clientSecret=process.env[opts.oauthClientSecretEnv];if(!clientSecret)throw new CliError("usage","The OAuth client secret environment variable is empty or unset.");if(!opts.oauthClientId)throw new CliError("usage","--oauth-client-secret-env requires --oauth-client-id.")}body.connectionId=await connectMcpOauth(client,org,{name:opts.name,url:spec.url,clientId:opts.oauthClientId,clientSecret},{browser:opts.browser,timeoutSeconds})}let result;try{result=await client.post(mcpPath(org),body)}catch(error){if(opts.oauth&&body.connectionId)console.error(`OAuth connected, but saving the MCP version failed. Retry with --connection ${body.connectionId} instead of --oauth.`);throw error}if(body.connectionId&&result.connectionId!==body.connectionId)throw new CliError("internal","The server did not confirm the MCP connection binding.",{hint:`Upgrade the Gauge API before retrying with --connection ${body.connectionId}.`});if(resolveOutput(cmd)==="json"){printJson(result);return}console.log(result.unchanged?`Already current: ${result.handle}@${result.label} carries this spec`:`Saved ${result.handle}@${result.label} \u2014 ${result.invocation}`);for(const w of result.warnings)console.error(`warning: ${w.message}`);console.error(`Attach it to an eval: gauge eval edit <id> --mcp ${result.handle}@${result.label}`)});group.command("edit <ref>").description("Rename or note a server (append versions with `gauge mcp add --name <handle>`)").option("--rename <handle>","new handle (name@label refs must be updated)").option("--note <text>","replace the note").option("--clear-note","remove the note").action(async(ref,opts,cmd)=>{const org=await resolveOrg(cmd);if(opts.note!==void 0&&opts.clearNote)throw new CliError("usage","--note and --clear-note conflict");const body={};if(opts.rename!==void 0)body.name=opts.rename;if(opts.clearNote)body.note=null;else if(opts.note!==void 0)body.note=opts.note;if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");const updated=await createClient().patch(serverPath(org,ref),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated MCP server ${updated.name}`)});group.command("rm <ref>").description("Remove a server; the handle stays reserved and finished runs keep their snapshot").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().delete(serverPath(org,ref));if(resolveOutput(cmd)==="json")printJson(result);else console.log(`Deleted MCP server ${ref} (${result.referencedRuns} finished run(s) referenced it)`)})}var members_exports={};__export(members_exports,{register:()=>register19});function register19(program2){const group=hostedOnly(program2.command("members")).description("Team members and pending invites (read-only)");group.command("list").description("List members and pending invites").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const base=`/api/v1/orgs/${encodeURIComponent(org)}`;const[members,pending]=await Promise.all([client.get(`${base}/members`),client.get(`${base}/invites`)]);if(resolveOutput(cmd)==="json"){printJson({members:members.items,invites:pending.items});return}printItems(members.items.map(m=>({email:m.email??"",name:m.name??"",joined:m.joinedAt})),"table");if(pending.items.length>0){console.log("\npending invites:");printItems(pending.items.map(i=>({id:i.id,email:i.email,"invited by":i.invitedBy.email??i.invitedBy.name??"",expires:i.expiresAt})),"table")}})}var models_exports={};__export(models_exports,{capabilityLabel:()=>capabilityLabel,executionTargetsPath:()=>executionTargetsPath,register:()=>register20});function executionTargetsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/execution-targets`}function capabilityLabel(capabilities){return Object.entries(capabilities).filter(([,supported])=>supported).map(([name])=>name==="agentContext"?"context":name).join(", ")}function register20(program2){const group=program2.command("models").description("Discover supported models, harnesses, and providers");group.command("list",{isDefault:true}).description("List execution targets available to the organization. Local: the agents local runs support, with default models").action(async(_opts,cmd)=>{if(gaugeState(cmd).mode==="local"){const agents=Object.entries(LOCAL_AGENTS).map(([agent,target])=>({agent,harness:target.kind,defaultModel:target.model}));if(resolveOutput(cmd)==="json")printJson(agents);else printItems(agents,"table");return}const org=await resolveOrg(cmd);const response=await createClient().get(executionTargetsPath(org));if(resolveOutput(cmd)==="json"){printJson(response.items);return}printItems(response.items.map(target=>({model:target.model,name:target.modelLabel,harness:target.harnessSlug,provider:target.providerSlug,capabilities:capabilityLabel(target.capabilities)})),"table")})}var localProfiles_exports={};__export(localProfiles_exports,{register:()=>register21});function register21(program2){const local=program2.commands.find(command=>command.name()==="local")??program2.command("local").description("Configure local execution");const profiles=local.command("profiles").description("Manage local agent credentials (references only; separate from Gauge account login)");profiles.command("list").description("List profiles and per-harness defaults; never print secrets").action((_opts,cmd)=>{const store=readProfiles();printItems(Object.entries(store.profiles).map(([name,p])=>({name,...p,active:store.defaults[p.agent]===name})),resolveOutput(cmd))});profiles.command("set <profile>").description("Create a profile or update it with --replace; --env takes a variable name, never a secret").requiredOption("--agent <agent>","CLAUDE_CODE, CODEX_CLI, PI, or OPENCODE").requiredOption("--provider <provider>","anthropic, openai, openai-codex, fireworks, or openrouter").requiredOption("--auth <mode>","access_token (native login), api_key, or setup_token (Claude only)").option("--env <variable>","environment variable holding the key/token (defaults to the provider's standard variable)").option("--replace","replace an existing profile").action((name,opts,cmd)=>{const{replace,...profile}=opts;const store=setProfile(name,profile,replace);const result={name,...store.profiles[name],active:store.defaults[profile.agent]===name};if(resolveOutput(cmd)==="json")printJson(result);else console.log(`Saved profile ${name}. Select with gauge local profiles use ${name}, or gauge run --local --profile ${name}.`)});profiles.command("use <profile>").description("Set the default for this profile's harness").action((name,_opts,cmd)=>{const store=readProfiles();const profile=requireProfile(store,name);store.defaults[profile.agent]=name;writeProfiles(store);if(resolveOutput(cmd)==="json")printJson({agent:profile.agent,profile:name});else console.log(`${profile.agent} now uses ${name}.`)});profiles.command("remove <profile>").description("Remove a reference and its default selection; keep native logins and environment variables").action((name,_opts,cmd)=>{const store=readProfiles();requireProfile(store,name);delete store.profiles[name];const cleared=Object.keys(store.defaults).filter(agent=>store.defaults[agent]===name);for(const agent of cleared)delete store.defaults[agent];writeProfiles(store);if(resolveOutput(cmd)==="json")printJson({removed:name,clearedDefaults:cleared});else console.log(`Removed ${name}.${cleared.length?` ${cleared.join(", ")} now uses automatic credential discovery.`:""}`)});profiles.command("inspect <profile>").description("Check local compatibility, credential availability and expiry; does not call the provider").option("--model <model>","check this model (default: the harness's default model)").action(async(name,opts,cmd)=>{const store=readProfiles();const p=requireProfile(store,name);if(!opts.model&&(p.agent==="PI"||p.agent==="OPENCODE")&&!LOCAL_TARGETS[p.agent].model.startsWith(`${p.provider}/`)){throw new CliError("usage",`Specify --model ${p.provider}/<model> to inspect this profile.`,{hint:`The ${p.agent} default model uses a different provider. Choose a model available through ${p.provider}.`})}const result=await inspectLocalCredential(p.agent,opts.model??LOCAL_TARGETS[p.agent].model,{profile:name},store);if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`${name}: ${result.ready?"ready locally":"not ready"} for ${result.agent} ${result.model}`);if(result.profile)console.log(`Source: ${result.profile.source}; auth: ${result.profile.auth}`);if(result.expiresAt)console.log(`Expires: ${result.expiresAt}`);if(result.problem)console.log(result.problem);console.log("Local checks do not verify provider acceptance or model access.")}if(!result.ready)process.exitCode=1})}var onboard_exports={};__export(onboard_exports,{register:()=>register22});var PERSONAL_DOMAINS=new Set(["gmail.com","googlemail.com","yahoo.com","hotmail.com","outlook.com","live.com","icloud.com","me.com","aol.com","proton.me","protonmail.com"]);function emailDomain(email){const domain=email?.split("@")[1]?.toLowerCase();return domain&&!PERSONAL_DOMAINS.has(domain)?domain:void 0}function orgNameFromEmail(email){const base=emailDomain(email)?.split(".")[0]??"";if(!base)return"My Workspace";return base.charAt(0).toUpperCase()+base.slice(1)}function websiteFromEmail(email){const domain=emailDomain(email);return domain?`https://${domain}`:void 0}async function ensureAccount(opts={}){const log=opts.log??console.log;let client;let me;if(resolveToken()){client=new ApiClient;me=await client.get("/api/v1/me");log(`Already signed in as ${me.user.email??me.user.id}.`)}else{const login=await deviceLogin({email:opts.email});await withCredentialLock(async()=>{if(login.credentials)saveCredentials(login.credentials);else saveToken(login.token)});useOnlyOrg(login.me,login.token.startsWith("gauge_cli_"),log);client=new ApiClient;me=await client.get("/api/v1/me");log(`
|
|
950
|
+
Waiting for consent\u2026`);if(options.browser)deps.openBrowser(url.toString());const deadline=Math.min(expiresAt,deps.now()+options.timeoutSeconds*1e3);while(deps.now()<deadline){const signal=AbortSignal.timeout(Math.max(1,deadline-deps.now()));let status;try{status=await client.get(`${path}/${encodeURIComponent(result.authorizationId)}`,void 0,signal)}catch(error){if(signal.aborted)break;throw error}if(status.status==="connected")return status.connectionId;if(status.status==="failed"||status.status==="expired")throw new CliError("invalid_request",status.message);await deps.sleep(Math.min(2e3,Math.max(0,deadline-deps.now())))}throw new CliError("invalid_request","Timed out waiting for MCP authorization.",{hint:"If consent succeeded, use `gauge connections list --hosted` and retry with --connection <id>. Otherwise retry with --oauth; incomplete saved grants can resume without new consent."})}function mcpPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/mcp-servers`}function serverPath(org,ref){return`${mcpPath(org)}/${encodeURIComponent(ref)}`}function specLabel(spec){return spec.transport==="stdio"?[spec.command,...spec.args??[]].join(" "):spec.url??""}function specFromOpts(opts){if(opts.specFile!==void 0){if(opts.command!==void 0||opts.url!==void 0)throw new CliError("usage","--spec-file conflicts with --command/--url");try{return JSON.parse(readFileSync8(opts.specFile==="-"?0:opts.specFile,"utf8"))}catch(e){throw new CliError("usage",`could not read --spec-file: ${e instanceof Error?e.message:e}`)}}if(opts.command!==void 0===(opts.url!==void 0))throw new CliError("usage","pass exactly one of --command (stdio) or --url (http), or --spec-file");const tools=opts.tool.map(line=>{const colon=line.indexOf(":");return colon>0?{name:line.slice(0,colon).trim(),description:line.slice(colon+1).trim()}:{name:line.trim()}});return{transport:opts.command!==void 0?"stdio":"http",...opts.command!==void 0?{command:opts.command,args:opts.arg,envNames:opts.env}:{url:opts.url,headerNames:opts.header},...opts.serverName?{serverName:opts.serverName}:{},...opts.description?{description:opts.description}:{},...tools.length?{tools}:{}}}function register18(program2){const group=hostedOnly(program2.command("mcp"),{hint:LOCAL_MCP_HINT}).description("Manage MCP servers used by hosted eval and preference runs");group.command("list").description("List the org's MCP servers with their newest version").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(mcpPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(s=>({name:s.name,latest:s.latestVersion?.label??"",versions:s.versionCount,transport:s.latestVersion?.spec.transport??"",invocation:s.latestVersion?specLabel(s.latestVersion.spec):"",note:s.note??""})),"table")});group.command("get <ref>").description("Show one MCP server: note and every version").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const server=await createClient().get(serverPath(org,ref));if(resolveOutput(cmd)==="json"){printJson(server);return}console.log(`name: ${server.name}`);console.log(`id: ${server.id}`);console.log(`note: ${server.note??"(none)"}`);console.log("versions:");for(const v of server.versions){console.log(` ${server.name}@${v.label} [${v.spec.transport}] ${specLabel(v.spec)} ${v.createdAt}`);if(v.connectionId)console.log(` connection: ${v.connectionId}`);for(const w of v.warnings)console.log(` warning: ${w.message}`)}});group.command("add").description("Add a server, or append a version to an existing --name (versions are immutable)").requiredOption("--name <handle>","the handle (the agent config key)").option("--command <cmd>","stdio launch command, e.g. npx").option("--arg <arg>","stdio argument (repeatable)",collect5,[]).option("--env <NAME>","required env-var NAME \u2014 never a value (repeatable)",collect5,[]).option("--url <url>","http endpoint").option("--header <NAME>","required header NAME \u2014 never a value (repeatable)",collect5,[]).option("--server-name <name>","agent-visible server name (default: handle)").option("--description <text>","agent-visible description").option("--tool <name[: description]>","declared tool (repeatable)",collect5,[]).option("--spec-file <path>",'whole spec as JSON ("-" for stdin)').option("--label <label>","version label (a typed label always mints; omitted dedups by spec digest)").option("--note <text>","note shown on the index").option("--connection <id>","bind an existing connection to this HTTP MCP server").option("--oauth","authorize this HTTP MCP in a browser, then save its connection binding").option("--oauth-client-id <id>","pre-registered OAuth client id, when the server requires one").option("--oauth-client-secret-env <NAME>","environment variable holding an optional OAuth client secret").option("--no-browser","print the OAuth URL without opening a browser").option("--oauth-timeout <seconds>","maximum time to wait for OAuth consent (1-600)","600").action(async(opts,cmd)=>{const org=await resolveOrg(cmd);if(opts.oauth&&opts.connection)throw new CliError("usage","--oauth conflicts with --connection; choose new consent or an existing connection.");if(!opts.oauth&&(opts.oauthClientId||opts.oauthClientSecretEnv||opts.browser===false||opts.oauthTimeout!=="600"))throw new CliError("usage","OAuth options require --oauth.");const timeoutSeconds=Number(opts.oauthTimeout);if(!Number.isInteger(timeoutSeconds)||timeoutSeconds<1||timeoutSeconds>600)throw new CliError("usage","--oauth-timeout must be an integer from 1 to 600 seconds.");let spec;let body;try{const raw=specFromOpts(opts);spec=parseMcpSpec(raw&&typeof raw==="object"&&!Array.isArray(raw)?{specVersion:1,serverName:opts.name,...raw}:raw);body=saveMcpServerBodySchema.parse({name:opts.name,spec,label:opts.label,note:opts.note,connectionId:opts.connection})}catch(error){if(error instanceof CliError)throw error;throw new CliError("usage",error instanceof Error?error.message:"Invalid MCP configuration.")}if((opts.oauth||opts.connection)&&spec.transport!=="http")throw new CliError("usage","--oauth and --connection require an HTTP MCP server.");const client=createClient();if(opts.oauth){let clientSecret;if(opts.oauthClientSecretEnv){if(!/^[A-Za-z_][A-Za-z0-9_]*$/.test(opts.oauthClientSecretEnv))throw new CliError("usage","--oauth-client-secret-env must name an environment variable.");clientSecret=process.env[opts.oauthClientSecretEnv];if(!clientSecret)throw new CliError("usage","The OAuth client secret environment variable is empty or unset.");if(!opts.oauthClientId)throw new CliError("usage","--oauth-client-secret-env requires --oauth-client-id.")}body.connectionId=await connectMcpOauth(client,org,{name:opts.name,url:spec.url,clientId:opts.oauthClientId,clientSecret},{browser:opts.browser,timeoutSeconds})}let result;try{result=await client.post(mcpPath(org),body)}catch(error){if(opts.oauth&&body.connectionId)console.error(`OAuth connected, but saving the MCP version failed. Retry with --connection ${body.connectionId} instead of --oauth.`);throw error}if(body.connectionId&&result.connectionId!==body.connectionId)throw new CliError("internal","The server did not confirm the MCP connection binding.",{hint:`Upgrade the Gauge API before retrying with --connection ${body.connectionId}.`});if(resolveOutput(cmd)==="json"){printJson(result);return}console.log(result.unchanged?`Already current: ${result.handle}@${result.label} carries this spec`:`Saved ${result.handle}@${result.label} \u2014 ${result.invocation}`);for(const w of result.warnings)console.error(`warning: ${w.message}`);console.error(`Attach it to an eval: gauge eval edit <id> --mcp ${result.handle}@${result.label}`)});group.command("edit <ref>").description("Rename or note a server (append versions with `gauge mcp add --name <handle>`)").option("--rename <handle>","new handle (name@label refs must be updated)").option("--note <text>","replace the note").option("--clear-note","remove the note").action(async(ref,opts,cmd)=>{const org=await resolveOrg(cmd);if(opts.note!==void 0&&opts.clearNote)throw new CliError("usage","--note and --clear-note conflict");const body={};if(opts.rename!==void 0)body.name=opts.rename;if(opts.clearNote)body.note=null;else if(opts.note!==void 0)body.note=opts.note;if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");const updated=await createClient().patch(serverPath(org,ref),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated MCP server ${updated.name}`)});group.command("rm <ref>").description("Remove a server; the handle stays reserved and finished runs keep their snapshot").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().delete(serverPath(org,ref));if(resolveOutput(cmd)==="json")printJson(result);else console.log(`Deleted MCP server ${ref} (${result.referencedRuns} finished run(s) referenced it)`)})}var members_exports={};__export(members_exports,{register:()=>register19});function register19(program2){const group=hostedOnly(program2.command("members")).description("Team members and pending invites (read-only)");group.command("list").description("List members and pending invites").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const base=`/api/v1/orgs/${encodeURIComponent(org)}`;const[members,pending]=await Promise.all([client.get(`${base}/members`),client.get(`${base}/invites`)]);if(resolveOutput(cmd)==="json"){printJson({members:members.items,invites:pending.items});return}printItems(members.items.map(m=>({email:m.email??"",name:m.name??"",joined:m.joinedAt})),"table");if(pending.items.length>0){console.log("\npending invites:");printItems(pending.items.map(i=>({id:i.id,email:i.email,"invited by":i.invitedBy.email??i.invitedBy.name??"",expires:i.expiresAt})),"table")}})}var models_exports={};__export(models_exports,{capabilityLabel:()=>capabilityLabel,executionTargetsPath:()=>executionTargetsPath,register:()=>register20});function executionTargetsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/execution-targets`}function capabilityLabel(capabilities){return Object.entries(capabilities).filter(([,supported])=>supported).map(([name])=>name==="agentContext"?"context":name).join(", ")}function register20(program2){const group=program2.command("models").description("Discover supported models, harnesses, and providers");group.command("list",{isDefault:true}).description("List execution targets available to the organization. Local: the agents local runs support, with default models").action(async(_opts,cmd)=>{if(gaugeState(cmd).mode==="local"){const agents=Object.entries(LOCAL_AGENTS).map(([agent,target])=>({agent,harness:target.kind,defaultModel:target.model}));if(resolveOutput(cmd)==="json")printJson(agents);else printItems(agents,"table");return}const org=await resolveOrg(cmd);const response=await createClient().get(executionTargetsPath(org));if(resolveOutput(cmd)==="json"){printJson(response.items);return}printItems(response.items.map(target=>({model:target.model,name:target.modelLabel,harness:target.harnessSlug,provider:target.providerSlug,capabilities:capabilityLabel(target.capabilities)})),"table")})}var localProfiles_exports={};__export(localProfiles_exports,{register:()=>register21});function register21(program2){const local=program2.commands.find(command=>command.name()==="local")??program2.command("local").description("Configure local execution");const profiles=local.command("profiles").description("Manage local agent credentials (references only; separate from Gauge account login)");profiles.command("list").description("List profiles and per-harness defaults; never print secrets").action((_opts,cmd)=>{const store=readProfiles();printItems(Object.entries(store.profiles).map(([name,p])=>({name,...p,active:store.defaults[p.agent]===name})),resolveOutput(cmd))});profiles.command("set <profile>").description("Create a profile or update it with --replace; --env takes a variable name, never a secret").requiredOption("--agent <agent>","CLAUDE_CODE, CODEX_CLI, PI, or OPENCODE").requiredOption("--provider <provider>","anthropic, openai, openai-codex, fireworks, or openrouter").requiredOption("--auth <mode>","access_token (native login), api_key, or setup_token (Claude only)").option("--env <variable>","environment variable holding the key/token (defaults to the provider's standard variable)").option("--replace","replace an existing profile").action((name,opts,cmd)=>{const{replace,...profile}=opts;const store=setProfile(name,profile,replace);const result={name,...store.profiles[name],active:store.defaults[profile.agent]===name};if(resolveOutput(cmd)==="json")printJson(result);else console.log(`Saved profile ${name}. Select with gauge local profiles use ${name}, or gauge run --local --profile ${name}.`)});profiles.command("use <profile>").description("Set the default for this profile's harness").action((name,_opts,cmd)=>{const store=readProfiles();const profile=requireProfile(store,name);store.defaults[profile.agent]=name;writeProfiles(store);if(resolveOutput(cmd)==="json")printJson({agent:profile.agent,profile:name});else console.log(`${profile.agent} now uses ${name}.`)});profiles.command("remove <profile>").description("Remove a reference and its default selection; keep native logins and environment variables").action((name,_opts,cmd)=>{const store=readProfiles();requireProfile(store,name);delete store.profiles[name];const cleared=Object.keys(store.defaults).filter(agent=>store.defaults[agent]===name);for(const agent of cleared)delete store.defaults[agent];writeProfiles(store);if(resolveOutput(cmd)==="json")printJson({removed:name,clearedDefaults:cleared});else console.log(`Removed ${name}.${cleared.length?` ${cleared.join(", ")} now uses automatic credential discovery.`:""}`)});profiles.command("inspect <profile>").description("Check local compatibility, credential availability and expiry; does not call the provider").option("--model <model>","check this model (default: the harness's default model)").action(async(name,opts,cmd)=>{const store=readProfiles();const p=requireProfile(store,name);if(!opts.model&&(p.agent==="PI"||p.agent==="OPENCODE")&&!LOCAL_TARGETS[p.agent].model.startsWith(`${p.provider}/`)){throw new CliError("usage",`Specify --model ${p.provider}/<model> to inspect this profile.`,{hint:`The ${p.agent} default model uses a different provider. Choose a model available through ${p.provider}.`})}const result=await inspectLocalCredential(p.agent,opts.model??LOCAL_TARGETS[p.agent].model,{profile:name},store);if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`${name}: ${result.ready?"ready locally":"not ready"} for ${result.agent} ${result.model}`);if(result.profile)console.log(`Source: ${result.profile.source}; auth: ${result.profile.auth}`);if(result.expiresAt)console.log(`Expires: ${result.expiresAt}`);if(result.problem)console.log(result.problem);console.log("Local checks do not verify provider acceptance or model access.")}if(!result.ready)process.exitCode=1})}var onboard_exports={};__export(onboard_exports,{register:()=>register22});var PERSONAL_DOMAINS=new Set(["gmail.com","googlemail.com","yahoo.com","hotmail.com","outlook.com","live.com","icloud.com","me.com","aol.com","proton.me","protonmail.com"]);function emailDomain(email){const domain=email?.split("@")[1]?.toLowerCase();return domain&&!PERSONAL_DOMAINS.has(domain)?domain:void 0}function orgNameFromEmail(email){const base=emailDomain(email)?.split(".")[0]??"";if(!base)return"My Workspace";return base.charAt(0).toUpperCase()+base.slice(1)}function websiteFromEmail(email){const domain=emailDomain(email);return domain?`https://${domain}`:void 0}async function ensureAccount(opts={}){const log=opts.log??console.log;let client;let me;if(resolveToken()){client=new ApiClient;me=await client.get("/api/v1/me");log(`Already signed in as ${me.user.email??me.user.id}.`)}else{const login=await deviceLogin({email:opts.email});await withCredentialLock(async()=>{if(login.credentials)saveCredentials(login.credentials);else saveToken(login.token)});useOnlyOrg(login.me,login.token.startsWith("gauge_cli_"),log);client=new ApiClient;me=await client.get("/api/v1/me");log(`
|
|
779
951
|
Signed in as ${me.user.email??me.user.id} \u2014 token saved to ${credentialsPath()}.`)}const directory=await fetchDirectory();const wantNewOrg=opts.new===true||opts.orgName!==void 0||me.orgs.length===0;let organization;let createdOrg=null;const website=opts.website??websiteFromEmail(me.user.email);if(wantNewOrg){const name=opts.orgName?.trim()||orgNameFromEmail(me.user.email);createdOrg=await client.post("/api/v1/orgs",{name,website});organization=await createdOrganization(createdOrg.slug)}else{const config=readConfig();if(!config.orgId&&config.defaultOrg)legacySelection("config");const linkedIds=new Set(me.orgs.map(org2=>org2.canonicalGaugeOrganizationId).filter(Boolean));const available=directory.filter(org2=>linkedIds.has(org2.id));const selected=config.orgId?available.find(org2=>org2.id===config.orgId):available.length===1?available[0]:void 0;if(selected)organization=selected;else throw new CliError("usage","Select a Gauge Agents organization before onboarding.",{hint:"Run `gauge orgs`, then `gauge orgs use <slug>`."})}const org=organization.id;updateConfig({orgId:org,defaultOrg:void 0});log(createdOrg?`Created workspace ${organization.name} (${organization.slug}).`:`Using workspace ${organization.name} (${organization.slug}).`);let seeded=null;if(opts.seed!==false&&createdOrg){log("Seeding starter measurements (scanning your site)\u2026");try{seeded=await client.post(`/api/v1/orgs/${encodeURIComponent(org)}/onboarding/seed`,website?{website}:{});log(`Seeded ${seeded.evalSets} eval set(s) and ${seeded.visibilityPrompts} preference prompt(s)`+(seeded.markets?` across ${seeded.markets} market(s).`:"."))}catch(e){console.error(`warning: seeding skipped \u2014 ${e instanceof Error?e.message:e}`)}}return{user:me.user,org,organization,createdOrg,seeded}}function register22(program2){hostedOnly(program2.command("onboard"),{account:true}).description("Sign in to hosted Gauge and set up your workspace. Then run `gauge init --hosted` in a repo to connect it").option("--email <email>","email a sign-in link (device-flow convenience)").option("--org-name <name>","workspace name (default: from your email)").option("--website <url>","company site, scanned to seed the workspace (default: from your email)").option("--new","create a new workspace even if you already belong to one").option("--no-seed","skip seeding starter eval sets and the Agent Preference portfolio").action(async(opts,command)=>{const output=resolveOutput(command);const account=await ensureAccount({...opts,log:output==="json"?()=>{}:console.log});if(output==="json"){printJson({user:account.user,org:account.organization,orgId:account.org,seeded:account.seeded});return}console.log(`
|
|
780
|
-
Default org set to ${account.organization.slug}. You're ready:`);console.log(" gauge status # the org at a glance");console.log(" gauge eval list # eval sets"," gauge preference list # preference prompts");console.log(" gauge run --eval <id> # launch one now");let inRepo=false;try{inRepo=gaugeState(command).inRepo}catch{}console.log(inRepo?" gauge init --hosted # connect this repo":" gauge init --hosted # connect a repo (run it inside the repo)")})}var optimizations_exports={};__export(optimizations_exports,{NEEDS_INPUT_EXIT:()=>NEEDS_INPUT_EXIT,awaitingInput:()=>awaitingInput,parseStatuses:()=>parseStatuses,parseSubject:()=>parseSubject,parseTarget:()=>parseTarget,readSkillDir:()=>readSkillDir,register:()=>register23,trialBodyFromOpts:()=>trialBodyFromOpts,trialTemplate:()=>trialTemplate,trialWarnings:()=>trialWarnings,watchStops:()=>watchStops});import{readdirSync,readFileSync as readFileSync8,statSync as statSync2}from"fs";import{join as join8,relative as relative3,sep}from"path";var TERMINAL=new Set(["DONE","ARCHIVED"]);function basePath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/optimizations`}function itemPath(org,id){return`${basePath(org)}/${encodeURIComponent(id)}`}function trialPath(org,id,trialId){return`${itemPath(org,id)}/trials/${encodeURIComponent(trialId)}`}var FAILED_RUN=new Set(["FAILED","TIMED_OUT"]);function readText(path){return readFileSync8(path==="-"?0:path,"utf8")}function sleep3(ms){return new Promise(resolve5=>setTimeout(resolve5,ms))}function parseSubject(opts){if(opts.eval&&opts.preference)throw new CliError("usage","pass --eval or --preference, not both");if(opts.eval)return{subjectKind:"EVAL_SET",subjectId:opts.eval};if(opts.preference)return{subjectKind:"VISIBILITY_PROMPT",subjectId:opts.preference};throw new CliError("usage","pass --eval <id> or --preference <id>")}function parseTarget(raw){const spec=parseAgentSpec(raw);if(spec.models.length>1)throw new CliError("usage",`--target "${raw}" names several models; an optimization pins exactly one`);return spec.models.length?{agent:spec.agent,model:spec.models[0]}:{agent:spec.agent}}function parseCount(raw,flag,max=100){if(raw===void 0)return void 0;const n=Number(raw);if(!Number.isInteger(n)||n<1||n>max)throw new CliError("usage",`${flag} must be a positive integer no greater than ${max}`);return n}var NEEDS_INPUT_EXIT=4;var NEXT_ACTION_HINT={START:"start the draft: `gauge optimizations start-draft <id>`",WAIT:"sessions are running; nothing to do",AUTHOR_TRIAL:"author the next trial: `gauge optimizations trials add <id>`",LAUNCH_ROUND:"launch the saved round: `gauge optimizations launch <id>`",ADOPT:"adopt the winning trial: `gauge optimizations adopt <id> <trial>`",RESUME:"resume it in the web app, or `gauge optimizations cancel <id>`",NONE:"nothing to do in Gauge"};function awaitingInput(detail){return detail.nextAction!=="WAIT"&&detail.nextAction!=="NONE"}function watchStops(detail,follow){return TERMINAL.has(detail.status)||detail.nextAction==="NONE"||!follow&&awaitingInput(detail)}function readObject(path){let value;try{value=JSON.parse(readText(path))}catch{throw new CliError("usage","the file must contain valid JSON")}if(!value||typeof value!=="object"||Array.isArray(value))throw new CliError("usage","the file must contain a JSON object");return value}function targetFromOpts(opts){return opts.target?{target:parseTarget(opts.target)}:{}}function parseStatuses(raw){if(!raw)return void 0;const parts=raw.split(",").map(part=>part.trim().toUpperCase()).filter(Boolean);for(const part of parts)if(!["DRAFT","ACTIVE","PAUSED","DONE","ARCHIVED"].includes(part))throw new CliError("usage",`unknown status "${part}" (expected DRAFT, ACTIVE, PAUSED, DONE, ARCHIVED)`);return parts.join(",")}function readSkillDir(dir,read2=readText,walk=listFiles){const paths=walk(dir).sort();if(!paths.length)throw new CliError("usage",`no files under ${dir} to send`);const files=paths.map(path=>({path:relative3(dir,path).split(sep).join("/"),body:read2(path)}));if(!files.some(file=>file.path==="SKILL.md"))throw new CliError("usage",`${dir} has no SKILL.md; a skill bundle must carry one`);return files}function listFiles(dir){let entries;try{entries=readdirSync(dir).map(String)}catch{throw new CliError("usage",`cannot read the directory ${dir}`)}return entries.flatMap(entry=>{const path=join8(dir,entry);return statSync2(path).isDirectory()?listFiles(path):[path]})}function trialBodyFromOpts(opts,read2=readText,walk){const launch2=opts.launch===true;const idempotencyKey=opts.idempotencyKey;if(opts.file){let parsed;try{parsed=JSON.parse(read2(opts.file))}catch{throw new CliError("usage","the trial file must contain valid JSON")}const pick=parseCount(opts.pick,"--pick")??1;let chosen;if(Array.isArray(parsed)){if(pick>parsed.length)throw new CliError("usage",`--pick ${pick} but the file holds ${parsed.length} proposal${parsed.length===1?"":"s"}`);chosen=parsed[pick-1]}else{if(opts.pick!==void 0)throw new CliError("usage","--pick applies to a proposals array");chosen=parsed}if(!chosen||typeof chosen!=="object"||Array.isArray(chosen))throw new CliError("usage","the trial file must contain an object or a non-empty array of objects");const record2=chosen;const body="trial"in record2?record2:{trial:record2};return{...body,...idempotencyKey?{idempotencyKey}:{},launch:launch2}}if(!opts.title)throw new CliError("usage","--title is required");const lever=opts.lever?.toUpperCase();let intervention;let files;if(opts.page||opts.bodyFile){if(!opts.page||!opts.bodyFile)throw new CliError("usage","--page and --body-file go together");intervention={kind:"REWRITE_DOCS",after:{pages:[{url:opts.page,body:read2(opts.bodyFile)}]}}}else if(opts.editSkill){if(!opts.skillDir)throw new CliError("usage","--edit-skill needs --skill-dir <path>: an edit sends the complete bundle");intervention={kind:"EDIT_SKILL",before:{skillVersionId:opts.editSkill},after:{skillVersionId:opts.editSkill}};files=readSkillDir(opts.skillDir,read2,walk)}else{throw new CliError("usage","describe the change: --edit-skill <id> --skill-dir <path>, --page <url> --body-file <path>, or --file <trial.json>")}const trial={title:opts.title,hypothesis:opts.hypothesis,lever:lever??(intervention.kind==="REWRITE_DOCS"?"DOCS":"SKILL"),intervention,authorMode:"MANUAL"};return{trial,...files?{files}:{},...idempotencyKey?{idempotencyKey}:{},launch:launch2}}var TRIAL_TEMPLATES={EDIT_SKILL:{trial:{title:"Name the env var in the first line",hypothesis:"The agent reads only the opening paragraph",lever:"SKILL",intervention:{kind:"EDIT_SKILL",before:{skillVersionId:"sv_REPLACE_ME"},after:{skillVersionId:"sv_REPLACE_ME"}},authorMode:"MANUAL"},files:[{path:"SKILL.md",body:"# Skill\n\nThe complete rewritten file.\n"}]},REWRITE_DOCS:{trial:{title:"Move the auth note above the example",hypothesis:"The agent copies the example before reading the note",lever:"DOCS",intervention:{kind:"REWRITE_DOCS",after:{pages:[{url:"https://example.com/docs/quickstart",body:"The complete rewritten page body."}]}},authorMode:"MANUAL"}}};function trialTemplate(kind){const template=TRIAL_TEMPLATES[kind.trim().toUpperCase()];if(!template)throw new CliError("usage",`--kind must be one of ${Object.keys(TRIAL_TEMPLATES).join(", ")} (got "${kind}")`);return template}function pct2(value){return value===null||value===void 0?"\u2014":`${Math.round(value*100)}%`}function subjectLabel(kind){return kind==="EVAL_SET"?"eval":"preference"}function toRow2(o){return{id:o.id,name:o.name.length>40?`${o.name.slice(0,37)}...`:o.name,status:o.status,subject:`${subjectLabel(o.subjectKind)} ${o.subjectName}`,round:o.round,baseline:pct2(o.before),now:pct2(o.after),delta:o.delta===null?"\u2014":`${o.delta>0?"+":""}${o.delta}pt`,credits:o.creditsSpent,updated:o.updatedAt}}function trialRow(t){return{key:t.key,round:t.roundNumber,title:t.title.length>48?`${t.title.slice(0,45)}...`:t.title,surface:t.lever??"",change:t.kind,status:t.status,score:pct2(t.score),result:t.verdict??"",sessions:t.sessions,failed:t.failedSessions||"",winner:t.adopted?"yes":"",id:t.id}}function printDetail2(o){console.log(`id: ${o.id}`);console.log(`name: ${o.name}`);console.log(`status: ${o.status}${o.stopReason?` (${o.stopReason})`:""}`);console.log(`mode: ${o.autonomous?"autonomous (legacy)":"interactive; launch each saved round explicitly"}`);console.log(`subject: ${subjectLabel(o.subjectKind)} ${o.subjectName} (${o.subjectId})`);console.log(`target: ${o.settings.pinnedAgent??"(subject roster)"}${o.settings.pinnedModel?`:${o.settings.pinnedModel}`:""}`);console.log(`sessions: ${o.settings.sampleCount} per trial`);console.log(`credits: ${o.creditsSpent} spent`);console.log(`score: ${pct2(o.before)} \u2192 ${pct2(o.after)}${o.delta===null?"":` (${o.delta>0?"+":""}${o.delta}pt)`}`);if(o.summary)console.log(`summary: ${o.summary}`);console.log(`next: ${o.nextAction} \u2014 ${NEXT_ACTION_HINT[o.nextAction]??""}`);console.log("trials:");const trials=o.rounds.flatMap(round=>round.trials);printItems(trials.map(trialRow),"table");for(const round of o.rounds)if(round.narrative)console.log(`round ${round.number}: ${round.narrative}`);for(const line of trials.flatMap(trialWarnings))console.error(line)}function runDetails(id){return`(details: \`gauge run get ${id}\`, \`gauge run get ${id} --logs\`)`}function trialWarnings(t){const total=`${t.runs.length} session${t.runs.length===1?"":"s"}`;const failed=t.runs.filter(run=>FAILED_RUN.has(run.status));const unproven=t.runs.filter(run=>!FAILED_RUN.has(run.status)&&run.cliCoverage?.instrumentationFailed);const lines=[];if(failed.length)lines.push(`WARNING ${t.key}: ${failed.length} of ${total} failed; the score and verdict leave them out. Tell the user before you trust this result.`,...failed.map(run=>` ${run.id}: ${run.status}: ${run.failureReason??run.exitReason??"no reason recorded"} ${runDetails(run.id)}`));if(unproven.length)lines.push(`WARNING ${t.key}: ${unproven.length} of ${total} have incomplete command evidence; the score counts them, but the CLI change may not have applied. Tell the user.`,...unproven.map(run=>` ${run.id}: ${runDetails(run.id)}`));return lines}function printTrial(t){console.log(`key: ${t.key} (round ${t.roundNumber}) \u2014 ${t.id}`);console.log(`title: ${t.title}`);if(t.hypothesis)console.log(`hypothesis: ${t.hypothesis}`);console.log(`change: ${t.kind}${t.lever?` on ${t.lever}`:""}`);console.log(`status: ${t.status}${t.verdict?` \xB7 ${t.verdict}`:""}${t.adopted?" \xB7 winner":""}`);console.log(`score: ${pct2(t.baselineScore)} \u2192 ${pct2(t.score)} over ${t.sessions} session${t.sessions===1?"":"s"}`);if(t.outcome)console.log(`outcome: ${t.outcome}`);if(t.runs.length){console.log("sessions:");for(const run of t.runs){const verdict=run.passed===null?"not assessed":run.passed?"pass":"fail";console.log(` ${run.id}: ${run.status} \xB7 ${verdict}`);if(FAILED_RUN.has(run.status))console.log(` error: ${run.failureReason??run.exitReason??"no reason recorded"}`);else if(run.cliCoverage?.instrumentationFailed)console.log(" warning: command evidence is incomplete");for(const criterion of run.criteria)console.log(` ${criterion.passed===null?"?":criterion.passed?"\u2713":"\u2717"} ${criterion.name}`)}}for(const line of trialWarnings(t))console.error(line)}function printChange(content){for(const doc of content.documents){console.log(`=== ${doc.path}`);console.log("--- before");console.log(doc.before||"(empty)");console.log("+++ after");console.log(doc.after||"(empty)")}}function printRoundPreview(preview){for(const trial of preview.trials){console.log(trial.title);for(const change of trial.changes??[]){console.log(`=== ${change.path}`);console.log(change.hunks)}for(const path of trial.droppedFiles??[])console.log(` ${path}: dropped`)}console.log(`previewDigest ${preview.previewDigest}`)}function printDiagnosis(d){console.log(`id: ${d.id}`);console.log(`status: ${d.status} over ${d.runCount} sessions`);console.log(`summary: ${d.summary??"(none)"}`);console.log(`recommended: ${d.recommendedLever??"no single lever"}`);if(d.unsupportedLevers.length)console.log(`not yet supported: ${d.unsupportedLevers.join(", ")}`);if(d.clusters.length){console.log("opportunities:");for(const cluster of d.clusters)console.log(` - ${cluster.title} [${cluster.lever??"?"} \xB7 ${Math.round(cluster.confidence*100)}%] ${cluster.failureMode}`)}}function addSubjectOptions(command){return command.option("--eval <id>","an Agent Experience eval").option("--preference <id>","an Agent Preference prompt")}function addTargetOption(command){return command.option("--target <agent[:model]>","the one harness/model every session runs on, e.g. claude-code:claude-opus-4-8")}function register23(program2){const group=hostedOnly(program2.command("optimizations")).description("Start, steer, and read optimizations: Gauge changes one owned surface per trial and reports the best measured change against the baseline");group.command("list").description("List the org's optimizations, newest activity first").option("--status <statuses>","comma-separated: DRAFT,ACTIVE,PAUSED,DONE,ARCHIVED").option("--no-drafts","leave out draft workspaces").option("--eval <id>","only optimizations on this eval").option("--preference <id>","only optimizations on this preference prompt").option("--limit <n>","max rows (default 50)").action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const subject=opts.eval||opts.preference?parseSubject(opts):void 0;const res=await createClient().get(basePath(org),{status:parseStatuses(opts.status),subjectKind:subject?.subjectKind,subjectId:subject?.subjectId,limit:parseCount(opts.limit,"--limit",200),includeDrafts:opts.drafts});if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(toRow2),"table")});addSubjectOptions(group.command("baselines").description("Discover which ordinary runs exposed which owned surfaces, and whether they can seed an optimization")).option("--limit <n>","page size (1\u2013100)","50").option("--cursor <cursor>","nextCursor from the previous page").action(async(opts,cmd)=>{const response=await createClient().get(`${basePath(await resolveOrg(cmd))}/baselines`,{...parseSubject(opts),limit:parseCount(opts.limit,"--limit"),cursor:opts.cursor});if(resolveOutput(cmd)==="json")printJson(response);else{printItems(response.items.flatMap(run=>(run.surfaces.length?run.surfaces:[{lever:"\u2014",label:"No recorded surface evidence"}]).map(surface=>({run:run.runId,model:run.model||run.resolvedModel||run.agent,date:run.createdAt.slice(0,10),criteria:run.outcome.totalCriteria?`${run.outcome.passedCriteria}/${run.outcome.totalCriteria}`:"\u2014",status:run.status,surface:surface.lever,target:surface.label,eligible:run.eligibility.eligible?"yes":"no"}))),"table");for(const run of response.items)if(run.eligibility.reason)console.log(`${run.runId}: ${run.eligibility.reason}`);console.log(response.note);if(response.nextCursor)console.log(`More: --cursor ${response.nextCursor}`)}});group.command("get <id>").description("One optimization: settings, score, and its trial ladder").action(async(id,_opts,cmd)=>{const org=await resolveOrg(cmd);const detail=await createClient().get(itemPath(org,id));if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail)});addSubjectOptions(group.command("diagnose").description("Improvement opportunities for an eval or preference prompt from its judged sessions")).action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const diagnosis=await createClient().post(`${basePath(org)}/diagnoses`,parseSubject(opts));if(resolveOutput(cmd)==="json")printJson(diagnosis);else printDiagnosis(diagnosis)});addTargetOption(addSubjectOptions(group.command("start").description("Reuse a selected baseline with --from-run, or establish one with --target"))).option("--from-run <runId>","reuse the eligible baseline containing this run; inherits its model").option("--name <name>","optimization name (default: from the subject)").option("--yes","skip the confirmation").action(async(opts,cmd)=>{const org=await resolveOrg(cmd);if(!!opts.target===!!opts.fromRun)throw new CliError("usage","pass --from-run <runId> or --target <agent[:model]>, not both; discover runs with `gauge optimizations baselines`");const subject=parseSubject(opts);await confirmOrAbort(opts.fromRun?`Start an optimization from the baseline containing ${opts.fromRun}?`:`Start an optimization on ${subjectLabel(subject.subjectKind)} ${subject.subjectId}? A new baseline may spend credits.`,opts.yes);const detail=await createClient().post(basePath(org),{...subject,...targetFromOpts(opts),...opts.fromRun?{fromRun:opts.fromRun}:{},name:opts.name});if(resolveOutput(cmd)==="json")printJson(detail);else{console.log(`Started ${detail.id} "${detail.name}". Each round waits for you to launch it. Follow the baseline with \`gauge optimizations watch ${detail.id}\`.`)}});group.command("cancel <id>").description("Cancel an optimization and stop its in-flight sessions").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);await confirmOrAbort(`Cancel optimization ${id}? In-flight sessions stop; completed usage remains billed.`,opts.yes);const detail=await createClient().post(`${itemPath(org,id)}/cancel`,{});if(resolveOutput(cmd)==="json")printJson(detail);else console.log(`${detail.id}: ${detail.status}${detail.stopReason?` (${detail.stopReason})`:""}`)});addTargetOption(addSubjectOptions(group.command("create").description("Create a draft workspace without running sessions"))).option("--name <name>","optimization name").action(async(opts,cmd)=>{const detail=await createClient().post(`${basePath(await resolveOrg(cmd))}/drafts`,{...parseSubject(opts),...targetFromOpts(opts),name:opts.name});if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail)});addTargetOption(group.command("edit <id>").description("Edit name/direction, or a draft's target")).option("--name <name>","optimization name").option("--file <path>","PATCH body JSON (name, direction, target); '-' for stdin").action(async(id,opts,cmd)=>{const body={...opts.file?readObject(opts.file):{},...targetFromOpts(opts),...opts.name?{name:opts.name}:{}};if(!Object.keys(body).length)throw new CliError("usage","pass --target, --name, or --file");const detail=await createClient().patch(itemPath(await resolveOrg(cmd),id),body);if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail)});addTargetOption(group.command("start-draft <id>").description("Start an existing draft's baseline (spends credits)")).option("--name <name>","optimization name").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const plan=targetFromOpts(opts);const current=await client.get(itemPath(org,id));if(current.status!=="DRAFT")throw new CliError("usage","only a draft can start; use continue instead");if(!plan.target&&!current.settings.pinnedAgent)throw new CliError("usage","--target is required until the draft has an agent/model");await confirmOrAbort(`Start the baseline for ${id}? It spends credits now.`,opts.yes);const detail=await client.post(`${itemPath(org,id)}/start`,{...plan,name:opts.name});if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail)});group.command("launch <id>").description("Launch all saved trials in the current round (spends credits)").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{await confirmOrAbort(`Launch the current round of ${id}? Spends credits.`,opts.yes);const result=await createClient().post(`${itemPath(await resolveOrg(cmd),id)}/rounds/current/launch`,{});if(resolveOutput(cmd)==="json")printJson(result);else console.log(result.alreadyLaunched?"Round already launched.":`Launched ${result.launched.length} trials in round ${result.roundNumber}.`)});group.command("continue <id>").description("Reopen a finished optimization for another round against its fixed baseline").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{await confirmOrAbort(`Reopen ${id} for another round? Saved trials still require launch.`,opts.yes);const detail=await createClient().post(`${itemPath(await resolveOrg(cmd),id)}/continue`,{});if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail)});group.command("activity <id>").description("Read the optimization's decision log").option("--after <timestamp>","entries after this ISO timestamp").option("--limit <n>","maximum entries (1\u2013500)").action(async(id,opts,cmd)=>{const res=await createClient().get(`${itemPath(await resolveOrg(cmd),id)}/activity`,{after:opts.after,limit:parseCount(opts.limit,"--limit",500)});if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(row2=>({at:row2.createdAt,kind:row2.kind,actor:row2.actorName??"Worker",summary:row2.summary})),"table")});group.command("chat <id>").description("Resolve your optimization chat and print its URL path").action(async(id,_opts,cmd)=>{const res=await createClient().post(`${itemPath(await resolveOrg(cmd),id)}/chat`,{});if(resolveOutput(cmd)==="json")printJson(res);else console.log(res.href)});group.command("rm <id>").description("Delete an optimization, retaining billed session history").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{await confirmOrAbort(`Delete optimization ${id}, its rounds, trials, and activity? Session history is kept.`,opts.yes);await createClient().delete(itemPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson({deleted:id});else console.log(`Deleted ${id}.`)});group.command("complete <id>").description("Complete an optimization: it reads as Finished, not canceled").option("--adopt <trialId>","mark this settled trial as the winner").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);await confirmOrAbort(`Complete optimization ${id}? No further rounds run; unspent credits are released.`,opts.yes);const detail=await createClient().post(`${itemPath(org,id)}/complete`,opts.adopt?{adoptTrialId:opts.adopt}:{});if(resolveOutput(cmd)==="json")printJson(detail);else console.log(`${detail.id}: ${detail.status} (${detail.stopReason})`)});group.command("watch <id>").description(`Follow until finished or awaiting input (exit ${NEEDS_INPUT_EXIT} when it needs you); --follow keeps waiting`).option("--interval <seconds>","poll interval (default 15)").option("--follow","keep polling while awaiting authoring or launch").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const interval=Number(opts.interval??15);if(!Number.isFinite(interval)||interval<=0)throw new CliError("usage","--interval must be a positive finite number");const intervalMs=Math.max(interval*1e3,2e3);let last="";for(;;){const detail=await client.get(itemPath(org,id));const trials2=detail.rounds.flatMap(round2=>round2.trials);const failed=trials2.reduce((sum,trial)=>sum+trial.failedSessions,0);const line=`${detail.status}${detail.stopReason?` (${detail.stopReason})`:""} \xB7 round ${detail.round} \xB7 ${pct2(detail.before)} \u2192 ${pct2(detail.after)} \xB7 ${detail.creditsSpent} credits spent${failed?` \xB7 ${failed} failed session${failed===1?"":"s"}`:""} \xB7 ${trials2.filter(t=>t.status==="RUNNING").map(t=>t.key).join(",")||"idle"}`;if(line!==last){last=line;console.error(line)}const needsInput=awaitingInput(detail);if(watchStops(detail,!!opts.follow)){if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail);if(needsInput)process.exitCode=NEEDS_INPUT_EXIT;return}await sleep3(intervalMs)}});group.command("authoring <id>").description("What the next round may change: exposed surfaces, CLI targets, baseline run evidence, revision").action(async(id,_opts,cmd)=>{printJson(await createClient().get(`${itemPath(await resolveOrg(cmd),id)}/authoring`))});group.command("cli-change <id>").description("Validate and seal v3 {schemaVersion:3, target, rules, shippingIntent}; use the output as intervention.after in a round file. Saves nothing").requiredOption("--file <path>","the change JSON; '-' for stdin").action(async(id,opts,cmd)=>{printJson(await createClient().post(`${itemPath(await resolveOrg(cmd),id)}/cli-changes`,readObject(opts.file)))});const round=group.command("round").description("Preview and save the round waiting to launch. A round file is {revision, trials: [{trial: {title, hypothesis, lever, intervention}, files?}]}; revision comes from `authoring`");round.command("preview <id>").description("Validate a round file and print its diff and previewDigest").requiredOption("--file <path>","the round JSON; '-' for stdin").action(async(id,opts,cmd)=>{const preview=await createClient().post(`${itemPath(await resolveOrg(cmd),id)}/rounds/current/preview`,readObject(opts.file));if(resolveOutput(cmd)==="json")printJson(preview);else printRoundPreview(preview)});round.command("save <id>").description("Save a previewed round as drafts, replacing every pending draft. Never launches").requiredOption("--file <path>","the round JSON you previewed").requiredOption("--digest <digest>","previewDigest from `round preview`").action(async(id,opts,cmd)=>{const result=await createClient().put(`${itemPath(await resolveOrg(cmd),id)}/rounds/current`,{...readObject(opts.file),previewDigest:opts.digest});if(resolveOutput(cmd)==="json")printJson(result);else for(const trial of result.trials)console.log(`Saved ${trial.key} "${trial.title}"`)});const trials=group.command("trials").description("List, author, and run trials");trials.command("list <id>").description("Every trial of an optimization with score and verdict").action(async(id,_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(`${itemPath(org,id)}/trials`);if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(trialRow),"table")});trials.command("add <id>").description("Save an authored draft trial; --launch also runs it (spends credits)").option("--file <path>","authored trial JSON, as `trials template` prints it: one {trial, files?} or an array; '-' for stdin").option("--pick <n>","which trial to take from an array (default 1)").option("--title <text>","what the change is").option("--hypothesis <text>","why it should help").option("--lever <lever>","DOCS, MARKETING_SITE, SKILL, SDK, CLI").option("--edit-skill <id>","skill version to rewrite (with --skill-dir); see `gauge optimizations skill-files <id>`").option("--skill-dir <path>","local directory holding the complete edited bundle, SKILL.md included").option("--page <url>","docs page to rewrite (with --body-file)").option("--body-file <path>","the rewritten page body; '-' for stdin").option("--launch","run the trial after saving (spends credits); otherwise save a draft").option("--idempotency-key <key>","a retry with this key returns the first call's trial instead of a second one").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);const body=trialBodyFromOpts(opts);if(body.launch!==false)await confirmOrAbort(`Run trial "${body.trial.title}" on ${id} now? Spends credits.`,opts.yes);const result=await createClient().post(`${itemPath(org,id)}/trials`,body);if(resolveOutput(cmd)==="json")printJson(result);else if(result.replayed)console.log(`Replayed ${result.trial.key} "${result.trial.title}" (${result.trial.status}) \u2014 this key already authored it`);else console.log(`${result.launch?"Running":"Saved"} ${result.trial.key} "${result.trial.title}"${result.launch?` \u2014 ${result.launch.runIds.length} session${result.launch.runIds.length===1?"":"s"}`:""}`)});trials.command("template").description("Print a trial JSON skeleton for `trials add --file`").requiredOption("--kind <kind>","EDIT_SKILL or REWRITE_DOCS").action(opts=>{printJson(trialTemplate(opts.kind))});trials.command("run <id> <trialId>").description("Run a draft trial now (spends credits)").option("--yes","skip the confirmation").action(async(id,trialId,opts,cmd)=>{const org=await resolveOrg(cmd);await confirmOrAbort(`Run trial ${trialId} now? Spends credits.`,opts.yes);const result=await createClient().post(`${trialPath(org,id,trialId)}/launch`,{});if(resolveOutput(cmd)==="json")printJson(result);else console.log(`Started ${result.runIds.length} session${result.runIds.length===1?"":"s"} for ${trialId}`)});trials.command("rm <id> <trialId>").description("Drop one draft trial from the round waiting to launch").option("--yes","skip the confirmation").action(async(id,trialId,opts,cmd)=>{await confirmOrAbort(`Remove draft ${trialId} from ${id}?`,opts.yes);const detail=await createClient().delete(trialPath(await resolveOrg(cmd),id,trialId));if(resolveOutput(cmd)==="json")printJson(detail);else console.log(`Removed ${trialId}.`)});group.command("trial <id> <trialId>").description("One trial by id or key (B0, R1.1, \u2026) with its sessions").action(async(id,trialId,_opts,cmd)=>{const org=await resolveOrg(cmd);const trial=await createClient().get(trialPath(org,id,trialId));if(resolveOutput(cmd)==="json")printJson(trial);else printTrial(trial)});group.command("change <id> <trialId>").description("The change a trial made: before and after text per document").option("--body","print only the pasteable page body of a docs rewrite").action(async(id,trialId,opts,cmd)=>{const org=await resolveOrg(cmd);const content=await createClient().get(`${trialPath(org,id,trialId)}/content`);if(resolveOutput(cmd)==="json")printJson(content);else if(opts.body){if(content.copyableBody===null)throw new CliError("usage","this trial is not a docs rewrite; drop --body");console.log(content.copyableBody)}else printChange(content)});group.command("adopt <id> <trialId>").description("Mark a settled trial as the winner without changing the baseline").option("--yes","skip the confirmation").action(async(id,trialId,opts,cmd)=>{const org=await resolveOrg(cmd);await confirmOrAbort(`Mark ${trialId} as the winner on ${id}?`,opts.yes);const detail=await createClient().post(`${trialPath(org,id,trialId)}/adopt`,{});if(resolveOutput(cmd)==="json")printJson(detail);else console.log(`Marked ${trialId} as the winner.`)});group.command("handoff <id>").description("The instructions a coding agent follows to land the winning trial in your source, as the web app's Copy instructions for agent").action(async(id,_opts,cmd)=>{const handoff=await createClient().get(`${itemPath(await resolveOrg(cmd),id)}/handoff`);if(resolveOutput(cmd)==="json")printJson(handoff);else process.stdout.write(handoff.prompt)});group.command("skill-files <versionId>").description("The text files of a skill version, the base a skill-edit trial edits").action(async(versionId,_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(`${basePath(org)}/skill-versions/${encodeURIComponent(versionId)}`);if(resolveOutput(cmd)==="json")printJson(res);else for(const file of res.files){console.log(`=== ${file.path}`);console.log(file.body)}});group.command("page <id>").description("The baseline body of a docs page, the base a docs-rewrite trial edits").requiredOption("--url <url>","the page URL").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(`${itemPath(org,id)}/page`,{url:opts.url});if(resolveOutput(cmd)==="json")printJson(res);else{if(res.source==="scrape")console.error("Source: fresh scrape; this is not the captured baseline response.");else if(res.source==="none")console.error("No page body is available.");console.log(res.body)}})}var org_exports={};__export(org_exports,{register:()=>register24});function printSettings(s,billingStatus){console.log(`slug: ${s.slug}`);console.log(`name: ${s.name}`);console.log(`website: ${s.website??"(none)"}`);console.log(`max spend/run: ${s.maxSpendUsd==null?"(server default)":`$${s.maxSpendUsd}`}`);console.log(`overage: ${fmtOverage(billingStatus)}`);console.log(`auto-join: ${s.autoJoinDomain??"off"}`)}function register24(program2){const group=hostedOnly(program2.command("org")).description("This org's settings, read-only (see `gauge orgs` for the org list)");group.command("get").description("Show the org's settings (edit them in the web app)").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const base=`/api/v1/orgs/${encodeURIComponent(org)}`;const[settings,billingStatus]=await Promise.all([client.get(base),client.get(`${base}/billing`)]);if(resolveOutput(cmd)==="json")printJson({...settings,paymentType:billingStatus.paymentType??null,creditLine:billingStatus.creditLine??null});else printSettings(settings,billingStatus)})}var orgs_exports={};__export(orgs_exports,{register:()=>register25});import{writeFileSync as writeFileSync2}from"fs";import{join as join9}from"path";function register25(program2){const orgs=hostedOnly(program2.command("orgs"),{account:true}).description("Organizations, and which Gauge products each one uses");orgs.command("list",{isDefault:true}).description("List your Gauge organizations (* marks the saved default)").action(async(_opts,command)=>{const rows=await fetchDirectory();const{orgId}=readConfig();if(resolveOutput(command)==="json"){printJson(rows.map(row2=>({...row2,isDefault:row2.id===orgId})));return}printItems(rows.map(row2=>({default:row2.id===orgId?"*":"",slug:row2.slug,name:row2.name,products:productsLabel(row2.products)})),"table")});orgs.command("create <name>").description("Create a Gauge Agents workspace and set its organization as the default").option("--website <url>","company site (used as the logo source)").option("--no-use","don't set the new org as the default").action(async(name,opts,command)=>{await fetchDirectory();const created=await createClient().post("/api/v1/orgs",{name,website:opts.website});const org=await createdOrganization(created.slug);if(opts.use!==false)updateConfig({orgId:org.id,defaultOrg:void 0});if(resolveOutput(command)==="json")return printJson(org);console.log(`Created workspace ${org.name} (${org.slug})`);if(opts.use!==false)console.log(`Default org set to ${org.slug}`)});orgs.command("use <org>").description("Select a Gauge organization by its public slug").option("--project","save the selection in this project's gauge.json").action(async(value,opts,command)=>{const org=await findOrganization(value);if(opts.project){const state=gaugeState(command);if(!state.root||!state.config)throw new CliError("usage","No valid gauge.json found. Run `gauge init` first.");const{org:_legacyOrg,...config}=state.config;writeFileSync2(join9(state.root,"gauge.json"),`${JSON.stringify({...config,orgId:org.id},null,2)}
|
|
781
|
-
`)}else updateConfig({orgId:org.id,defaultOrg:void 0});if(resolveOutput(command)==="json")return printJson({orgId:org.id,slug:org.slug,name:org.name});console.log(`${opts.project?"Project":"Default"} org set to ${org.name} (${org.slug})`)})}var personas_exports={};__export(personas_exports,{register:()=>register26});function personasPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/personas`}function register26(program2){const group=hostedOnly(program2.command("personas")).description("Manage simulated-user personas for eval and preference runs");group.command("list").description("List the org's personas").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(personasPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(p=>({id:p.id,name:p.name,description:p.description.length>60?`${p.description.slice(0,57)}...`:p.description})),"table")});group.command("get <id>").description("Show one persona").action(async(id,_opts,cmd)=>{const org=await resolveOrg(cmd);const persona=await createClient().get(`${personasPath(org)}/${encodeURIComponent(id)}`);if(resolveOutput(cmd)==="json"){printJson(persona);return}console.log(`id: ${persona.id}`);console.log(`name: ${persona.name}`);console.log(`model: ${persona.model}`);console.log(`description: ${persona.description}`)});group.command("add").description("Create a persona").requiredOption("--name <name>","persona name").requiredOption("--description <text>",'memory-style notes ("The user is a solo founder\u2026")').action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const created=await createClient().post(personasPath(org),{name:opts.name,description:opts.description});if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created persona ${created.id} (${created.name})`)});group.command("edit <id>").description("Update a persona").option("--name <name>","rename the persona").option("--description <text>","replace the notes").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);if(opts.name===void 0&&opts.description===void 0)throw new CliError("usage","nothing to change");const updated=await createClient().patch(`${personasPath(org)}/${encodeURIComponent(id)}`,{name:opts.name,description:opts.description});if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated persona ${updated.id}`)});group.command("rm <id>").description("Delete a persona (run configurations fall back to the default persona)").action(async(id,_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(`${personasPath(org)}/${encodeURIComponent(id)}`);if(resolveOutput(cmd)!=="json")console.log(`Deleted persona ${id}`)})}var preference_exports={};__export(preference_exports,{register:()=>register27});function visibilityPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/visibility-prompts`}function promptPath(org,id){return`${visibilityPath(org)}/${encodeURIComponent(id)}`}function scenariosPath(org,id){return`${promptPath(org,id)}/scenarios`}function scenarioRow(scenario){return{id:scenario.id,repo:scenario.repoUrl??"(empty project)",ref:scenario.repoRef??"",persona:scenario.profile?.name??"(default)"}}var KIND_CHOICES=["head-to-head","open-ended"];function parseKind(raw){const upper=raw==="open-ended"?"RECOMMEND":raw.trim().toUpperCase().replace(/-/g,"_");if(upper!=="HEAD_TO_HEAD"&&upper!=="RECOMMEND")throw new CliError("usage",`invalid --kind "${raw}" (expected: ${KIND_CHOICES.join(" | ")})`);return upper}function brandsLabel(v){if(v.kind==="HEAD_TO_HEAD")return`${v.brandA} vs ${v.brandB}`;return v.branded?v.brandA??"branded":"organic"}function preferencePromptRow(v){return{id:v.id,name:v.name,kind:v.kind==="RECOMMEND"?"open-ended":"head-to-head",brands:brandsLabel(v),market:v.market?.slug??"",tags:v.tags.join(", "),runs:v.runCount,"last run":v.lastRunAt??"-","created by":creatorLabel(v.createdBy)}}function printDetail3(v){console.log(`id: ${v.id}`);console.log(`name: ${v.name}`);console.log(`kind: ${v.kind==="RECOMMEND"?"open-ended":"head-to-head"}`);console.log(`brands: ${brandsLabel(v)}`);console.log(`market: ${v.market?`${v.market.name} (${v.market.slug})`:"(none)"}`);console.log(`tags: ${v.tags.join(", ")||"(none)"}`);console.log(`runs: ${v.runCount} (last: ${v.lastRunAt??"never"})`);console.log(`creator: ${creatorLabel(v.createdBy)}`);console.log(`prompt: ${v.promptText}`);console.log(`cadence: ${cadenceLabel(v.cadence)} \u2014 ${v.sampleCount} sample(s)`);console.log(`agents: ${v.agents.map(a=>a.models.length?`${a.agent}:${a.models.join(",")}`:a.agent).join(", ")}`);console.log(`skills: ${v.skillRefs.join(", ")||"(none)"}`);console.log(`mcp: ${v.mcpRefs.join(", ")||"(none)"}`);console.log(`conns: ${v.connectionIds.join(", ")||"(none)"}`);console.log(`scenarios: ${v.scenarios.length}`);for(const scenario of v.scenarios)console.log(` [${scenario.id}] ${scenario.repoUrl??"(empty project)"}${scenario.repoRef?`@${scenario.repoRef}`:""} \u2014 ${scenario.profile?.name??"default persona"}`)}async function listPreferencePrompts(cmd){if(isLocal(cmd))return listLocalMeasurements(cmd,"preference");const res=await createClient().get(visibilityPath(await resolveOrg(cmd)));return res.items}async function getPreferencePrompt(cmd,id){const vp=isLocal(cmd)?localMeasurementView(getLocalMeasurement(cmd,"preference",id)):await createClient().get(promptPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson(vp);else printDetail3(vp)}async function createPreferencePrompt(cmd,opts){if(opts.kind===void 0)throw new CliError("usage",`--kind is required with --preference (${KIND_CHOICES.join(" | ")})`);if(opts.ref!==void 0&&opts.repo===void 0)throw new CliError("usage","--ref requires --repo");const hasScenario=opts.repo!==void 0||opts.persona!==void 0;const body={name:opts.name,promptText:readPromptText(opts),kind:parseKind(opts.kind),branded:opts.branded,brandA:opts.brandA??null,brandB:opts.brandB??null,marketId:opts.market??null,tags:opts.tag.length>0?opts.tag:void 0,...hasRunSettingsFlags(opts)?runSettingsFromOpts(opts):{},scenarios:hasScenario?[{repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:opts.persona??null}]:[]};const created=isLocal(cmd)?createLocalMeasurement(cmd,"preference",body):await createClient().post(visibilityPath(await resolveOrg(cmd)),body);if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created preference prompt ${created.id}`)}async function editPreferencePrompt(cmd,id,opts){const body={...runSettingsPatchFromOpts(opts)};if(opts.name!==void 0)body.name=opts.name;if(opts.text!==void 0||opts.file!==void 0)body.promptText=readPromptText(opts);if(opts.brandA!==void 0)body.brandA=opts.brandA;if(opts.brandB!==void 0)body.brandB=opts.brandB;if(opts.branded&&opts.organic)throw new CliError("usage","--branded and --organic conflict");if(opts.branded)body.branded=true;else if(opts.organic)body.branded=false;if(opts.market!==void 0&&opts.clearMarket)throw new CliError("usage","--market and --clear-market conflict");if(opts.clearMarket)body.marketId=null;else if(opts.market!==void 0)body.marketId=opts.market;if(opts.tag.length>0&&opts.clearTags)throw new CliError("usage","--tag and --clear-tags conflict");if(opts.clearTags)body.tags=[];else if(opts.tag.length>0)body.tags=opts.tag;if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");const updated=isLocal(cmd)?editLocalMeasurement(cmd,"preference",id,body):await createClient().patch(promptPath(await resolveOrg(cmd),id),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated preference prompt ${updated.id}`)}async function removePreferencePrompt(cmd,id){if(isLocal(cmd))removeLocalMeasurement(cmd,"preference",id);else await createClient().delete(promptPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)!=="json")console.log(`Deleted preference prompt ${id}`)}async function listScenarios(cmd,id){const vp=isLocal(cmd)?localMeasurementView(getLocalMeasurement(cmd,"preference",id)):await createClient().get(promptPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson(vp.scenarios);else printItems(vp.scenarios.map(scenarioRow),"table")}async function addScenario(cmd,id,opts){if(opts.ref!==void 0&&opts.repo===void 0)throw new CliError("usage","--ref requires --repo");if(isLocal(cmd)){if(opts.persona)throw new CliError("usage","Personas require hosted Gauge");const row2=getLocalMeasurement(cmd,"preference",id);const updated=editLocalMeasurement(cmd,"preference",id,{scenarios:[...row2.scenarios,{repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:null}]});const scenario=updated.scenarios.at(-1);if(resolveOutput(cmd)==="json")printJson(scenario);else console.log(`Added scenario ${scenario.id}`);return}const created=await createClient().post(scenariosPath(await resolveOrg(cmd),id),{repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:opts.persona??null});if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Added scenario ${created.id}`)}async function removeScenario(cmd,id,scenarioId){if(isLocal(cmd)){const row2=getLocalMeasurement(cmd,"preference",id);const scenarios=row2.scenarios;if(!scenarios.some(s=>s.id===scenarioId))throw new CliError("usage",`Unknown scenario: ${scenarioId}`);editLocalMeasurement(cmd,"preference",id,{scenarios:scenarios.filter(s=>s.id!==scenarioId)});if(resolveOutput(cmd)!=="json")console.log(`Removed scenario ${scenarioId}`);return}await createClient().delete(`${scenariosPath(await resolveOrg(cmd),id)}/${encodeURIComponent(scenarioId)}`);if(resolveOutput(cmd)!=="json")console.log(`Removed scenario ${scenarioId}`)}async function runPreferencePrompt(cmd,id,opts){const org=await resolveOrg(cmd);const overrides=runOverridesFromOpts(opts);const settings=Object.keys(overrides).length>0?"these overrides (stored settings stay unchanged)":"the stored settings";await confirmOrAbort(opts.scenario.length>0?`Run ${opts.scenario.length} selected scenario${opts.scenario.length===1?"":"s"} with ${settings} now?`:`Run every scenario of this prompt with ${settings} now?`,opts.yes);const result=await createClient().post(`${promptPath(org,id)}/run`,{visibilityScenarioIds:opts.scenario.length>0?opts.scenario:void 0,...overrides});if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`Started ${result.runIds.length} run${result.runIds.length===1?"":"s"} across ${result.batchIds.length} scenario${result.batchIds.length===1?"":"s"}`);for(const batch of result.batchIds)console.log(`Runs: gauge run list --batch ${batch}`)}}async function listPreferencePromptRuns(cmd,id,opts){const res=await createClient().get(`${promptPath(await resolveOrg(cmd),id)}/runs`,{limit:opts.limit,cursor:opts.cursor});if(resolveOutput(cmd)==="json"){printJson(res);return}printItems(res.items.map(r=>({id:r.id,status:r.status,agent:r.agent,model:r.resolvedModel??r.model??"",source:r.triggeredBy?.source??"",credits:r.creditsCharged??"",created:r.createdAt,verdict:r.verdict?.status??"-"})),"table");if(res.nextCursor)console.error(`(more results \u2014 rerun with --cursor ${res.nextCursor})`)}async function markets(cmd,opts){if(isLocal(cmd))throw new CliError("usage","Markets require hosted Gauge. Pass --hosted.");const client=createClient();const base=`/api/v1/orgs/${encodeURIComponent(await resolveOrg(cmd))}/markets`;if(opts.add!==void 0){const market=await client.post(base,{marketId:opts.add});if(resolveOutput(cmd)==="json")printJson(market);else console.log(`Tracking market ${market.name} (${market.slug})`);return}const res=await client.get(base);if(resolveOutput(cmd)==="json"){printJson(res);return}printItems([...res.tracked.map(m=>({id:m.marketId,slug:m.slug,name:m.name,tracked:"yes",source:m.source})),...res.addable.map(m=>({id:m.marketId,slug:m.slug,name:m.name,tracked:"no",source:""}))],"table")}function register27(program2){const preference=program2.command("preference").description("Manage preference prompts: which product agents choose. Launch one with `gauge run --preference <id>`");preference.command("list").description("List preference prompts").option("--created-by <email|me>","only prompts created by this user").action(async(opts,cmd)=>{const items=await filterByCreator(await listPreferencePrompts(cmd),opts.createdBy);if(resolveOutput(cmd)==="json")printJson(items);else printItems(items.map(preferencePromptRow),"table")});preference.command("get <id>").description("Show one preference prompt").action(async(id,_opts,cmd)=>{await getPreferencePrompt(cmd,id)});addRunSettingsCreateOptions(preference.command("create").description("Create a preference prompt").requiredOption("--name <name>","name").option("--text <text>","prompt text inline").option("--file <path>",'prompt text from a file ("-" for stdin)').option("--repo <url>","the first scenario's repository (default: empty project)").option("--ref <ref>","branch, tag, or commit (with --repo)").option("--persona <profileId>","persona profile id")).option("--kind <kind>",KIND_CHOICES.join(" | ")).option("--brand-a <name>","first brand (head-to-head)").option("--brand-b <name>","second brand (head-to-head)").option("--branded","an open-ended prompt that names a brand").option("--market <marketId>","tracked market id (see `gauge preference markets`)").option("--tag <name>","user tag name, create-on-type (repeatable)",collect5,[]).action(async(opts,cmd)=>{await createPreferencePrompt(cmd,opts)});addRunSettingsEditOptions(preference.command("edit <id>").description("Update a preference prompt. Omitted flags keep their value.").option("--name <name>","rename it").option("--text <text>","replace the prompt text inline").option("--file <path>",'replace the prompt text from a file ("-" stdin)')).option("--brand-a <name>","first brand").option("--brand-b <name>","second brand").option("--branded","mark branded (open-ended only)").option("--organic","mark organic (open-ended only)").option("--market <marketId>","set the measured market").option("--clear-market","unscope from any market").option("--tag <name>","replace the user tag set (repeatable)",collect5,[]).option("--clear-tags","remove all user tags").action(async(id,opts,cmd)=>{await editPreferencePrompt(cmd,id,opts)});preference.command("rm <id>").description("Delete a preference prompt (run history stays)").action(async(id,_opts,cmd)=>{await removePreferencePrompt(cmd,id)});const scenarios=preference.command("scenarios").description("The repo \xD7 persona scenarios a preference prompt runs in");scenarios.command("list <id>").description("List a preference prompt's scenarios").action(async(id,_opts,cmd)=>{await listScenarios(cmd,id)});scenarios.command("add <id>").description("Add one repo \xD7 persona scenario").option("--repo <url>","repository URL (default: empty project)").option("--ref <ref>","branch, tag, or commit (with --repo)").option("--persona <profileId>","persona profile id (default persona if omitted)").action(async(id,opts,cmd)=>{await addScenario(cmd,id,opts)});scenarios.command("rm <id>").description("Remove one scenario").requiredOption("--scenario <scenarioId>","scenario id to remove").action(async(id,opts,cmd)=>{await removeScenario(cmd,id,opts.scenario)});preference.command("markets").description("List tracked and addable markets for preference prompts, or track one with --add").option("--add <marketId>","track a published market").action(async(opts,cmd)=>{await markets(cmd,opts)})}var repos_exports={};__export(repos_exports,{register:()=>register28});function reposPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/repos`}async function fetchAllRepos(client,org){return fetchAllPages(client,reposPath(org))}function toRow3(repo){return{name:repo.name,url:repo.url,languages:repo.languages.join(", "),frameworks:repo.frameworks.join(", "),tags:repo.tags.join(", "),size:repo.sizeBucket??"",stars:repo.stars,system:repo.isSystem?"yes":""}}function syncWarning(repo){return`warning: GitHub metadata sync failed; retry with \`gauge repos resync ${repo.url}\``}function register28(program2){const repos=hostedOnly(program2.command("repos")).description("Manage repositories");repos.command("list").description("List repositories (yours + read-only Gauge-provided ones)").action(async(_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);const items=await fetchAllRepos(createClient(),org);if(output==="json"){printJson(items);return}printItems(items.map(toRow3),"table")});repos.command("add <url>").description("Add a repository (pulls languages/size/stars from GitHub)").action(async(url,_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);const result=await createClient().post(reposPath(org),{url});if(output==="json"){printJson(result);return}console.log(`Added ${result.repo.name} (${result.repo.url})`);if(!result.synced)console.error(syncWarning(result.repo))});repos.command("edit <url>").description("Update a repository's display name, frameworks, or tags").option("--name <name>","display name").option("--framework <name>","REPLACE the framework list (repeatable)",collect5,[]).option("--tag <name>","REPLACE the tag list (repeatable)",collect5,[]).option("--clear-frameworks","remove all frameworks").option("--clear-tags","remove all tags").action(async(url,opts,command)=>{const org=await resolveOrg(command);if(opts.framework.length>0&&opts.clearFrameworks)throw new CliError("usage","--framework and --clear-frameworks conflict");if(opts.tag.length>0&&opts.clearTags)throw new CliError("usage","--tag and --clear-tags conflict");const body={url};if(opts.name!==void 0)body.name=opts.name;if(opts.clearFrameworks)body.frameworks=[];else if(opts.framework.length>0)body.frameworks=opts.framework;if(opts.clearTags)body.tags=[];else if(opts.tag.length>0)body.tags=opts.tag;if(Object.keys(body).length===1)throw new CliError("usage","nothing to change");const repo=await createClient().patch(reposPath(org),body);if(resolveOutput(command)==="json")printJson(repo);else console.log(`Updated ${repo.name} (${repo.url})`)});repos.command("resync <url>").description("Re-fetch a repository's GitHub metadata").action(async(url,_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);const result=await createClient().post(`${reposPath(org)}/resync`,{url});if(output==="json"){printJson(result);return}if(!result.synced){throw new CliError("internal","GitHub metadata sync failed (is the URL a public GitHub repository?)")}console.log(`Re-synced ${result.repo.name} (${result.repo.url})`)});repos.command("rm <url>").description("Remove a repository from the catalog").action(async(url,_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);await createClient().delete(reposPath(org),{url});if(output!=="json")console.log(`Removed ${url}`)})}var run_exports={};__export(run_exports,{register:()=>register29});import{readFileSync as readFileSync11,existsSync as existsSync4,mkdirSync as mkdirSync2,writeFileSync as writeFileSync3}from"fs";import{z as z14}from"zod";import{join as join12,relative as relative6}from"path";import{z as z7}from"zod";var preferenceDefinitionSchema=z7.discriminatedUnion("kind",[z7.object({kind:z7.literal("open-ended"),branded:z7.boolean().default(false)}).strict(),z7.object({kind:z7.literal("head-to-head"),brandA:z7.string().trim().min(1),brandB:z7.string().trim().min(1)}).strict().refine(p=>p.brandA.toLowerCase()!==p.brandB.toLowerCase(),"A head-to-head needs two different brands")]);var preferenceSnapshotSchema=z7.object({version:z7.literal(1),preference:preferenceDefinitionSchema}).strict();var headToHeadSchema=z7.object({kind:z7.literal("HEAD_TO_HEAD"),assessability:z7.enum(["assessed","partial","unassessable"]).optional(),selectionStatus:z7.enum(["measured","not_measured","unassessable"]).optional(),winner:z7.enum(["A","B","both","neither"]),brandA:z7.string(),brandB:z7.string(),rationale:z7.string(),brandAReasoning:z7.string(),brandBReasoning:z7.string(),brandAWhyRaised:z7.string().optional(),brandBWhyRaised:z7.string().optional()});var recommendPickSchema=z7.object({brand:z7.string(),rank:z7.number().int(),adopted:z7.boolean(),reasoning:z7.string().optional(),whyRaised:z7.string().optional()});var recommendSchema=z7.object({kind:z7.literal("RECOMMEND"),selectedBrands:z7.array(z7.string()).optional(),assessability:z7.enum(["assessed","partial","unassessable"]).optional(),selectionStatus:z7.enum(["measured","not_measured","unassessable"]).optional(),picks:z7.array(recommendPickSchema).default([]),singlePick:z7.boolean(),complianceNote:z7.string()});var visibilityVerdictSchema=z7.discriminatedUnion("kind",[headToHeadSchema,recommendSchema]);function preferenceSelectionLabel(verdict){if(verdict.assessability==="unassessable"||verdict.selectionStatus==="unassessable")return"Not assessable";if(verdict.kind==="HEAD_TO_HEAD")return verdict.winner==="A"?verdict.brandA:verdict.winner==="B"?verdict.brandB:verdict.winner==="both"?`${verdict.brandA} and ${verdict.brandB}`:"Neither";if(verdict.selectedBrands!==void 0)return verdict.selectedBrands.length?verdict.selectedBrands.join(", "):"No selection";if(!verdict.singlePick)return verdict.picks.length?"No clear selection":"No selection";return verdict.picks.find(pick=>pick.rank===1)?.brand??"No selection"}import{Option as Option4}from"commander";import{readFileSync as readFileSync9}from"fs";import{z as z8}from"zod";function assertRequiredExperiment(spec,required){if(required.length===0){throw new CliError("invalid_request","--require-experiment requires a nonempty label")}const mismatches=[];for(const[promptIndex,prompt]of spec.prompts.entries()){const promptExperiment=prompt.experiment??spec.experiment;const schedules=prompt.schedules??[{}];if(schedules.length===0){mismatches.push(`prompts.${promptIndex}.schedules`)}for(const[scheduleIndex,schedule]of schedules.entries()){if((schedule.experiment??promptExperiment)!==required){mismatches.push(`prompts.${promptIndex}.schedules.${scheduleIndex}`)}}}if(mismatches.length>0){throw new CliError("invalid_request",`--require-experiment expected every target to resolve to ${JSON.stringify(required)}; mismatched: ${mismatches.join(", ")}`)}}var EXAMPLE_SPEC={version:1,expand:true,experiment:"my-experiment",defaults:{agents:["CLAUDE_CODE"],samplesPerCycle:3},prompts:[{key:"add-healthcheck",text:"Add a /health endpoint that returns 200 OK."},{key:"migrate-to-vitest",text:"Migrate the test suite from Jest to Vitest.",schedules:[{repoUrl:"https://github.com/acme/api",repoRef:"main",model:"claude-opus-4-8"}]}]};async function readSpecJson(source){let raw;if(source==="-"){if(process.stdin.isTTY)throw new CliError("usage","--apply - needs a run spec piped to stdin");const chunks2=[];for await(const chunk of process.stdin)chunks2.push(chunk);raw=Buffer.concat(chunks2).toString("utf8")}else{try{raw=readFileSync9(source,"utf8")}catch(e){throw new CliError("usage",`cannot read spec file ${source}: ${e instanceof Error?e.message:String(e)}`)}}try{return JSON.parse(raw)}catch(e){throw new CliError("invalid_request",`spec is not valid JSON: ${e instanceof Error?e.message:String(e)}`)}}function parseSpec(parsed){const result=batteries_exports.batterySpecSchema.safeParse(parsed);if(!result.success){const details=result.error.issues.map(i=>` ${i.path.join(".")||"(spec)"}: ${i.message}`).join("\n");throw new CliError("invalid_request",`not a valid battery spec:
|
|
782
|
-
${details}`,{hint:"Run `gauge run --schema --hosted` for the full schema, or `gauge run --example --hosted` for a starter."})}return result.data}async function resolveApplyOrg(command,spec){return resolveOrg(command,spec.org)}function repoLabel2(repoUrl,repoRef){if(!repoUrl)return"greenfield";return repoRef?`${repoUrl}@${repoRef}`:repoUrl}function printPlan(plan){console.log(`Plan for org ${plan.org}`+(plan.experiment?` (experiment: ${plan.experiment})`:""));for(const prompt of plan.prompts){console.log(` ${prompt.action} prompt "${prompt.key}"`);for(const s of prompt.schedules){console.log(` ${s.action} schedule ${repoLabel2(s.repoUrl,s.repoRef)}`+(plan.expand?` \u2192 ${s.runs} run(s)`:""))}}const sum=plan.summary;console.log(`Prompts: ${sum.promptsToCreate} to create, ${sum.promptsToUpdate} to update; schedules: ${sum.schedulesToCreate} to create, ${sum.schedulesToUpdate} to update`);if(plan.expand){const byAgent=Object.entries(sum.runsByAgent).map(([agent,n])=>`${agent}: ${n}`).join(", ");console.log(`Runs to enqueue: ${sum.runsToEnqueue}${byAgent?` (${byAgent})`:""} \u2014 estimated ${sum.estimatedCredits} credits`)}else{console.log('No runs will be enqueued (config only \u2014 set "expand": true to run).')}console.log("Dry run \u2014 nothing was applied.")}function printApplied(res){console.log(`Applied${res.experiment?` (experiment: ${res.experiment})`:""} \u2014 ${res.batchIds.length} batch(es), ${res.runIds.length} run(s) enqueued`);for(const id of res.runIds)console.log(` run ${id}`);if(res.runIds.length>0){console.error(`(watch one live: gauge run watch ${res.runIds[0]})`)}if(res.blocked){console.error(`warning: some schedules were not expanded (${res.blocked}) \u2014 re-apply after resolving to enqueue the rest`)}}var BATTERY_SCHEMA=()=>z8.toJSONSchema(batteries_exports.batterySpecSchema);async function applySpec(command,spec,opts){if(opts.requireExperiment!==void 0)assertRequiredExperiment(spec,opts.requireExperiment);const org=await resolveApplyOrg(command,spec);const result=await createClient().post(`/api/v1/orgs/${encodeURIComponent(org)}/batteries`,spec,opts.dryRun?{dryRun:1}:void 0);if(resolveOutput(command)==="json")printJson(result);else if(opts.dryRun)printPlan(result);else printApplied(result)}import{createHash as createHash3}from"crypto";import{z as z12}from"zod";import{z as z9}from"zod";var runAddonsSchema=z9.object({browser:z9.object({session:z9.string().min(1),projectId:z9.number().int().positive(),region:z9.literal("us").default("us")}).strict().optional()}).strict();import{z as z11}from"zod";import{z as z10}from"zod";var inputPathSchema=z10.string().min(1).max(256).regex(/^(?![!/])(?!.*[\\\x00-\x1f\x7f])(?!.*(?:^|\/)\.{1,2}(?:\/|$))(?!.*\/\/)(?!.*\/$).+$/,"Expected a repository-relative path or glob without traversal");var reservedBinaryCommands=["bash","sh","zsh","env","node","nodejs","npm","npx","bun","python","python3","git","tar","curl","cp","chmod","mkdir","rm","ln","test","sha256sum","printenv","uname","claude","codex","opencode","pi","wizard"];var binaryCommandSchema=z10.string().max(128).regex(/^[A-Za-z0-9_][A-Za-z0-9_.+-]*$/,"Expected a command name without paths or shell syntax").refine(name=>!reservedBinaryCommands.includes(name),"Command is reserved for Gauge workspace setup or the agent harness; choose another command name");var binaryUseSchema=z10.object({type:z10.literal("binary"),paths:z10.array(inputPathSchema).min(1).max(32),platform:z10.literal("linux-x64").describe("Linux x86-64 ELF executables compatible with the consumer sandbox. No system packages are installed."),commands:z10.record(binaryCommandSchema,inputPathSchema.regex(/^[^*?{}[\]]+$/,"Command targets must be explicit relative file paths")).refine(commands=>Object.keys(commands).length>=1&&Object.keys(commands).length<=32,"Declare between 1 and 32 commands").meta({minProperties:1,maxProperties:32})}).strict();var inputUseSchema=z10.discriminatedUnion("type",[binaryUseSchema,websiteUseSchema,z10.object({type:z10.literal("web-fixture"),path:inputPathSchema.regex(/^[^*?{}[\]]+$/)}).strict(),z10.object({type:z10.literal("files"),paths:z10.array(inputPathSchema).min(1).max(32).optional()}).strict(),z10.object({type:z10.literal("npm-cli"),package:inputPathSchema,install:z10.enum(["environment","workspace"]).optional()}).strict(),z10.object({type:z10.literal("npm-package"),package:inputPathSchema,install:z10.enum(["environment","workspace"]).optional()}).strict(),z10.object({type:z10.literal("skill"),root:z10.union([z10.literal("."),inputPathSchema]).refine(path=>!/[*?{}[\]]/.test(path),"Skill root must be an explicit directory")}).strict()]);var inputSourceSchema=z10.discriminatedUnion("type",[z10.object({type:z10.literal("git"),paths:z10.array(inputPathSchema).min(1).max(32)}).strict(),z10.object({type:z10.literal("github-actions"),artifact:z10.string().min(1).max(256).optional()}).strict()]);var runPlanResourceSchema=z10.object({name:z10.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64),type:z10.literal("input"),source:z10.discriminatedUnion("type",[z10.object({type:z10.literal("upload"),sha256:z10.string().regex(/^sha256:[a-f0-9]{64}$/),bytes:z10.number().int().positive().max(512*1024*1024),repoUrl:z10.string().nullable(),commit:z10.string().regex(/^[a-f0-9]{40}$/i).nullable()}).strict(),commentSourceSchema.extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i)}).strict(),deploymentSourceSchema.extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i)}).strict(),inputSourceSchema.options[0].extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i)}).strict(),inputSourceSchema.options[1].extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i),workflow:inputPathSchema}).strict()]),use:inputUseSchema}).strict().refine(r=>r.use.type==="website"===(r.source.type==="github-deployment"||r.source.type==="github-pr-comment"),"Website inputs require a deployment or PR comment source; file inputs require Git, Actions, or an upload");var definitionInputSourceSchema=z11.discriminatedUnion("type",[z11.object({type:z11.literal("git")}).strict(),z11.object({type:z11.literal("github-actions"),workflow:inputPathSchema.regex(/^\.github\/workflows\/[^/*?{}[\]]+\.ya?ml$/).describe("Workflow file in this repository, for example .github/workflows/build.yml."),artifact:z11.string().min(1).max(256).optional().describe("Upload name or pattern. Omit only when the workflow produces exactly one artifact.")}).strict()]);var skillInputPathSchema=z11.union([z11.literal("."),inputPathSchema.regex(/^[^*?{}[\]]+$/)]).describe("Explicit Skill directory or ZIP/tar.gz file containing SKILL.md at its root.");var runDefinitionInputSchema=z11.discriminatedUnion("type",[binaryUseSchema.extend({source:definitionInputSourceSchema.default({type:"git"})}).strict(),websiteUseSchema.extend({source:websiteSourceSchema}).strict(),z11.object({type:z11.literal("web-fixture"),source:definitionInputSourceSchema.default({type:"git"}),path:inputPathSchema.regex(/^[^*?{}[\]]+$/).describe("Explicit path to a version 1 web fixture JSON file.")}).strict(),z11.object({type:z11.literal("npm-package"),source:definitionInputSourceSchema.default({type:"git"}),path:inputPathSchema.describe("Path or glob selecting exactly one packed npm .tgz file."),install:z11.enum(["environment","workspace"]).default("environment").describe("environment exposes the package's commands on PATH; workspace adds a consumer application dependency.")}).strict(),z11.object({type:z11.literal("skill"),source:definitionInputSourceSchema.default({type:"git"}),path:skillInputPathSchema}).strict(),z11.object({type:z11.literal("files"),source:definitionInputSourceSchema.default({type:"git"}),paths:z11.array(inputPathSchema).min(1).max(32).describe("Files or globs selected from the committed Git tree or the uploaded artifact.")}).strict()]);function isSkillArchive(path){return/\.(?:zip|tar\.gz|tgz)$/i.test(path)}function compileDefinitionInput(name,input,repoUrl,commit){if(input.type==="website")return{name,type:"input",source:{...input.source,repoUrl,commit},use:{type:"website",url:input.url}};const paths=input.type==="files"||input.type==="binary"?input.paths:input.type==="skill"&&!isSkillArchive(input.path)?[input.path==="."?"**":`${input.path}/**`]:[input.path];return{name,type:"input",source:input.source.type==="git"?{type:"git",paths,repoUrl,commit}:{...input.source,repoUrl,commit},use:input.type==="binary"?{type:"binary",paths:input.paths,platform:input.platform,commands:input.commands}:input.type==="npm-package"?{type:"npm-package",package:input.path,install:input.install}:input.type==="web-fixture"?{type:"web-fixture",path:input.path}:input.type==="skill"?{type:"skill",root:input.path}:{type:"files",paths:input.paths}}}var inputUploadSpecSchema=z12.object({sha256:z12.string().regex(/^sha256:[a-f0-9]{64}$/),bytes:z12.number().int().positive().max(512*1024*1024)}).strict();var uncredentialedRepoUrlSchema=z12.string().min(1).refine(raw=>{if(!/^[a-z][a-z0-9+.-]*:\/\//i.test(raw))return true;try{const url=new URL(raw);return(url.protocol==="ssh:"||!url.username)&&!url.password&&!url.search&&!url.hash}catch{return false}},"Repository URL must not contain embedded credentials or query parameters");var runDefinitionChecksSchema=z12.object({paths:z12.array(inputPathSchema).min(1).max(32)}).strict();var runDefinitionConfigSchema=z12.object({agents:z12.array(agentConfigInputSchema).min(1),sampleCount:z12.number().int().min(1).max(50).optional(),repoUrl:uncredentialedRepoUrlSchema.nullable().optional(),repoRef:z12.string().min(1).nullable().optional(),profileId:z12.string().min(1).nullable().optional(),skillRefs:assetRefListSchema.optional(),mcpRefs:assetRefListSchema.optional(),connectionIds:z12.array(z12.string().min(1)).max(32).optional(),addons:runAddonsSchema.optional(),inputs:z12.array(z12.string()).optional(),checks:runDefinitionChecksSchema.optional()}).strict();var runDefinitionFileConfigSchema=z12.object({version:z12.literal(2).default(2),org:z12.string().min(1).optional(),orgId:z12.string().uuid().optional(),checks:runDefinitionChecksSchema.optional(),cases:z12.array(inputPathSchema).min(1).max(32).default(["gauge-cases/**/*.md"]),inputs:z12.record(z12.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64),runDefinitionInputSchema).refine(inputs=>Object.keys(inputs).length<=32,"At most 32 inputs are allowed").meta({maxProperties:32}).default({})}).strict();var runDefinitionCriterionSchema=z12.object({name:z12.string().min(1),rubric:z12.string().min(1)}).strict();var caseMeasurement={criteria:z12.array(runDefinitionCriterionSchema).min(1).optional(),preference:preferenceDefinitionSchema.optional()};function exactlyOneMeasurement(item){return item.criteria!==void 0!==(item.preference!==void 0)}var runDefinitionFrontmatterSchema=z12.object({...caseMeasurement,config:runDefinitionConfigSchema}).strict().refine(exactlyOneMeasurement,"Specify exactly one of criteria or preference");var runDefinitionCaseSchema=z12.object({path:z12.string().min(1),prompt:z12.string().min(1),...caseMeasurement,config:runDefinitionConfigSchema}).strict().refine(exactlyOneMeasurement,"Specify exactly one of criteria or preference");var runDefinitionSubmissionSchema=z12.object({version:z12.literal(2),source:z12.object({repoUrl:uncredentialedRepoUrlSchema.nullable(),commit:z12.string().regex(/^[a-f0-9]{40}$/i).nullable(),digest:z12.string().regex(/^sha256:[a-f0-9]{64}$/),verification:z12.enum(["SUBMITTED_UNVERIFIED","GITHUB_VERIFIED","LOCAL"])}).strict(),resources:z12.array(runPlanResourceSchema).max(32).default([]),cases:z12.array(runDefinitionCaseSchema).min(1).max(20)}).strict().superRefine((definition,ctx)=>{if(definition.source.verification!=="LOCAL"&&(!definition.source.repoUrl||!definition.source.commit))ctx.addIssue({code:"custom",path:["source"],message:"Committed definitions require a repository and commit"});if(definition.source.verification==="GITHUB_VERIFIED"&&definition.resources.some(r=>r.source.type==="upload"))ctx.addIssue({code:"custom",path:["resources"],message:"GitHub definitions cannot contain uploaded inputs"});const paths=new Set;for(const[index,item]of definition.cases.entries()){if(paths.has(item.path))ctx.addIssue({code:"custom",path:["cases",index,"path"],message:"Case paths must be unique"});paths.add(item.path)}const names=definition.resources.map(resource=>resource.name);if(new Set(names).size!==names.length)ctx.addIssue({code:"custom",path:["resources"],message:"Input names must be unique"});for(const[index,item]of definition.cases.entries())for(const name of item.config.inputs??names)if(!names.includes(name))ctx.addIssue({code:"custom",path:["cases",index,"config","inputs"],message:`Unknown input: ${name}`});for(const[index,resource]of definition.resources.entries()){if(resource.source.repoUrl!==definition.source.repoUrl||resource.source.commit!==definition.source.commit)ctx.addIssue({code:"custom",path:["resources",index,"source"],message:"Project preparation must use the definition repository and commit"})}});var runDefinitionPublicSubmissionSchema=runDefinitionSubmissionSchema.refine(definition=>definition.source.verification!=="GITHUB_VERIFIED",{path:["source","verification"],message:"Only the GitHub App may verify committed bytes"});var runDefinitionPlanResponseSchema=z12.object({source:runDefinitionSubmissionSchema.shape.source,measurementMode:z12.enum(["AGENT_EXPERIENCE","AGENT_PREFERENCE","MIXED"]),cases:z12.array(z12.object({path:z12.string(),agents:z12.array(agentConfigSchema),samples:z12.number().int().nonnegative(),expectedRuns:z12.number().int().nonnegative(),workspace:z12.object({repoUrl:z12.string().nullable(),repoRef:z12.string().nullable()}),criteriaCount:z12.number().int().nonnegative(),preference:preferenceDefinitionSchema.optional(),inputs:z12.array(z12.string())})),resources:z12.array(z12.object({name:z12.string(),type:z12.literal("input"),source:z12.enum(["git","github-actions","github-deployment","github-pr-comment","upload"]),use:inputUseSchema,readiness:z12.enum(["READY","WAITING","FAILED"]),detail:z12.string().optional(),commands:z12.array(z12.string()).optional()})),expectedRuns:z12.number().int().nonnegative(),canRun:z12.boolean()});var runDefinitionRunResponseSchema=z12.object({runRequestId:z12.string(),batchIds:z12.array(z12.string()),runIds:z12.array(z12.string()),requestedRuns:z12.number().int().nonnegative(),createdRuns:z12.number().int().nonnegative(),aggregateUrl:z12.string()});import{z as z13}from"zod";var caseSpecSchema=z13.object({version:z13.literal(2),cases:z13.array(z13.object({key:z13.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64).describe("Unique case id, lowercase-kebab"),prompt:z13.string().trim().min(1).describe("The user's request"),criteria:z13.array(runDefinitionCriterionSchema).min(1).optional().describe("Optional for local runs; hosted needs criteria or preference"),preference:preferenceDefinitionSchema.optional(),config:runDefinitionConfigSchema}).strict().refine(item=>!(item.criteria&&item.preference),"Specify criteria or preference, not both")).min(1).max(20)}).strict().refine(spec=>new Set(spec.cases.map(item=>item.key)).size===spec.cases.length,"Case keys must be unique");var EXAMPLE_CASE_SPEC={version:2,cases:[{key:"release-notes",prompt:"Write the release notes for version 1.2.0. We added dark mode and fixed a login crash.",criteria:[{name:"Team format",rubric:"RELEASE-NOTES.md lists both changes in the team's format."}],config:{agents:[{agent:"CODEX_CLI"}],inputs:["release-notes"]}}]};var CASE_SPEC_SCHEMA=()=>z13.toJSONSchema(caseSpecSchema);function isCaseSpec(value){return value!==null&&typeof value==="object"&&"cases"in value&&!("prompts"in value)}function parseCaseSpec(value){const result=caseSpecSchema.safeParse(value);if(!result.success)throw new CliError("usage",`not a valid case spec:
|
|
783
|
-
${result.error.issues.map(i=>` ${i.path.join(".")||"(spec)"}: ${i.message}`).join("\n")}`,{hint:"Run `gauge run --schema --local` for the full schema, or `gauge run --example --local` for a starter."});return result.data}function caseSpecCases(spec){return spec.cases.map(({key,...item})=>({path:key,...item}))}function caseSpecDigest(spec){return`sha256:${createHash3("sha256").update("gauge-case-spec-v2\0").update(JSON.stringify(spec)).digest("hex")}`}import{execFileSync as execFileSync2}from"child_process";import{createHash as createHash5}from"crypto";import{createReadStream as createReadStream2,globSync as globSync2}from"fs";import{lstat,mkdtemp,readdir,rm,stat}from"fs/promises";import{tmpdir}from"os";import{basename,dirname,isAbsolute as isAbsolute4,join as join11,relative as relative5,resolve as resolve4,sep as sep3}from"path";import{setTimeout as delay2}from"timers/promises";import{createHash as createHash4}from"crypto";import{parseDocument}from"yaml";function utf8(bytes,path){try{return new TextDecoder("utf-8",{fatal:true}).decode(bytes)}catch{throw new Error(`${path} is not UTF-8 text`)}}function parseRunDefinitionCase(path,bytes){const source=utf8(bytes,path);const match=/^---\r?\n([\s\S]*?)\r?\n---(?:\r?\n|$)([\s\S]*)$/.exec(source);if(!match)throw new Error(`${path}: expected YAML frontmatter between --- lines`);const document=parseDocument(match[1],{uniqueKeys:true});if(document.errors.length)throw new Error(`${path}: ${document.errors.map(error=>error.message).join("; ")}`);const parsed=runDefinitionFrontmatterSchema.safeParse(document.toJS());if(!parsed.success)throw new Error(`${path}: ${parsed.error.message}`);const prompt=match[2].trim();if(!prompt)throw new Error(`${path}: Markdown body is empty`);return{path,prompt,...parsed.data}}function compileRunDefinitionFromBlobs(input){if(input.repoUrl!==null&&!uncredentialedRepoUrlSchema.safeParse(input.repoUrl).success)throw new Error("Git origin must not contain embedded credentials");if(input.commit!==null&&!/^[a-f0-9]{40}$/i.test(input.commit))throw new Error("Expected a full Git commit SHA");if(input.casePaths.length===0)throw new Error("At least one Markdown case is required");if(new Set(input.casePaths).size!==input.casePaths.length)throw new Error("Case paths must be unique");const configBytes=input.blobs.get("gauge.json");let config;if(configBytes){try{config=runDefinitionFileConfigSchema.parse(JSON.parse(utf8(configBytes,"gauge.json")))}catch(error){throw new Error(`gauge.json: ${error instanceof Error?error.message:error}`)}}const cases=input.casePaths.map(path=>{const bytes=input.blobs.get(path);if(!path.endsWith(".md")||!bytes)throw new Error(`Missing Markdown case: ${path}`);return parseRunDefinitionCase(path,bytes)});const hash=createHash4("sha256");hash.update("gauge-run-definition-v2\0");for(const[path,bytes]of[...input.blobs.entries()].sort(([a],[b])=>a<b?-1:a>b?1:0)){hash.update(path);hash.update("\0");hash.update(String(bytes.length));hash.update("\0");hash.update(bytes);hash.update("\0")}const submission=runDefinitionSubmissionSchema.parse({version:2,source:{repoUrl:input.repoUrl,commit:input.commit,digest:`sha256:${hash.digest("hex")}`,verification:input.verification},resources:input.resources??Object.entries(config?.inputs??{}).map(([name,item])=>{if(!input.repoUrl||!input.commit)throw new Error(`Supply --input ${name}=<path> outside a Git checkout with an origin`);return compileDefinitionInput(name,item,input.repoUrl,input.commit)}),cases});return{definition:submission,org:config?.orgId??config?.org}}import{minimatch}from"minimatch";function matchInputPaths(paths,patterns,options={}){for(const pattern of patterns)inputPathSchema.parse(pattern);const selections=patterns.map(pattern=>({pattern,matches:paths.filter(path=>minimatch(path,pattern,{dot:true,nocase:false,nonegate:true}))}));if(options.allowEmptySelection&&selections.every(({matches:matches2})=>matches2.length===0))return[];const selected=new Set;for(const{pattern,matches:matches2}of selections){if(!matches2.length)throw new Error(`No committed files match ${pattern}`);for(const path of matches2)selected.add(path)}return[...selected].sort()}import{create as tar}from"tar";import{execFileSync}from"child_process";import{globSync,lstatSync,readdirSync as readdirSync2,readFileSync as readFileSync10,realpathSync as realpathSync3}from"fs";import{isAbsolute as isAbsolute3,join as join10,relative as relative4,resolve as resolve3,sep as sep2}from"path";import{minimatch as minimatch2}from"minimatch";import{parseDocument as parseDocument2}from"yaml";import{CommanderError}from"commander";import{ZodError}from"zod";function errorReport(error,command,context={}){const base=error instanceof CliError&&error.needsAccount&&context.local?localModeError(command,context.local,error.hint):describe(error);const quiet=base.message==="aborted"||command==="feedback";const isChat=command==="chat"||command.startsWith("chat ");const docs=base.docs??(isChat?void 0:DOCS_URL);return{...base,...docs?{docs}:{},...quiet?{}:{feedback:`gauge feedback -m "<what happened>" --kind friction${command?` --command "${command}"`:""}`}}}function localModeError(command,source,hint){return{code:"usage",message:`\`gauge ${command}\` requires hosted Gauge, but ${source==="flag"?"you passed --local":source==="workspace"?"this project is set to run locally":source==="user"?"your saved setting is local":"the CLI is using local mode"}.`,hint:hint??"Pass --hosted for this command, or run `gauge init --hosted` to use hosted Gauge in this project. Sign in with `gauge auth login` if needed.",exitCode:exitCodeFor("usage")}}function describe(error){if(error instanceof CliError)return{code:error.apiCode??error.code,message:readable(error.message),...error.details?{details:error.details}:{},hint:error.hint,docs:error.docs,exitCode:error.exitCode};if(error instanceof CommanderError)return{code:"usage",message:error.message.replace(/^error: /,""),hint:"Run the command with --help for its usage.",exitCode:exitCodeFor("usage")};if(error instanceof ZodError)return{code:"usage",message:issueLines("Invalid input",error.issues),exitCode:exitCodeFor("usage")};const node=error;if(node?.code==="ENOENT"&&node.path)return{code:"usage",message:`File not found: ${node.path}`,exitCode:exitCodeFor("usage")};if(node?.code==="EACCES"&&node.path)return{code:"usage",message:`Permission denied: ${node.path}`,exitCode:exitCodeFor("usage")};return{code:"internal",message:`Unexpected error: ${error instanceof Error?readable(error.message):String(error)}`,hint:process.env.GAUGE_DEBUG?void 0:"Set GAUGE_DEBUG=1 to print the stack trace.",exitCode:1}}function readable(message2){const start=message2.search(/\[\s*\{/);if(start<0)return message2;try{const issues=JSON.parse(message2.slice(start));if(!Array.isArray(issues)||!issues.every(issue=>typeof issue?.message==="string"))return message2;return issueLines(message2.slice(0,start).replace(/[:\s]+$/,""),issues)}catch{return message2}}function issueLines(prefix,issues){return[prefix,...issues.map(issue=>` ${issue.path?.length?`${issue.path.map(String).join(".")}: `:""}${issue.message}`)].filter(Boolean).join("\n")}function renderError(report,json){if(json)return`${JSON.stringify({error:report},null,2)}
|
|
952
|
+
Default org set to ${account.organization.slug}. You're ready:`);console.log(" gauge status # the org at a glance");console.log(" gauge eval list # eval sets"," gauge preference list # preference prompts");console.log(" gauge run --eval <id> # launch one now");let inRepo=false;try{inRepo=gaugeState(command).inRepo}catch{}console.log(inRepo?" gauge init --hosted # connect this repo":" gauge init --hosted # connect a repo (run it inside the repo)")})}var optimizations_exports={};__export(optimizations_exports,{fileEditFromOpts:()=>fileEditFromOpts,fileSourceFromOpts:()=>fileSourceFromOpts,latestLaunch:()=>latestLaunch,launchSettled:()=>launchSettled,parseEdits:()=>parseEdits,parseRestoreBefore:()=>parseRestoreBefore,parseStatus:()=>parseStatus,parseSubjectRef:()=>parseSubjectRef,register:()=>register23});import{readFileSync as readFileSync9,writeFileSync as writeFileSync2}from"fs";var STATUSES2=["OPEN","DONE","ARCHIVED"];var TERMINAL_RUN=new Set(["SUCCEEDED","FAILED","TIMED_OUT","CANCELED"]);function basePath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/optimizations`}function itemPath(org,id){return`${basePath(org)}/${encodeURIComponent(id)}`}function readText(path){return readFileSync9(path==="-"?0:path,"utf8")}function readJson2(path){try{return JSON.parse(readText(path))}catch{throw new CliError("usage",`${path} must contain valid JSON`)}}function sleep3(ms){return new Promise(resolve5=>setTimeout(resolve5,ms))}function plural(n,noun){return`${n} ${noun}${n===1?"":"s"}`}function parseCount(raw,flag,max){if(raw===void 0)return void 0;const n=Number(raw);if(!Number.isInteger(n)||n<1||n>max)throw new CliError("usage",`${flag} must be a positive integer no greater than ${max}`);return n}function parseStatus(raw){if(!raw)return void 0;const status=raw.trim().toUpperCase();if(!STATUSES2.includes(status))throw new CliError("usage",`unknown status "${raw}" (expected ${STATUSES2.join(", ")})`);return status}function parseSubjectRef(opts){if([opts.issue,opts.finding,opts.goal].filter(Boolean).length!==1)throw new CliError("usage","pass one of --issue <id>, --finding <id> --run <runId>, or --goal <text>");if(opts.run&&!opts.finding)throw new CliError("usage","--run only applies with --finding");if(opts.issue)return{kind:"issue",issueId:opts.issue};if(opts.finding){if(!opts.run)throw new CliError("usage","--finding needs --run <runId>");return{kind:"finding",observationId:opts.finding,runId:opts.run}}return{kind:"free",text:opts.goal}}function fileEditFromOpts(opts,read2=readText){if([opts.page,opts.skill,opts.cli].filter(Boolean).length!==1)throw new CliError("usage","pass one of --page <url>, --skill <name> --path <path>, or --cli <package>");if(opts.delete&&!opts.skill)throw new CliError("usage","--delete removes one path from a skill");if(opts.path&&!opts.skill)throw new CliError("usage","--path only applies with --skill");if(opts.contentType&&!opts.page)throw new CliError("usage","--content-type only applies with --page");if(opts.delete&&opts.from)throw new CliError("usage","pass --delete or --from, not both");if(!opts.delete&&!opts.from)throw new CliError("usage","--from <file> is required ('-' for stdin)");const from=opts.from;if(opts.page)return{kind:"page",url:opts.page,body:read2(from),...opts.contentType?{contentType:opts.contentType}:{}};if(opts.skill){if(!opts.path)throw new CliError("usage","--skill needs --path <path inside the skill>");return{kind:"skill",name:opts.skill,path:opts.path,body:opts.delete?null:read2(from)}}let rules;try{rules=JSON.parse(read2(from))}catch{throw new CliError("usage","--from must contain CLI rules as JSON")}return{kind:"cli",package:opts.cli,rules}}function fileSourceFromOpts(opts){if([opts.page,opts.skill,opts.cli].filter(Boolean).length!==1)throw new CliError("usage","pass one of --page <url>, --skill <name>, or --cli");if(opts.page)return{kind:"page",url:opts.page};if(opts.skill)return{kind:"skill",name:opts.skill};return{kind:"cli"}}function latestLaunch(launches){return launches.reduce((latest,launch2)=>!latest||launch2.number>latest.number?launch2:latest,void 0)}function launchSettled(launch2){return launch2.runs.every(run=>TERMINAL_RUN.has(run.status)&&(run.status!=="SUCCEEDED"||run.analyzed))}function subjectLine(subject){switch(subject.kind){case"issue":return`issue "${subject.title}" (${subject.issueId})`;case"finding":return`finding "${subject.label}" on run ${subject.runId}`;case"free":return`goal "${subject.text}"`}}function toRow2(o){return{id:o.id,title:o.title.length>48?`${o.title.slice(0,45)}...`:o.title,status:o.status,subject:o.subject.kind,changed:o.changed,launches:o.launches,updated:o.updatedAt}}function fileRow(key,file,change){return{key,change:change??"",size:file.kind==="page"?`${file.body.length} chars`:file.kind==="skill"?plural(Object.keys(file.files).length,"file"):plural(file.rules.rules.length,"rule")}}function launchProgress(launch2){const done2=launch2.runs.filter(run=>TERMINAL_RUN.has(run.status));const analyzed=launch2.runs.filter(run=>run.analyzed);return`${done2.length}/${launch2.runs.length} finished, ${analyzed.length} analyzed`}function launchRow(launch2){return{id:launch2.id,number:launch2.number,reruns:launchProgress(launch2),changed:launch2.changed.length,note:launch2.note??"",created:launch2.createdAt}}function printDetail2(o){console.log(`id: ${o.id}`);console.log(`title: ${o.title}`);console.log(`status: ${o.status}`);console.log(`subject: ${subjectLine(o.subject)}`);console.log(`sources: ${o.sourceRunIds.join(", ")||"(none)"}`);const changes=new Map(o.changes.map(c=>[c.key,c.status]));const files=Object.entries(o.head.files);console.log(`files: ${files.length?"":"(empty tree)"}`);if(files.length)printItems(files.map(([key,file])=>fileRow(key,file,changes.get(key))),"table");for(const[key,change]of changes)if(change==="removed")console.log(` ${key}: removed`);console.log(`launches: ${o.launchList.length?"":"(none)"}`);if(o.launchList.length)printItems(o.launchList.map(launchRow),"table");if(o.result)console.log(`result: ${o.result.summary}`)}function printFile(file,path){if(file.kind==="page")return void console.log(file.body);if(file.kind==="cli")return printJson(file.rules);if(path){const body=file.files[path];if(body===void 0)throw new CliError("not_found",`no ${path} in skill ${file.name}`);return void console.log(body)}for(const[name,body]of Object.entries(file.files)){console.log(`=== ${name}`);console.log(body)}}function printLaunch(launch2){console.log(`launch ${launch2.number} (${launch2.id})${launch2.note?`: ${launch2.note}`:""}`);console.log(`changed: ${launch2.changed.join(", ")||"(nothing)"}`);for(const run of launch2.runs){console.log(`
|
|
953
|
+
${run.id} rerun of ${run.sourceRunId??"?"}: ${run.status}, ${run.agent} ${run.model}`);if(run.restoredFrom)console.log(` restored at seq ${run.restoredFrom.seq} (${run.restoredFrom.checkpointId}); cost and time cover only the part after the restore`);if(!run.findings){console.log(" not analyzed");continue}const now=new Set(run.findings.map(f=>f.label));const before=new Set(run.sourceFindings.map(f=>f.label));for(const f of run.sourceFindings)console.log(` ${now.has(f.label)?"returned":"resolved"} ${f.label}`);for(const f of run.findings)if(!before.has(f.label))console.log(` introduced ${f.label}`)}}function parseEdits(value){const ok=Array.isArray(value)&&value.length>0&&value.every(edit=>edit&&typeof edit==="object"&&typeof edit.find==="string"&&edit.find.length>0&&typeof edit.replace==="string");if(!ok)throw new CliError("usage","--edits must be a JSON array of {find, replace} with non-empty find");return value.map(({find,replace})=>({find,replace}))}function printWrite(res){if(!res.change){console.log(`Wrote ${res.key}`);return}const{linesBefore,linesAfter,retainedPercent,hunks}=res.change;console.log(`Wrote ${res.key}: ${linesBefore} \u2192 ${linesAfter} lines, ${retainedPercent??"\u2014"}% of the original kept`);console.log(hunks)}function printPageView(view){if(view.text!==void 0){console.log(view.text);return}console.log(`${view.matches??0} matches for ${JSON.stringify(view.query)}`);for(const excerpt of view.excerpts??[]){console.log(`=== offset ${excerpt.offset}`);console.log(excerpt.raw)}if(view.truncated)console.log("\u2026 more matches; search a longer phrase");if(view.hint)console.log(view.hint)}function parseRestoreBefore(raw){const at=raw.lastIndexOf(":");const rerunId=raw.slice(0,at);const seq=raw.slice(at+1);if(at<=0||!/^-?\d+$/.test(seq))throw new CliError("usage",`--restore-before takes <rerunId>:<seq>, e.g. run_1:140 (got ${raw})`);return{rerunId,seq:Number(seq)}}function checkpointRow(cp){const exposure=[...cp.exposure.surfaces,...cp.exposure.skillChanged?["skill changed"]:[],...cp.exposureIncomplete?["trace incomplete"]:[]];return{checkpoint:cp.checkpointId,source:cp.sourceRunId,rerun:cp.rerunId,seq:`${cp.seq}/${cp.rerunEventCount}`,exposure:exposure.join(", ")||"clean","last step":cp.lastSteps.at(-1)??"",captured:cp.capturedAt}}function printLaunchResult(id,res){console.log(`Launched ${plural(res.batches.flatMap(b=>b.runIds).length,"rerun")} (${res.launchId}) with ${plural(res.changed.length,"changed file")}:`);for(const key of res.changed)console.log(` ${key}`);for(const batch of res.batches){const from=batch.restoredFrom;console.log(from?`${batch.sourceRunId}: restored at seq ${from.seq} of rerun ${from.rerunId} (${from.checkpointId})`:`${batch.sourceRunId}: fresh`);for(const warning of batch.warnings)console.log(` ! ${warning}`)}console.log(`Follow with \`gauge optimizations watch ${id}\`.`)}function register23(program2){const group=hostedOnly(program2.command("optimizations")).description("Hill climb your agent surfaces: edit the pages, skills, and CLI rules agent sessions used, rerun those sessions, and compare findings");group.command("list").description("List the org's optimizations, newest activity first").option("--status <status>",STATUSES2.join(", ")).option("--limit <n>","max rows (default 50)").action(async(opts,cmd)=>{const res=await createClient().get(basePath(await resolveOrg(cmd)),{status:parseStatus(opts.status),limit:parseCount(opts.limit,"--limit",200)});if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(toRow2),"table")});group.command("create").description("Open an optimization on an issue, one finding, or a goal; the tree starts empty").option("--issue <id>","an issue id from `gauge issues list`").option("--finding <id>","a finding id from `gauge run insight <run> -o json`; needs --run").option("--run <runId>","the run the finding is on").option("--goal <text>","free text; starts with no sessions, attach them with `sessions add`").option("--title <title>","name (default: from the subject)").action(async(opts,cmd)=>{const detail=await createClient().post(basePath(await resolveOrg(cmd)),{subject:parseSubjectRef(opts),...opts.title?{title:opts.title}:{}});if(resolveOutput(cmd)==="json")printJson(detail);else console.log(detail.sourceRunIds.length?`Opened ${detail.id} "${detail.title}" on ${plural(detail.sourceRunIds.length,"source session")}. Next: \`gauge optimizations authoring ${detail.id}\`.`:`Opened ${detail.id} "${detail.title}" with no sessions. Next: \`gauge optimizations sessions add ${detail.id} <runId...>\`.`)});group.command("get <id>").description("One optimization: subject, source runs, tree, and launches").action(async(id,_opts,cmd)=>{const detail=await createClient().get(itemPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail)});group.command("authoring <id>").description("Authoring guidance: the source sessions' findings, the tree, and the editing practices to follow").action(async(id,_opts,cmd)=>{const res=await createClient().get(`${itemPath(await resolveOrg(cmd),id)}/authoring`);if(resolveOutput(cmd)==="json")printJson(res);else console.log(res.guidance)});group.command("sessions").description("Attach sessions an optimization can rerun").command("add <id> <runIds...>").description("Attach sessions to rerun, up to 12 in all; a goal starts with none. Optimization reruns cannot be sources").action(async(id,runIds,_opts,cmd)=>{const res=await createClient().post(`${itemPath(await resolveOrg(cmd),id)}/sessions`,{runIds});if(resolveOutput(cmd)==="json")printJson(res);else console.log(`Attached ${plural(res.added.length,"session")}; ${id} has ${plural(res.sourceRunIds.length,"source session")}.`)});const file=group.command("file").description("Read and edit the files in an optimization's tree");file.command("read <id> <key>").description("Print one file by tree key: a page's exact source (--text for readable text, --query for the source around a match), a skill (or one --path in it), or CLI rules").option("--text","page only: readable text instead of the source").option("--query <text>","page only: the exact source around each match, to copy into file edit").option("--path <path>","one file inside a skill").option("--out <file>","save the page source or skill --path to a file").action(async(id,key,opts,cmd)=>{const path=`${itemPath(await resolveOrg(cmd),id)}/files`;if(opts.text||opts.query){if(opts.out)throw new CliError("usage","--out saves the stored file only");const view=await createClient().get(path,{key,...opts.query?{query:opts.query}:{view:"text"}});if(resolveOutput(cmd)==="json")printJson(view);else printPageView(view);return}const res=await createClient().get(path,{key});if(opts.out){const body=res.kind==="page"?res.body:res.kind==="skill"&&opts.path?res.files[opts.path]:void 0;if(body===void 0)throw new CliError("usage","--out saves a page source or one skill --path");writeFileSync2(opts.out,body);console.error(`Saved ${opts.out}`)}else if(resolveOutput(cmd)==="json")printJson(res);else printFile(res,opts.path)});file.command("edit <id>").description("Edit a page in the tree with find/replace on its exact source; every byte outside the edits stays. Each find must appear once; all edits apply or none do").requiredOption("--page <url>","the page URL").requiredOption("--edits <file>","JSON [{find, replace}]; copy find text from `file read --query`; '-' for stdin").action(async(id,opts,cmd)=>{const res=await createClient().patch(`${itemPath(await resolveOrg(cmd),id)}/files`,{url:opts.page,edits:parseEdits(readJson2(opts.edits))});if(resolveOutput(cmd)==="json")printJson(res);else printWrite(res)});file.command("add <id>").description("Copy an original into the tree before you edit it: a page a source session read (else the live page), a staged skill, or the CLI rules").option("--page <url>","a page by URL").option("--skill <name>","a skill a source session staged").option("--cli","the CLI rules a source session ran with").action(async(id,opts,cmd)=>{const res=await createClient().post(`${itemPath(await resolveOrg(cmd),id)}/files`,fileSourceFromOpts(opts));if(resolveOutput(cmd)==="json")printJson(res);else console.log(`Added ${res.key} (from ${res.origin})`)});file.command("write <id>").description("Write a new page or rewrite a whole one, one skill file, or the CLI rules; to change part of a page, use file edit").option("--page <url>","the page URL; --from holds the complete body").option("--content-type <type>","the page content type (default: keep)").option("--skill <name>","the skill; needs --path").option("--path <path>","the file inside the skill, e.g. SKILL.md").option("--delete","remove --path from the skill").option("--cli <package>","the CLI package; --from holds {schemaVersion: 3, target, rules, shippingIntent}").option("--from <file>","the new content; '-' for stdin").action(async(id,opts,cmd)=>{const res=await createClient().put(`${itemPath(await resolveOrg(cmd),id)}/files`,fileEditFromOpts(opts));if(resolveOutput(cmd)==="json")printJson(res);else printWrite(res)});file.command("rm <id> <key>").description("Drop a file from the tree; reruns use what the source session had").action(async(id,key,_opts,cmd)=>{const res=await createClient().delete(`${itemPath(await resolveOrg(cmd),id)}/files`,{key});if(resolveOutput(cmd)==="json")printJson(res);else console.log(`Removed ${res.removed}`)});group.command("notes <id>").description("Print the working notes, or replace them with --from: markdown the user reads beside the chat").option("--from <file>","the complete notes; '-' for stdin").action(async(id,opts,cmd)=>{const path=itemPath(await resolveOrg(cmd),id);const client=createClient();const detail=opts.from?await client.patch(path,{notes:readText(opts.from)}):await client.get(path);if(resolveOutput(cmd)==="json")printJson({notes:detail.notes});else if(opts.from)console.log(`Saved notes (${detail.notes.length} chars)`);else console.log(detail.notes||"(no notes)")});group.command("launch <id>").description("Rerun the source sessions with the tree injected; each rerun restores from the deepest checkpoint before any changed surface, or starts fresh").option("--run <runId...>","rerun only these source runs").option("--note <text>","one line on what this launch tests").option("--restore-before <rerunId:seq...>","restore from the deepest checkpoint at or before this seq of an earlier rerun").option("--checkpoint <checkpointId...>","restore from this checkpoint (see `checkpoints`)").option("--fresh [sourceRunId...]","start from zero: these source runs, or every one").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{const path=itemPath(await resolveOrg(cmd),id);await confirmOrAbort(`Rerun ${opts.run?.length?plural(opts.run.length,"source session"):"every source session"} of ${id} with the tree?`,opts.yes);const client=createClient();const restoreAt=[...(opts.checkpoint??[]).map(checkpointId=>({checkpointId})),...(opts.restoreBefore??[]).map(parseRestoreBefore)];const fresh=opts.fresh===true?opts.run??(await client.get(path)).sourceRunIds:opts.fresh??[];const res=await client.post(`${path}/launch`,{...opts.run?.length?{runIds:opts.run}:{},...opts.note?{note:opts.note}:{},...fresh.length?{restore:Object.fromEntries(fresh.map(runId=>[runId,"none"]))}:{},...restoreAt.length?{restoreAt}:{}});if(resolveOutput(cmd)==="json")printJson(res);else printLaunchResult(id,res)});group.command("checkpoints <id>").description("Checkpoints captured on fresh reruns, with how far in each is and whether its prefix saw a changed surface").option("--run <sourceRunId>","only this source run").action(async(id,opts,cmd)=>{const res=await createClient().get(`${itemPath(await resolveOrg(cmd),id)}/checkpoints`,{sourceRunId:opts.run});if(resolveOutput(cmd)==="json")printJson(res.items);else if(!res.items.length)console.log("(no checkpoints)");else printItems(res.items.map(checkpointRow),"table")});group.command("launches <id> [launchId]").description("List launches, or one launch's reruns with their findings beside the source findings").action(async(id,launchId,_opts,cmd)=>{const path=itemPath(await resolveOrg(cmd),id);const client=createClient();if(!launchId){const detail=await client.get(path);if(resolveOutput(cmd)==="json")printJson(detail.launchList);else printItems(detail.launchList.map(launchRow),"table");return}const launch2=await client.get(`${path}/launches/${encodeURIComponent(launchId)}`);if(resolveOutput(cmd)==="json")printJson(launch2);else printLaunch(launch2)});group.command("watch <id>").description("Wait until the latest launch's reruns finish and are analyzed, then print them").option("--interval <seconds>","poll interval (default 15)").option("--timeout <minutes>","give up after this long (default 60)").action(async(id,opts,cmd)=>{const path=itemPath(await resolveOrg(cmd),id);const client=createClient();const interval=(parseCount(opts.interval,"--interval",600)??15)*1e3;const timeoutMinutes=parseCount(opts.timeout,"--timeout",1440)??60;const deadline=Date.now()+timeoutMinutes*6e4;let last="";for(;;){const detail=await client.get(path);const latest=latestLaunch(detail.launchList);if(!latest)throw new CliError("usage",`${id} has no launches yet`);if(launchSettled(latest)){const launch2=await client.get(`${path}/launches/${encodeURIComponent(latest.id)}`);if(resolveOutput(cmd)==="json")printJson(launch2);else printLaunch(launch2);return}const progress=launchProgress(latest);if(progress!==last)console.error(`launch ${latest.number}: ${progress}`);last=progress;if(Date.now()>=deadline)throw new CliError("network",`Timed out after ${timeoutMinutes} minutes (launch ${latest.number}: ${progress}). Re-run watch, or check \`gauge optimizations launches ${id} ${latest.id}\`.`);await sleep3(interval)}});group.command("finish <id>").description("Close the optimization with its result: summary, changed keys, and findings resolved, returned, and introduced").requiredOption("--from <file>","result JSON: {summary, changed, resolved, returned, introduced}").option("--yes","skip the confirmation").action(async(id,opts,cmd)=>{const path=itemPath(await resolveOrg(cmd),id);const result=readJson2(opts.from);await confirmOrAbort(`Finish ${id}? Its tree can no longer change.`,opts.yes);const detail=await createClient().post(`${path}/finish`,result);if(resolveOutput(cmd)==="json")printJson(detail);else console.log(`Finished ${detail.id} "${detail.title}"`)})}var org_exports={};__export(org_exports,{register:()=>register24});function printSettings(s,billingStatus){console.log(`slug: ${s.slug}`);console.log(`name: ${s.name}`);console.log(`website: ${s.website??"(none)"}`);console.log(`max spend/run: ${s.maxSpendUsd==null?"(server default)":`$${s.maxSpendUsd}`}`);console.log(`overage: ${fmtOverage(billingStatus)}`);console.log(`auto-join: ${s.autoJoinDomain??"off"}`)}function register24(program2){const group=hostedOnly(program2.command("org")).description("This org's settings, read-only (see `gauge orgs` for the org list)");group.command("get").description("Show the org's settings (edit them in the web app)").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const base=`/api/v1/orgs/${encodeURIComponent(org)}`;const[settings,billingStatus]=await Promise.all([client.get(base),client.get(`${base}/billing`)]);if(resolveOutput(cmd)==="json")printJson({...settings,paymentType:billingStatus.paymentType??null,creditLine:billingStatus.creditLine??null});else printSettings(settings,billingStatus)})}var orgs_exports={};__export(orgs_exports,{register:()=>register25});import{writeFileSync as writeFileSync3}from"fs";import{join as join8}from"path";function register25(program2){const orgs=hostedOnly(program2.command("orgs"),{account:true}).description("Organizations, and which Gauge products each one uses");orgs.command("list",{isDefault:true}).description("List your Gauge organizations (* marks the saved default)").action(async(_opts,command)=>{const rows=await fetchDirectory();const{orgId}=readConfig();if(resolveOutput(command)==="json"){printJson(rows.map(row2=>({...row2,isDefault:row2.id===orgId})));return}printItems(rows.map(row2=>({default:row2.id===orgId?"*":"",slug:row2.slug,name:row2.name,products:productsLabel(row2.products)})),"table")});orgs.command("create <name>").description("Create a Gauge Agents workspace and set its organization as the default").option("--website <url>","company site (used as the logo source)").option("--no-use","don't set the new org as the default").action(async(name,opts,command)=>{await fetchDirectory();const created=await createClient().post("/api/v1/orgs",{name,website:opts.website});const org=await createdOrganization(created.slug);if(opts.use!==false)updateConfig({orgId:org.id,defaultOrg:void 0});if(resolveOutput(command)==="json")return printJson(org);console.log(`Created workspace ${org.name} (${org.slug})`);if(opts.use!==false)console.log(`Default org set to ${org.slug}`)});orgs.command("use <org>").description("Select a Gauge organization by its public slug").option("--project","save the selection in this project's gauge.json").action(async(value,opts,command)=>{const org=await findOrganization(value);if(opts.project){const state=gaugeState(command);if(!state.root||!state.config)throw new CliError("usage","No valid gauge.json found. Run `gauge init` first.");const{org:_legacyOrg,...config}=state.config;writeFileSync3(join8(state.root,"gauge.json"),`${JSON.stringify({...config,orgId:org.id},null,2)}
|
|
954
|
+
`)}else updateConfig({orgId:org.id,defaultOrg:void 0});if(resolveOutput(command)==="json")return printJson({orgId:org.id,slug:org.slug,name:org.name});console.log(`${opts.project?"Project":"Default"} org set to ${org.name} (${org.slug})`)})}var personas_exports={};__export(personas_exports,{register:()=>register26});function personasPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/personas`}function register26(program2){const group=hostedOnly(program2.command("personas")).description("Manage simulated-user personas for eval and preference runs");group.command("list").description("List the org's personas").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(personasPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(p=>({id:p.id,name:p.name,description:p.description.length>60?`${p.description.slice(0,57)}...`:p.description})),"table")});group.command("get <id>").description("Show one persona").action(async(id,_opts,cmd)=>{const org=await resolveOrg(cmd);const persona=await createClient().get(`${personasPath(org)}/${encodeURIComponent(id)}`);if(resolveOutput(cmd)==="json"){printJson(persona);return}console.log(`id: ${persona.id}`);console.log(`name: ${persona.name}`);console.log(`model: ${persona.model}`);console.log(`description: ${persona.description}`)});group.command("add").description("Create a persona").requiredOption("--name <name>","persona name").requiredOption("--description <text>",'memory-style notes ("The user is a solo founder\u2026")').action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const created=await createClient().post(personasPath(org),{name:opts.name,description:opts.description});if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created persona ${created.id} (${created.name})`)});group.command("edit <id>").description("Update a persona").option("--name <name>","rename the persona").option("--description <text>","replace the notes").action(async(id,opts,cmd)=>{const org=await resolveOrg(cmd);if(opts.name===void 0&&opts.description===void 0)throw new CliError("usage","nothing to change");const updated=await createClient().patch(`${personasPath(org)}/${encodeURIComponent(id)}`,{name:opts.name,description:opts.description});if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated persona ${updated.id}`)});group.command("rm <id>").description("Delete a persona (run configurations fall back to the default persona)").action(async(id,_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(`${personasPath(org)}/${encodeURIComponent(id)}`);if(resolveOutput(cmd)!=="json")console.log(`Deleted persona ${id}`)})}var preference_exports={};__export(preference_exports,{register:()=>register27});function visibilityPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/visibility-prompts`}function promptPath(org,id){return`${visibilityPath(org)}/${encodeURIComponent(id)}`}function scenariosPath(org,id){return`${promptPath(org,id)}/scenarios`}function scenarioRow(scenario){return{id:scenario.id,repo:scenario.repoUrl??"(empty project)",ref:scenario.repoRef??"",persona:scenario.profile?.name??"(default)"}}var KIND_CHOICES=["head-to-head","open-ended"];function parseKind(raw){const upper=raw==="open-ended"?"RECOMMEND":raw.trim().toUpperCase().replace(/-/g,"_");if(upper!=="HEAD_TO_HEAD"&&upper!=="RECOMMEND")throw new CliError("usage",`invalid --kind "${raw}" (expected: ${KIND_CHOICES.join(" | ")})`);return upper}function brandsLabel(v){if(v.kind==="HEAD_TO_HEAD")return`${v.brandA} vs ${v.brandB}`;return v.branded?v.brandA??"branded":"organic"}function preferencePromptRow(v){return{id:v.id,name:v.name,kind:v.kind==="RECOMMEND"?"open-ended":"head-to-head",brands:brandsLabel(v),market:v.market?.slug??"",tags:v.tags.join(", "),runs:v.runCount,"last run":v.lastRunAt??"-","created by":creatorLabel(v.createdBy)}}function printDetail3(v){console.log(`id: ${v.id}`);console.log(`name: ${v.name}`);console.log(`kind: ${v.kind==="RECOMMEND"?"open-ended":"head-to-head"}`);console.log(`brands: ${brandsLabel(v)}`);console.log(`market: ${v.market?`${v.market.name} (${v.market.slug})`:"(none)"}`);console.log(`tags: ${v.tags.join(", ")||"(none)"}`);console.log(`runs: ${v.runCount} (last: ${v.lastRunAt??"never"})`);console.log(`creator: ${creatorLabel(v.createdBy)}`);console.log(`prompt: ${v.promptText}`);console.log(`cadence: ${cadenceLabel(v.cadence)} \u2014 ${v.sampleCount} sample(s)`);console.log(`agents: ${v.agents.map(a=>a.models.length?`${a.agent}:${a.models.join(",")}`:a.agent).join(", ")}`);console.log(`skills: ${v.skillRefs.join(", ")||"(none)"}`);console.log(`mcp: ${v.mcpRefs.join(", ")||"(none)"}`);console.log(`conns: ${v.connectionIds.join(", ")||"(none)"}`);console.log(`scenarios: ${v.scenarios.length}`);for(const scenario of v.scenarios)console.log(` [${scenario.id}] ${scenario.repoUrl??"(empty project)"}${scenario.repoRef?`@${scenario.repoRef}`:""} \u2014 ${scenario.profile?.name??"default persona"}`)}async function listPreferencePrompts(cmd){if(isLocal(cmd))return listLocalMeasurements(cmd,"preference");const res=await createClient().get(visibilityPath(await resolveOrg(cmd)));return res.items}async function getPreferencePrompt(cmd,id){const vp=isLocal(cmd)?localMeasurementView(getLocalMeasurement(cmd,"preference",id)):await createClient().get(promptPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson(vp);else printDetail3(vp)}async function createPreferencePrompt(cmd,opts){if(hasLifecycleFlags(opts)&&isLocal(cmd))throw new CliError("usage","Lifecycle scripts require hosted execution; pass --hosted");if(opts.kind===void 0)throw new CliError("usage",`--kind is required with --preference (${KIND_CHOICES.join(" | ")})`);if(opts.ref!==void 0&&opts.repo===void 0)throw new CliError("usage","--ref requires --repo");const hasScenario=opts.repo!==void 0||opts.persona!==void 0;const body={name:opts.name,promptText:readPromptText(opts),kind:parseKind(opts.kind),branded:opts.branded,brandA:opts.brandA??null,brandB:opts.brandB??null,marketId:opts.market??null,tags:opts.tag.length>0?opts.tag:void 0,...hasRunSettingsFlags(opts)?runSettingsFromOpts(opts):{},scenarios:hasScenario?[{repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:opts.persona??null}]:[]};const created=isLocal(cmd)?createLocalMeasurement(cmd,"preference",body):await createClient().post(visibilityPath(await resolveOrg(cmd)),body);if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Created preference prompt ${created.id}`)}async function editPreferencePrompt(cmd,id,opts){if(hasLifecycleFlags(opts)&&isLocal(cmd))throw new CliError("usage","Lifecycle scripts require hosted execution; pass --hosted");const lifecyclePath=hasLifecycleFlags(opts)&&!opts.clearLifecycle?promptPath(await resolveOrg(cmd),id):void 0;const existing=lifecyclePath?await createClient().get(lifecyclePath):void 0;const body={...runSettingsPatchFromOpts(opts,existing?.lifecycle)};if(opts.name!==void 0)body.name=opts.name;if(opts.text!==void 0||opts.file!==void 0)body.promptText=readPromptText(opts);if(opts.brandA!==void 0)body.brandA=opts.brandA;if(opts.brandB!==void 0)body.brandB=opts.brandB;if(opts.branded&&opts.organic)throw new CliError("usage","--branded and --organic conflict");if(opts.branded)body.branded=true;else if(opts.organic)body.branded=false;if(opts.market!==void 0&&opts.clearMarket)throw new CliError("usage","--market and --clear-market conflict");if(opts.clearMarket)body.marketId=null;else if(opts.market!==void 0)body.marketId=opts.market;if(opts.tag.length>0&&opts.clearTags)throw new CliError("usage","--tag and --clear-tags conflict");if(opts.clearTags)body.tags=[];else if(opts.tag.length>0)body.tags=opts.tag;if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");const updated=isLocal(cmd)?editLocalMeasurement(cmd,"preference",id,body):await createClient().patch(lifecyclePath??promptPath(await resolveOrg(cmd),id),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated preference prompt ${updated.id}`)}async function removePreferencePrompt(cmd,id){if(isLocal(cmd))removeLocalMeasurement(cmd,"preference",id);else await createClient().delete(promptPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)!=="json")console.log(`Deleted preference prompt ${id}`)}async function listScenarios(cmd,id){const vp=isLocal(cmd)?localMeasurementView(getLocalMeasurement(cmd,"preference",id)):await createClient().get(promptPath(await resolveOrg(cmd),id));if(resolveOutput(cmd)==="json")printJson(vp.scenarios);else printItems(vp.scenarios.map(scenarioRow),"table")}async function addScenario(cmd,id,opts){if(opts.ref!==void 0&&opts.repo===void 0)throw new CliError("usage","--ref requires --repo");if(isLocal(cmd)){if(opts.persona)throw new CliError("usage","Personas require hosted Gauge");const row2=getLocalMeasurement(cmd,"preference",id);const updated=editLocalMeasurement(cmd,"preference",id,{scenarios:[...row2.scenarios,{repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:null}]});const scenario=updated.scenarios.at(-1);if(resolveOutput(cmd)==="json")printJson(scenario);else console.log(`Added scenario ${scenario.id}`);return}const created=await createClient().post(scenariosPath(await resolveOrg(cmd),id),{repoUrl:opts.repo??null,repoRef:opts.ref??null,profileId:opts.persona??null});if(resolveOutput(cmd)==="json")printJson(created);else console.log(`Added scenario ${created.id}`)}async function removeScenario(cmd,id,scenarioId){if(isLocal(cmd)){const row2=getLocalMeasurement(cmd,"preference",id);const scenarios=row2.scenarios;if(!scenarios.some(s=>s.id===scenarioId))throw new CliError("usage",`Unknown scenario: ${scenarioId}`);editLocalMeasurement(cmd,"preference",id,{scenarios:scenarios.filter(s=>s.id!==scenarioId)});if(resolveOutput(cmd)!=="json")console.log(`Removed scenario ${scenarioId}`);return}await createClient().delete(`${scenariosPath(await resolveOrg(cmd),id)}/${encodeURIComponent(scenarioId)}`);if(resolveOutput(cmd)!=="json")console.log(`Removed scenario ${scenarioId}`)}async function runPreferencePrompt(cmd,id,opts){const org=await resolveOrg(cmd);const overrides={...runOverridesFromOpts(opts),...hasLifecycleFlags(opts)?lifecyclePatchFromOpts(opts,(await createClient().get(promptPath(org,id))).lifecycle):{}};const settings=Object.keys(overrides).length>0?"these overrides (stored settings stay unchanged)":"the stored settings";await confirmOrAbort(opts.scenario.length>0?`Run ${opts.scenario.length} selected scenario${opts.scenario.length===1?"":"s"} with ${settings} now?`:`Run every scenario of this prompt with ${settings} now?`,opts.yes);const result=await createClient().post(`${promptPath(org,id)}/run`,{visibilityScenarioIds:opts.scenario.length>0?opts.scenario:void 0,...overrides});if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`Started ${result.runIds.length} run${result.runIds.length===1?"":"s"} across ${result.batchIds.length} scenario${result.batchIds.length===1?"":"s"}`);for(const batch of result.batchIds)console.log(`Runs: gauge run list --batch ${batch}`)}return result}async function listPreferencePromptRuns(cmd,id,opts){const res=await createClient().get(`${promptPath(await resolveOrg(cmd),id)}/runs`,{limit:opts.limit,cursor:opts.cursor});if(resolveOutput(cmd)==="json"){printJson(res);return}printItems(res.items.map(r=>({id:r.id,status:r.status,agent:r.agent,model:r.resolvedModel??r.model??"",source:r.triggeredBy?.source??"",credits:r.creditsCharged??"",created:r.createdAt,verdict:r.verdict?.status??"-"})),"table");if(res.nextCursor)console.error(`(more results \u2014 rerun with --cursor ${res.nextCursor})`)}async function markets(cmd,opts){if(isLocal(cmd))throw new CliError("usage","Markets require hosted Gauge. Pass --hosted.");const client=createClient();const base=`/api/v1/orgs/${encodeURIComponent(await resolveOrg(cmd))}/markets`;if(opts.add!==void 0){const market=await client.post(base,{marketId:opts.add});if(resolveOutput(cmd)==="json")printJson(market);else console.log(`Tracking market ${market.name} (${market.slug})`);return}const res=await client.get(base);if(resolveOutput(cmd)==="json"){printJson(res);return}printItems([...res.tracked.map(m=>({id:m.marketId,slug:m.slug,name:m.name,tracked:"yes",source:m.source})),...res.addable.map(m=>({id:m.marketId,slug:m.slug,name:m.name,tracked:"no",source:""}))],"table")}function register27(program2){const preference=program2.command("preference").description("Manage preference prompts: which product agents choose. Launch one with `gauge run --preference <id>`");preference.command("list").description("List preference prompts").option("--created-by <email|me>","only prompts created by this user").action(async(opts,cmd)=>{const items=await filterByCreator(await listPreferencePrompts(cmd),opts.createdBy);if(resolveOutput(cmd)==="json")printJson(items);else printItems(items.map(preferencePromptRow),"table")});preference.command("get <id>").description("Show one preference prompt").action(async(id,_opts,cmd)=>{await getPreferencePrompt(cmd,id)});addRunSettingsCreateOptions(preference.command("create").description("Create a preference prompt").requiredOption("--name <name>","name").option("--text <text>","prompt text inline").option("--file <path>",'prompt text from a file ("-" for stdin)').option("--repo <url>","the first scenario's repository (default: empty project)").option("--ref <ref>","branch, tag, or commit (with --repo)").option("--persona <profileId>","persona profile id")).option("--kind <kind>",KIND_CHOICES.join(" | ")).option("--brand-a <name>","first brand (head-to-head)").option("--brand-b <name>","second brand (head-to-head)").option("--branded","an open-ended prompt that names a brand").option("--market <marketId>","tracked market id (see `gauge preference markets`)").option("--tag <name>","user tag name, create-on-type (repeatable)",collect5,[]).action(async(opts,cmd)=>{await createPreferencePrompt(cmd,opts)});addRunSettingsEditOptions(preference.command("edit <id>").description("Update a preference prompt. Omitted flags keep their value.").option("--name <name>","rename it").option("--text <text>","replace the prompt text inline").option("--file <path>",'replace the prompt text from a file ("-" stdin)')).option("--brand-a <name>","first brand").option("--brand-b <name>","second brand").option("--branded","mark branded (open-ended only)").option("--organic","mark organic (open-ended only)").option("--market <marketId>","set the measured market").option("--clear-market","unscope from any market").option("--tag <name>","replace the user tag set (repeatable)",collect5,[]).option("--clear-tags","remove all user tags").action(async(id,opts,cmd)=>{await editPreferencePrompt(cmd,id,opts)});preference.command("rm <id>").description("Delete a preference prompt (run history stays)").action(async(id,_opts,cmd)=>{await removePreferencePrompt(cmd,id)});const scenarios=preference.command("scenarios").description("The repo \xD7 persona scenarios a preference prompt runs in");scenarios.command("list <id>").description("List a preference prompt's scenarios").action(async(id,_opts,cmd)=>{await listScenarios(cmd,id)});scenarios.command("add <id>").description("Add one repo \xD7 persona scenario").option("--repo <url>","repository URL (default: empty project)").option("--ref <ref>","branch, tag, or commit (with --repo)").option("--persona <profileId>","persona profile id (default persona if omitted)").action(async(id,opts,cmd)=>{await addScenario(cmd,id,opts)});scenarios.command("rm <id>").description("Remove one scenario").requiredOption("--scenario <scenarioId>","scenario id to remove").action(async(id,opts,cmd)=>{await removeScenario(cmd,id,opts.scenario)});preference.command("markets").description("List tracked and addable markets for preference prompts, or track one with --add").option("--add <marketId>","track a published market").action(async(opts,cmd)=>{await markets(cmd,opts)})}var repos_exports={};__export(repos_exports,{register:()=>register28});function reposPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/repos`}async function fetchAllRepos(client,org){return fetchAllPages(client,reposPath(org))}function toRow3(repo){return{name:repo.name,url:repo.url,languages:repo.languages.join(", "),frameworks:repo.frameworks.join(", "),tags:repo.tags.join(", "),size:repo.sizeBucket??"",stars:repo.stars,system:repo.isSystem?"yes":""}}function syncWarning(repo){return`warning: GitHub metadata sync failed; retry with \`gauge repos resync ${repo.url}\``}function register28(program2){const repos=hostedOnly(program2.command("repos")).description("Manage repositories");repos.command("list").description("List repositories (yours + read-only Gauge-provided ones)").action(async(_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);const items=await fetchAllRepos(createClient(),org);if(output==="json"){printJson(items);return}printItems(items.map(toRow3),"table")});repos.command("add <url>").description("Add a repository (pulls languages/size/stars from GitHub)").action(async(url,_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);const result=await createClient().post(reposPath(org),{url});if(output==="json"){printJson(result);return}console.log(`Added ${result.repo.name} (${result.repo.url})`);if(!result.synced)console.error(syncWarning(result.repo))});repos.command("edit <url>").description("Update a repository's display name, frameworks, or tags").option("--name <name>","display name").option("--framework <name>","REPLACE the framework list (repeatable)",collect5,[]).option("--tag <name>","REPLACE the tag list (repeatable)",collect5,[]).option("--clear-frameworks","remove all frameworks").option("--clear-tags","remove all tags").action(async(url,opts,command)=>{const org=await resolveOrg(command);if(opts.framework.length>0&&opts.clearFrameworks)throw new CliError("usage","--framework and --clear-frameworks conflict");if(opts.tag.length>0&&opts.clearTags)throw new CliError("usage","--tag and --clear-tags conflict");const body={url};if(opts.name!==void 0)body.name=opts.name;if(opts.clearFrameworks)body.frameworks=[];else if(opts.framework.length>0)body.frameworks=opts.framework;if(opts.clearTags)body.tags=[];else if(opts.tag.length>0)body.tags=opts.tag;if(Object.keys(body).length===1)throw new CliError("usage","nothing to change");const repo=await createClient().patch(reposPath(org),body);if(resolveOutput(command)==="json")printJson(repo);else console.log(`Updated ${repo.name} (${repo.url})`)});repos.command("resync <url>").description("Re-fetch a repository's GitHub metadata").action(async(url,_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);const result=await createClient().post(`${reposPath(org)}/resync`,{url});if(output==="json"){printJson(result);return}if(!result.synced){throw new CliError("internal","GitHub metadata sync failed (is the URL a public GitHub repository?)")}console.log(`Re-synced ${result.repo.name} (${result.repo.url})`)});repos.command("rm <url>").description("Remove a repository from the catalog").action(async(url,_opts,command)=>{const org=await resolveOrg(command);const output=resolveOutput(command);await createClient().delete(reposPath(org),{url});if(output!=="json")console.log(`Removed ${url}`)})}var run_exports={};__export(run_exports,{register:()=>register29});import{readFileSync as readFileSync12,existsSync as existsSync4,mkdirSync as mkdirSync2,writeFileSync as writeFileSync4}from"fs";import{z as z14}from"zod";import{join as join11,relative as relative5}from"path";import{z as z7}from"zod";var preferenceDefinitionSchema=z7.discriminatedUnion("kind",[z7.object({kind:z7.literal("open-ended"),branded:z7.boolean().default(false)}).strict(),z7.object({kind:z7.literal("head-to-head"),brandA:z7.string().trim().min(1),brandB:z7.string().trim().min(1)}).strict().refine(p=>p.brandA.toLowerCase()!==p.brandB.toLowerCase(),"A head-to-head needs two different brands")]);var preferenceSnapshotSchema=z7.object({version:z7.literal(1),preference:preferenceDefinitionSchema}).strict();var headToHeadSchema=z7.object({kind:z7.literal("HEAD_TO_HEAD"),assessability:z7.enum(["assessed","partial","unassessable"]).optional(),selectionStatus:z7.enum(["measured","not_measured","unassessable"]).optional(),winner:z7.enum(["A","B","both","neither"]),brandA:z7.string(),brandB:z7.string(),rationale:z7.string(),brandAReasoning:z7.string(),brandBReasoning:z7.string(),brandAWhyRaised:z7.string().optional(),brandBWhyRaised:z7.string().optional()});var recommendPickSchema=z7.object({brand:z7.string(),rank:z7.number().int(),adopted:z7.boolean(),reasoning:z7.string().optional(),whyRaised:z7.string().optional()});var recommendSchema=z7.object({kind:z7.literal("RECOMMEND"),selectedBrands:z7.array(z7.string()).optional(),assessability:z7.enum(["assessed","partial","unassessable"]).optional(),selectionStatus:z7.enum(["measured","not_measured","unassessable"]).optional(),picks:z7.array(recommendPickSchema).default([]),singlePick:z7.boolean(),complianceNote:z7.string()});var visibilityVerdictSchema=z7.discriminatedUnion("kind",[headToHeadSchema,recommendSchema]);function preferenceSelectionLabel(verdict){if(verdict.assessability==="unassessable"||verdict.selectionStatus==="unassessable")return"Not assessable";if(verdict.kind==="HEAD_TO_HEAD")return verdict.winner==="A"?verdict.brandA:verdict.winner==="B"?verdict.brandB:verdict.winner==="both"?`${verdict.brandA} and ${verdict.brandB}`:"Neither";if(verdict.selectedBrands!==void 0)return verdict.selectedBrands.length?verdict.selectedBrands.join(", "):"No selection";if(!verdict.singlePick)return verdict.picks.length?"No clear selection":"No selection";return verdict.picks.find(pick=>pick.rank===1)?.brand??"No selection"}import{Option as Option5}from"commander";import{readFileSync as readFileSync10}from"fs";import{z as z8}from"zod";function assertRequiredExperiment(spec,required){if(required.length===0){throw new CliError("invalid_request","--require-experiment requires a nonempty label")}const mismatches=[];for(const[promptIndex,prompt]of spec.prompts.entries()){const promptExperiment=prompt.experiment??spec.experiment;const schedules=prompt.schedules??[{}];if(schedules.length===0){mismatches.push(`prompts.${promptIndex}.schedules`)}for(const[scheduleIndex,schedule]of schedules.entries()){if((schedule.experiment??promptExperiment)!==required){mismatches.push(`prompts.${promptIndex}.schedules.${scheduleIndex}`)}}}if(mismatches.length>0){throw new CliError("invalid_request",`--require-experiment expected every target to resolve to ${JSON.stringify(required)}; mismatched: ${mismatches.join(", ")}`)}}var EXAMPLE_SPEC={version:1,expand:true,experiment:"my-experiment",defaults:{agents:["CLAUDE_CODE"],samplesPerCycle:3},prompts:[{key:"add-healthcheck",text:"Add a /health endpoint that returns 200 OK."},{key:"migrate-to-vitest",text:"Migrate the test suite from Jest to Vitest.",schedules:[{repoUrl:"https://github.com/acme/api",repoRef:"main",model:"claude-opus-4-8"}]}]};async function readSpecJson(source){let raw;if(source==="-"){if(process.stdin.isTTY)throw new CliError("usage","--apply - needs a run spec piped to stdin");const chunks2=[];for await(const chunk of process.stdin)chunks2.push(chunk);raw=Buffer.concat(chunks2).toString("utf8")}else{try{raw=readFileSync10(source,"utf8")}catch(e){throw new CliError("usage",`cannot read spec file ${source}: ${e instanceof Error?e.message:String(e)}`)}}try{return JSON.parse(raw)}catch(e){throw new CliError("invalid_request",`spec is not valid JSON: ${e instanceof Error?e.message:String(e)}`)}}function parseSpec(parsed){const result=batteries_exports.batterySpecSchema.safeParse(parsed);if(!result.success){const details=result.error.issues.map(i=>` ${i.path.join(".")||"(spec)"}: ${i.message}`).join("\n");throw new CliError("invalid_request",`not a valid battery spec:
|
|
955
|
+
${details}`,{hint:"Run `gauge run --schema --hosted` for the full schema, or `gauge run --example --hosted` for a starter."})}return result.data}async function resolveApplyOrg(command,spec){return resolveOrg(command,spec.org)}function repoLabel2(repoUrl,repoRef){if(!repoUrl)return"greenfield";return repoRef?`${repoUrl}@${repoRef}`:repoUrl}function printPlan(plan){console.log(`Plan for org ${plan.org}`+(plan.experiment?` (experiment: ${plan.experiment})`:""));for(const prompt of plan.prompts){console.log(` ${prompt.action} prompt "${prompt.key}"`);for(const s of prompt.schedules){console.log(` ${s.action} schedule ${repoLabel2(s.repoUrl,s.repoRef)}`+(plan.expand?` \u2192 ${s.runs} run(s)`:""))}}const sum=plan.summary;console.log(`Prompts: ${sum.promptsToCreate} to create, ${sum.promptsToUpdate} to update; schedules: ${sum.schedulesToCreate} to create, ${sum.schedulesToUpdate} to update`);if(plan.expand){const byAgent=Object.entries(sum.runsByAgent).map(([agent,n])=>`${agent}: ${n}`).join(", ");console.log(`Runs to enqueue: ${sum.runsToEnqueue}${byAgent?` (${byAgent})`:""} \u2014 estimated ${sum.estimatedCredits} credits`)}else{console.log('No runs will be enqueued (config only \u2014 set "expand": true to run).')}console.log("Dry run \u2014 nothing was applied.")}function printApplied(res){console.log(`Applied${res.experiment?` (experiment: ${res.experiment})`:""} \u2014 ${res.batchIds.length} batch(es), ${res.runIds.length} run(s) enqueued`);for(const id of res.runIds)console.log(` run ${id}`);if(res.runIds.length>0){console.error(`(watch one live: gauge run watch ${res.runIds[0]})`)}if(res.blocked){console.error(`warning: some schedules were not expanded (${res.blocked}) \u2014 re-apply after resolving to enqueue the rest`)}}var BATTERY_SCHEMA=()=>z8.toJSONSchema(batteries_exports.batterySpecSchema);async function applySpec(command,spec,opts){if(opts.requireExperiment!==void 0)assertRequiredExperiment(spec,opts.requireExperiment);const org=await resolveApplyOrg(command,spec);const result=await createClient().post(`/api/v1/orgs/${encodeURIComponent(org)}/batteries`,spec,opts.dryRun?{dryRun:1}:void 0);if(resolveOutput(command)==="json")printJson(result);else if(opts.dryRun)printPlan(result);else printApplied(result);return result}import{createHash as createHash3}from"crypto";import{z as z12}from"zod";import{z as z9}from"zod";var runAddonsSchema=z9.object({browser:z9.object({session:z9.string().min(1),projectId:z9.number().int().positive(),region:z9.literal("us").default("us")}).strict().optional()}).strict();import{z as z11}from"zod";import{z as z10}from"zod";var inputPathSchema=z10.string().min(1).max(256).regex(/^(?![!/])(?!.*[\\\x00-\x1f\x7f])(?!.*(?:^|\/)\.{1,2}(?:\/|$))(?!.*\/\/)(?!.*\/$).+$/,"Expected a repository-relative path or glob without traversal");var reservedBinaryCommands=["bash","sh","zsh","env","node","nodejs","npm","npx","bun","python","python3","git","tar","curl","cp","chmod","mkdir","rm","ln","test","sha256sum","printenv","uname","claude","codex","opencode","pi","wizard"];var binaryCommandSchema=z10.string().max(128).regex(/^[A-Za-z0-9_][A-Za-z0-9_.+-]*$/,"Expected a command name without paths or shell syntax").refine(name=>!reservedBinaryCommands.includes(name),"Command is reserved for Gauge workspace setup or the agent harness; choose another command name");var binaryUseSchema=z10.object({type:z10.literal("binary"),paths:z10.array(inputPathSchema).min(1).max(32),platform:z10.literal("linux-x64").describe("Linux x86-64 ELF executables compatible with the consumer sandbox. No system packages are installed."),commands:z10.record(binaryCommandSchema,inputPathSchema.regex(/^[^*?{}[\]]+$/,"Command targets must be explicit relative file paths")).refine(commands=>Object.keys(commands).length>=1&&Object.keys(commands).length<=32,"Declare between 1 and 32 commands").meta({minProperties:1,maxProperties:32})}).strict();var inputUseSchema=z10.discriminatedUnion("type",[binaryUseSchema,websiteUseSchema,z10.object({type:z10.literal("web-fixture"),path:inputPathSchema.regex(/^[^*?{}[\]]+$/)}).strict(),z10.object({type:z10.literal("files"),paths:z10.array(inputPathSchema).min(1).max(32).optional()}).strict(),z10.object({type:z10.literal("npm-cli"),package:inputPathSchema,install:z10.enum(["environment","workspace"]).optional()}).strict(),z10.object({type:z10.literal("npm-package"),package:inputPathSchema,install:z10.enum(["environment","workspace"]).optional()}).strict(),z10.object({type:z10.literal("skill"),root:z10.union([z10.literal("."),inputPathSchema]).refine(path=>!/[*?{}[\]]/.test(path),"Skill root must be an explicit directory")}).strict()]);var inputSourceSchema=z10.discriminatedUnion("type",[z10.object({type:z10.literal("git"),paths:z10.array(inputPathSchema).min(1).max(32)}).strict(),z10.object({type:z10.literal("github-actions"),artifact:z10.string().min(1).max(256).optional()}).strict()]);var runPlanResourceSchema=z10.object({name:z10.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64),type:z10.literal("input"),source:z10.discriminatedUnion("type",[z10.object({type:z10.literal("upload"),sha256:z10.string().regex(/^sha256:[a-f0-9]{64}$/),bytes:z10.number().int().positive().max(512*1024*1024),repoUrl:z10.string().nullable(),commit:z10.string().regex(/^[a-f0-9]{40}$/i).nullable()}).strict(),commentSourceSchema.extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i)}).strict(),deploymentSourceSchema.extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i)}).strict(),inputSourceSchema.options[0].extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i)}).strict(),inputSourceSchema.options[1].extend({repoUrl:z10.string(),commit:z10.string().regex(/^[a-f0-9]{40}$/i),workflow:inputPathSchema}).strict()]),use:inputUseSchema}).strict().refine(r=>r.use.type==="website"===(r.source.type==="github-deployment"||r.source.type==="github-pr-comment"),"Website inputs require a deployment or PR comment source; file inputs require Git, Actions, or an upload");var definitionInputSourceSchema=z11.discriminatedUnion("type",[z11.object({type:z11.literal("git")}).strict(),z11.object({type:z11.literal("github-actions"),workflow:inputPathSchema.regex(/^\.github\/workflows\/[^/*?{}[\]]+\.ya?ml$/).describe("Workflow file in this repository, for example .github/workflows/build.yml."),artifact:z11.string().min(1).max(256).optional().describe("Upload name or pattern. Omit only when the workflow produces exactly one artifact.")}).strict()]);var skillInputPathSchema=z11.union([z11.literal("."),inputPathSchema.regex(/^[^*?{}[\]]+$/)]).describe("Explicit Skill directory or ZIP/tar.gz file containing SKILL.md at its root.");var runDefinitionInputSchema=z11.discriminatedUnion("type",[binaryUseSchema.extend({source:definitionInputSourceSchema.default({type:"git"})}).strict(),websiteUseSchema.extend({source:websiteSourceSchema}).strict(),z11.object({type:z11.literal("web-fixture"),source:definitionInputSourceSchema.default({type:"git"}),path:inputPathSchema.regex(/^[^*?{}[\]]+$/).describe("Explicit path to a version 1 web fixture JSON file.")}).strict(),z11.object({type:z11.literal("npm-package"),source:definitionInputSourceSchema.default({type:"git"}),path:inputPathSchema.describe("Path or glob selecting exactly one packed npm .tgz file."),install:z11.enum(["environment","workspace"]).default("environment").describe("environment exposes the package's commands on PATH; workspace adds a consumer application dependency.")}).strict(),z11.object({type:z11.literal("skill"),source:definitionInputSourceSchema.default({type:"git"}),path:skillInputPathSchema}).strict(),z11.object({type:z11.literal("files"),source:definitionInputSourceSchema.default({type:"git"}),paths:z11.array(inputPathSchema).min(1).max(32).describe("Files or globs selected from the committed Git tree or the uploaded artifact.")}).strict()]);function isSkillArchive(path){return/\.(?:zip|tar\.gz|tgz)$/i.test(path)}function compileDefinitionInput(name,input,repoUrl,commit){if(input.type==="website")return{name,type:"input",source:{...input.source,repoUrl,commit},use:{type:"website",url:input.url}};const paths=input.type==="files"||input.type==="binary"?input.paths:input.type==="skill"&&!isSkillArchive(input.path)?[input.path==="."?"**":`${input.path}/**`]:[input.path];return{name,type:"input",source:input.source.type==="git"?{type:"git",paths,repoUrl,commit}:{...input.source,repoUrl,commit},use:input.type==="binary"?{type:"binary",paths:input.paths,platform:input.platform,commands:input.commands}:input.type==="npm-package"?{type:"npm-package",package:input.path,install:input.install}:input.type==="web-fixture"?{type:"web-fixture",path:input.path}:input.type==="skill"?{type:"skill",root:input.path}:{type:"files",paths:input.paths}}}var inputUploadSpecSchema=z12.object({sha256:z12.string().regex(/^sha256:[a-f0-9]{64}$/),bytes:z12.number().int().positive().max(512*1024*1024)}).strict();var uncredentialedRepoUrlSchema=z12.string().min(1).refine(raw=>{if(!/^[a-z][a-z0-9+.-]*:\/\//i.test(raw))return true;try{const url=new URL(raw);return(url.protocol==="ssh:"||!url.username)&&!url.password&&!url.search&&!url.hash}catch{return false}},"Repository URL must not contain embedded credentials or query parameters");var runDefinitionChecksSchema=z12.object({paths:z12.array(inputPathSchema).min(1).max(32)}).strict();var runDefinitionConfigSchema=z12.object({agents:z12.array(agentConfigInputSchema).min(1),sampleCount:z12.number().int().min(1).max(50).optional(),repoUrl:uncredentialedRepoUrlSchema.nullable().optional(),repoRef:z12.string().min(1).nullable().optional(),profileId:z12.string().min(1).nullable().optional(),skillRefs:assetRefListSchema.optional(),mcpRefs:assetRefListSchema.optional(),connectionIds:z12.array(z12.string().min(1)).max(32).optional(),addons:runAddonsSchema.optional(),inputs:z12.array(z12.string()).optional(),checks:runDefinitionChecksSchema.optional()}).strict();var runDefinitionFileConfigSchema=z12.object({version:z12.literal(2).default(2),org:z12.string().min(1).optional(),orgId:z12.string().uuid().optional(),checks:runDefinitionChecksSchema.optional(),cases:z12.array(inputPathSchema).min(1).max(32).default(["gauge-cases/**/*.md"]),inputs:z12.record(z12.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64),runDefinitionInputSchema).refine(inputs=>Object.keys(inputs).length<=32,"At most 32 inputs are allowed").meta({maxProperties:32}).default({})}).strict();var runDefinitionCriterionSchema=z12.object({name:z12.string().min(1),rubric:z12.string().min(1)}).strict();var caseMeasurement={criteria:z12.array(runDefinitionCriterionSchema).min(1).optional(),preference:preferenceDefinitionSchema.optional()};function exactlyOneMeasurement(item){return item.criteria!==void 0!==(item.preference!==void 0)}var runDefinitionFrontmatterSchema=z12.object({...caseMeasurement,config:runDefinitionConfigSchema}).strict().refine(exactlyOneMeasurement,"Specify exactly one of criteria or preference");var runDefinitionCaseSchema=z12.object({path:z12.string().min(1),prompt:z12.string().min(1),...caseMeasurement,config:runDefinitionConfigSchema}).strict().refine(exactlyOneMeasurement,"Specify exactly one of criteria or preference");var runDefinitionSubmissionSchema=z12.object({version:z12.literal(2),lifecycle:lifecycleSchema.nullish(),source:z12.object({repoUrl:uncredentialedRepoUrlSchema.nullable(),commit:z12.string().regex(/^[a-f0-9]{40}$/i).nullable(),digest:z12.string().regex(/^sha256:[a-f0-9]{64}$/),verification:z12.enum(["SUBMITTED_UNVERIFIED","GITHUB_VERIFIED","LOCAL"])}).strict(),resources:z12.array(runPlanResourceSchema).max(32).default([]),cases:z12.array(runDefinitionCaseSchema).min(1).max(20)}).strict().superRefine((definition,ctx)=>{if(definition.source.verification==="GITHUB_VERIFIED"&&definition.lifecycle!==void 0)ctx.addIssue({code:"custom",path:["lifecycle"],message:"GitHub definitions cannot contain launch-only lifecycle scripts"});if(definition.source.verification!=="LOCAL"&&(!definition.source.repoUrl||!definition.source.commit))ctx.addIssue({code:"custom",path:["source"],message:"Committed definitions require a repository and commit"});if(definition.source.verification==="GITHUB_VERIFIED"&&definition.resources.some(r=>r.source.type==="upload"))ctx.addIssue({code:"custom",path:["resources"],message:"GitHub definitions cannot contain uploaded inputs"});const paths=new Set;for(const[index,item]of definition.cases.entries()){if(paths.has(item.path))ctx.addIssue({code:"custom",path:["cases",index,"path"],message:"Case paths must be unique"});paths.add(item.path)}const names=definition.resources.map(resource=>resource.name);if(new Set(names).size!==names.length)ctx.addIssue({code:"custom",path:["resources"],message:"Input names must be unique"});for(const[index,item]of definition.cases.entries())for(const name of item.config.inputs??names)if(!names.includes(name))ctx.addIssue({code:"custom",path:["cases",index,"config","inputs"],message:`Unknown input: ${name}`});for(const[index,resource]of definition.resources.entries()){if(resource.source.repoUrl!==definition.source.repoUrl||resource.source.commit!==definition.source.commit)ctx.addIssue({code:"custom",path:["resources",index,"source"],message:"Project preparation must use the definition repository and commit"})}});var runDefinitionPublicSubmissionSchema=runDefinitionSubmissionSchema.refine(definition=>definition.source.verification!=="GITHUB_VERIFIED",{path:["source","verification"],message:"Only the GitHub App may verify committed bytes"});var runDefinitionPlanResponseSchema=z12.object({source:runDefinitionSubmissionSchema.shape.source,lifecycle:z12.object({setupTimeoutSeconds:z12.number().int().min(1).max(600).optional(),teardownTimeoutSeconds:z12.number().int().min(1).max(600).optional()}).optional(),measurementMode:z12.enum(["AGENT_EXPERIENCE","AGENT_PREFERENCE","MIXED"]),cases:z12.array(z12.object({path:z12.string(),agents:z12.array(agentConfigSchema),samples:z12.number().int().nonnegative(),expectedRuns:z12.number().int().nonnegative(),workspace:z12.object({repoUrl:z12.string().nullable(),repoRef:z12.string().nullable()}),criteriaCount:z12.number().int().nonnegative(),preference:preferenceDefinitionSchema.optional(),inputs:z12.array(z12.string())})),resources:z12.array(z12.object({name:z12.string(),type:z12.literal("input"),source:z12.enum(["git","github-actions","github-deployment","github-pr-comment","upload"]),use:inputUseSchema,readiness:z12.enum(["READY","WAITING","FAILED"]),detail:z12.string().optional(),commands:z12.array(z12.string()).optional()})),expectedRuns:z12.number().int().nonnegative(),canRun:z12.boolean()});var runDefinitionRunResponseSchema=z12.object({runRequestId:z12.string(),batchIds:z12.array(z12.string()),runIds:z12.array(z12.string()),requestedRuns:z12.number().int().nonnegative(),createdRuns:z12.number().int().nonnegative(),aggregateUrl:z12.string()});import{z as z13}from"zod";var caseSpecSchema=z13.object({version:z13.literal(2),cases:z13.array(z13.object({key:z13.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64).describe("Unique case id, lowercase-kebab"),prompt:z13.string().trim().min(1).describe("The user's request"),criteria:z13.array(runDefinitionCriterionSchema).min(1).optional().describe("Optional for local runs; hosted needs criteria or preference"),preference:preferenceDefinitionSchema.optional(),config:runDefinitionConfigSchema}).strict().refine(item=>!(item.criteria&&item.preference),"Specify criteria or preference, not both")).min(1).max(20)}).strict().refine(spec=>new Set(spec.cases.map(item=>item.key)).size===spec.cases.length,"Case keys must be unique");var EXAMPLE_CASE_SPEC={version:2,cases:[{key:"release-notes",prompt:"Write the release notes for version 1.2.0. We added dark mode and fixed a login crash.",criteria:[{name:"Team format",rubric:"RELEASE-NOTES.md lists both changes in the team's format."}],config:{agents:[{agent:"CODEX_CLI"}],inputs:["release-notes"]}}]};var CASE_SPEC_SCHEMA=()=>z13.toJSONSchema(caseSpecSchema);function isCaseSpec(value){return value!==null&&typeof value==="object"&&"cases"in value&&!("prompts"in value)}function parseCaseSpec(value){const result=caseSpecSchema.safeParse(value);if(!result.success)throw new CliError("usage",`not a valid case spec:
|
|
956
|
+
${result.error.issues.map(i=>` ${i.path.join(".")||"(spec)"}: ${i.message}`).join("\n")}`,{hint:"Run `gauge run --schema --local` for the full schema, or `gauge run --example --local` for a starter."});return result.data}function caseSpecCases(spec){return spec.cases.map(({key,...item})=>({path:key,...item}))}function caseSpecDigest(spec){return`sha256:${createHash3("sha256").update("gauge-case-spec-v2\0").update(JSON.stringify(spec)).digest("hex")}`}import{execFileSync as execFileSync2}from"child_process";import{createHash as createHash5}from"crypto";import{createReadStream as createReadStream2,globSync as globSync2}from"fs";import{lstat,mkdtemp,readdir,rm,stat}from"fs/promises";import{tmpdir}from"os";import{basename,dirname,isAbsolute as isAbsolute4,join as join10,relative as relative4,resolve as resolve4,sep as sep2}from"path";import{setTimeout as delay2}from"timers/promises";import{createHash as createHash4}from"crypto";import{parseDocument}from"yaml";function utf8(bytes,path){try{return new TextDecoder("utf-8",{fatal:true}).decode(bytes)}catch{throw new Error(`${path} is not UTF-8 text`)}}function parseRunDefinitionCase(path,bytes){const source=utf8(bytes,path);const match=/^---\r?\n([\s\S]*?)\r?\n---(?:\r?\n|$)([\s\S]*)$/.exec(source);if(!match)throw new Error(`${path}: expected YAML frontmatter between --- lines`);const document=parseDocument(match[1],{uniqueKeys:true});if(document.errors.length)throw new Error(`${path}: ${document.errors.map(error=>error.message).join("; ")}`);const parsed=runDefinitionFrontmatterSchema.safeParse(document.toJS());if(!parsed.success)throw new Error(`${path}: ${parsed.error.message}`);const prompt=match[2].trim();if(!prompt)throw new Error(`${path}: Markdown body is empty`);return{path,prompt,...parsed.data}}function compileRunDefinitionFromBlobs(input){if(input.repoUrl!==null&&!uncredentialedRepoUrlSchema.safeParse(input.repoUrl).success)throw new Error("Git origin must not contain embedded credentials");if(input.commit!==null&&!/^[a-f0-9]{40}$/i.test(input.commit))throw new Error("Expected a full Git commit SHA");if(input.casePaths.length===0)throw new Error("At least one Markdown case is required");if(new Set(input.casePaths).size!==input.casePaths.length)throw new Error("Case paths must be unique");const configBytes=input.blobs.get("gauge.json");let config;if(configBytes){try{config=runDefinitionFileConfigSchema.parse(JSON.parse(utf8(configBytes,"gauge.json")))}catch(error){throw new Error(`gauge.json: ${error instanceof Error?error.message:error}`)}}const cases=input.casePaths.map(path=>{const bytes=input.blobs.get(path);if(!path.endsWith(".md")||!bytes)throw new Error(`Missing Markdown case: ${path}`);return parseRunDefinitionCase(path,bytes)});const hash=createHash4("sha256");hash.update("gauge-run-definition-v2\0");for(const[path,bytes]of[...input.blobs.entries()].sort(([a],[b])=>a<b?-1:a>b?1:0)){hash.update(path);hash.update("\0");hash.update(String(bytes.length));hash.update("\0");hash.update(bytes);hash.update("\0")}const submission=runDefinitionSubmissionSchema.parse({version:2,source:{repoUrl:input.repoUrl,commit:input.commit,digest:`sha256:${hash.digest("hex")}`,verification:input.verification},resources:input.resources??Object.entries(config?.inputs??{}).map(([name,item])=>{if(!input.repoUrl||!input.commit)throw new Error(`Supply --input ${name}=<path> outside a Git checkout with an origin`);return compileDefinitionInput(name,item,input.repoUrl,input.commit)}),cases});return{definition:submission,org:config?.orgId??config?.org}}import{minimatch}from"minimatch";function matchInputPaths(paths,patterns,options={}){for(const pattern of patterns)inputPathSchema.parse(pattern);const selections=patterns.map(pattern=>({pattern,matches:paths.filter(path=>minimatch(path,pattern,{dot:true,nocase:false,nonegate:true}))}));if(options.allowEmptySelection&&selections.every(({matches:matches2})=>matches2.length===0))return[];const selected=new Set;for(const{pattern,matches:matches2}of selections){if(!matches2.length)throw new Error(`No committed files match ${pattern}`);for(const path of matches2)selected.add(path)}return[...selected].sort()}import{create as tar}from"tar";import{execFileSync}from"child_process";import{globSync,lstatSync,readdirSync,readFileSync as readFileSync11,realpathSync as realpathSync3}from"fs";import{isAbsolute as isAbsolute3,join as join9,relative as relative3,resolve as resolve3,sep}from"path";import{minimatch as minimatch2}from"minimatch";import{parseDocument as parseDocument2}from"yaml";import{CommanderError}from"commander";import{ZodError}from"zod";function errorReport(error,command,context={}){const base=error instanceof CliError&&error.needsAccount&&context.local?localModeError(command,context.local,error.hint):describe(error);const quiet=base.message==="aborted"||command==="feedback";const isChat=command==="chat"||command.startsWith("chat ");const docs=base.docs??(isChat?void 0:DOCS_URL);return{...base,...docs?{docs}:{},...quiet?{}:{feedback:`gauge feedback -m "<what happened>" --kind friction${command?` --command "${command}"`:""}`}}}function localModeError(command,source,hint){return{code:"usage",message:`\`gauge ${command}\` requires hosted Gauge, but ${source==="flag"?"you passed --local":source==="workspace"?"this project is set to run locally":source==="user"?"your saved setting is local":"the CLI is using local mode"}.`,hint:hint??"Pass --hosted for this command, or run `gauge init --hosted` to use hosted Gauge in this project. Sign in with `gauge auth login` if needed.",exitCode:exitCodeFor("usage")}}function describe(error){if(error instanceof CliError)return{code:error.apiCode??error.code,message:readable(error.message),...error.details?{details:error.details}:{},hint:error.hint,docs:error.docs,exitCode:error.exitCode};if(error instanceof CommanderError)return{code:"usage",message:error.message.replace(/^error: /,""),hint:"Run the command with --help for its usage.",exitCode:exitCodeFor("usage")};if(error instanceof ZodError)return{code:"usage",message:issueLines("Invalid input",error.issues),exitCode:exitCodeFor("usage")};const node=error;if(node?.code==="ENOENT"&&node.path)return{code:"usage",message:`File not found: ${node.path}`,exitCode:exitCodeFor("usage")};if(node?.code==="EACCES"&&node.path)return{code:"usage",message:`Permission denied: ${node.path}`,exitCode:exitCodeFor("usage")};return{code:"internal",message:`Unexpected error: ${error instanceof Error?readable(error.message):String(error)}`,hint:process.env.GAUGE_DEBUG?void 0:"Set GAUGE_DEBUG=1 to print the stack trace.",exitCode:1}}function readable(message2){const start=message2.search(/\[\s*\{/);if(start<0)return message2;try{const issues=JSON.parse(message2.slice(start));if(!Array.isArray(issues)||!issues.every(issue=>typeof issue?.message==="string"))return message2;return issueLines(message2.slice(0,start).replace(/[:\s]+$/,""),issues)}catch{return message2}}function issueLines(prefix,issues){return[prefix,...issues.map(issue=>` ${issue.path?.length?`${issue.path.map(String).join(".")}: `:""}${issue.message}`)].filter(Boolean).join("\n")}function renderError(report,json){if(json)return`${JSON.stringify({error:report},null,2)}
|
|
784
957
|
`;const red=style("31",process.stderr);const dim=style("2",process.stderr);return`${[`${red("error:")} ${report.message}`,...report.hint?[report.hint]:[],...report.docs?[dim(`Docs: ${report.docs}`)]:[],...report.feedback?[dim(`Hit friction? ${report.feedback}`)]:[]].join("\n")}
|
|
785
|
-
`}function commandPath(program2,argv){const names=[];let command=program2;for(const token of argv){if(token.startsWith("-"))continue;const next=command.commands.find(sub=>sub.name()===token||sub.aliases().includes(token));if(next){names.push(token);command=next}else if(command.commands.length===0)break}return names.join(" ")}function wantsJson(argv){return argv.some((token,i)=>token==="--output=json"||token==="-ojson"||(token==="-o"||token==="--output")&&argv[i+1]==="json")}var record=value=>value!==null&&typeof value==="object"&&!Array.isArray(value)?value:{};var message=error=>oneLine(error instanceof CliError&&error.hint?`${error.message} ${error.hint}`:error instanceof Error?error.message:String(error));var oneLine=text=>readable(text).trim().replace(/\n\s*/g,"; ");var check=(id,state,detail)=>({id,state,detail});var dynamic=value=>typeof value!=="string"||value.includes("${{");var condition=value=>{if(value===void 0||value===true)return true;if(value===false)return false;if(typeof value==="string"){const compact2=value.replace(/\s/g,"");if(compact2==="true"||compact2===`\${{true}}`)return true;if(compact2==="false"||compact2===`\${{false}}`)return false}return void 0};function git(cwd,args){return execFileSync("git",args,{cwd,encoding:"utf8",stdio:["ignore","pipe","pipe"],maxBuffer:16*1024*1024})}function rootDirectory(cwd){return findWorkspace(cwd)??resolve3(cwd)}function regularFile(root,path){const full=join10(root,path);try{const rel=relative4(realpathSync3(root),realpathSync3(full));return rel!==".."&&!rel.startsWith(`..${sep2}`)&&!isAbsolute3(rel)&&lstatSync(full).isFile()}catch{return false}}function read(root,path){if(!regularFile(root,path))throw new Error(`${path}: missing regular file inside the working tree`);if(lstatSync(join10(root,path)).size>1024*1024)throw new Error(`${path}: exceeds the 1 MiB static inspection limit`);return readFileSync10(join10(root,path))}function ignoredByGit(root){try{git(root,["check-ignore","-q","."]);return true}catch{return false}}function workingFiles(root){let paths;try{if(ignoredByGit(root))throw new Error("workspace is ignored by Git");paths=git(root,["ls-files","--cached","--others","--exclude-standard","-z"]).split("\0").filter(Boolean)}catch{paths=[];const walk=directory=>{for(const entry of readdirSync2(join10(root,directory),{withFileTypes:true})){if([".git",".gauge","node_modules",".next",".venv","venv"].includes(entry.name))continue;const path=directory?`${directory}/${entry.name}`:entry.name;if(entry.isDirectory())walk(path);else if(entry.isFile())paths.push(path);if(paths.length>5e4)throw new Error("Working tree exceeds 50,000 files; initialize Git and ignore generated files before verifying")}};walk("")}return[...new Set(paths)].filter(path=>regularFile(root,path)).sort()}var skippedInputEntry=name=>[".git",".gauge","node_modules",".aws",".ssh",".npmrc",".pypirc"].includes(name)||name===".env"||name.startsWith(".env.");function diskFiles(root,base){const paths=[];const walk=directory=>{for(const entry of readdirSync2(join10(root,directory),{withFileTypes:true})){if(skippedInputEntry(entry.name))continue;const path=directory?`${directory}/${entry.name}`:entry.name;if(entry.isDirectory())walk(path);else if(entry.isFile())paths.push(path);if(paths.length>2e4)throw new Error("Input exceeds 20,000 files")}};walk(base==="."?"":base);return paths.sort()}function workflowChecks(root,input){if(input.source.type!=="github-actions")return[];const source=input.source;const checks=[];let workflow;try{const document=parseDocument2(read(root,source.workflow).toString("utf8"),{uniqueKeys:true});if(document.errors.length)throw new Error(document.errors.map(message).join("; "));workflow=record(document.toJS());if(!Object.keys(record(workflow.jobs)).length)throw new Error("Expected a workflow jobs mapping");checks.push(check("workflow","yes",`${source.workflow} exists with a parseable YAML jobs mapping`))}catch(error){return[check("workflow","no",message(error))]}const jobs=Object.entries(record(workflow.jobs)).map(([name,value])=>({name,job:record(value)})).filter(({job})=>condition(job.if)!==false);const steps=jobs.flatMap(({name,job})=>Array.isArray(job.steps)?job.steps.map(value=>({job:name,jobCondition:condition(job.if),step:record(value)})).filter(({step})=>condition(step.if)!==false):[]);const opaque=jobs.some(({job})=>typeof job.uses==="string")||steps.some(({step})=>typeof step.uses==="string"&&!/^actions\/(checkout|upload-artifact|setup-[^@]+)@/.test(step.uses));const uploads=steps.filter(({step})=>typeof step.uses==="string"&&step.uses.startsWith("actions/upload-artifact@"));const candidates=uploads.filter(({step})=>{const name=record(step.with).name??"artifact";return!source.artifact||dynamic(name)||minimatch2(String(name),source.artifact,{nonegate:true})});const recognizedUpload=candidates.length===1&&!dynamic(record(candidates[0].step.with).name??"artifact");checks.push(check("artifact-upload",recognizedUpload?"yes":"unknown",recognizedUpload?`One direct upload step declares the matching artifact ${String(record(candidates[0].step.with).name??"artifact")}`:candidates.length?`Found ${candidates.length} potentially matching upload step(s); resolve dynamic artifact names or overlapping uploads to establish a unique selection`:`No direct upload step matches ${source.artifact??"the configured artifact"}; ${opaque?"inspect the referenced reusable workflows or custom actions":"inspect build scripts or declare a matching actions/upload-artifact step"}`));for(const[index,{job,jobCondition,step}]of candidates.entries()){const options=record(step.with);const version=/^actions\/upload-artifact@v(\d+)(?:\.\d+){0,2}$/.exec(String(step.uses));const major=version?Number(version[1]):void 0;let state=source.artifact&&dynamic(options.name??"artifact")?"unknown":major===7?options.archive===true||options.archive==="true"?"yes":options.archive===void 0||options.archive===false||options.archive==="false"?"no":"unknown":major!==void 0&&major>=4&&major<=6?"yes":"unknown";const conditional=jobCondition===void 0||condition(step.if)===void 0;if(state==="no"&&conditional)state="unknown";const detail=source.artifact&&dynamic(options.name??"artifact")?"Resolve the artifact name expression to determine whether this upload is selected":major===7?options.archive===true||options.archive==="true"?"archive: true declares a ZIP artifact":options.archive===void 0||options.archive===false||options.archive==="false"?`archive: true is required for Gauge's ZIP download${conditional?"; resolve the upload condition to determine whether this incompatible upload is selected":""}`:"Resolve the archive expression; Gauge requires archive: true":major!==void 0&&major>=4&&major<=6?"This action version produces ZIP artifacts":"Resolve the pinned action version; ZIP format is recognized for upload-artifact v4\u2013v7 (v7 requires archive: true)";checks.push(check(`artifact-format.${index}`,state,`${job}: ${String(step.uses)}; ${detail}`))}const checkouts=steps.filter(({step})=>typeof step.uses==="string"&&step.uses.startsWith("actions/checkout@"));const events=typeof workflow.on==="string"?[workflow.on]:Array.isArray(workflow.on)?workflow.on:Object.keys(record(workflow.on));const pr=events.includes("pull_request");const recognizedCheckout=step=>{const options=record(step.with);if(options.repository!==void 0)return false;if(!events.length||options.ref!==void 0&&typeof options.ref!=="string")return false;const ref=typeof options.ref==="string"?options.ref.replace(/\s/g,""):"";return events.every(event=>["pull_request","push","workflow_dispatch"].includes(event))&&(!ref||ref===`\${{github.sha}}`)||ref===`\${{github.event.pull_request.head.sha||github.sha}}`||pr&&events.every(event=>event==="pull_request")&&ref===`\${{github.event.pull_request.head.sha}}`};const checkoutState=checkouts.length>0&&checkouts.every(({step})=>recognizedCheckout(step))?"yes":"unknown";checks.push(check("candidate-checkout",checkoutState,checkoutState==="yes"?"Workflow checkout recognized, including the default PR merge checkout":"Could not inspect checkout settings; check that your workflow builds the intended code"));const filtered=Object.values(record(workflow.on)).some(value=>["paths","paths-ignore","branches","branches-ignore","types"].some(key=>record(value)[key]!==void 0));const eligible=events.filter(event=>["pull_request","push","workflow_dispatch"].includes(String(event)));const knownEvents=events.length>0&&events.every(event=>!dynamic(event));checks.push(check("workflow-trigger",eligible.length?"yes":knownEvents?"no":"unknown",eligible.length?`Eligible trigger(s) declared: ${eligible.join(", ")}${filtered?"; event filters are present":""}`:knownEvents?`Declared event(s) ${events.join(", ")} are not eligible artifact sources; declare push, pull_request, or workflow_dispatch`:"Could not read workflow events; declare push, pull_request, or workflow_dispatch"));return checks}function inputChecks(root,files,input,local){const checks=[check("configuration","yes",`Supported ${input.type} input from ${input.source.type==="git"?"project files":input.source.type}`)];if(input.type==="website")return checks;if(input.type==="binary"){const excluded=Object.entries(input.commands).filter(([,target])=>!input.paths.some(pattern=>minimatch2(target,pattern,{dot:true,nocase:false,nonegate:true})));checks.push(check("command-selection",excluded.length?"no":"yes",excluded.length?`Declared command targets are excluded by paths: ${excluded.map(([name,target])=>`${name}: ${target}`).join(", ")}`:"Declared command targets match selected paths"))}if(input.source.type==="github-actions"&&!local)checks.push(...workflowChecks(root,input));else{try{const patterns=input.type==="files"||input.type==="binary"?input.paths:input.type==="skill"&&!isSkillArchive(input.path)?[input.path==="."?"**":`${input.path}/**`]:[input.path];const selected=local&&input.type==="npm-package"?globSync(input.path,{cwd:root}):matchInputPaths(local?diskFiles(root,input.type==="skill"&&!isSkillArchive(input.path)?input.path:""):files,patterns);if(input.type==="binary"){for(const[command,target]of Object.entries(input.commands)){if(!selected.includes(target))throw new Error(`Binary command ${command} target is missing or excluded by paths: ${target}`)}}if(input.type==="web-fixture")parseWebFixtureInput(read(root,input.path).toString("utf8"));if(input.type==="npm-package"&&(selected.length!==1||!selected[0].endsWith(".tgz")))throw new Error("npm-package must select exactly one .tgz file");if(input.type==="skill"&&!isSkillArchive(input.path)&&!selected.includes(input.path==="."?"SKILL.md":`${input.path}/SKILL.md`))throw new Error("Skill directory must contain SKILL.md at its root");checks.push(check("local-paths","yes",`${selected.length} ${local?"local":"working-tree"} file(s) selected`))}catch(error){checks.push(check("local-paths","no",message(error).replace("committed files",local?"local files":"working-tree files")))}}return checks}function verifyEvals(cwd=process.cwd(),{local=false}={}){const root=rootDirectory(cwd);const result={version:1,mode:"local-static",root,status:"passed",summary:{yes:0,no:0,unknown:0},facts:{configFilePresent:null,caseCount:null},info:[],notChecked:["github-app-activation","account-settings","mcp-server-access","mcp-server-reachability","live-input-delivery","artifact-contents","sandbox-installation","workflow-execution"],global:[],inputs:[]};const finish=()=>{for(const item of[...result.global,...result.inputs.flatMap(input=>input.checks)])result.summary[item.state]++;result.status=result.summary.no>0?"failed":result.summary.unknown>0?"incomplete":"passed";return result};let raw={};let files;try{files=workingFiles(root)}catch(error){result.global.push(check("working-tree","no",message(error)));return finish()}try{result.facts.configFilePresent=files.includes("gauge.json")||regularFile(root,"gauge.json");if(result.facts.configFilePresent)raw=JSON.parse(read(root,"gauge.json").toString("utf8"));else result.info.push({id:"config-file-absent",detail:"Optional gauge.json is absent; using default case discovery and no inputs"})}catch(error){result.global.push(check("configuration","no",message(error)));return finish()}const config=runDefinitionFileConfigSchema.safeParse(raw);result.global.push(check("configuration",config.success?"yes":"no",config.success?"Configuration matches the supported gauge.json schema":oneLine(config.error.message)));const rawInputs=record(record(raw).inputs);for(const[name,value]of Object.entries(rawInputs).sort(([a],[b])=>a.localeCompare(b))){const parsed=runDefinitionInputSchema.safeParse(value);result.inputs.push({name,checks:parsed.success?inputChecks(root,files,parsed.data,local):[check("configuration","no",oneLine(parsed.error.message))]})}if(config.success){try{const paths=matchInputPaths(files,config.data.cases,{allowEmptySelection:true});if(paths.length>20||paths.some(path=>!path.endsWith(".md")))throw new Error("Expected at most 20 Markdown cases");result.facts.caseCount=paths.length;if(paths.length)result.global.push(check("case-discovery","yes",`${paths.length} working-tree case(s) selected`));else result.info.push({id:"no-cases",detail:"No eval cases found; input-only setup is valid, but no evaluations will run"});for(const path of paths){try{const item=parseRunDefinitionCase(path,read(root,path));if(local)assertLocalCapabilities(item.config,`${path}: config.`);const missing=(item.config.inputs??[]).filter(name=>!Object.hasOwn(config.data.inputs,name));if(missing.length)throw new Error(`${path}: Unknown input(s): ${missing.join(", ")}`);const commandOwners=new Map;mergeWebFixtures((item.config.inputs??Object.keys(config.data.inputs)).flatMap(name=>{const input=config.data.inputs[name];return input.type==="web-fixture"&&(local||input.source.type==="git")?[parseWebFixtureInput(read(root,input.path).toString("utf8"))]:[]}));for(const name of new Set(item.config.inputs??Object.keys(config.data.inputs))){const input=config.data.inputs[name];if(input.type!=="binary")continue;for(const command of Object.keys(input.commands)){const owner=commandOwners.get(command);if(owner)throw new Error(`${path}: Inputs ${owner} and ${name} both expose command ${command}; choose distinct commands or separate cases`);commandOwners.set(command,name)}}result.global.push(check(`case.${path}`,"yes","Case syntax, input references, and declared command names are valid"))}catch(error){result.global.push(check(`case.${path}`,"no",message(error)))}}}catch(error){result.global.push(check("case-discovery","no",message(error).replace("committed files","working-tree files")))}}else result.info.push({id:"case-validation-skipped",detail:"Fix gauge.json before validating cases and input references"});return finish()}function printEvalVerification(result){console.log(`Local static verification: ${result.root}`);const print2=checks=>{for(const item of checks)console.log(` [${item.state}] ${item.id}: ${item.detail}`)};console.log("Global");print2(result.global);for(const input of result.inputs){console.log(`Input: ${input.name}`);print2(input.checks)}for(const item of result.info)console.log(`[info] ${item.id}: ${item.detail}`);console.log(`Static verification ${result.status}: ${result.summary.yes} yes, ${result.summary.no} no, ${result.summary.unknown} unknown.`);console.log("Scope: repository setup only. Live delivery, account settings, and sandbox execution were not tested; no sessions were launched.")}var LIMIT=512*1024*1024;var SNAPSHOT_FILES=100;function parseInputBindings(values){const bindings=new Map;for(const value of values){const split=value.indexOf("=");const name=split<0?value:value.slice(0,split);const path=split<0?void 0:value.slice(split+1);if(!/^[a-z0-9]+(?:-[a-z0-9]+)*$/.test(name)||path==="")throw new CliError("usage","--input expects name or name=path");if(bindings.has(name))throw new CliError("usage",`Duplicate --input: ${name}`);bindings.set(name,path)}return bindings}var isGlob=path=>/[*?[\]{}]/.test(path);function rejectSkippedPatterns(name,patterns){for(const pattern of patterns){const skipped=pattern.split("/").find(skippedInputEntry);if(skipped===".gauge")throw new CliError("usage",`Input ${name}: ${pattern} is inside .gauge/. Inputs never include .gauge/, because it holds run results and local config.`,{hint:"Move the files outside .gauge/ and update the glob."});if(skipped)throw new CliError("usage",`Input ${name}: ${pattern} is inside ${skipped}, which inputs never include`,{hint:`Move the files outside ${skipped} and update the glob.`})}}async function walkInput(root,depth=Infinity){const paths=[];const walk=async(directory,level)=>{for(const entry of await readdir(join11(root,directory),{withFileTypes:true})){if(skippedInputEntry(entry.name)||!entry.isFile()&&!entry.isDirectory())continue;const next=directory?`${directory}/${entry.name}`:entry.name;if(entry.isDirectory()){if(level<depth)await walk(next,level+1)}else paths.push(next);if(paths.length>2e4)throw new CliError("usage","Input exceeds 20,000 files")}};await walk("",1);return paths.sort()}function selectPaths(name,paths,patterns){rejectSkippedPatterns(name,patterns);try{return matchInputPaths(paths,patterns)}catch(error){throw new CliError("usage",`Input ${name}: ${message(error).replace("committed files","files")}`)}}async function expandOverrideGlob(name,glob){const segments=glob.split(sep3);const first=segments.findIndex(isGlob);const root=segments.slice(0,first).join(sep3)||sep3;const pattern=segments.slice(first).join("/");const info=await stat(root).catch(()=>void 0);if(!info?.isDirectory())throw new CliError("usage",`Input ${name}: ${root} is not a directory`);const depth=pattern.includes("**")?Infinity:pattern.split("/").length;return{root,pattern,files:selectPaths(name,await walkInput(root,depth),[pattern])}}async function selectLocalInput(resource,supplied,override){let path=resolve4(supplied);let info=await lstat(path).catch(()=>void 0);if(!info&&override&&isGlob(path)){if(resource.use.type==="skill")throw new CliError("usage",`Input ${resource.name}: Skill inputs take an explicit directory or archive, not a glob`);const{root:root2,pattern,files}=await expandOverrideGlob(resource.name,path);if(resource.use.type==="files")return{use:{type:"files"},root:root2,files,skillDirectory:false};if(resource.use.type==="binary")return{use:{...resource.use,paths:[pattern]},root:root2,files,skillDirectory:false};if(files.length!==1)throw new CliError("usage",`Input ${resource.name}: ${supplied} matches ${files.length} files; keep one`);path=join11(root2,files[0]);info=await lstat(path)}if(!info)throw new CliError("usage",`Input path does not exist: ${supplied}`,{hint:"Pass --input <name>=<path>, or fix the input's path in gauge.json."});if(!info.isFile()&&!info.isDirectory())throw new CliError("usage",`Input must be an ordinary file or directory: ${supplied}`);let use=resource.use;if(use.type==="web-fixture"){if(!info.isFile())throw new CliError("usage","Supply a web fixture JSON file");use={type:"web-fixture",path:basename(path)}}else if(use.type==="skill"){if(info.isFile()&&!isSkillArchive(path))throw new CliError("usage","Supply a Skill directory, .zip, or .tar.gz archive");use={...use,root:info.isDirectory()?".":basename(path)}}else if(use.type==="npm-package"||use.type==="npm-cli"){if(!info.isFile()||!path.endsWith(".tgz"))throw new CliError("usage","Supply the packed npm .tgz file");use={...use,package:basename(path)}}else if(override&&use.type==="files")use={type:"files"};else if(override&&use.type==="binary")use={...use,paths:[info.isDirectory()?"**":basename(path)]};const root=info.isDirectory()?path:dirname(path);const paths=info.isDirectory()?await walkInput(root):[basename(path)];const patterns=!override&&(use.type==="files"||use.type==="binary")?use.paths:void 0;return{use,root,files:patterns?selectPaths(resource.name,paths,patterns):paths,skillDirectory:use.type==="skill"&&info.isDirectory()}}async function packLocalInput(resource,supplied,output,{override=false}={}){if(resource.use.type==="website")throw new CliError("usage",`Input ${resource.name}: website inputs cannot be supplied through --input.`,{hint:"Use the configured public preview with --hosted. For content-only alternatives and their discovery limits, read `gauge instructions hillclimb`."});const{use,root,files,skillDirectory}=await selectLocalInput(resource,supplied,override);if(!files.length)throw new CliError("usage","Input contains no selected files");if(use.type==="binary"){for(const[command,target]of Object.entries(use.commands))if(!files.includes(target))throw new CliError("usage",`Input ${resource.name}: binary command ${command} target is missing: ${target}`,override?{hint:"Command targets are relative to the supplied path."}:void 0)}if(skillDirectory&&!files.includes("SKILL.md"))throw new CliError("usage","Skill directory must contain SKILL.md at its root");let total=0;for(const file of files){inputPathSchema.parse(file);const entry=await lstat(join11(root,file));if(!entry.isFile())throw new CliError("usage",`Input file changed or is not an ordinary file: ${file}`);total+=entry.size;if(total>LIMIT)throw new CliError("usage","Input exceeds 512 MiB")}await tar({cwd:root,file:output,gzip:true,portable:true,noMtime:true,follow:false,noDirRecurse:true},files.map(file=>`./${file}`));const bytes=(await stat(output)).size;if(bytes>LIMIT)throw new CliError("usage","Input archive exceeds 512 MiB");const hash=createHash5("sha256");for await(const chunk of createReadStream2(output))hash.update(chunk);const sha256=`sha256:${hash.digest("hex")}`;if(use.type==="web-fixture")await readWebFixtureArchive(output,use.path,sha256);return{use,upload:{path:output,bytes,sha256},files}}async function loadLocalRunDefinition(files,values,cwd=process.cwd(),{local=false,spec}={}){const directory=await mkdtemp(join11(tmpdir(),"gauge-inputs-"));const cleanup=()=>rm(directory,{recursive:true,force:true});try{const root=rootDirectory(cwd);const paths=workingFiles(root);const blobs=new Map;if(paths.includes("gauge.json"))blobs.set("gauge.json",read(root,"gauge.json"));const configBytes=blobs.get("gauge.json");let config;try{config=runDefinitionFileConfigSchema.parse(configBytes?JSON.parse(configBytes.toString("utf8")):{})}catch(error){throw new CliError("usage",`gauge.json is not valid: ${error instanceof Error?error.message:error}`,{hint:"Run `gauge run --verify` for each problem."})}const patterns=spec?[]:files.length?files.map(file=>{const path=relative5(root,resolve4(cwd,file));if(!path||path===".."||path.startsWith(`..${sep3}`)||isAbsolute4(path))throw new CliError("usage","Case paths must be inside the definition directory");return path.split(sep3).join("/")}):config.cases;const cases=spec?[]:files.length?matchInputPaths(paths,patterns):matchInputPaths(paths,patterns,{allowEmptySelection:true});if(!spec&&cases.length===0)throw noCasesError();if(cases.length>20||cases.some(path=>!path.endsWith(".md")))throw new CliError("usage","Expected 1\u201320 Markdown cases");for(const path of cases)blobs.set(path,read(root,path));const parsed=spec?caseSpecCases(spec):cases.map(path=>parseRunDefinitionCase(path,blobs.get(path)??Buffer.alloc(0)));const unmeasured=parsed.filter(item=>!item.criteria&&!item.preference);if(!local&&unmeasured.length)throw new CliError("usage",`Hosted cases need criteria or preference: ${unmeasured.map(item=>item.path).join(", ")}`,{hint:"Add criteria, or run locally with --local."});const bindings=parseInputBindings(values);for(const name of bindings.keys())if(!(name in config.inputs))throw new CliError("usage",`Unknown input: ${name}`,{hint:`gauge.json defines: ${Object.keys(config.inputs).join(", ")||"none"}`});for(const item of parsed)for(const name of item.config.inputs??[])if(!(name in config.inputs))throw new CliError("usage",`${item.path}: unknown input "${name}". gauge.json defines: ${Object.keys(config.inputs).join(", ")||"none"}`);const selected=new Set(parsed.flatMap(item=>item.config.inputs??Object.keys(config.inputs)));for(const[name,override]of bindings){if(!selected.has(name))throw new CliError("usage",`Input is not used by the selected cases: ${name}`);if(override!==void 0)rejectSkippedPatterns(name,[relative5(root,resolve4(cwd,override)).split(sep3).join("/")])}const git2=args=>{try{return execFileSync2("git",args,{cwd:root,encoding:"utf8",stdio:["ignore","pipe","pipe"]}).trim()}catch{return null}};const commit=git2(["rev-parse","--verify","HEAD^{commit}"]);const origin=git2(["remote","get-url","origin"]);const repoUrl=origin&&uncredentialedRepoUrlSchema.safeParse(origin).success?origin:null;const unattachedInputs=Object.keys(config.inputs).filter(name=>!selected.has(name));const resources=[];const uploads=[];const inputSnapshots=[];for(const[name,input]of Object.entries(config.inputs)){if(!selected.has(name))continue;if(local&&input.type==="website")throw new CliError("usage",`Website input "${name}" requires hosted Gauge to replace the original site with your preview.`,{hint:"Use --hosted with a public preview. For local runs, supply documentation files or a preview URL in the prompt. These give the agent your content directly rather than testing whether it finds your site. See `gauge instructions hillclimb`."});const original=compileDefinitionInput(name,input,repoUrl??"",commit??"");const override=bindings.get(name);const supplied=override??(original.source.type==="git"?workingTreeInput(root,name,input):void 0);if(supplied!==void 0){const{use,upload,files:files2}=await packLocalInput(original,resolve4(cwd,supplied),join11(directory,`${name}.tar.gz`),{override:override!==void 0});uploads.push(upload);inputSnapshots.push({name,source:override!==void 0?"override":"working-tree",path:resolve4(cwd,supplied),gaugeArchive:{sha256:upload.sha256,bytes:upload.bytes},files:files2.slice(0,SNAPSHOT_FILES),fileCount:files2.length});resources.push({...original,use,source:{type:"upload",sha256:upload.sha256,bytes:upload.bytes,repoUrl,commit}})}else{if(local)throw new CliError("usage",`Input ${name} has a remote source. Supply --input ${name}=<path> for local execution, or use --hosted.`);if(!repoUrl||!commit)throw new CliError("usage",`Input ${name}'s configured ${original.source.type} source requires a Git checkout with an origin.`,{hint:input.type==="website"?"Configure the repository origin and publish a matching public preview through the declared source. Website inputs cannot use --input; see `gauge instructions hillclimb` for content-only alternatives.":`Supply --input ${name}=<path> to use local files.`});resources.push(original)}}if(!spec){const compiled=compileRunDefinitionFromBlobs({repoUrl,commit,resources,blobs,casePaths:cases,verification:"LOCAL"});const webFixtures=await freezeWebFixtures({...compiled.definition,resources:resources.filter(r=>r.source.type==="upload")},uploads);return{...compiled,webFixtures,uploads,inputSnapshots,unattachedInputs,cleanup}}const definition={version:2,source:{repoUrl,commit,digest:caseSpecDigest(spec),verification:"LOCAL"},resources,cases:parsed};if(!local)runDefinitionSubmissionSchema.parse(definition);return{definition,webFixtures:await freezeWebFixtures({...definition,resources:resources.filter(r=>r.source.type==="upload")},uploads),org:config.orgId??config.org,uploads,inputSnapshots,unattachedInputs,cleanup}}catch(error){await cleanup();if(error instanceof CliError||!(error instanceof Error))throw error;throw new CliError("usage",error.message,{hint:"Check the case format in `gauge instructions run`, then run `gauge run --verify`."})}}function workingTreeInput(root,name,input){switch(input.type){case"website":throw new CliError("usage",`Input ${name} is a website; website inputs run on hosted Gauge only`);case"files":case"binary":return root;case"web-fixture":case"skill":return join11(root,input.path);case"npm-package":{const matches2=globSync2(input.path,{cwd:root});if(matches2.length!==1)throw new CliError("usage",matches2.length?`Input ${name}: ${input.path} matches ${matches2.length} files; keep one`:`Input ${name}: no file matches ${input.path}. Run \`npm pack\` first`);return join11(root,matches2[0])}}}async function uploadLocalInputs(client,org,uploads){for(const input of uploads){const signed2=await client.post(`/api/v1/orgs/${encodeURIComponent(org)}/run-definitions/uploads`,{sha256:input.sha256,bytes:input.bytes});const signal=AbortSignal.timeout(10*60*1e3);for(let attempt=0;attempt<3;attempt++){const body=createReadStream2(input.path);let response;try{response=await fetch(signed2.url,{method:"PUT",headers:signed2.headers,body,duplex:"half",redirect:"error",signal})}finally{body.destroy()}await response.body?.cancel();if(response.ok||response.status===412)break;if(response.status!==409||attempt===2)throw new CliError("network",`Input upload failed (${response.status})`);await delay2(100*2**attempt,void 0,{signal})}}}var SIDE_KEYS=["request","batch","eval","revision","experiment","trial","agent","model","since","until"];var SELECTING_KEYS=new Set(["request","batch","eval","experiment","trial"]);function csv4(value){return value.split(",").map(part=>part.trim()).filter(Boolean)}function parseCompareSide(flag,pairs){const side={};const seen=new Set;for(const pair2 of pairs){const eq=pair2.indexOf("=");const key=eq<0?"":pair2.slice(0,eq).trim();const value=eq<0?"":pair2.slice(eq+1).trim();if(!SIDE_KEYS.includes(key)||!value)throw new CliError("usage",`invalid ${flag} "${pair2}" (expected key=value with key one of ${SIDE_KEYS.join(", ")})`);if(seen.has(key)&&key!=="batch")throw new CliError("usage",`${flag} ${key}= given twice`);seen.add(key);if(key==="request")side.runRequest=value;else if(key==="batch")side.batch=[...side.batch??[],...csv4(value)];else if(key==="eval")side.evalSet=value;else if(key==="revision")side.evalRevision=value;else if(key==="experiment")side.experiment=value;else if(key==="trial")side.trial=value;else if(key==="agent")side.agent=parseAgents(value);else if(key==="model")side.model=csv4(value);else if(key==="since")side.since=value;else side.until=value}if(![...seen].some(key=>SELECTING_KEYS.has(key)))throw new CliError("usage",`${flag} needs request=, batch=, eval=, experiment=, or trial= to select runs`);return side}function pct3(rate){return rate===null?"":`${Math.round(rate*100)}%`}function passCell(stats){return stats.judged===0?"":`${pct3(stats.passRate)} (${stats.passed}/${stats.judged})`}function signed(value,format){if(value===null)return"";return`${value>0?"+":value<0?"-":"\xB1"}${format(Math.abs(value))}`}var points=rate=>`${Math.round(rate*100)}pp`;var turns=n=>String(Math.round(n*10)/10);var usd=n=>`$${n.toFixed(2)}`;var secs=ms=>`${Math.round(ms/1e3)}s`;function pair(before,after,delta3,format){if(before===null&&after===null)return"";const side=v=>v===null?"\u2014":format(v);return`${side(before)} \u2192 ${side(after)}${delta3===null?"":` (${signed(delta3,format)})`}`}function row(target,item){const{baseline:a,candidate:b,delta:delta3}=item;return{target,runs:`${a.runs} \u2192 ${b.runs}`,completed:`${a.succeeded}/${a.runs} \u2192 ${b.succeeded}/${b.runs}`,baseline:passCell(a),candidate:passCell(b),"\u0394 pass":signed(delta3.passRate,points),turns:pair(a.medianTurns,b.medianTurns,delta3.medianTurns,turns),cost:pair(a.medianUsdCost,b.medianUsdCost,delta3.medianUsdCost,usd),duration:pair(a.medianDurationMs,b.medianDurationMs,delta3.medianDurationMs,secs)}}function compareWarnings(result){const warnings=[];for(const name of["baseline","candidate"]){const side=result.sides[name];const stats=result.overall[name];if(stats.runs===0)warnings.push(`${name} matched no runs. Check its selectors; optimization trial runs need trial=<trialId>.`);else if(stats.finished<stats.runs)warnings.push(`${name} has ${stats.runs-stats.finished} of ${stats.runs} runs still in progress, so its numbers are partial. Run \`gauge run wait <request-id>\` first.`);if(stats.judged<stats.finished)warnings.push(`${name} has ${stats.finished-stats.judged} finished runs without a verdict (not judged yet, or no criteria). Its pass rate covers judged runs only.`);for(const mixed of side.mixedRevisions??[])warnings.push(`${name}: ${mixed.evalSetId?`eval ${mixed.evalSetId}`:`case prompt ${mixed.promptId}`} ran under ${mixed.revisions.length} revisions (${mixed.revisions.join(", ")}): it was edited between runs. Add --${name} revision=<hash> to pick one.`);if(side.truncated)warnings.push(`${name} matched too many runs; only the newest were compared. Narrow it with since=, until=, or batch=.`)}return warnings}function printComparison(result){const revisions=name=>result.sides[name].evalRevisions.join(", ")||"\u2014";console.log(`Baseline revision: ${revisions("baseline")}. Candidate revision: ${revisions("candidate")}.`);for(const name of["baseline","candidate"]){for(const trial of result.sides[name].trials??[])console.log(`${name}: ${trial.runs} run${trial.runs===1?"":"s"} from optimization trial ${trial.key} "${trial.title}" (${trial.kind}). Inspect: gauge optimizations trial ${trial.optimizationId} ${trial.trialId}`)}console.log("Medians are baseline \u2192 candidate (difference).");console.log("");printItems([row("all",result.overall),...result.targets.map(target=>row(`${target.agent} ${target.model||"(default)"}`,target))],"table");const before=new Map(result.overall.baseline.criteria.map(c=>[c.name,c]));const after=new Map(result.overall.candidate.criteria.map(c=>[c.name,c]));if(result.overall.delta.criteria.length>0){console.log("");printItems(result.overall.delta.criteria.map(c=>{const a=before.get(c.name);const b=after.get(c.name);return{criterion:c.name,baseline:a&&a.judged>0?`${pct3(a.passRate)} (${a.passed}/${a.judged})`:"",candidate:b&&b.judged>0?`${pct3(b.passRate)} (${b.passed}/${b.judged})`:"","\u0394 pass":signed(c.passRate,points)}}),"table")}}var LOCAL_AGENTS2=new Set(["CLAUDE_CODE","CODEX_CLI","CURSOR","PI","OPENCODE"]);var LOCAL_TERMINAL_STATES=new Set(["completed","failed","canceled","interrupted"]);function localCompareFilter(flag,side){const hostedOnly2=[side.runRequest!==void 0&&"request=",side.experiment!==void 0&&"experiment=",side.trial!==void 0&&"trial="].filter(Boolean);if(hostedOnly2.length>0)throw new CliError("usage",`${flag} ${hostedOnly2.join(", ")} select hosted runs; local sides take batch=, eval=<saved id>, revision=<number>, agent=, model=, since=, until=`);if((side.batch?.length??0)>1)throw new CliError("usage",`${flag} takes one batch= locally`);if((side.agent?.length??0)>1||(side.model?.length??0)>1)throw new CliError("usage",`${flag} takes one agent= and one model= locally`);if(side.evalRevision!==void 0&&!/^[1-9]\d*$/.test(side.evalRevision))throw new CliError("usage",`${flag} local revision= must be a positive integer`);return{batch:side.batch?.[0],evalId:side.evalSet,measurementRevision:side.evalRevision===void 0?void 0:Number(side.evalRevision),agent:side.agent?.[0],model:side.model?.[0],since:side.since?parseSince(side.since,Date.now(),"since="):void 0,until:side.until?parseSince(side.until,Date.now(),"until="):void 0}}function median(values){if(values.length===0)return null;const sorted=[...values].sort((a,b)=>a-b);const mid=sorted.length>>1;return sorted.length%2?sorted[mid]:(sorted[mid-1]+sorted[mid])/2}function localStats(runs){const judged=runs.filter(run=>run.criterionResults?.length&&run.criterionResults.every(criterion=>criterion.outcome!=="UNASSESSABLE"));const passed=judged.filter(run=>run.criterionResults.every(criterion=>criterion.outcome==="PASS")).length;const byCriterion=new Map;for(const run of runs)for(const criterion of run.criterionResults??[]){const current=byCriterion.get(criterion.name)??{name:criterion.name,judged:0,passed:0};if(criterion.outcome!=="UNASSESSABLE"){current.judged++;if(criterion.outcome==="PASS")current.passed++}byCriterion.set(criterion.name,current)}const recorded=values=>values.filter(value=>typeof value==="number"&&Number.isFinite(value));return{runs:runs.length,finished:runs.filter(run=>LOCAL_TERMINAL_STATES.has(run.outcome)).length,succeeded:runs.filter(run=>run.outcome==="completed").length,judged:judged.length,passed,passRate:judged.length?passed/judged.length:null,medianTurns:median(recorded(runs.map(run=>run.turns))),medianUsdCost:null,medianDurationMs:median(recorded(runs.map(run=>run.durationMs))),criteria:[...byCriterion.values()].map(criterion=>({...criterion,passRate:criterion.judged?criterion.passed/criterion.judged:null}))}}function diff(a,b){return a===null||b===null?null:b-a}function localRow(baseline,candidate){const a=localStats(baseline);const b=localStats(candidate);return{baseline:a,candidate:b,delta:{passRate:diff(a.passRate,b.passRate),medianTurns:diff(a.medianTurns,b.medianTurns),medianUsdCost:null,medianDurationMs:diff(a.medianDurationMs,b.medianDurationMs),criteria:[...new Set([...a.criteria,...b.criteria].map(criterion=>criterion.name))].map(name=>({name,passRate:diff(a.criteria.find(criterion=>criterion.name===name)?.passRate??null,b.criteria.find(criterion=>criterion.name===name)?.passRate??null)}))}}}function compareLocalRuns(baseline,candidate){const targets=new Map;for(const run of[...baseline,...candidate]){const agent=LOCAL_AGENTS2.has(run.agent)?run.agent:void 0;if(agent)targets.set(`${run.agent} ${run.model}`,{agent,model:run.model})}const forTarget=(runs,key)=>runs.filter(run=>`${run.agent} ${run.model}`===key);return{overall:localRow(baseline,candidate),targets:[...targets].map(([key,target])=>({...target,...localRow(forTarget(baseline,key),forTarget(candidate,key))})),sides:{baseline:localRevisionMetadata(baseline),candidate:localRevisionMetadata(candidate)}}}function localRevisionMetadata(runs){const byMeasurement=new Map;const allRevisions=new Set;for(const run of runs){if(run.measurementRevision===void 0||run.measurementRevision===null)continue;const revision=String(run.measurementRevision);allRevisions.add(revision);if(run.evalId){const revisions=byMeasurement.get(run.evalId)??new Set;revisions.add(revision);byMeasurement.set(run.evalId,revisions)}}return{evalRevisions:[...allRevisions].sort((a,b)=>Number(a)-Number(b)),mixedRevisions:[...byMeasurement].filter(([,revisions])=>revisions.size>1).map(([evalSetId,revisions])=>({evalSetId,promptId:null,revisions:[...revisions]})),truncated:false}}function localCompareWarnings(result){const warnings=compareWarnings(result).filter(warning=>!warning.includes("matched no runs")).map(warning=>warning.replace("gauge run wait <request-id>","gauge run wait --batch <batch-id> --local").replace("revision=<hash>","revision=<number>").replace("not judged yet, or no criteria","analysis missing, unassessable criteria, or no criteria"));for(const name of["baseline","candidate"])if(result.overall[name].runs===0)warnings.unshift(`${name} matched no local runs. Check its selectors: batch=, eval=<saved id>, revision=<number>, agent=, model=, since=, until=. \`gauge run list --local\` shows the values.`);return warnings}function runPath(id){return`/api/v1/runs/${encodeURIComponent(id)}`}function streamPath(id){return`/api/runs/${encodeURIComponent(id)}/stream`}function fmtDuration(ms){if(ms==null)return"";const s=Math.round(ms/1e3);return s>=60?`${Math.floor(s/60)}m${s%60}s`:`${s}s`}function fmtUsd(v){return v==null?"":`$${v.toFixed(2)}`}function compact(value,max=140){const s=typeof value==="string"?value:JSON.stringify(value)??"null";const one=s.replace(/\s+/g," ").trim();return one.length>max?`${one.slice(0,max-1)}\u2026`:one}function toRow4(run){const selected=run.selectedProvider?.slug??"";const observed=run.observedProvider?.slug;return{id:run.id,status:run.status,agent:run.agent,model:run.modelDisplayName??run.resolvedModel??run.model,provider:observed&&observed!==selected?`${selected||"?"} \u2192 ${observed}`:selected||observed||"",experiment:run.experiment??"",revision:run.evalRevision??"",trial:run.optimizationTrial?`${run.optimizationTrial.key} ${run.optimizationTrial.trialId}`:"",source:run.triggeredBy?.source??"",turns:run.turns??"",cost:fmtUsd(run.usdCost),credits:run.creditsCharged??"",duration:fmtDuration(run.durationMs),created:run.createdAt,prompt:compact(run.promptSnippet,48)}}function toLocalRow(run){return{id:run.id,status:run.outcome,agent:run.agent,model:run.model,batch:run.batchId,eval:run.evalId??"",turns:run.turns,analyzed:run.analyzed?"yes":"",ended:run.endedAt}}function detailLines(run){return[["Id",run.id],["Status",run.status],["Agent",run.agent],["Model",run.modelDisplayName||run.resolvedModel||run.model||"(platform default)"],["Requested model",run.model||"(platform default)"],["Selected provider",run.selectedProvider?`${run.selectedProvider.displayName} (${run.selectedProvider.slug})`:"\u2014"],["Observed provider",run.observedProvider?`${run.observedProvider.slug} (${run.observedProvider.source.toLowerCase().replaceAll("_"," ")})`:"\u2014"],["Prompt",compact(run.promptText,120)],["Prompt id",run.promptId],["Batch",run.batchId],["Experiment",run.experiment??"\u2014"],["Eval set",run.evalSetId??"\u2014"],["Eval revision",run.evalRevision??"\u2014"],["Optimization trial",run.optimizationTrial?`${run.optimizationTrial.key} "${run.optimizationTrial.title}" (${run.optimizationTrial.kind}), trial ${run.optimizationTrial.trialId} in optimization ${run.optimizationTrial.optimizationId}`:"\u2014"],["Scenario",run.scenario],["Funding",run.funding],["Repo",run.repoUrl?`${run.repoUrl}${run.repoRef?` @ ${run.repoRef}`:""}`:"\u2014"],["Attempt",String(run.attempt)],["Turns",run.turns==null?"\u2014":String(run.turns)],["Triggered by",triggeredByLabel(run.triggeredBy)||"\u2014"],["Cost",fmtUsd(run.usdCost)||"\u2014"],["Credits",run.creditsCharged==null?"\u2014":String(run.creditsCharged)],["Duration",fmtDuration(run.durationMs)||"\u2014"],["Exit reason",run.exitReason??"\u2014"],["Failure reason",run.failureReason??"\u2014"],["Cancel requested",run.cancelRequested?"yes":"no"],["Diff",run.hasDiff?`yes (gauge run diff ${run.id})`:"no"],...run.criterionResults.map(c=>[`Criterion ${c.name}`,c.passed===null?"not assessed":c.passed?"passed":"failed"]),["Created",run.createdAt],["Started",run.startedAt??"\u2014"],["Finished",run.finishedAt??"\u2014"]]}function passLine(label,pass){const bits=[pass.status];if(pass.model)bits.push(pass.model);const cost=fmtUsd(pass.usdCost);if(cost)bits.push(cost);if(pass.error)bits.push(`error: ${compact(pass.error,60)}`);return`${label.padEnd(20)} ${bits.join(" ")}`}var PASS_LABELS={ANALYSIS:"Session analysis",CLI_DISCOVERY:"CLI discovery",LEVER:"Improvement levers",EXPERIMENT_DELTA:"Experiment delta"};var INSTALLED={NONE:"",ATTEMPTED:"attempted",CONFIGURED:"configured",EXERCISED:"exercised"};var check2=value=>value?"yes":"";function printInsight(insight){for(const pass of insight.passes){console.log(passLine(PASS_LABELS[pass.kind]??pass.kind,pass))}const analysis=insight.analysis;if(!analysis){console.log(insight.decision.status==="SKIPPED"?"\n(skipped)":"\n(no analysis yet)");return}if(analysis.limitation)console.log(`
|
|
958
|
+
`}function commandPath(program2,argv){const names=[];let command=program2;for(const token of argv){if(token.startsWith("-"))continue;const next=command.commands.find(sub=>sub.name()===token||sub.aliases().includes(token));if(next){names.push(token);command=next}else if(command.commands.length===0)break}return names.join(" ")}function wantsJson(argv){return argv.some((token,i)=>token==="--output=json"||token==="-ojson"||(token==="-o"||token==="--output")&&argv[i+1]==="json")}var record=value=>value!==null&&typeof value==="object"&&!Array.isArray(value)?value:{};var message=error=>oneLine(error instanceof CliError&&error.hint?`${error.message} ${error.hint}`:error instanceof Error?error.message:String(error));var oneLine=text=>readable(text).trim().replace(/\n\s*/g,"; ");var check=(id,state,detail)=>({id,state,detail});var dynamic=value=>typeof value!=="string"||value.includes("${{");var condition=value=>{if(value===void 0||value===true)return true;if(value===false)return false;if(typeof value==="string"){const compact2=value.replace(/\s/g,"");if(compact2==="true"||compact2===`\${{true}}`)return true;if(compact2==="false"||compact2===`\${{false}}`)return false}return void 0};function git(cwd,args){return execFileSync("git",args,{cwd,encoding:"utf8",stdio:["ignore","pipe","pipe"],maxBuffer:16*1024*1024})}function rootDirectory(cwd){return findWorkspace(cwd)??resolve3(cwd)}function regularFile(root,path){const full=join9(root,path);try{const rel=relative3(realpathSync3(root),realpathSync3(full));return rel!==".."&&!rel.startsWith(`..${sep}`)&&!isAbsolute3(rel)&&lstatSync(full).isFile()}catch{return false}}function read(root,path){if(!regularFile(root,path))throw new Error(`${path}: missing regular file inside the working tree`);if(lstatSync(join9(root,path)).size>1024*1024)throw new Error(`${path}: exceeds the 1 MiB static inspection limit`);return readFileSync11(join9(root,path))}function ignoredByGit(root){try{git(root,["check-ignore","-q","."]);return true}catch{return false}}function workingFiles(root){let paths;try{if(ignoredByGit(root))throw new Error("workspace is ignored by Git");paths=git(root,["ls-files","--cached","--others","--exclude-standard","-z"]).split("\0").filter(Boolean)}catch{paths=[];const walk=directory=>{for(const entry of readdirSync(join9(root,directory),{withFileTypes:true})){if([".git",".gauge","node_modules",".next",".venv","venv"].includes(entry.name))continue;const path=directory?`${directory}/${entry.name}`:entry.name;if(entry.isDirectory())walk(path);else if(entry.isFile())paths.push(path);if(paths.length>5e4)throw new Error("Working tree exceeds 50,000 files; initialize Git and ignore generated files before verifying")}};walk("")}return[...new Set(paths)].filter(path=>regularFile(root,path)).sort()}var skippedInputEntry=name=>[".git",".gauge","node_modules",".aws",".ssh",".npmrc",".pypirc"].includes(name)||name===".env"||name.startsWith(".env.");function diskFiles(root,base){const paths=[];const walk=directory=>{for(const entry of readdirSync(join9(root,directory),{withFileTypes:true})){if(skippedInputEntry(entry.name))continue;const path=directory?`${directory}/${entry.name}`:entry.name;if(entry.isDirectory())walk(path);else if(entry.isFile())paths.push(path);if(paths.length>2e4)throw new Error("Input exceeds 20,000 files")}};walk(base==="."?"":base);return paths.sort()}function workflowChecks(root,input){if(input.source.type!=="github-actions")return[];const source=input.source;const checks=[];let workflow;try{const document=parseDocument2(read(root,source.workflow).toString("utf8"),{uniqueKeys:true});if(document.errors.length)throw new Error(document.errors.map(message).join("; "));workflow=record(document.toJS());if(!Object.keys(record(workflow.jobs)).length)throw new Error("Expected a workflow jobs mapping");checks.push(check("workflow","yes",`${source.workflow} exists with a parseable YAML jobs mapping`))}catch(error){return[check("workflow","no",message(error))]}const jobs=Object.entries(record(workflow.jobs)).map(([name,value])=>({name,job:record(value)})).filter(({job})=>condition(job.if)!==false);const steps=jobs.flatMap(({name,job})=>Array.isArray(job.steps)?job.steps.map(value=>({job:name,jobCondition:condition(job.if),step:record(value)})).filter(({step})=>condition(step.if)!==false):[]);const opaque=jobs.some(({job})=>typeof job.uses==="string")||steps.some(({step})=>typeof step.uses==="string"&&!/^actions\/(checkout|upload-artifact|setup-[^@]+)@/.test(step.uses));const uploads=steps.filter(({step})=>typeof step.uses==="string"&&step.uses.startsWith("actions/upload-artifact@"));const candidates=uploads.filter(({step})=>{const name=record(step.with).name??"artifact";return!source.artifact||dynamic(name)||minimatch2(String(name),source.artifact,{nonegate:true})});const recognizedUpload=candidates.length===1&&!dynamic(record(candidates[0].step.with).name??"artifact");checks.push(check("artifact-upload",recognizedUpload?"yes":"unknown",recognizedUpload?`One direct upload step declares the matching artifact ${String(record(candidates[0].step.with).name??"artifact")}`:candidates.length?`Found ${candidates.length} potentially matching upload step(s); resolve dynamic artifact names or overlapping uploads to establish a unique selection`:`No direct upload step matches ${source.artifact??"the configured artifact"}; ${opaque?"inspect the referenced reusable workflows or custom actions":"inspect build scripts or declare a matching actions/upload-artifact step"}`));for(const[index,{job,jobCondition,step}]of candidates.entries()){const options=record(step.with);const version=/^actions\/upload-artifact@v(\d+)(?:\.\d+){0,2}$/.exec(String(step.uses));const major=version?Number(version[1]):void 0;let state=source.artifact&&dynamic(options.name??"artifact")?"unknown":major===7?options.archive===true||options.archive==="true"?"yes":options.archive===void 0||options.archive===false||options.archive==="false"?"no":"unknown":major!==void 0&&major>=4&&major<=6?"yes":"unknown";const conditional=jobCondition===void 0||condition(step.if)===void 0;if(state==="no"&&conditional)state="unknown";const detail=source.artifact&&dynamic(options.name??"artifact")?"Resolve the artifact name expression to determine whether this upload is selected":major===7?options.archive===true||options.archive==="true"?"archive: true declares a ZIP artifact":options.archive===void 0||options.archive===false||options.archive==="false"?`archive: true is required for Gauge's ZIP download${conditional?"; resolve the upload condition to determine whether this incompatible upload is selected":""}`:"Resolve the archive expression; Gauge requires archive: true":major!==void 0&&major>=4&&major<=6?"This action version produces ZIP artifacts":"Resolve the pinned action version; ZIP format is recognized for upload-artifact v4\u2013v7 (v7 requires archive: true)";checks.push(check(`artifact-format.${index}`,state,`${job}: ${String(step.uses)}; ${detail}`))}const checkouts=steps.filter(({step})=>typeof step.uses==="string"&&step.uses.startsWith("actions/checkout@"));const events=typeof workflow.on==="string"?[workflow.on]:Array.isArray(workflow.on)?workflow.on:Object.keys(record(workflow.on));const pr=events.includes("pull_request");const recognizedCheckout=step=>{const options=record(step.with);if(options.repository!==void 0)return false;if(!events.length||options.ref!==void 0&&typeof options.ref!=="string")return false;const ref=typeof options.ref==="string"?options.ref.replace(/\s/g,""):"";return events.every(event=>["pull_request","push","workflow_dispatch"].includes(event))&&(!ref||ref===`\${{github.sha}}`)||ref===`\${{github.event.pull_request.head.sha||github.sha}}`||pr&&events.every(event=>event==="pull_request")&&ref===`\${{github.event.pull_request.head.sha}}`};const checkoutState=checkouts.length>0&&checkouts.every(({step})=>recognizedCheckout(step))?"yes":"unknown";checks.push(check("candidate-checkout",checkoutState,checkoutState==="yes"?"Workflow checkout recognized, including the default PR merge checkout":"Could not inspect checkout settings; check that your workflow builds the intended code"));const filtered=Object.values(record(workflow.on)).some(value=>["paths","paths-ignore","branches","branches-ignore","types"].some(key=>record(value)[key]!==void 0));const eligible=events.filter(event=>["pull_request","push","workflow_dispatch"].includes(String(event)));const knownEvents=events.length>0&&events.every(event=>!dynamic(event));checks.push(check("workflow-trigger",eligible.length?"yes":knownEvents?"no":"unknown",eligible.length?`Eligible trigger(s) declared: ${eligible.join(", ")}${filtered?"; event filters are present":""}`:knownEvents?`Declared event(s) ${events.join(", ")} are not eligible artifact sources; declare push, pull_request, or workflow_dispatch`:"Could not read workflow events; declare push, pull_request, or workflow_dispatch"));return checks}function inputChecks(root,files,input,local){const checks=[check("configuration","yes",`Supported ${input.type} input from ${input.source.type==="git"?"project files":input.source.type}`)];if(input.type==="website")return checks;if(input.type==="binary"){const excluded=Object.entries(input.commands).filter(([,target])=>!input.paths.some(pattern=>minimatch2(target,pattern,{dot:true,nocase:false,nonegate:true})));checks.push(check("command-selection",excluded.length?"no":"yes",excluded.length?`Declared command targets are excluded by paths: ${excluded.map(([name,target])=>`${name}: ${target}`).join(", ")}`:"Declared command targets match selected paths"))}if(input.source.type==="github-actions"&&!local)checks.push(...workflowChecks(root,input));else{try{const patterns=input.type==="files"||input.type==="binary"?input.paths:input.type==="skill"&&!isSkillArchive(input.path)?[input.path==="."?"**":`${input.path}/**`]:[input.path];const selected=local&&input.type==="npm-package"?globSync(input.path,{cwd:root}):matchInputPaths(local?diskFiles(root,input.type==="skill"&&!isSkillArchive(input.path)?input.path:""):files,patterns);if(input.type==="binary"){for(const[command,target]of Object.entries(input.commands)){if(!selected.includes(target))throw new Error(`Binary command ${command} target is missing or excluded by paths: ${target}`)}}if(input.type==="web-fixture")parseWebFixtureInput(read(root,input.path).toString("utf8"));if(input.type==="npm-package"&&(selected.length!==1||!selected[0].endsWith(".tgz")))throw new Error("npm-package must select exactly one .tgz file");if(input.type==="skill"&&!isSkillArchive(input.path)&&!selected.includes(input.path==="."?"SKILL.md":`${input.path}/SKILL.md`))throw new Error("Skill directory must contain SKILL.md at its root");checks.push(check("local-paths","yes",`${selected.length} ${local?"local":"working-tree"} file(s) selected`))}catch(error){checks.push(check("local-paths","no",message(error).replace("committed files",local?"local files":"working-tree files")))}}return checks}function verifyEvals(cwd=process.cwd(),{local=false}={}){const root=rootDirectory(cwd);const result={version:1,mode:"local-static",root,status:"passed",summary:{yes:0,no:0,unknown:0},facts:{configFilePresent:null,caseCount:null},info:[],notChecked:["github-app-activation","account-settings","mcp-server-access","mcp-server-reachability","live-input-delivery","artifact-contents","sandbox-installation","workflow-execution"],global:[],inputs:[]};const finish=()=>{for(const item of[...result.global,...result.inputs.flatMap(input=>input.checks)])result.summary[item.state]++;result.status=result.summary.no>0?"failed":result.summary.unknown>0?"incomplete":"passed";return result};let raw={};let files;try{files=workingFiles(root)}catch(error){result.global.push(check("working-tree","no",message(error)));return finish()}try{result.facts.configFilePresent=files.includes("gauge.json")||regularFile(root,"gauge.json");if(result.facts.configFilePresent)raw=JSON.parse(read(root,"gauge.json").toString("utf8"));else result.info.push({id:"config-file-absent",detail:"Optional gauge.json is absent; using default case discovery and no inputs"})}catch(error){result.global.push(check("configuration","no",message(error)));return finish()}const config=runDefinitionFileConfigSchema.safeParse(raw);result.global.push(check("configuration",config.success?"yes":"no",config.success?"Configuration matches the supported gauge.json schema":oneLine(config.error.message)));const rawInputs=record(record(raw).inputs);for(const[name,value]of Object.entries(rawInputs).sort(([a],[b])=>a.localeCompare(b))){const parsed=runDefinitionInputSchema.safeParse(value);result.inputs.push({name,checks:parsed.success?inputChecks(root,files,parsed.data,local):[check("configuration","no",oneLine(parsed.error.message))]})}if(config.success){try{const paths=matchInputPaths(files,config.data.cases,{allowEmptySelection:true});if(paths.length>20||paths.some(path=>!path.endsWith(".md")))throw new Error("Expected at most 20 Markdown cases");result.facts.caseCount=paths.length;if(paths.length)result.global.push(check("case-discovery","yes",`${paths.length} working-tree case(s) selected`));else result.info.push({id:"no-cases",detail:"No eval cases found; input-only setup is valid, but no evaluations will run"});for(const path of paths){try{const item=parseRunDefinitionCase(path,read(root,path));if(local)assertLocalCapabilities(item.config,`${path}: config.`);const missing=(item.config.inputs??[]).filter(name=>!Object.hasOwn(config.data.inputs,name));if(missing.length)throw new Error(`${path}: Unknown input(s): ${missing.join(", ")}`);const commandOwners=new Map;mergeWebFixtures((item.config.inputs??Object.keys(config.data.inputs)).flatMap(name=>{const input=config.data.inputs[name];return input.type==="web-fixture"&&(local||input.source.type==="git")?[parseWebFixtureInput(read(root,input.path).toString("utf8"))]:[]}));for(const name of new Set(item.config.inputs??Object.keys(config.data.inputs))){const input=config.data.inputs[name];if(input.type!=="binary")continue;for(const command of Object.keys(input.commands)){const owner=commandOwners.get(command);if(owner)throw new Error(`${path}: Inputs ${owner} and ${name} both expose command ${command}; choose distinct commands or separate cases`);commandOwners.set(command,name)}}result.global.push(check(`case.${path}`,"yes","Case syntax, input references, and declared command names are valid"))}catch(error){result.global.push(check(`case.${path}`,"no",message(error)))}}}catch(error){result.global.push(check("case-discovery","no",message(error).replace("committed files","working-tree files")))}}else result.info.push({id:"case-validation-skipped",detail:"Fix gauge.json before validating cases and input references"});return finish()}function printEvalVerification(result){console.log(`Local static verification: ${result.root}`);const print2=checks=>{for(const item of checks)console.log(` [${item.state}] ${item.id}: ${item.detail}`)};console.log("Global");print2(result.global);for(const input of result.inputs){console.log(`Input: ${input.name}`);print2(input.checks)}for(const item of result.info)console.log(`[info] ${item.id}: ${item.detail}`);console.log(`Static verification ${result.status}: ${result.summary.yes} yes, ${result.summary.no} no, ${result.summary.unknown} unknown.`);console.log("Scope: repository setup only. Live delivery, account settings, and sandbox execution were not tested; no sessions were launched.")}var LIMIT=512*1024*1024;var SNAPSHOT_FILES=100;function parseInputBindings(values){const bindings=new Map;for(const value of values){const split=value.indexOf("=");const name=split<0?value:value.slice(0,split);const path=split<0?void 0:value.slice(split+1);if(!/^[a-z0-9]+(?:-[a-z0-9]+)*$/.test(name)||path==="")throw new CliError("usage","--input expects name or name=path");if(bindings.has(name))throw new CliError("usage",`Duplicate --input: ${name}`);bindings.set(name,path)}return bindings}var isGlob=path=>/[*?[\]{}]/.test(path);function rejectSkippedPatterns(name,patterns){for(const pattern of patterns){const skipped=pattern.split("/").find(skippedInputEntry);if(skipped===".gauge")throw new CliError("usage",`Input ${name}: ${pattern} is inside .gauge/. Inputs never include .gauge/, because it holds run results and local config.`,{hint:"Move the files outside .gauge/ and update the glob."});if(skipped)throw new CliError("usage",`Input ${name}: ${pattern} is inside ${skipped}, which inputs never include`,{hint:`Move the files outside ${skipped} and update the glob.`})}}async function walkInput(root,depth=Infinity){const paths=[];const walk=async(directory,level)=>{for(const entry of await readdir(join10(root,directory),{withFileTypes:true})){if(skippedInputEntry(entry.name)||!entry.isFile()&&!entry.isDirectory())continue;const next=directory?`${directory}/${entry.name}`:entry.name;if(entry.isDirectory()){if(level<depth)await walk(next,level+1)}else paths.push(next);if(paths.length>2e4)throw new CliError("usage","Input exceeds 20,000 files")}};await walk("",1);return paths.sort()}function selectPaths(name,paths,patterns){rejectSkippedPatterns(name,patterns);try{return matchInputPaths(paths,patterns)}catch(error){throw new CliError("usage",`Input ${name}: ${message(error).replace("committed files","files")}`)}}async function expandOverrideGlob(name,glob){const segments=glob.split(sep2);const first=segments.findIndex(isGlob);const root=segments.slice(0,first).join(sep2)||sep2;const pattern=segments.slice(first).join("/");const info=await stat(root).catch(()=>void 0);if(!info?.isDirectory())throw new CliError("usage",`Input ${name}: ${root} is not a directory`);const depth=pattern.includes("**")?Infinity:pattern.split("/").length;return{root,pattern,files:selectPaths(name,await walkInput(root,depth),[pattern])}}async function selectLocalInput(resource,supplied,override){let path=resolve4(supplied);let info=await lstat(path).catch(()=>void 0);if(!info&&override&&isGlob(path)){if(resource.use.type==="skill")throw new CliError("usage",`Input ${resource.name}: Skill inputs take an explicit directory or archive, not a glob`);const{root:root2,pattern,files}=await expandOverrideGlob(resource.name,path);if(resource.use.type==="files")return{use:{type:"files"},root:root2,files,skillDirectory:false};if(resource.use.type==="binary")return{use:{...resource.use,paths:[pattern]},root:root2,files,skillDirectory:false};if(files.length!==1)throw new CliError("usage",`Input ${resource.name}: ${supplied} matches ${files.length} files; keep one`);path=join10(root2,files[0]);info=await lstat(path)}if(!info)throw new CliError("usage",`Input path does not exist: ${supplied}`,{hint:"Pass --input <name>=<path>, or fix the input's path in gauge.json."});if(!info.isFile()&&!info.isDirectory())throw new CliError("usage",`Input must be an ordinary file or directory: ${supplied}`);let use=resource.use;if(use.type==="web-fixture"){if(!info.isFile())throw new CliError("usage","Supply a web fixture JSON file");use={type:"web-fixture",path:basename(path)}}else if(use.type==="skill"){if(info.isFile()&&!isSkillArchive(path))throw new CliError("usage","Supply a Skill directory, .zip, or .tar.gz archive");use={...use,root:info.isDirectory()?".":basename(path)}}else if(use.type==="npm-package"||use.type==="npm-cli"){if(!info.isFile()||!path.endsWith(".tgz"))throw new CliError("usage","Supply the packed npm .tgz file");use={...use,package:basename(path)}}else if(override&&use.type==="files")use={type:"files"};else if(override&&use.type==="binary")use={...use,paths:[info.isDirectory()?"**":basename(path)]};const root=info.isDirectory()?path:dirname(path);const paths=info.isDirectory()?await walkInput(root):[basename(path)];const patterns=!override&&(use.type==="files"||use.type==="binary")?use.paths:void 0;return{use,root,files:patterns?selectPaths(resource.name,paths,patterns):paths,skillDirectory:use.type==="skill"&&info.isDirectory()}}async function packLocalInput(resource,supplied,output,{override=false}={}){if(resource.use.type==="website")throw new CliError("usage",`Input ${resource.name}: website inputs cannot be supplied through --input.`,{hint:"Use the configured public preview with --hosted. For content-only alternatives and their discovery limits, read `gauge instructions hillclimb`."});const{use,root,files,skillDirectory}=await selectLocalInput(resource,supplied,override);if(!files.length)throw new CliError("usage","Input contains no selected files");if(use.type==="binary"){for(const[command,target]of Object.entries(use.commands))if(!files.includes(target))throw new CliError("usage",`Input ${resource.name}: binary command ${command} target is missing: ${target}`,override?{hint:"Command targets are relative to the supplied path."}:void 0)}if(skillDirectory&&!files.includes("SKILL.md"))throw new CliError("usage","Skill directory must contain SKILL.md at its root");let total=0;for(const file of files){inputPathSchema.parse(file);const entry=await lstat(join10(root,file));if(!entry.isFile())throw new CliError("usage",`Input file changed or is not an ordinary file: ${file}`);total+=entry.size;if(total>LIMIT)throw new CliError("usage","Input exceeds 512 MiB")}await tar({cwd:root,file:output,gzip:true,portable:true,noMtime:true,follow:false,noDirRecurse:true},files.map(file=>`./${file}`));const bytes=(await stat(output)).size;if(bytes>LIMIT)throw new CliError("usage","Input archive exceeds 512 MiB");const hash=createHash5("sha256");for await(const chunk of createReadStream2(output))hash.update(chunk);const sha256=`sha256:${hash.digest("hex")}`;if(use.type==="web-fixture")await readWebFixtureArchive(output,use.path,sha256);return{use,upload:{path:output,bytes,sha256},files}}async function loadLocalRunDefinition(files,values,cwd=process.cwd(),{local=false,spec}={}){const directory=await mkdtemp(join10(tmpdir(),"gauge-inputs-"));const cleanup=()=>rm(directory,{recursive:true,force:true});try{const root=rootDirectory(cwd);const paths=workingFiles(root);const blobs=new Map;if(paths.includes("gauge.json"))blobs.set("gauge.json",read(root,"gauge.json"));const configBytes=blobs.get("gauge.json");let config;try{config=runDefinitionFileConfigSchema.parse(configBytes?JSON.parse(configBytes.toString("utf8")):{})}catch(error){throw new CliError("usage",`gauge.json is not valid: ${error instanceof Error?error.message:error}`,{hint:"Run `gauge run --verify` for each problem."})}const patterns=spec?[]:files.length?files.map(file=>{const path=relative4(root,resolve4(cwd,file));if(!path||path===".."||path.startsWith(`..${sep2}`)||isAbsolute4(path))throw new CliError("usage","Case paths must be inside the definition directory");return path.split(sep2).join("/")}):config.cases;const cases=spec?[]:files.length?matchInputPaths(paths,patterns):matchInputPaths(paths,patterns,{allowEmptySelection:true});if(!spec&&cases.length===0)throw noCasesError();if(cases.length>20||cases.some(path=>!path.endsWith(".md")))throw new CliError("usage","Expected 1\u201320 Markdown cases");for(const path of cases)blobs.set(path,read(root,path));const parsed=spec?caseSpecCases(spec):cases.map(path=>parseRunDefinitionCase(path,blobs.get(path)??Buffer.alloc(0)));const unmeasured=parsed.filter(item=>!item.criteria&&!item.preference);if(!local&&unmeasured.length)throw new CliError("usage",`Hosted cases need criteria or preference: ${unmeasured.map(item=>item.path).join(", ")}`,{hint:"Add criteria, or run locally with --local."});const bindings=parseInputBindings(values);for(const name of bindings.keys())if(!(name in config.inputs))throw new CliError("usage",`Unknown input: ${name}`,{hint:`gauge.json defines: ${Object.keys(config.inputs).join(", ")||"none"}`});for(const item of parsed)for(const name of item.config.inputs??[])if(!(name in config.inputs))throw new CliError("usage",`${item.path}: unknown input "${name}". gauge.json defines: ${Object.keys(config.inputs).join(", ")||"none"}`);const selected=new Set(parsed.flatMap(item=>item.config.inputs??Object.keys(config.inputs)));for(const[name,override]of bindings){if(!selected.has(name))throw new CliError("usage",`Input is not used by the selected cases: ${name}`);if(override!==void 0)rejectSkippedPatterns(name,[relative4(root,resolve4(cwd,override)).split(sep2).join("/")])}const git2=args=>{try{return execFileSync2("git",args,{cwd:root,encoding:"utf8",stdio:["ignore","pipe","pipe"]}).trim()}catch{return null}};const commit=git2(["rev-parse","--verify","HEAD^{commit}"]);const origin=git2(["remote","get-url","origin"]);const repoUrl=origin&&uncredentialedRepoUrlSchema.safeParse(origin).success?origin:null;const unattachedInputs=Object.keys(config.inputs).filter(name=>!selected.has(name));const resources=[];const uploads=[];const inputSnapshots=[];for(const[name,input]of Object.entries(config.inputs)){if(!selected.has(name))continue;if(local&&input.type==="website")throw new CliError("usage",`Website input "${name}" requires hosted Gauge to replace the original site with your preview.`,{hint:"Use --hosted with a public preview. For local runs, supply documentation files or a preview URL in the prompt. These give the agent your content directly rather than testing whether it finds your site. See `gauge instructions hillclimb`."});const original=compileDefinitionInput(name,input,repoUrl??"",commit??"");const override=bindings.get(name);const supplied=override??(original.source.type==="git"?workingTreeInput(root,name,input):void 0);if(supplied!==void 0){const{use,upload,files:files2}=await packLocalInput(original,resolve4(cwd,supplied),join10(directory,`${name}.tar.gz`),{override:override!==void 0});uploads.push(upload);inputSnapshots.push({name,source:override!==void 0?"override":"working-tree",path:resolve4(cwd,supplied),gaugeArchive:{sha256:upload.sha256,bytes:upload.bytes},files:files2.slice(0,SNAPSHOT_FILES),fileCount:files2.length});resources.push({...original,use,source:{type:"upload",sha256:upload.sha256,bytes:upload.bytes,repoUrl,commit}})}else{if(local)throw new CliError("usage",`Input ${name} has a remote source. Supply --input ${name}=<path> for local execution, or use --hosted.`);if(!repoUrl||!commit)throw new CliError("usage",`Input ${name}'s configured ${original.source.type} source requires a Git checkout with an origin.`,{hint:input.type==="website"?"Configure the repository origin and publish a matching public preview through the declared source. Website inputs cannot use --input; see `gauge instructions hillclimb` for content-only alternatives.":`Supply --input ${name}=<path> to use local files.`});resources.push(original)}}if(!spec){const compiled=compileRunDefinitionFromBlobs({repoUrl,commit,resources,blobs,casePaths:cases,verification:"LOCAL"});const webFixtures=await freezeWebFixtures({...compiled.definition,resources:resources.filter(r=>r.source.type==="upload")},uploads);return{...compiled,webFixtures,uploads,inputSnapshots,unattachedInputs,cleanup}}const definition={version:2,source:{repoUrl,commit,digest:caseSpecDigest(spec),verification:"LOCAL"},resources,cases:parsed};if(!local)runDefinitionSubmissionSchema.parse(definition);return{definition,webFixtures:await freezeWebFixtures({...definition,resources:resources.filter(r=>r.source.type==="upload")},uploads),org:config.orgId??config.org,uploads,inputSnapshots,unattachedInputs,cleanup}}catch(error){await cleanup();if(error instanceof CliError||!(error instanceof Error))throw error;throw new CliError("usage",error.message,{hint:"Check the case format in `gauge instructions run`, then run `gauge run --verify`."})}}function workingTreeInput(root,name,input){switch(input.type){case"website":throw new CliError("usage",`Input ${name} is a website; website inputs run on hosted Gauge only`);case"files":case"binary":return root;case"web-fixture":case"skill":return join10(root,input.path);case"npm-package":{const matches2=globSync2(input.path,{cwd:root});if(matches2.length!==1)throw new CliError("usage",matches2.length?`Input ${name}: ${input.path} matches ${matches2.length} files; keep one`:`Input ${name}: no file matches ${input.path}. Run \`npm pack\` first`);return join10(root,matches2[0])}}}async function uploadLocalInputs(client,org,uploads){for(const input of uploads){const signed2=await client.post(`/api/v1/orgs/${encodeURIComponent(org)}/run-definitions/uploads`,{sha256:input.sha256,bytes:input.bytes});const signal=AbortSignal.timeout(10*60*1e3);for(let attempt=0;attempt<3;attempt++){const body=createReadStream2(input.path);let response;try{response=await fetch(signed2.url,{method:"PUT",headers:signed2.headers,body,duplex:"half",redirect:"error",signal})}finally{body.destroy()}await response.body?.cancel();if(response.ok||response.status===412)break;if(response.status!==409||attempt===2)throw new CliError("network",`Input upload failed (${response.status})`);await delay2(100*2**attempt,void 0,{signal})}}}var SIDE_KEYS=["request","batch","eval","revision","experiment","trial","agent","model","since","until"];var SELECTING_KEYS=new Set(["request","batch","eval","experiment","trial"]);function csv4(value){return value.split(",").map(part=>part.trim()).filter(Boolean)}function parseCompareSide(flag,pairs){const side={};const seen=new Set;for(const pair2 of pairs){const eq=pair2.indexOf("=");const key=eq<0?"":pair2.slice(0,eq).trim();const value=eq<0?"":pair2.slice(eq+1).trim();if(!SIDE_KEYS.includes(key)||!value)throw new CliError("usage",`invalid ${flag} "${pair2}" (expected key=value with key one of ${SIDE_KEYS.join(", ")})`);if(seen.has(key)&&key!=="batch")throw new CliError("usage",`${flag} ${key}= given twice`);seen.add(key);if(key==="request")side.runRequest=value;else if(key==="batch")side.batch=[...side.batch??[],...csv4(value)];else if(key==="eval")side.evalSet=value;else if(key==="revision")side.evalRevision=value;else if(key==="experiment")side.experiment=value;else if(key==="trial")side.trial=value;else if(key==="agent")side.agent=parseAgents(value);else if(key==="model")side.model=csv4(value);else if(key==="since")side.since=value;else side.until=value}if(![...seen].some(key=>SELECTING_KEYS.has(key)))throw new CliError("usage",`${flag} needs request=, batch=, eval=, experiment=, or trial= to select runs`);return side}function pct2(rate){return rate===null?"":`${Math.round(rate*100)}%`}function passCell(stats){return stats.judged===0?"":`${pct2(stats.passRate)} (${stats.passed}/${stats.judged})`}function signed(value,format){if(value===null)return"";return`${value>0?"+":value<0?"-":"\xB1"}${format(Math.abs(value))}`}var points=rate=>`${Math.round(rate*100)}pp`;var turns=n=>String(Math.round(n*10)/10);var usd=n=>`$${n.toFixed(2)}`;var secs=ms=>`${Math.round(ms/1e3)}s`;function pair(before,after,delta3,format){if(before===null&&after===null)return"";const side=v=>v===null?"\u2014":format(v);return`${side(before)} \u2192 ${side(after)}${delta3===null?"":` (${signed(delta3,format)})`}`}function row(target,item){const{baseline:a,candidate:b,delta:delta3}=item;return{target,runs:`${a.runs} \u2192 ${b.runs}`,completed:`${a.succeeded}/${a.runs} \u2192 ${b.succeeded}/${b.runs}`,baseline:passCell(a),candidate:passCell(b),"\u0394 pass":signed(delta3.passRate,points),turns:pair(a.medianTurns,b.medianTurns,delta3.medianTurns,turns),cost:pair(a.medianUsdCost,b.medianUsdCost,delta3.medianUsdCost,usd),duration:pair(a.medianDurationMs,b.medianDurationMs,delta3.medianDurationMs,secs)}}function compareWarnings(result){const warnings=[];for(const name of["baseline","candidate"]){const side=result.sides[name];const stats=result.overall[name];if(stats.runs===0)warnings.push(`${name} matched no runs. Check its selectors; optimization trial runs need trial=<trialId>.`);else if(stats.finished<stats.runs)warnings.push(`${name} has ${stats.runs-stats.finished} of ${stats.runs} runs still in progress, so its numbers are partial. Run \`gauge run wait <request-id>\` first.`);if(stats.judged<stats.finished)warnings.push(`${name} has ${stats.finished-stats.judged} finished runs without a verdict (not judged yet, or no criteria). Its pass rate covers judged runs only.`);for(const mixed of side.mixedRevisions??[])warnings.push(`${name}: ${mixed.evalSetId?`eval ${mixed.evalSetId}`:`case prompt ${mixed.promptId}`} ran under ${mixed.revisions.length} revisions (${mixed.revisions.join(", ")}): it was edited between runs. Add --${name} revision=<hash> to pick one.`);if(side.truncated)warnings.push(`${name} matched too many runs; only the newest were compared. Narrow it with since=, until=, or batch=.`)}return warnings}function printComparison(result){const revisions=name=>result.sides[name].evalRevisions.join(", ")||"\u2014";console.log(`Baseline revision: ${revisions("baseline")}. Candidate revision: ${revisions("candidate")}.`);for(const name of["baseline","candidate"]){for(const trial of result.sides[name].trials??[])console.log(`${name}: ${trial.runs} run${trial.runs===1?"":"s"} from optimization trial ${trial.key} "${trial.title}" (${trial.kind}).`)}console.log("Medians are baseline \u2192 candidate (difference).");console.log("");printItems([row("all",result.overall),...result.targets.map(target=>row(`${target.agent} ${target.model||"(default)"}`,target))],"table");const before=new Map(result.overall.baseline.criteria.map(c=>[c.name,c]));const after=new Map(result.overall.candidate.criteria.map(c=>[c.name,c]));if(result.overall.delta.criteria.length>0){console.log("");printItems(result.overall.delta.criteria.map(c=>{const a=before.get(c.name);const b=after.get(c.name);return{criterion:c.name,baseline:a&&a.judged>0?`${pct2(a.passRate)} (${a.passed}/${a.judged})`:"",candidate:b&&b.judged>0?`${pct2(b.passRate)} (${b.passed}/${b.judged})`:"","\u0394 pass":signed(c.passRate,points)}}),"table")}}var LOCAL_AGENTS2=new Set(["CLAUDE_CODE","CODEX_CLI","CURSOR","PI","OPENCODE"]);var LOCAL_TERMINAL_STATES=new Set(["completed","failed","canceled","interrupted"]);function localCompareFilter(flag,side){const hostedOnly2=[side.runRequest!==void 0&&"request=",side.experiment!==void 0&&"experiment=",side.trial!==void 0&&"trial="].filter(Boolean);if(hostedOnly2.length>0)throw new CliError("usage",`${flag} ${hostedOnly2.join(", ")} select hosted runs; local sides take batch=, eval=<saved id>, revision=<number>, agent=, model=, since=, until=`);if((side.batch?.length??0)>1)throw new CliError("usage",`${flag} takes one batch= locally`);if((side.agent?.length??0)>1||(side.model?.length??0)>1)throw new CliError("usage",`${flag} takes one agent= and one model= locally`);if(side.evalRevision!==void 0&&!/^[1-9]\d*$/.test(side.evalRevision))throw new CliError("usage",`${flag} local revision= must be a positive integer`);return{batch:side.batch?.[0],evalId:side.evalSet,measurementRevision:side.evalRevision===void 0?void 0:Number(side.evalRevision),agent:side.agent?.[0],model:side.model?.[0],since:side.since?parseSince(side.since,Date.now(),"since="):void 0,until:side.until?parseSince(side.until,Date.now(),"until="):void 0}}function median(values){if(values.length===0)return null;const sorted=[...values].sort((a,b)=>a-b);const mid=sorted.length>>1;return sorted.length%2?sorted[mid]:(sorted[mid-1]+sorted[mid])/2}function localStats(runs){const judged=runs.filter(run=>run.criterionResults?.length&&run.criterionResults.every(criterion=>criterion.outcome!=="UNASSESSABLE"));const passed=judged.filter(run=>run.criterionResults.every(criterion=>criterion.outcome==="PASS")).length;const byCriterion=new Map;for(const run of runs)for(const criterion of run.criterionResults??[]){const current=byCriterion.get(criterion.name)??{name:criterion.name,judged:0,passed:0};if(criterion.outcome!=="UNASSESSABLE"){current.judged++;if(criterion.outcome==="PASS")current.passed++}byCriterion.set(criterion.name,current)}const recorded=values=>values.filter(value=>typeof value==="number"&&Number.isFinite(value));return{runs:runs.length,finished:runs.filter(run=>LOCAL_TERMINAL_STATES.has(run.outcome)).length,succeeded:runs.filter(run=>run.outcome==="completed").length,judged:judged.length,passed,passRate:judged.length?passed/judged.length:null,medianTurns:median(recorded(runs.map(run=>run.turns))),medianUsdCost:null,medianDurationMs:median(recorded(runs.map(run=>run.durationMs))),criteria:[...byCriterion.values()].map(criterion=>({...criterion,passRate:criterion.judged?criterion.passed/criterion.judged:null}))}}function diff(a,b){return a===null||b===null?null:b-a}function localRow(baseline,candidate){const a=localStats(baseline);const b=localStats(candidate);return{baseline:a,candidate:b,delta:{passRate:diff(a.passRate,b.passRate),medianTurns:diff(a.medianTurns,b.medianTurns),medianUsdCost:null,medianDurationMs:diff(a.medianDurationMs,b.medianDurationMs),criteria:[...new Set([...a.criteria,...b.criteria].map(criterion=>criterion.name))].map(name=>({name,passRate:diff(a.criteria.find(criterion=>criterion.name===name)?.passRate??null,b.criteria.find(criterion=>criterion.name===name)?.passRate??null)}))}}}function compareLocalRuns(baseline,candidate){const targets=new Map;for(const run of[...baseline,...candidate]){const agent=LOCAL_AGENTS2.has(run.agent)?run.agent:void 0;if(agent)targets.set(`${run.agent} ${run.model}`,{agent,model:run.model})}const forTarget=(runs,key)=>runs.filter(run=>`${run.agent} ${run.model}`===key);return{overall:localRow(baseline,candidate),targets:[...targets].map(([key,target])=>({...target,...localRow(forTarget(baseline,key),forTarget(candidate,key))})),sides:{baseline:localRevisionMetadata(baseline),candidate:localRevisionMetadata(candidate)}}}function localRevisionMetadata(runs){const byMeasurement=new Map;const allRevisions=new Set;for(const run of runs){if(run.measurementRevision===void 0||run.measurementRevision===null)continue;const revision=String(run.measurementRevision);allRevisions.add(revision);if(run.evalId){const revisions=byMeasurement.get(run.evalId)??new Set;revisions.add(revision);byMeasurement.set(run.evalId,revisions)}}return{evalRevisions:[...allRevisions].sort((a,b)=>Number(a)-Number(b)),mixedRevisions:[...byMeasurement].filter(([,revisions])=>revisions.size>1).map(([evalSetId,revisions])=>({evalSetId,promptId:null,revisions:[...revisions]})),truncated:false}}function localCompareWarnings(result){const warnings=compareWarnings(result).filter(warning=>!warning.includes("matched no runs")).map(warning=>warning.replace("gauge run wait <request-id>","gauge run wait --batch <batch-id> --local").replace("revision=<hash>","revision=<number>").replace("not judged yet, or no criteria","analysis missing, unassessable criteria, or no criteria"));for(const name of["baseline","candidate"])if(result.overall[name].runs===0)warnings.unshift(`${name} matched no local runs. Check its selectors: batch=, eval=<saved id>, revision=<number>, agent=, model=, since=, until=. \`gauge run list --local\` shows the values.`);return warnings}function runPath(id){return`/api/v1/runs/${encodeURIComponent(id)}`}function streamPath(id){return`/api/runs/${encodeURIComponent(id)}/stream`}function fmtDuration(ms){if(ms==null)return"";const s=Math.round(ms/1e3);return s>=60?`${Math.floor(s/60)}m${s%60}s`:`${s}s`}function fmtUsd(v){return v==null?"":`$${v.toFixed(2)}`}function compact(value,max=140){const s=typeof value==="string"?value:JSON.stringify(value)??"null";const one=s.replace(/\s+/g," ").trim();return one.length>max?`${one.slice(0,max-1)}\u2026`:one}function toRow4(run){const selected=run.selectedProvider?.slug??"";const observed=run.observedProvider?.slug;return{id:run.id,status:run.status,agent:run.agent,model:run.modelDisplayName??run.resolvedModel??run.model,provider:observed&&observed!==selected?`${selected||"?"} \u2192 ${observed}`:selected||observed||"",experiment:run.experiment??"",revision:run.evalRevision??"",trial:run.optimizationTrial?`${run.optimizationTrial.key} ${run.optimizationTrial.trialId}`:"",source:run.triggeredBy?.source??"",turns:run.turns??"",cost:fmtUsd(run.usdCost),credits:run.creditsCharged??"",duration:fmtDuration(run.durationMs),created:run.createdAt,prompt:compact(run.promptSnippet,48)}}function toLocalRow(run){return{id:run.id,status:run.outcome,agent:run.agent,model:run.model,batch:run.batchId,eval:run.evalId??"",turns:run.turns,analyzed:run.analyzed?"yes":"",ended:run.endedAt}}function detailLines(run){return[["Id",run.id],["Status",run.status],["Agent",run.agent],["Model",run.modelDisplayName||run.resolvedModel||run.model||"(platform default)"],["Requested model",run.model||"(platform default)"],["Selected provider",run.selectedProvider?`${run.selectedProvider.displayName} (${run.selectedProvider.slug})`:"\u2014"],["Observed provider",run.observedProvider?`${run.observedProvider.slug} (${run.observedProvider.source.toLowerCase().replaceAll("_"," ")})`:"\u2014"],["Prompt",compact(run.promptText,120)],["Prompt id",run.promptId],["Batch",run.batchId],["Experiment",run.experiment??"\u2014"],["Eval set",run.evalSetId??"\u2014"],["Eval revision",run.evalRevision??"\u2014"],["Optimization trial",run.optimizationTrial?`${run.optimizationTrial.key} "${run.optimizationTrial.title}" (${run.optimizationTrial.kind}), trial ${run.optimizationTrial.trialId} in optimization ${run.optimizationTrial.optimizationId}`:"\u2014"],["Scenario",run.scenario],["Funding",run.funding],["Repo",run.repoUrl?`${run.repoUrl}${run.repoRef?` @ ${run.repoRef}`:""}`:"\u2014"],["Attempt",String(run.attempt)],["Turns",run.turns==null?"\u2014":String(run.turns)],["Triggered by",triggeredByLabel(run.triggeredBy)||"\u2014"],["Cost",fmtUsd(run.usdCost)||"\u2014"],["Credits",run.creditsCharged==null?"\u2014":String(run.creditsCharged)],["Duration",fmtDuration(run.durationMs)||"\u2014"],["Exit reason",run.exitReason??"\u2014"],["Failure reason",run.failureReason??"\u2014"],["Cancel requested",run.cancelRequested?"yes":"no"],["Diff",run.hasDiff?`yes (gauge run diff ${run.id})`:"no"],...run.criterionResults.map(c=>[`Criterion ${c.name}`,c.passed===null?"not assessed":c.passed?"passed":"failed"]),["Created",run.createdAt],["Started",run.startedAt??"\u2014"],["Finished",run.finishedAt??"\u2014"]]}function passLine(label,pass){const bits=[pass.status];if(pass.model)bits.push(pass.model);const cost=fmtUsd(pass.usdCost);if(cost)bits.push(cost);if(pass.error)bits.push(`error: ${compact(pass.error,60)}`);return`${label.padEnd(20)} ${bits.join(" ")}`}var PASS_LABELS={ANALYSIS:"Session analysis",CLI_DISCOVERY:"CLI discovery",LEVER:"Improvement levers",EXPERIMENT_DELTA:"Experiment delta"};var INSTALLED={NONE:"",ATTEMPTED:"attempted",CONFIGURED:"configured",EXERCISED:"exercised"};var check2=value=>value?"yes":"";function printInsight(insight){for(const pass of insight.passes){console.log(passLine(PASS_LABELS[pass.kind]??pass.kind,pass))}const analysis=insight.analysis;if(!analysis){console.log(insight.decision.status==="SKIPPED"?"\n(skipped)":"\n(no analysis yet)");return}if(analysis.limitation)console.log(`
|
|
786
959
|
${compact(analysis.limitation,200)}`);console.log(`
|
|
787
|
-
Findings (${analysis.observations.length})`);for(const finding of analysis.observations){console.log(` ${finding.kind}: ${finding.label}`);console.log(` ${finding.detail}`);console.log(` Evidence: ${JSON.stringify(finding.evidence)}`)}console.log(`
|
|
788
|
-
Products (${analysis.products.length})`);if(analysis.products.length>0)printItems(analysis.products.map(p=>({product:p.name,considered:check2(p.considered),chosen:check2(p.chosen),rejected:check2(p.rejected),installed:INSTALLED[p.installPhase]??p.installPhase,rank:p.rank??"",source:p.source.toLowerCase()})),"table");const delta3=insight.experimentDelta;if(delta3&&typeof delta3==="object"){console.log("\nExperiment delta");console.log([delta3.changed==null?"":delta3.changed?"changed":"no change",delta3.hypothesisSupported??""].filter(Boolean).join(" "));if(delta3.summary)console.log(compact(delta3.summary,300))}console.error("\n(full detail: rerun with -o json)")}function parseEvent(data){try{const parsed=runEventSchema.safeParse(JSON.parse(data));return parsed.success?parsed.data:null}catch{return null}}function formatEventLine(e){const rec=e;const head=`[t${e.turn}#${e.seq}]`;switch(e.type){case"run_meta":{const agent=rec.agent;const model=agent?.model?` model=${agent.model}`:"";return`${head} run_meta agent=${agent?.name??"?"}${model}`}case"agent_message":return`${head} agent: ${compact(rec.text)}`;case"reasoning":return`${head} reasoning: ${compact(rec.text)}`;case"tool_call":return`${head} tool_call ${String(rec.name)}: ${compact(rec.input)}`;case"tool_result":return`${head} tool_result${rec.is_error?" ERROR":""}: ${compact(rec.output,100)}`;case"run_result":return`${head} run_result ${String(rec.exit_reason)} turns=${String(rec.turns)} cost=${fmtUsd(rec.cost_usd)} duration=${fmtDuration(rec.duration_ms)}`;default:{const{run_id:_runId,seq:_seq,turn:_turn,ts:_ts,type:_type,...payload}=rec;return`${head} ${e.type} ${compact(payload)}`}}}var TERMINAL_STATUSES=new Set(["SUCCEEDED","FAILED","TIMED_OUT","CANCELED"]);var STATUS_POLL_MS=5e3;var TERMINAL_DRAIN_MS=2e3;async function streamRun(id,json,opts){const client=createClient();const controller=new AbortController;let timer;let poller;let drainTimer;let gotResult=false;let sawData=false;const arm=ms=>{if(timer)clearTimeout(timer);if(ms>0)timer=setTimeout(()=>controller.abort(),ms)};arm(opts.idleMs>0?opts.firstEventMs:0);if(opts.pollStatus){let inFlight=false;poller=setInterval(()=>{if(inFlight||drainTimer)return;inFlight=true;client.get(runPath(id)).then(run=>{if(drainTimer||!TERMINAL_STATUSES.has(run.status))return;if(poller)clearInterval(poller);drainTimer=setTimeout(()=>{if(!gotResult){console.error(`Run is ${run.status}; no more events.`)}controller.abort()},TERMINAL_DRAIN_MS)}).catch(()=>{}).finally(()=>{inFlight=false})},STATUS_POLL_MS)}try{await client.sse(streamPath(id),frame=>{const event=parseEvent(frame.data);if(!event)return;sawData=true;if(json)console.log(frame.data);else console.log(formatEventLine(event));if(event.type==="run_result"){gotResult=true;controller.abort()}else if(opts.stopAtLiveMs!==void 0&&Date.parse(event.ts)>=opts.stopAtLiveMs){gotResult=true;controller.abort()}else arm(opts.idleMs)},{signal:controller.signal,onComment:()=>{if(!sawData&&opts.idleMs>0)arm(opts.firstEventMs)}})}finally{if(timer)clearTimeout(timer);if(poller)clearInterval(poller);if(drainTimer)clearTimeout(drainTimer)}}async function loadWorkingTree(cmd,opts,upload,spec){const loaded=await loadLocalRunDefinition(opts.files??[],opts.input,void 0,{spec});try{const org=await resolveOrg(cmd,loaded.org);if(upload)await confirmOrAbort("Run these cases now?",opts.yes);await uploadLocalInputs(createClient(),org,loaded.uploads);return{definition:loaded.definition,org,inputSnapshots:loaded.inputSnapshots,unattachedInputs:loaded.unattachedInputs}}finally{await loaded.cleanup()}}function printPlan2(result){console.log(`Definition: ${result.source.verification==="LOCAL"?"current files":result.source.commit}`);console.log(`Expected sessions: ${result.expectedRuns}`);for(const item of result.cases)console.log(` ${item.path}: ${item.expectedRuns} sessions, ${item.preference?`${item.preference.kind} preference (advisory)`:`${item.criteriaCount} criteria`}`);for(const resource of result.resources){const use=resource.use;if(use.type==="npm-package"||use.type==="npm-cli"){const install=use.install??"environment";console.log(` ${resource.name}: npm package \u2192 ${install}`);const label=install==="environment"?"Commands on PATH":"Package commands";console.log(` ${label}: ${resource.commands===void 0?"determined when the input is prepared":resource.commands.join(", ")||"none (library)"}`)}else if(use.type==="binary"){console.log(` ${resource.name}: native executable \u2192 ${use.platform}`);console.log(` Commands on PATH: ${Object.keys(use.commands).join(", ")}`);console.log(" Requires compatible system libraries; Gauge does not install OS packages.")}else console.log(` ${resource.name}: ${use.type}`);if(resource.detail)console.log(` ${resource.detail}`)}if(result.resources.length){console.log(`Inputs: ${result.resources.map(r=>`${r.name} (${r.source}, ${r.readiness})`).join(", ")}`);if(!result.canRun)console.log("Resolve the input errors before launching.")}printInputSnapshots(result.inputSnapshots);printUnattachedInputs(result.unattachedInputs,false)}function printInputSnapshots(snapshots){for(const input of snapshots){console.log(` ${input.name}: ${input.source==="working-tree"?"working-tree snapshot":"local path override snapshot"}`);console.log(` Source path: ${input.path}`);const shown=input.files.slice(0,10);console.log(` Files (${input.fileCount}): ${shown.join(", ")}${input.fileCount>shown.length?`, and ${input.fileCount-shown.length} more`:""}`);console.log(` Gauge archive digest: ${input.gaugeArchive.sha256}`);console.log(` Gauge archive bytes: ${input.gaugeArchive.bytes}`)}if(snapshots.length)console.log("Gauge archive hashes identify delivery archives containing the selected files. For npm inputs, the archive contains the .tgz; its hash and size differ from the .tgz itself. Launching again snapshots current files.")}function unattachedNote(names,saved){return`Not attached: ${names.join(", ")}. ${saved?`Saved measurements attach only inputs passed at launch; add ${names.map(name=>`--input ${name}`).join(" ")} to attach.`:`The selected cases do not use ${names.length===1?"it":"them"}.`}`}function printUnattachedInputs(names,saved){if(names.length)console.log(unattachedNote(names,saved))}async function launch(cmd,opts){if(opts.profile&&gaugeState(cmd).mode!=="local")throw new CliError("usage","--profile applies only to local runs. Pass --local.");const sources=[opts.eval!==void 0&&"--eval",opts.preference!==void 0&&"--preference",opts.apply!==void 0&&"--apply"].filter(flag=>Boolean(flag));const source=sources[0];if(opts.schema||opts.example){const flag=opts.schema?"--schema":"--example";if(source||opts.schema&&opts.example)throw new CliError("usage",`${flag} prints and exits; pass it alone`);const cases=opts.format!=="battery";if(opts.schema)printJson(cases?CASE_SPEC_SCHEMA():BATTERY_SCHEMA());else{printJson(cases?EXAMPLE_CASE_SPEC:EXAMPLE_SPEC);console.error(cases?"\n# Case spec: eval cases as JSON. Save to cases.json, edit, then: gauge run --apply cases.json":"\n# Battery spec: saved prompts and schedules. Save to battery.json, edit, then: gauge run --apply battery.json\n# For eval cases as JSON (criteria, inputs): gauge run --example --format case")}return}if(opts.format)throw new CliError("usage","--format applies with --schema or --example");if(sources.length>1)throw new CliError("usage",`pass only one of ${sources.join(", ")}`);if(opts.verify){const extra=[source,opts.dryRun&&"--dry-run",opts.files!==void 0&&"--files",opts.input.length>0&&"--input"].filter(Boolean);if(extra.length>0)throw new CliError("usage",`--verify checks the eval files in this repo; drop ${extra.join(", ")}`);const result2=verifyEvals(void 0,{local:gaugeState(cmd).mode==="local"});if(resolveOutput(cmd)==="json")printJson(result2);else printEvalVerification(result2);if(result2.summary.no>0)process.exitCode=1;return}if(opts.scenario.length>0&&source!=="--preference")throw new CliError("usage","--scenario applies with --preference only");if((opts.agent.length>0||opts.samples!==void 0)&&source!=="--eval"&&source!=="--preference")throw new CliError("usage","--agent and --samples apply with --eval or --preference; eval case files set them in their front matter");if(opts.files!==void 0&&source)throw new CliError("usage",`--files applies to this repo's eval cases; drop ${source}`);if(opts.input.length>0&&(opts.eval!==void 0||opts.preference!==void 0)&&gaugeState(cmd).mode!=="local")throw new CliError("usage",`Hosted saved measurements do not take project inputs; drop --input or pass --local`,{hint:"For hosted project inputs, use eval case files or a case spec."});if(opts.requireExperiment!==void 0&&!opts.apply)throw new CliError("usage","--require-experiment applies with --apply");let spec;if(opts.apply!==void 0){const json=await readSpecJson(opts.apply);if(!isCaseSpec(json)){if(gaugeState(cmd).mode==="local")throw new CliError("usage","Battery specs (with `prompts`) run on hosted Gauge only.",{hint:"For local runs, pass a case spec (with `cases`): `gauge run --example --format case`. Or pass --hosted."});if(opts.input.length>0)throw new CliError("usage","--input applies to case specs, not battery specs");await applySpec(cmd,parseSpec(json),{dryRun:opts.dryRun,requireExperiment:opts.requireExperiment});return}if(opts.requireExperiment!==void 0)throw new CliError("usage","--require-experiment applies to battery specs, not case specs");spec=parseCaseSpec(json)}if(gaugeState(cmd).mode==="local"&&(opts.eval||opts.preference)){const saved=localMeasurementSpec(cmd,opts.eval?"eval":"preference",opts.eval??opts.preference,{...opts,inputs:[...parseInputBindings(opts.input).keys()]});await runLocalCases(cmd,opts,saved.spec,saved.measurement);return}if(opts.dryRun&&source&&!spec)throw new CliError("usage",`--dry-run previews this repo's eval cases or an --apply spec; drop ${source}`);if(opts.eval!==void 0||opts.preference!==void 0)requireHosted(cmd);if(opts.eval!==void 0){await runEvalSet(cmd,opts.eval,opts);return}if(opts.preference!==void 0){await runPreferencePrompt(cmd,opts.preference,opts);return}if(gaugeState(cmd).mode==="local"){await runLocalCases(cmd,opts,spec);return}const{definition,org,inputSnapshots,unattachedInputs}=await loadWorkingTree(cmd,opts,!opts.dryRun,spec);const orgPath=`/api/v1/orgs/${encodeURIComponent(org)}/run-definitions`;if(opts.dryRun){const result2=await createClient().post(`${orgPath}/plan`,definition);const preview={...result2,inputSnapshots,unattachedInputs};if(resolveOutput(cmd)==="json")printJson(preview);else printPlan2(preview);return}const result=await createClient().post(`${orgPath}/run`,definition);if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`Started ${result.createdRuns}/${result.requestedRuns} sessions in request ${result.runRequestId}`);console.log(`Wait: gauge --org ${org} run wait ${result.runRequestId}`)}}async function runLocalCases(cmd,opts,spec,measurement){const{root}=gaugeState(cmd);if(!root)throw new CliError("usage","Local runs need a Gauge workspace. Run `gauge init` in this directory first.");const local=await loadLocalRunDefinition(opts.files??[],opts.input,process.cwd(),{local:true,spec});const json=resolveOutput(cmd)==="json";if(measurement&&local.unattachedInputs.length&&!opts.dryRun)console.error(`note: ${unattachedNote(local.unattachedInputs,true)}`);if(opts.dryRun){try{const plan={...await planLocal(local.definition,{profile:opts.profile,runtimeImage:opts.runtimeImage,gatewayImage:opts.gatewayImage}),inputSnapshots:local.inputSnapshots,webFixtures:Object.fromEntries(Object.entries(local.webFixtures).map(([path,{fixture,digest}])=>[path,{digest,pages:fixture.pages.map(p=>p.url),scopes:fixture.scopes??[],origins:fixture.origins??[]}])),unattachedInputs:local.unattachedInputs};if(json)printJson(plan);else printLocalPlan(plan,measurement!==void 0)}finally{await local.cleanup()}return}if(opts.detach){try{const detached=await detachLocal(local.definition,local.uploads,join12(root,".gauge"),{profile:opts.profile,runtimeImage:opts.runtimeImage,gatewayImage:opts.gatewayImage,measurement});if(json)printJson(detached);else console.log(`Batch ${detached.batchId} detached.
|
|
960
|
+
Findings (${analysis.observations.length})`);for(const finding of analysis.observations){console.log(` ${finding.kind}: ${finding.label}`);console.log(` id ${finding.id}${finding.issueId?`, issue ${finding.issueId}`:""}`);console.log(` ${finding.detail}`);console.log(` Evidence: ${JSON.stringify(finding.evidence)}`)}console.log(`
|
|
961
|
+
Products (${analysis.products.length})`);if(analysis.products.length>0)printItems(analysis.products.map(p=>({product:p.name,considered:check2(p.considered),chosen:check2(p.chosen),rejected:check2(p.rejected),installed:INSTALLED[p.installPhase]??p.installPhase,rank:p.rank??"",source:p.source.toLowerCase()})),"table");const delta3=insight.experimentDelta;if(delta3&&typeof delta3==="object"){console.log("\nExperiment delta");console.log([delta3.changed==null?"":delta3.changed?"changed":"no change",delta3.hypothesisSupported??""].filter(Boolean).join(" "));if(delta3.summary)console.log(compact(delta3.summary,300))}console.error("\n(full detail: rerun with -o json)")}function parseEvent(data){try{const parsed=runEventSchema.safeParse(JSON.parse(data));return parsed.success?parsed.data:null}catch{return null}}function formatEventLine(e){const rec=e;const head=`[t${e.turn}#${e.seq}]`;switch(e.type){case"run_meta":{const agent=rec.agent;const model=agent?.model?` model=${agent.model}`:"";return`${head} run_meta agent=${agent?.name??"?"}${model}`}case"agent_message":return`${head} agent: ${compact(rec.text)}`;case"reasoning":return`${head} reasoning: ${compact(rec.text)}`;case"tool_call":return`${head} tool_call ${String(rec.name)}: ${compact(rec.input)}`;case"tool_result":return`${head} tool_result${rec.is_error?" ERROR":""}: ${compact(rec.output,100)}`;case"run_result":return`${head} run_result ${String(rec.exit_reason)} turns=${String(rec.turns)} cost=${fmtUsd(rec.cost_usd)} duration=${fmtDuration(rec.duration_ms)}`;default:{const{run_id:_runId,seq:_seq,turn:_turn,ts:_ts,type:_type,...payload}=rec;return`${head} ${e.type} ${compact(payload)}`}}}var TERMINAL_STATUSES=new Set(["SUCCEEDED","FAILED","TIMED_OUT","CANCELED"]);var STATUS_POLL_MS=5e3;var TERMINAL_DRAIN_MS=2e3;async function streamRun(id,json,opts){const client=createClient();const controller=new AbortController;let timer;let poller;let drainTimer;let gotResult=false;let sawData=false;const arm=ms=>{if(timer)clearTimeout(timer);if(ms>0)timer=setTimeout(()=>controller.abort(),ms)};arm(opts.idleMs>0?opts.firstEventMs:0);if(opts.pollStatus){let inFlight=false;poller=setInterval(()=>{if(inFlight||drainTimer)return;inFlight=true;client.get(runPath(id)).then(run=>{if(drainTimer||!TERMINAL_STATUSES.has(run.status))return;if(poller)clearInterval(poller);drainTimer=setTimeout(()=>{if(!gotResult){console.error(`Run is ${run.status}; no more events.`)}controller.abort()},TERMINAL_DRAIN_MS)}).catch(()=>{}).finally(()=>{inFlight=false})},STATUS_POLL_MS)}try{await client.sse(streamPath(id),frame=>{const event=parseEvent(frame.data);if(!event)return;sawData=true;if(json)console.log(frame.data);else console.log(formatEventLine(event));if(event.type==="run_result"){gotResult=true;controller.abort()}else if(opts.stopAtLiveMs!==void 0&&Date.parse(event.ts)>=opts.stopAtLiveMs){gotResult=true;controller.abort()}else arm(opts.idleMs)},{signal:controller.signal,onComment:()=>{if(!sawData&&opts.idleMs>0)arm(opts.firstEventMs)}})}finally{if(timer)clearTimeout(timer);if(poller)clearInterval(poller);if(drainTimer)clearTimeout(drainTimer)}}async function loadWorkingTree(cmd,opts,upload,spec){const lifecycle=opts.prompt!==void 0?lifecyclePatchFromOpts(opts):{};const loaded=await loadLocalRunDefinition(opts.files??[],opts.input,void 0,{spec});try{const org=await resolveOrg(cmd,loaded.org);if(upload)await confirmOrAbort("Run these cases now?",opts.yes);await uploadLocalInputs(createClient(),org,loaded.uploads);return{definition:{...loaded.definition,...lifecycle},org,inputSnapshots:loaded.inputSnapshots,unattachedInputs:loaded.unattachedInputs}}finally{await loaded.cleanup()}}function printPlan2(result){console.log(`Definition: ${result.source.verification==="LOCAL"?"current files":result.source.commit}`);console.log(`Expected sessions: ${result.expectedRuns}`);if(result.lifecycle?.setupTimeoutSeconds!==void 0)console.log(`Setup hook: ${result.lifecycle.setupTimeoutSeconds}s deadline`);if(result.lifecycle?.teardownTimeoutSeconds!==void 0)console.log(`Teardown hook: ${result.lifecycle.teardownTimeoutSeconds}s deadline`);for(const item of result.cases)console.log(` ${item.path}: ${item.expectedRuns} sessions, ${item.preference?`${item.preference.kind} preference (advisory)`:`${item.criteriaCount} criteria`}`);for(const resource of result.resources){const use=resource.use;if(use.type==="npm-package"||use.type==="npm-cli"){const install=use.install??"environment";console.log(` ${resource.name}: npm package \u2192 ${install}`);const label=install==="environment"?"Commands on PATH":"Package commands";console.log(` ${label}: ${resource.commands===void 0?"determined when the input is prepared":resource.commands.join(", ")||"none (library)"}`)}else if(use.type==="binary"){console.log(` ${resource.name}: native executable \u2192 ${use.platform}`);console.log(` Commands on PATH: ${Object.keys(use.commands).join(", ")}`);console.log(" Requires compatible system libraries; Gauge does not install OS packages.")}else console.log(` ${resource.name}: ${use.type}`);if(resource.detail)console.log(` ${resource.detail}`)}if(result.resources.length){console.log(`Inputs: ${result.resources.map(r=>`${r.name} (${r.source}, ${r.readiness})`).join(", ")}`);if(!result.canRun)console.log("Resolve the input errors before launching.")}printInputSnapshots(result.inputSnapshots);printUnattachedInputs(result.unattachedInputs,false)}function printInputSnapshots(snapshots){for(const input of snapshots){console.log(` ${input.name}: ${input.source==="working-tree"?"working-tree snapshot":"local path override snapshot"}`);console.log(` Source path: ${input.path}`);const shown=input.files.slice(0,10);console.log(` Files (${input.fileCount}): ${shown.join(", ")}${input.fileCount>shown.length?`, and ${input.fileCount-shown.length} more`:""}`);console.log(` Gauge archive digest: ${input.gaugeArchive.sha256}`);console.log(` Gauge archive bytes: ${input.gaugeArchive.bytes}`)}if(snapshots.length)console.log("Gauge archive hashes identify delivery archives containing the selected files. For npm inputs, the archive contains the .tgz; its hash and size differ from the .tgz itself. Launching again snapshots current files.")}function unattachedNote(names,saved){return`Not attached: ${names.join(", ")}. ${saved?`Saved measurements attach only inputs passed at launch; add ${names.map(name=>`--input ${name}`).join(" ")} to attach.`:`The selected cases do not use ${names.length===1?"it":"them"}.`}`}function printUnattachedInputs(names,saved){if(names.length)console.log(unattachedNote(names,saved))}function callerAgent(){const caller=detectSkillAgent(process.env);return caller==="codex"||caller==="opencode"?caller:"claude-code"}function promptSpec(prompt,agents){return parseCaseSpec({version:2,cases:[{key:"prompt",prompt,preference:{kind:"open-ended"},config:{agents:buildAgentRoster(agents)}}]})}async function launch(cmd,opts){if(opts.profile&&gaugeState(cmd).mode!=="local")throw new CliError("usage","--profile applies only to local runs. Pass --local.");const sources=[opts.prompt!==void 0&&"a prompt",opts.eval!==void 0&&"--eval",opts.preference!==void 0&&"--preference",opts.apply!==void 0&&"--apply"].filter(flag=>Boolean(flag));const source=sources[0];if(hasLifecycleFlags(opts)){if(gaugeState(cmd).mode!=="hosted")throw new CliError("usage","Lifecycle hooks require hosted Gauge; pass --hosted.");if(source!=="a prompt"&&source!=="--eval"&&source!=="--preference"||opts.apply!==void 0||opts.schema||opts.example||opts.verify||opts.dryRun&&source!=="a prompt")throw new CliError("usage","Lifecycle flags apply only to a hosted prompt, --eval, or --preference launch; they do not apply to case files or --apply. Only prompt launches support --dry-run with hooks.")}if(opts.schema||opts.example){const flag=opts.schema?"--schema":"--example";if(source||opts.schema&&opts.example)throw new CliError("usage",`${flag} prints and exits; pass it alone`);const cases=opts.format!=="battery";if(opts.schema)printJson(cases?CASE_SPEC_SCHEMA():BATTERY_SCHEMA());else{printJson(cases?EXAMPLE_CASE_SPEC:EXAMPLE_SPEC);console.error(cases?"\n# Case spec: eval cases as JSON. Save to cases.json, edit, then: gauge run --apply cases.json":"\n# Battery spec: saved prompts and schedules. Save to battery.json, edit, then: gauge run --apply battery.json\n# For eval cases as JSON (criteria, inputs): gauge run --example --format case")}return}if(opts.format)throw new CliError("usage","--format applies with --schema or --example");if(sources.length>1)throw new CliError("usage",`pass only one of ${sources.join(", ")}`);if(opts.verify){const extra=[source,opts.dryRun&&"--dry-run",opts.files!==void 0&&"--files",opts.input.length>0&&"--input"].filter(Boolean);if(extra.length>0)throw new CliError("usage",`--verify checks the eval files in this repo; drop ${extra.join(", ")}`);const result2=verifyEvals(void 0,{local:gaugeState(cmd).mode==="local"});if(resolveOutput(cmd)==="json")printJson(result2);else printEvalVerification(result2);if(result2.summary.no>0)process.exitCode=1;return}if(opts.scenario.length>0&&source!=="--preference")throw new CliError("usage","--scenario applies with --preference only");if(source==="a prompt"){if(opts.samples!==void 0||opts.input.length>0)throw new CliError("usage","--samples and --input do not apply to prompt runs; use `gauge preference` or `gauge eval` for more settings")}else if((opts.agent.length>0||opts.samples!==void 0)&&source!=="--eval"&&source!=="--preference")throw new CliError("usage","--agent and --samples apply with --eval or --preference; eval case files set them in their front matter");if(opts.files!==void 0&&source)throw new CliError("usage",`--files applies to this repo's eval cases; drop ${source}`);if(opts.input.length>0&&(opts.eval!==void 0||opts.preference!==void 0)&&gaugeState(cmd).mode!=="local")throw new CliError("usage",`Hosted saved measurements do not take project inputs; drop --input or pass --local`,{hint:"For hosted project inputs, use eval case files or a case spec."});if(opts.requireExperiment!==void 0&&!opts.apply)throw new CliError("usage","--require-experiment applies with --apply");let spec;if(opts.apply!==void 0){const json=await readSpecJson(opts.apply);if(!isCaseSpec(json)){if(gaugeState(cmd).mode==="local")throw new CliError("usage","Battery specs (with `prompts`) run on hosted Gauge only.",{hint:"For local runs, pass a case spec (with `cases`): `gauge run --example --format case`. Or pass --hosted."});if(opts.input.length>0)throw new CliError("usage","--input applies to case specs, not battery specs");const result2=await applySpec(cmd,parseSpec(json),{dryRun:opts.dryRun,requireExperiment:opts.requireExperiment});if(!opts.dryRun&&"runIds"in result2&&result2.runIds.length>0)console.error(RUN_PRESENTATION_REMINDER);return}if(opts.requireExperiment!==void 0)throw new CliError("usage","--require-experiment applies to battery specs, not case specs");spec=parseCaseSpec(json)}if(opts.prompt!==void 0)spec=promptSpec(opts.prompt,opts.agent.length>0?opts.agent:[callerAgent()]);if(gaugeState(cmd).mode==="local"&&(opts.eval||opts.preference)){const saved=localMeasurementSpec(cmd,opts.eval?"eval":"preference",opts.eval??opts.preference,{...opts,inputs:[...parseInputBindings(opts.input).keys()]});await runLocalCases(cmd,opts,saved.spec,saved.measurement);return}if(opts.dryRun&&source&&!spec)throw new CliError("usage",`--dry-run previews this repo's eval cases or an --apply spec; drop ${source}`);if(opts.eval!==void 0||opts.preference!==void 0)requireHosted(cmd);if(opts.eval!==void 0){const result2=await runEvalSet(cmd,opts.eval,opts);if(result2.runIds.length>0||result2.runRequestId)console.error(RUN_PRESENTATION_REMINDER);return}if(opts.preference!==void 0){const result2=await runPreferencePrompt(cmd,opts.preference,opts);if(result2.runIds.length>0)console.error(RUN_PRESENTATION_REMINDER);return}if(gaugeState(cmd).mode==="local"){await runLocalCases(cmd,opts,spec);return}const{definition,org,inputSnapshots,unattachedInputs}=await loadWorkingTree(cmd,opts,!opts.dryRun,spec);const orgPath=`/api/v1/orgs/${encodeURIComponent(org)}/run-definitions`;if(opts.dryRun){const result2=await createClient().post(`${orgPath}/plan`,definition);const preview={...result2,inputSnapshots,unattachedInputs};if(resolveOutput(cmd)==="json")printJson(preview);else printPlan2(preview);return}const result=await createClient().post(`${orgPath}/run`,definition);if(resolveOutput(cmd)==="json")printJson(result);else{console.log(`Started ${result.createdRuns}/${result.requestedRuns} sessions in request ${result.runRequestId}`);console.log(`Wait: gauge --org ${org} run wait ${result.runRequestId}`)}if(result.requestedRuns>0)console.error(RUN_PRESENTATION_REMINDER)}async function runLocalCases(cmd,opts,spec,measurement){const{root}=gaugeState(cmd);if(!root)throw new CliError("usage","Local runs need a Gauge workspace. Run `gauge init` in this directory first.");const local=await loadLocalRunDefinition(opts.files??[],opts.input,process.cwd(),{local:true,spec});const json=resolveOutput(cmd)==="json";if(measurement&&local.unattachedInputs.length&&!opts.dryRun)console.error(`note: ${unattachedNote(local.unattachedInputs,true)}`);if(opts.dryRun){try{const plan={...await planLocal(local.definition,{profile:opts.profile,runtimeImage:opts.runtimeImage,gatewayImage:opts.gatewayImage}),inputSnapshots:local.inputSnapshots,webFixtures:Object.fromEntries(Object.entries(local.webFixtures).map(([path,{fixture,digest}])=>[path,{digest,pages:fixture.pages.map(p=>p.url),scopes:fixture.scopes??[],origins:fixture.origins??[]}])),unattachedInputs:local.unattachedInputs};if(json)printJson(plan);else printLocalPlan(plan,measurement!==void 0)}finally{await local.cleanup()}return}if(opts.detach){try{const detached=await detachLocal(local.definition,local.uploads,join11(root,".gauge"),{profile:opts.profile,runtimeImage:opts.runtimeImage,gatewayImage:opts.gatewayImage,measurement});if(json)printJson(detached);else console.log(`Batch ${detached.batchId} detached.
|
|
789
962
|
Status: ${detached.status}
|
|
790
963
|
Wait: ${detached.wait}
|
|
791
|
-
Stop: ${detached.stop}`)}finally{await local.cleanup()}return}let batch;try{batch=await runLocal(local.definition,local.uploads,
|
|
964
|
+
Stop: ${detached.stop}`);console.error(RUN_PRESENTATION_REMINDER)}finally{await local.cleanup()}return}let batch;try{batch=await runLocal(local.definition,local.uploads,join11(root,".gauge"),run=>{if(!json)console.error(`${run.id} ${run.agent} ${run.model} ${run.case}: ${run.outcome}${run.detail?` (${run.detail})`:""}`)},{measurement,profile:opts.profile,runtimeImage:opts.runtimeImage,gatewayImage:opts.gatewayImage,onStart:batchId=>{console.error(`Batch ${batchId}
|
|
792
965
|
Status: gauge run status --batch ${batchId} --local
|
|
793
966
|
Wait: gauge run wait --batch ${batchId} --local
|
|
794
|
-
Stop: gauge run cancel --batch ${batchId} --local`)})}finally{await local.cleanup()}if(batch.canceled&&!process.exitCode)process.exitCode=130;if(!process.exitCode&&batch.runs.some(run=>run.outcome!=="completed"))process.exitCode=1;if(json){printJson(batch);return}const counts=new Map;for(const run of batch.runs)counts.set(run.outcome,(counts.get(run.outcome)??0)+1);console.log(`Batch ${batch.batchId}: ${batch.runs.length} runs. ${[...counts].map(([outcome2,n])=>`${n} ${outcome2}`).join(", ")}.`);console.log(`Results: .gauge/runs (gauge run list --local --batch ${batch.batchId})`);console.log("");console.log(batch.next)}function printLocalPlan(plan,saved){console.log(`Local plan: ${plan.expectedSessions} session${plan.expectedSessions===1?"":"s"} in local Docker. Nothing started.`);console.log(`Runtime: ${plan.runtime.reference}${plan.runtime.imageId?` (${plan.runtime.imageId})`:""}. Compatibility probe runs at launch.`);if(plan.gateway)console.log(`Gateway: ${plan.gateway.reference}${plan.gateway.imageId?` (${plan.gateway.imageId})`:""}`);for(const item of plan.cases)console.log(` ${item.path}: ${item.sessions.map(s=>`${s.agent} ${s.model} \xD7${s.samples}`).join(", ")}; ${item.criteria} criteria`);console.log(`Inputs: ${plan.inputs.map(i=>`${i.name} (${i.use})`).join(", ")||"none"}`);printInputSnapshots(plan.inputSnapshots);printUnattachedInputs(plan.unattachedInputs,saved);for(const warning of plan.warnings)console.log(`Warning: ${warning}`);console.log("Run it: the same command without --dry-run.")}var WAIT_POLL_MS=5e3;var LOCAL_WAIT_POLL_MS=2e3;async function runRequestPath(cmd,id){return`/api/v1/orgs/${encodeURIComponent(await resolveOrg(cmd))}/run-requests/${encodeURIComponent(id)}`}function printRunRequest(result){const{counts}=result;const measured=counts.measuredRuns??0;if(!result.settled)console.log(`${result.status}: ${counts.terminalRuns}/${counts.createdRuns} sessions finished, ${counts.judgedRuns} judged, ${counts.pendingRuns} pending`);else console.log(measured>0?`${result.result}: ${measured} preference sessions measured${counts.passedRuns?`; ${counts.passedRuns} experience sessions passed`:""}`:`${result.result}: ${counts.passedRuns}/${counts.createdRuns} passed sessions (${counts.judgedRuns} judged)`);for(const batch of result.batches)for(const run of batch.runs)console.log(` ${run.id} ${run.agent} ${run.model||"(default)"}: ${run.preference?`${preferenceSelectionLabel(run.preference)} (advisory)`:run.result}`);console.log(`Request: ${result.id}`)}function failsCheck(result){return result.result!=="PASSED"&&result.result!=="MEASURED"}async function waitForRequest(cmd,id,opts){const path=await runRequestPath(cmd,id);const client=createClient();const json=resolveOutput(cmd)==="json";let lastStatus="";while(true){const result=await client.get(path);if(!json&&result.status!==lastStatus){console.error(`Run request ${id}: ${result.status}`);lastStatus=result.status}if(result.settled){if(json)printJson(result);else printRunRequest(result);if(opts.check&&failsCheck(result))process.exitCode=1;return}await new Promise(resolve5=>setTimeout(resolve5,WAIT_POLL_MS))}}async function waitForLocalBatch(cmd,gaugeDir,batchId,opts){const json=resolveOutput(cmd)==="json";let lastFinished=-1;while(true){let batch;try{batch=await localBatchStatus(gaugeDir,batchId)}catch(error){if(error.code==="ENOENT")throw new CliError("not_found","Local batch not found");throw error}const settled=batch.state!=="queued"&&batch.state!=="running";if(!json&&!settled&&batch.finishedSessions!==lastFinished){console.error(`Batch ${batchId}: ${batch.finishedSessions??0}/${batch.expectedSessions} finished`);lastFinished=batch.finishedSessions}if(settled){const runs=await localStatus(gaugeDir,{batch:batchId});if(json)printJson({batch,runs,next:analysisNext(batchId)});else{printLocalStatus(batch,runs);console.log(analysisNext(batchId))}if(opts.check&&(batch.state!=="completed"||runs.some(run=>run.state!=="completed")))process.exitCode=1;return}await new Promise(resolve5=>setTimeout(resolve5,LOCAL_WAIT_POLL_MS))}}function printLocalStatus(batch,runs){if(batch)console.log(`Batch ${batch.batchId}: ${batch.state}; ${batch.finishedSessions??0}/${batch.expectedSessions} finished`);for(const run of runs){console.log(`${run.id}: ${run.state}${run.cleanupPending?" (cleanup pending)":""}${run.detail?` (${run.detail})`:""}`);if(run.stderrLog)console.log(` agent stderr: ${relative6(process.cwd(),run.stderrLog)}`)}}function register29(program2){const run=program2.command("run").description("Start runs and read them. With no flag: run this repo's eval cases (gauge-cases/) (local: in local Docker)").argument("[command]").usage("[options] [command]").option("--eval <id>","run a saved eval set (see `gauge eval`)",once("--eval")).option("--preference <id>","run a saved preference prompt",once("--preference")).option("--scenario <scenarioId>","with --preference: run only this scenario (repeatable)",collect5,[]).option("--agent <agent[:m1,m2]>","with --eval/--preference: run these agents and model pins for this launch only; the stored roster stays (repeatable)",collect5,[]).option("--samples <n>","with --eval/--preference: sessions per agent and model for this launch only (1-50)").option("--apply <file>",'run a JSON spec ("-" for stdin): a case spec (`cases`, both modes) or a battery spec (`prompts`, hosted). See --example').option("--files <paths...>","eval case Markdown paths or globs (default gauge-cases/**/*.md)").option("--input <name[=path]>","attach a gauge.json input; =path or =glob replaces its configured files (repeatable)",collect5,[]).option("--detach","run a local batch in the background; print status and stop commands").option("--profile <profile>","local credential profile; overrides the default for every target").option("--runtime-image <image>","use a compatible custom Docker image for local runs (tag, digest, or image ID)").option("--gateway-image <image>","use a compatible custom gateway image for local web fixtures").option("--verify","check this repo's eval files; start no sessions").option("--dry-run","preview this repo's eval cases or an --apply spec; start no sessions").option("--require-experiment <label>","with --apply: fail unless every target resolves to this experiment label").option("--schema","print the JSON Schema of an --apply spec and exit").option("--example","print an example --apply spec and exit").addOption(new Option4("--format <format>","schema/example format (default: case)").choices(["case","battery"])).option("--yes","skip the confirmation").action(async(arg,opts,cmd)=>{if(arg!==void 0)throw new CliError("usage",`unknown run command "${arg}". Launch a saved eval set or preference prompt with \`gauge run --eval <id>\` or \`--preference <id>\`; read runs with \`gauge run list\` or \`gauge run get <id> [--logs]\`.`);if(opts.detach&&(gaugeState(cmd).mode!=="local"||opts.dryRun||opts.verify||opts.schema||opts.example))throw new CliError("usage","--detach requires a local run");if((opts.runtimeImage||opts.gatewayImage)&&gaugeState(cmd).mode!=="local")throw new CliError("usage","--runtime-image and --gateway-image work only on local runs; pass --local");await launch(cmd,opts)});run.hook("preAction",(parent,action)=>{if(action===parent)return;const allowed=action.name()==="list"?["eval","preference","agent"]:[];const stray=Object.entries(parent.opts()).filter(([key,value])=>!allowed.includes(key)&&value!==void 0&&!(Array.isArray(value)&&value.length===0)).map(([key])=>`--${key.replace(/[A-Z]/g,c=>`-${c.toLowerCase()}`)}`);if(stray.length>0)throw new CliError("usage",`${stray.join(", ")} ${stray.length===1?"is a launch flag":"are launch flags"}; \`gauge run ${action.name()}\` does not take ${stray.length===1?"it":"them"}`)});run.command("list").description("List runs (newest first). Filters combine. --preference <id> lists one prompt's runs with verdicts and takes no other filter. Local: runs in .gauge/").option("--eval <id>","filter by eval set (local: by eval)").option("--preference <id>","one preference prompt's runs with verdicts (no other filter)").option("--status <status>","filter by status, e.g. running,queued (local: completed,failed)").option("--batch <id>","filter by batch").option("--prompt <id>","filter by prompt").option("--experiment <experiment>","filter by experiment label").option("--since <window>","e.g. 24h, 7d, or an ISO date").option("--agent <agent>","filter by agent (local only)").option("--model <model>","filter by model (local only)").option("--unanalyzed","only runs without a published analysis (local only)").option("--limit <n>","page size (1-200)","50").option("--cursor <cursor>","resume from a previous page's nextCursor").action(async(_opts,command)=>{const opts=command.optsWithGlobals();const agents=[opts.agent??[]].flat();if(agents.length>1)throw new CliError("usage","pass one --agent to filter by");const agent=agents[0];const output=resolveOutput(command);const state=gaugeState(command);if(state.mode==="local"){if(opts.prompt||opts.experiment||opts.cursor)throw new CliError("usage","Local run list does not support --prompt, --experiment, or --cursor. Use --eval or --preference to select a saved measurement.");const items=state.root?listLocalRuns(join12(state.root,".gauge"),{batch:opts.batch,evalId:opts.eval,preferenceId:opts.preference,agent:agent?parseAgents(agent)[0]:void 0,model:opts.model,outcome:opts.status,since:opts.since?parseSince(opts.since):void 0,analyzed:opts.unanalyzed?false:void 0}).slice(0,Number(opts.limit)):[];if(output==="json")printJson({items});else printItems(items.map(toLocalRow),"table");return}const localOnly=[agent!==void 0&&"--agent",opts.model!==void 0&&"--model",opts.unanalyzed&&"--unanalyzed"].filter(Boolean);if(localOnly.length>0)throw new CliError("usage",`${localOnly.join(", ")} ${localOnly.length===1?"works":"work"} only on local runs; pass --local`);if(opts.preference!==void 0){const filters=[opts.eval!==void 0&&"--eval",opts.status!==void 0&&"--status",opts.batch!==void 0&&"--batch",opts.prompt!==void 0&&"--prompt",opts.experiment!==void 0&&"--experiment",opts.since!==void 0&&"--since"].filter(Boolean);if(filters.length>0)throw new CliError("usage",`--preference lists one prompt's runs and takes no other filter; drop ${filters.join(", ")}`);await listPreferencePromptRuns(command,opts.preference,opts);return}const org=await resolveOrg(command);const page=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/runs`,{status:opts.status,batch:opts.batch,prompt:opts.prompt,experiment:opts.experiment,evalSet:opts.eval,since:opts.since,limit:opts.limit,cursor:opts.cursor});if(output==="json"){printJson(page);return}printItems(page.items.map(toRow4),"table");if(page.nextCursor){console.error(`(more results \u2014 rerun with --cursor ${page.nextCursor})`)}});run.command("get <id>").description("Show a run (status, trigger source, cost, credits, turns, exit reason). --logs prints its trace events instead").option("--logs","print the trace events so far, then exit").action(async(id,opts,command)=>{if(gaugeState(command).mode==="local"){const dir=localDirectory2(command);const session=withLocalStore(command,store=>store.getSession(id));if(!session)throw new CliError("usage",`Local run ${id} was not found. If its files are still in .gauge/runs, run \`gauge run repair --local\` to restore it to the run list.`);if(opts.logs){const path=join12(dir,"runs",session.id,"attempts",String(session.attempt),"session","events.ndjson");if(!existsSync4(path))throw new CliError("usage","This run has no saved transcript yet.");process.stdout.write(readFileSync11(path,"utf8"))}else printJson({...session,analysis:getLocalAnalysis(dir,id)});return}if(opts.logs){await streamRun(id,resolveOutput(command)==="json",{idleMs:2e3,firstEventMs:15e3,stopAtLiveMs:Date.now()-2e3});return}const run2=await createClient().get(runPath(id));if(resolveOutput(command)==="json"){printJson(run2);return}const lines=detailLines(run2);const width=Math.max(...lines.map(([label])=>label.length));for(const[label,value]of lines){console.log(`${label.padEnd(width)} ${value}`)}});run.command("insight <id>").description("Show the run's findings, criteria, and product decisions").action(async(id,_opts,command)=>{if(gaugeState(command).mode==="local"){const dir=localDirectory2(command);if(!withLocalStore(command,store=>store.getSession(id)))throw new CliError("usage",`Local run ${id} was not found.`);const analysis2=getLocalAnalysis(dir,id);if(resolveOutput(command)==="json")printJson({runId:id,analysis:analysis2});else console.log(analysis2?renderLocalAnalysisMarkdown(analysis2):"No analysis published. Follow gauge instructions session-analysis --local.");return}const insight=await createClient().get(`${runPath(id)}/insight`);if(resolveOutput(command)==="json"){printJson(insight);return}printInsight(insight)});const analysis=run.command("analysis").description("Publish structured analysis from your coding agent");analysis.command("schema").description("Print the versioned analysis submission schema").action(()=>printJson(z14.toJSONSchema(analysisSubmissionSchema)));analysis.command("save <id>").description("Validate and publish analysis for a local session attempt").requiredOption("--file <path>","analysis JSON file").action((id,opts,command)=>{if(gaugeState(command).mode!=="local")throw new CliError("usage","Analysis submission is available locally; hosted Gauge publishes its own analysis. Use --local.");let input;try{input=JSON.parse(readFileSync11(opts.file,"utf8"))}catch(error){throw new CliError("usage",`Cannot read analysis JSON: ${String(error)}`)}const dir=localDirectory2(command);const saved=saveLocalAnalysis(dir,id,input);const results=join12(dir,"runs",saved.runId,"results");try{mkdirSync2(results,{recursive:true});writeFileSync3(join12(results,"session-analysis.md"),renderLocalAnalysisMarkdown(saved))}catch(error){console.error(`Analysis ${saved.id} was published in SQLite, but its Markdown report could not be written: ${String(error)}. Read it with gauge run insight ${id} --local.`)}if(resolveOutput(command)==="json")printJson(saved);else console.log(`Published analysis ${saved.id} (revision ${saved.revision}); ${saved.findings.length} findings. Read with gauge run insight ${id} --local.`)});run.command("repair").description("Restore local runs to the run list from their saved files, keeping measurements and analysis").action((_opts,command)=>{if(gaugeState(command).mode!=="local")throw new CliError("usage","Session repair is local only; pass --local.");printJson(withLocalStore(command,store=>store.repairSessionSummaries()))});hostedOnly(run.command("regrade <id>")).description("Re-judge using current eval criteria or the run's criteria snapshot (no new session)").action(async(id,_opts,command)=>{const result=await createClient().post(`${runPath(id)}/regrade`);if(resolveOutput(command)==="json"){printJson(result);return}console.log(`Re-judging run ${result.runId} against ${result.criteria} ${result.criteria===1?"criterion":"criteria"} ${result.evalSetId?`of eval ${result.evalSetId}`:"from its launch-time snapshot"}.`);console.log("Verdicts land on the run; `gauge run get <id>` to check.")});hostedOnly(run.command("watch <id>")).description("Stream normalized trace events live (SSE; exits when the run finishes)").action(async(id,_opts,command)=>{await streamRun(id,resolveOutput(command)==="json",{idleMs:0,firstEventMs:0,pollStatus:true})});run.command("diff <id>").description("Print the run's working-tree diff to stdout").action(async(id,_opts,command)=>{if(gaugeState(command).mode==="local"){const session=withLocalStore(command,store=>store.getSession(id));if(!session)throw new CliError("usage",`Local run ${id} was not found.`);const path=join12(localDirectory2(command),"runs",session.id,"attempts",String(session.attempt),"session","diff.patch");if(existsSync4(path))process.stdout.write(readFileSync11(path,"utf8"));else console.error("(no diff)");return}const diff2=await createClient().get(`${runPath(id)}/diff`);if(!diff2){console.error("(no diff)");return}process.stdout.write(diff2)});run.command("status [id]").description("Print a launch request's progress once, without waiting; local: live session status and batch progress").option("--batch <batch-id>","show this local batch").action(async(id,opts,command)=>{const state=gaugeState(command);if(state.mode!=="local"){if(!id||opts.batch)throw new CliError("usage","Hosted status requires a request ID; --batch is local only");const result=await createClient().get(await runRequestPath(command,id));if(resolveOutput(command)==="json")printJson(result);else printRunRequest(result);return}if(!state.root)throw new CliError("usage","Run this command inside the batch's Gauge workspace");if(id&&opts.batch)throw new CliError("usage","Pass a run ID or --batch, not both");const gaugeDir=join12(state.root,".gauge");const runs=await localStatus(gaugeDir,{id,batch:opts.batch});let batch;if(opts.batch){try{batch=await localBatchStatus(gaugeDir,opts.batch)}catch(error){if(error.code!=="ENOENT")throw error}}if((id||opts.batch)&&!runs.length&&!batch)throw new CliError("not_found","Local run or batch not found");if(resolveOutput(command)==="json")printJson({batch,runs});else printLocalStatus(batch,runs)});run.command("cancel [id]").option("--batch <batch-id>","cancel an entire local batch, including queued sessions").description("Request cancellation of a run").action(async(id,opts,command)=>{const state=gaugeState(command);if(state.mode==="local"){if(!state.root)throw new CliError("usage","Run this command inside the batch's Gauge workspace");const results=await cancelLocal(join12(state.root,".gauge"),{id,batch:opts.batch});if(results.some(r=>r.status==="pending"))process.exitCode=1;if(resolveOutput(command)==="json")printJson({cancelRequested:true,runs:results});else{console.log(`Cancellation requested for ${opts.batch??id}.`);for(const result2 of results)console.log(`${result2.id}: ${result2.status}${"detail"in result2?` (${result2.detail})`:""}`)}return}if(!id||opts.batch)throw new CliError("usage","Hosted cancellation requires a run ID; --batch is local only");const result=await createClient().post(`${runPath(id)}/cancel`);if(resolveOutput(command)==="json"){printJson(result);return}console.log(`Run ${result.id}: ${result.status}${result.cancelRequested?" (cancel requested)":""}`);if(result.note)console.error(result.note)});run.command("compare").description("Compare two sets of runs: pass rate, criteria, and median turns, cost, and duration, per agent and model. Local: completed runs and median turns").requiredOption("--baseline <key=value>","baseline runs (repeatable): request=, batch=, eval=, revision=, experiment=, trial=, agent=, model=, since=, until= (local: batch=, eval=<saved id>, revision=<number>, agent=, model=, since=, until=)",collect5,[]).requiredOption("--candidate <key=value>","candidate runs (repeatable), same keys as --baseline",collect5,[]).addHelpText("after",`
|
|
967
|
+
Stop: gauge run cancel --batch ${batchId} --local`);console.error(RUN_PRESENTATION_REMINDER)}})}finally{await local.cleanup()}if(batch.canceled&&!process.exitCode)process.exitCode=130;if(!process.exitCode&&batch.runs.some(run=>run.outcome!=="completed"))process.exitCode=1;if(json){printJson(batch);return}const counts=new Map;for(const run of batch.runs)counts.set(run.outcome,(counts.get(run.outcome)??0)+1);console.log(`Batch ${batch.batchId}: ${batch.runs.length} runs. ${[...counts].map(([outcome2,n])=>`${n} ${outcome2}`).join(", ")}.`);console.log(`Results: .gauge/runs (gauge run list --local --batch ${batch.batchId})`);console.log("");console.log(batch.next)}function printLocalPlan(plan,saved){console.log(`Local plan: ${plan.expectedSessions} session${plan.expectedSessions===1?"":"s"} in local Docker. Nothing started.`);console.log(`Runtime: ${plan.runtime.reference}${plan.runtime.imageId?` (${plan.runtime.imageId})`:""}. Compatibility probe runs at launch.`);if(plan.gateway)console.log(`Gateway: ${plan.gateway.reference}${plan.gateway.imageId?` (${plan.gateway.imageId})`:""}`);for(const item of plan.cases)console.log(` ${item.path}: ${item.sessions.map(s=>`${s.agent} ${s.model} \xD7${s.samples}`).join(", ")}; ${item.criteria} criteria`);console.log(`Inputs: ${plan.inputs.map(i=>`${i.name} (${i.use})`).join(", ")||"none"}`);printInputSnapshots(plan.inputSnapshots);printUnattachedInputs(plan.unattachedInputs,saved);for(const warning of plan.warnings)console.log(`Warning: ${warning}`);console.log("Run it: the same command without --dry-run.")}var WAIT_POLL_MS=5e3;var LOCAL_WAIT_POLL_MS=2e3;async function runRequestPath(cmd,id){return`/api/v1/orgs/${encodeURIComponent(await resolveOrg(cmd))}/run-requests/${encodeURIComponent(id)}`}function printRunRequest(result){const{counts}=result;const measured=counts.measuredRuns??0;if(!result.settled)console.log(`${result.status}: ${counts.terminalRuns}/${counts.createdRuns} sessions finished, ${counts.judgedRuns} judged, ${counts.pendingRuns} pending`);else console.log(measured>0?`${result.result}: ${measured} preference sessions measured${counts.passedRuns?`; ${counts.passedRuns} experience sessions passed`:""}`:`${result.result}: ${counts.passedRuns}/${counts.createdRuns} passed sessions (${counts.judgedRuns} judged)`);for(const batch of result.batches)for(const run of batch.runs)console.log(` ${run.id} ${run.agent} ${run.model||"(default)"}: ${run.preference?`${preferenceSelectionLabel(run.preference)} (advisory)`:run.result}`);console.log(`Request: ${result.id}`)}function failsCheck(result){return result.result!=="PASSED"&&result.result!=="MEASURED"}async function waitForRequest(cmd,id,opts){const path=await runRequestPath(cmd,id);const client=createClient();const json=resolveOutput(cmd)==="json";let lastStatus="";while(true){const result=await client.get(path);if(!json&&result.status!==lastStatus){console.error(`Run request ${id}: ${result.status}`);lastStatus=result.status}if(result.settled){if(json)printJson(result);else printRunRequest(result);if(opts.check&&failsCheck(result))process.exitCode=1;return}await new Promise(resolve5=>setTimeout(resolve5,WAIT_POLL_MS))}}async function waitForLocalBatch(cmd,gaugeDir,batchId,opts){const json=resolveOutput(cmd)==="json";let lastFinished=-1;while(true){let batch;try{batch=await localBatchStatus(gaugeDir,batchId)}catch(error){if(error.code==="ENOENT")throw new CliError("not_found","Local batch not found");throw error}const settled=["completed","canceled","failed","interrupted"].includes(batch.state);if(!json&&!settled&&batch.finishedSessions!==lastFinished){console.error(`Batch ${batchId}: ${batch.finishedSessions??0}/${batch.expectedSessions} finished`);lastFinished=batch.finishedSessions}if(settled){const runs=await localStatus(gaugeDir,{batch:batchId});if(json)printJson({batch,runs,next:analysisNext(batchId)});else{printLocalStatus(batch,runs);console.log(analysisNext(batchId))}if(opts.check&&(batch.state!=="completed"||runs.some(run=>run.state!=="completed")))process.exitCode=1;return}await new Promise(resolve5=>setTimeout(resolve5,LOCAL_WAIT_POLL_MS))}}function printLocalStatus(batch,runs){if(batch)console.log(`Batch ${batch.batchId}: ${batch.state}${batch.lastKnownState?` (last known: ${batch.lastKnownState})`:""}; ${batch.finishedSessions??0}/${batch.expectedSessions} finished`);for(const run of runs){console.log(`${run.id}: ${run.state}${"lastKnownState"in run&&run.lastKnownState?` (last known: ${run.lastKnownState})`:""}${run.cleanupPending?" (cleanup pending)":""}${run.detail?` (${run.detail})`:""}`);if(run.stderrLog)console.log(` agent stderr: ${relative5(process.cwd(),run.stderrLog)}`)}}function register29(program2){const run=addLifecycleEditOptions(program2.command("run")).description('Start runs and read them. `gauge run "<prompt>"` runs one open-ended preference prompt and saves no measurement. With no flag: run this repo\'s eval cases (gauge-cases/) (local: in local Docker). Lifecycle flags attach hooks to hosted prompts or override saved --eval/--preference hooks for this launch only.').argument("[prompt]").usage('[options] ["<prompt>"]').option("--eval <id>","run a saved eval set (see `gauge eval`)",once("--eval")).option("--preference <id>","run a saved preference prompt",once("--preference")).option("--scenario <scenarioId>","with --preference: run only this scenario (repeatable)",collect5,[]).option("--agent <agent[:m1,m2]>","with a prompt or --eval/--preference: run these agents and model pins for this launch only (repeatable)",collect5,[]).option("--samples <n>","with --eval/--preference: sessions per agent and model for this launch only (1-50)").option("--apply <file>",'run a JSON spec ("-" for stdin): a case spec (`cases`, both modes) or a battery spec (`prompts`, hosted). See --example').option("--files <paths...>","eval case Markdown paths or globs (default gauge-cases/**/*.md)").option("--input <name[=path]>","attach a gauge.json input; =path or =glob replaces its configured files (repeatable)",collect5,[]).option("--detach","run a local batch in the background; print status and stop commands").option("--profile <profile>","local credential profile; overrides the default for every target").option("--runtime-image <image>","use a compatible custom Docker image for local runs (tag, digest, or image ID)").option("--gateway-image <image>","use a compatible custom gateway image for local web fixtures").option("--verify","check this repo's eval files; start no sessions").option("--dry-run","preview a prompt, this repo's eval cases, or an --apply spec; start no sessions").option("--require-experiment <label>","with --apply: fail unless every target resolves to this experiment label").option("--schema","print the JSON Schema of an --apply spec and exit").option("--example","print an example --apply spec and exit").addOption(new Option5("--format <format>","schema/example format (default: case)").choices(["case","battery"])).option("--yes","skip the confirmation").action(async(arg,opts,cmd)=>{if(arg!==void 0&&!/\s/.test(arg.trim()))throw new CliError("usage",`unknown run command "${arg}". Run a prompt with \`gauge run "<prompt>"\`. Launch a saved eval set or preference prompt with \`gauge run --eval <id>\` or \`--preference <id>\`; read runs with \`gauge run list\` or \`gauge run get <id> [--logs]\`.`);if(opts.detach&&(gaugeState(cmd).mode!=="local"||opts.dryRun||opts.verify||opts.schema||opts.example))throw new CliError("usage","--detach requires a local run");if((opts.runtimeImage||opts.gatewayImage)&&gaugeState(cmd).mode!=="local")throw new CliError("usage","--runtime-image and --gateway-image work only on local runs; pass --local");await launch(cmd,{...opts,prompt:arg?.trim()})});run.hook("preAction",(parent,action)=>{if(action===parent)return;const allowed=action.name()==="list"?["eval","preference","agent"]:[];const stray=Object.entries(parent.opts()).filter(([key,value])=>!allowed.includes(key)&&value!==void 0&&!(Array.isArray(value)&&value.length===0)).map(([key])=>`--${key.replace(/[A-Z]/g,c=>`-${c.toLowerCase()}`)}`);if(stray.length>0)throw new CliError("usage",`${stray.join(", ")} ${stray.length===1?"is a launch flag":"are launch flags"}; \`gauge run ${action.name()}\` does not take ${stray.length===1?"it":"them"}`)});run.command("list").description("List runs (newest first). Filters combine. --preference <id> lists one prompt's runs with verdicts and takes no other filter. Local: runs in .gauge/").option("--eval <id>","filter by eval set (local: by eval)").option("--preference <id>","one preference prompt's runs with verdicts (no other filter)").option("--status <status>","filter by status, e.g. running,queued (local: completed,failed)").option("--batch <id>","filter by batch").option("--prompt <id>","filter by prompt").option("--experiment <experiment>","filter by experiment label").option("--since <window>","e.g. 24h, 7d, or an ISO date").option("--agent <agent>","filter by agent (local only)").option("--model <model>","filter by model (local only)").option("--unanalyzed","only runs without a published analysis (local only)").option("--limit <n>","page size (1-200)","50").option("--cursor <cursor>","resume from a previous page's nextCursor").action(async(_opts,command)=>{const opts=command.optsWithGlobals();const agents=[opts.agent??[]].flat();if(agents.length>1)throw new CliError("usage","pass one --agent to filter by");const agent=agents[0];const output=resolveOutput(command);const state=gaugeState(command);if(state.mode==="local"){if(opts.prompt||opts.experiment||opts.cursor)throw new CliError("usage","Local run list does not support --prompt, --experiment, or --cursor. Use --eval or --preference to select a saved measurement.");const items=state.root?listLocalRuns(join11(state.root,".gauge"),{batch:opts.batch,evalId:opts.eval,preferenceId:opts.preference,agent:agent?parseAgents(agent)[0]:void 0,model:opts.model,outcome:opts.status,since:opts.since?parseSince(opts.since):void 0,analyzed:opts.unanalyzed?false:void 0}).slice(0,Number(opts.limit)):[];if(output==="json")printJson({items});else printItems(items.map(toLocalRow),"table");return}const localOnly=[agent!==void 0&&"--agent",opts.model!==void 0&&"--model",opts.unanalyzed&&"--unanalyzed"].filter(Boolean);if(localOnly.length>0)throw new CliError("usage",`${localOnly.join(", ")} ${localOnly.length===1?"works":"work"} only on local runs; pass --local`);if(opts.preference!==void 0){const filters=[opts.eval!==void 0&&"--eval",opts.status!==void 0&&"--status",opts.batch!==void 0&&"--batch",opts.prompt!==void 0&&"--prompt",opts.experiment!==void 0&&"--experiment",opts.since!==void 0&&"--since"].filter(Boolean);if(filters.length>0)throw new CliError("usage",`--preference lists one prompt's runs and takes no other filter; drop ${filters.join(", ")}`);await listPreferencePromptRuns(command,opts.preference,opts);return}const org=await resolveOrg(command);const page=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/runs`,{status:opts.status,batch:opts.batch,prompt:opts.prompt,experiment:opts.experiment,evalSet:opts.eval,since:opts.since,limit:opts.limit,cursor:opts.cursor});if(output==="json"){printJson(page);return}printItems(page.items.map(toRow4),"table");if(page.nextCursor){console.error(`(more results \u2014 rerun with --cursor ${page.nextCursor})`)}});run.command("get <id>").description("Show a run (status, trigger source, cost, credits, turns, exit reason). --logs prints its trace events instead").option("--logs","print the trace events so far, then exit").action(async(id,opts,command)=>{if(gaugeState(command).mode==="local"){const dir=localDirectory2(command);const session=withLocalStore(command,store=>store.getSession(id));if(!session)throw new CliError("usage",`Local run ${id} was not found. If its files are still in .gauge/runs, run \`gauge run repair --local\` to restore it to the run list.`);if(opts.logs){const path=join11(dir,"runs",session.id,"attempts",String(session.attempt),"session","events.ndjson");if(!existsSync4(path))throw new CliError("usage","This run has no saved transcript yet.");process.stdout.write(readFileSync12(path,"utf8"))}else printJson({...session,analysis:getLocalAnalysis(dir,id)});return}if(opts.logs){await streamRun(id,resolveOutput(command)==="json",{idleMs:2e3,firstEventMs:15e3,stopAtLiveMs:Date.now()-2e3});return}const run2=await createClient().get(runPath(id));if(resolveOutput(command)==="json"){printJson(run2);return}const lines=detailLines(run2);const width=Math.max(...lines.map(([label])=>label.length));for(const[label,value]of lines){console.log(`${label.padEnd(width)} ${value}`)}});run.command("insight <id>").description("Show the run's findings, criteria, and product decisions").action(async(id,_opts,command)=>{if(gaugeState(command).mode==="local"){const dir=localDirectory2(command);if(!withLocalStore(command,store=>store.getSession(id)))throw new CliError("usage",`Local run ${id} was not found.`);const analysis2=getLocalAnalysis(dir,id);if(resolveOutput(command)==="json")printJson({runId:id,analysis:analysis2});else console.log(analysis2?renderLocalAnalysisMarkdown(analysis2):"No analysis published. Follow gauge instructions session-analysis --local.");return}const insight=await createClient().get(`${runPath(id)}/insight`);if(resolveOutput(command)==="json"){printJson(insight);return}printInsight(insight)});const analysis=run.command("analysis").description("Publish structured analysis from your coding agent");analysis.command("schema").description("Print the versioned analysis submission schema").action(()=>printJson(z14.toJSONSchema(analysisSubmissionSchema)));analysis.command("save <id>").description("Validate and publish analysis for a local session attempt").requiredOption("--file <path>","analysis JSON file").action((id,opts,command)=>{if(gaugeState(command).mode!=="local")throw new CliError("usage","Analysis submission is available locally; hosted Gauge publishes its own analysis. Use --local.");let input;try{input=JSON.parse(readFileSync12(opts.file,"utf8"))}catch(error){throw new CliError("usage",`Cannot read analysis JSON: ${String(error)}`)}const dir=localDirectory2(command);const saved=saveLocalAnalysis(dir,id,input);const results=join11(dir,"runs",saved.runId,"results");try{mkdirSync2(results,{recursive:true});writeFileSync4(join11(results,"session-analysis.md"),renderLocalAnalysisMarkdown(saved))}catch(error){console.error(`Analysis ${saved.id} was published in SQLite, but its Markdown report could not be written: ${String(error)}. Read it with gauge run insight ${id} --local.`)}if(resolveOutput(command)==="json")printJson(saved);else console.log(`Published analysis ${saved.id} (revision ${saved.revision}); ${saved.findings.length} findings. Read with gauge run insight ${id} --local.`)});run.command("repair").description("Restore local runs to the run list from their saved files, keeping measurements and analysis").action((_opts,command)=>{if(gaugeState(command).mode!=="local")throw new CliError("usage","Session repair is local only; pass --local.");printJson(withLocalStore(command,store=>store.repairSessionSummaries()))});hostedOnly(run.command("regrade <id>")).description("Re-judge using current eval criteria or the run's criteria snapshot (no new session)").action(async(id,_opts,command)=>{const result=await createClient().post(`${runPath(id)}/regrade`);if(resolveOutput(command)==="json"){printJson(result);return}console.log(`Re-judging run ${result.runId} against ${result.criteria} ${result.criteria===1?"criterion":"criteria"} ${result.evalSetId?`of eval ${result.evalSetId}`:"from its launch-time snapshot"}.`);console.log("Verdicts land on the run; `gauge run get <id>` to check.")});hostedOnly(run.command("watch <id>")).description("Stream normalized trace events live (SSE; exits when the run finishes)").action(async(id,_opts,command)=>{await streamRun(id,resolveOutput(command)==="json",{idleMs:0,firstEventMs:0,pollStatus:true})});hostedOnly(run.command("trace <id>")).description("A window of the run's trace around a seq, one line per event: web fetches, tool calls, results, and messages").requiredOption("--seq <n>","the seq at the center, e.g. a finding's evidence seq").option("--before <n>","events before the seq (default 30)").option("--after <n>","events after the seq (default 10)").action(async(id,opts,command)=>{const res=await createClient().get(`${runPath(id)}/trace`,{seq:opts.seq,before:opts.before,after:opts.after});if(resolveOutput(command)==="json")printJson(res);else for(const line of res.lines)console.log(line)});hostedOnly(run.command("web <id>")).description("Every URL the run's agent fetched, with the seqs, and its searches").action(async(id,_opts,command)=>{const res=await createClient().get(`${runPath(id)}/web`);if(resolveOutput(command)==="json")return printJson(res);for(const page of res.pages)console.log(`${page.seqs.join(",")} ${page.url}`);for(const search of res.searches)console.log(`${search.seq} search: ${search.query}`);if(!res.pages.length&&!res.searches.length)console.log("No web activity recorded.")});hostedOnly(run.command("commands <id>")).description("Shell commands the run's agent executed: without --seq, commands matching --query with their first output; with --seq, that call's full output").option("--query <text>","only commands whose text contains this").option("--seq <n>","one call from the listing").option("--offset <n>","page into a long output").action(async(id,opts)=>{printJson(await createClient().get(`${runPath(id)}/commands`,{query:opts.query,seq:opts.seq,offset:opts.offset}))});run.command("diff <id>").description("Print the run's working-tree diff to stdout").action(async(id,_opts,command)=>{if(gaugeState(command).mode==="local"){const session=withLocalStore(command,store=>store.getSession(id));if(!session)throw new CliError("usage",`Local run ${id} was not found.`);const path=join11(localDirectory2(command),"runs",session.id,"attempts",String(session.attempt),"session","diff.patch");if(existsSync4(path))process.stdout.write(readFileSync12(path,"utf8"));else console.error("(no diff)");return}const diff2=await createClient().get(`${runPath(id)}/diff`);if(!diff2){console.error("(no diff)");return}process.stdout.write(diff2)});run.command("status [id]").description("Print a launch request's progress once, without waiting; local: live session status and batch progress").option("--batch <batch-id>","show this local batch").action(async(id,opts,command)=>{const state=gaugeState(command);if(state.mode!=="local"){if(!id||opts.batch)throw new CliError("usage","Hosted status requires a request ID; --batch is local only");const result=await createClient().get(await runRequestPath(command,id));if(resolveOutput(command)==="json")printJson(result);else printRunRequest(result);return}if(!state.root)throw new CliError("usage","Run this command inside the batch's Gauge workspace");if(id&&opts.batch)throw new CliError("usage","Pass a run ID or --batch, not both");const gaugeDir=join11(state.root,".gauge");const runs=await localStatus(gaugeDir,{id,batch:opts.batch});let batch;if(opts.batch){try{batch=await localBatchStatus(gaugeDir,opts.batch)}catch(error){if(error.code!=="ENOENT")throw error}}if((id||opts.batch)&&!runs.length&&!batch)throw new CliError("not_found","Local run or batch not found");if(resolveOutput(command)==="json")printJson({batch,runs});else printLocalStatus(batch,runs)});run.command("cancel [id]").option("--batch <batch-id>","cancel an entire local batch, including queued sessions").description("Request cancellation of a run").action(async(id,opts,command)=>{const state=gaugeState(command);if(state.mode==="local"){if(!state.root)throw new CliError("usage","Run this command inside the batch's Gauge workspace");const results=await cancelLocal(join11(state.root,".gauge"),{id,batch:opts.batch});if(results.some(r=>r.status==="pending"))process.exitCode=1;if(resolveOutput(command)==="json")printJson({cancelRequested:true,runs:results});else{console.log(`Cancellation requested for ${opts.batch??id}.`);for(const result2 of results)console.log(`${result2.id}: ${result2.status}${"detail"in result2?` (${result2.detail})`:""}`)}return}if(!id||opts.batch)throw new CliError("usage","Hosted cancellation requires a run ID; --batch is local only");const result=await createClient().post(`${runPath(id)}/cancel`);if(resolveOutput(command)==="json"){printJson(result);return}console.log(`Run ${result.id}: ${result.status}${result.cancelRequested?" (cancel requested)":""}`);if(result.note)console.error(result.note)});run.command("compare").description("Compare two sets of runs: pass rate, criteria, and median turns, cost, and duration, per agent and model. Local: completed runs and median turns").requiredOption("--baseline <key=value>","baseline runs (repeatable): request=, batch=, eval=, revision=, experiment=, trial=, agent=, model=, since=, until= (local: batch=, eval=<saved id>, revision=<number>, agent=, model=, since=, until=)",collect5,[]).requiredOption("--candidate <key=value>","candidate runs (repeatable), same keys as --baseline",collect5,[]).addHelpText("after",`
|
|
795
968
|
${style("1")(" EXAMPLES")}
|
|
796
969
|
gauge run compare --baseline request=<id> --candidate request=<id>
|
|
797
970
|
gauge run compare --baseline eval=<id> --candidate trial=<trialId>
|
|
798
971
|
gauge run compare --baseline eval=<id> --baseline revision=<hash> \\
|
|
799
972
|
--candidate eval=<id> --candidate since=2h
|
|
800
|
-
gauge run compare --local --baseline batch=<id> --candidate batch=<id>`).action(async(opts,cmd)=>{const body={baseline:parseCompareSide("--baseline",opts.baseline),candidate:parseCompareSide("--candidate",opts.candidate)};const state=gaugeState(cmd);let result;let warnings;if(state.mode==="local"){const baseline=localCompareFilter("--baseline",body.baseline);const candidate=localCompareFilter("--candidate",body.candidate);const root=state.root;if(!root)throw new CliError("usage","Local runs need a Gauge workspace. Run `gauge init` in this directory first.");const gaugeDir=join12(root,".gauge");result=compareLocalRuns(listLocalRuns(gaugeDir,baseline),listLocalRuns(gaugeDir,candidate));warnings=localCompareWarnings}else{result=await createClient().post(`/api/v1/orgs/${encodeURIComponent(await resolveOrg(cmd))}/runs/compare`,body);warnings=compareWarnings}if(resolveOutput(cmd)==="json"){printJson(result);return}for(const warning of warnings(result))console.error(`Warning: ${warning}`);printComparison(result)});run.command("wait [request-id]").description("Wait for every session and judge in a launch request, then print the verdict. Exits 0 once settled, whatever the verdict. Local: wait for a batch with --batch").option("--batch <batch-id>","wait for this local batch").option("--check","exit 1 unless the verdict is PASSED or MEASURED (for CI gates); local: unless every session completed").action(async(id,opts,cmd)=>{const state=gaugeState(cmd);if(state.mode==="local"){if(!opts.batch||id)throw new CliError("usage","Local wait takes --batch <batch-id>, not a request ID");if(!state.root)throw new CliError("usage","Run this command inside the batch's Gauge workspace");return waitForLocalBatch(cmd,join12(state.root,".gauge"),opts.batch,opts)}if(!id||opts.batch)throw new CliError("usage","Hosted wait requires a request ID; --batch is local only");return waitForRequest(cmd,id,opts)})}var skills_exports={};__export(skills_exports,{buildSkillEditBody:()=>buildSkillEditBody,register:()=>register30});function skillsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/skills`}function skillPath(org,ref){return`${skillsPath(org)}/${encodeURIComponent(ref)}`}function sourceLabel(s){if(!s)return"";const path=s.subpath?`/${s.subpath}`:"";return`${s.repoFullName}${path}#${s.ref}${s.autoIngest?" (auto)":""}`}async function buildSkillEditBody(opts,fetchCurrentSource){if(opts.note!==void 0&&opts.clearNote)throw new CliError("usage","--note and --clear-note conflict");if(opts.source!==void 0&&opts.clearSource)throw new CliError("usage","--source and --clear-source conflict");if(opts.autoIngest!==void 0&&opts.clearSource)throw new CliError("usage","--auto-ingest and --clear-source conflict");let autoIngest;if(opts.autoIngest!==void 0){if(opts.autoIngest!=="true"&&opts.autoIngest!=="false")throw new CliError("usage","--auto-ingest expects true or false");autoIngest=opts.autoIngest==="true"}const body={};if(opts.rename!==void 0)body.name=opts.rename;if(opts.clearNote)body.note=null;else if(opts.note!==void 0)body.note=opts.note;if(opts.clearSource)body.source=null;else if(opts.source!==void 0){const m=opts.source.match(/^([^#]+\/[^#/]+)#([^/]+)(?:\/(.+))?$/);if(!m)throw new CliError("usage",'invalid --source (expected "owner/repo#branch" or "owner/repo#branch/subpath")');body.source={repoFullName:m[1],ref:m[2],subpath:m[3],autoIngest}}else if(autoIngest!==void 0){const current=await fetchCurrentSource();if(!current)throw new CliError("usage","this skill has no tracked source \u2014 set one first with --source owner/repo#branch");body.source={repoFullName:current.repoFullName,ref:current.ref,subpath:current.subpath||void 0,autoIngest}}if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");return body}function register30(program2){const group=hostedOnly(program2.command("skills")).description("Manage agent skills used by eval and preference runs");group.command("list").description("List the org's skills with their newest version").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(skillsPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(s=>({name:s.name,latest:s.latestVersion?.label??"",versions:s.versionCount,source:sourceLabel(s.source),note:s.note??""})),"table")});group.command("get <ref>").description("Show one skill: source, note, every version").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const skill=await createClient().get(skillPath(org,ref));if(resolveOutput(cmd)==="json"){printJson(skill);return}console.log(`name: ${skill.name}`);console.log(`id: ${skill.id}`);console.log(`note: ${skill.note??"(none)"}`);console.log(`source: ${sourceLabel(skill.source)||"(untracked)"}`);console.log("versions:");for(const v of skill.versions)console.log(` ${skill.name}@${v.label} ${v.contentDigest.slice(0,19)} ${v.createdAt}`)});group.command("add <reference>").description("Ingest a skill from owner/repo[/skill][#ref] or a GitHub URL").option("--as <handle>","the handle (default: the directory basename)").option("--label <label>","version label (a typed label always mints; omitted dedups by digest)").option("--skill <name>","select one skill from a source holding several").option("--note <text>","note shown on the index").option("--track","track the source branch for auto re-ingest (reference must name a branch)").action(async(reference,opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().post(skillsPath(org),{reference,handle:opts.as,label:opts.label,skill:opts.skill,note:opts.note,trackSource:opts.track});if(resolveOutput(cmd)==="json"){printJson(result);return}console.log(result.unchanged?`Already current: ${result.handle}@${result.label} carries this content`:`Ingested ${result.handle}@${result.label}`);console.error(`Attach it to an eval: gauge eval edit <id> --skill ${result.handle}@${result.label}`)});group.command("edit <ref>").description("Rename, note, or repoint a skill's tracked source").option("--rename <handle>","new handle (name@label refs must be updated)").option("--note <text>","replace the note").option("--clear-note","remove the note").option("--source <owner/repo#branch[/subpath]>","set the tracked source branch").option("--clear-source","stop tracking the source").option("--auto-ingest <bool>","true|false \u2014 auto re-ingest on new commits").action(async(ref,opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const body=await buildSkillEditBody(opts,async()=>{const skill=await client.get(skillPath(org,ref));return skill.source});const updated=await client.patch(skillPath(org,ref),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated skill ${updated.name}`)});group.command("reingest <ref>").description("Pull the tracked source now (no-op when already current)").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().post(`${skillPath(org,ref)}/reingest`,{});if(resolveOutput(cmd)==="json")printJson(result);else console.log(result.unchanged?"Already current \u2014 nothing new upstream":`Ingested new version ${ref}@${result.label}`)});group.command("rm <ref>").description("Remove a skill; the handle stays reserved and finished runs keep their snapshot").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().delete(skillPath(org,ref));if(resolveOutput(cmd)==="json")printJson(result);else console.log(`Deleted skill ${ref} (${result.referencedRuns} finished run(s) referenced it)`)});group.command("runs <ref>").description("Recent sessions that staged this skill").option("--limit <n>","max runs (default 10, cap 50)").action(async(ref,opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(`${skillPath(org,ref)}/runs`,{limit:opts.limit});if(resolveOutput(cmd)==="json"){printJson(res.items);return}printItems(res.items.map(r=>({id:r.runId,agent:r.agent,version:r.versionLabel??"",turns:r.turns??"",cost:r.costUsd==null?"":`$${r.costUsd.toFixed(2)}`,finished:r.finishedAt??"(running)"})),"table")})}var stats_exports={};__export(stats_exports,{register:()=>register31});function addFilterOptions(cmd){return cmd.option("--prompt <id>","filter to a single prompt").option("--tag <ids>","comma-separated tag ids").option("--repo <urls>","comma-separated base repo urls").option("--language <langs>","comma-separated repo languages").option("--framework <names>","comma-separated repo frameworks").option("--size <buckets>","comma-separated codebase-size buckets").option("--branded","only branded prompts").option("--unbranded","only non-branded prompts").option("--limit <n>","max rows (default 50)").option("--full","return all rows (up to 1000)")}function filterQuery(opts){if(opts.branded&&opts.unbranded){throw new CliError("usage","Pass at most one of --branded / --unbranded.")}return{prompt:opts.prompt,tag:opts.tag,repo:opts.repo,language:opts.language,framework:opts.framework,size:opts.size,branded:opts.branded?"true":opts.unbranded?"false":void 0,limit:opts.full?"1000":opts.limit}}var pct4=rate=>`${(rate*100).toFixed(1)}%`;var delta=d=>d===null||d===0?"\u2014":d>0?`+${d}`:String(d);function noteFreshness2(f){if(!f.available){console.error("warning: analytics store is temporarily unavailable; results may be incomplete.")}else if(f.pending>0){console.error(`note: ${f.pending} recent run(s) still ingesting; results may not reflect them yet.`)}}async function runStats(command,resource,opts,toRow5){const org=await resolveOrg(command);const output=resolveOutput(command);const res=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/stats/${resource}`,filterQuery(opts));if(output==="json"){printJson(res);return}printItems(res.items.map(toRow5),"table");noteFreshness2(res.freshness)}var rankingRow=r=>({rank:r.rank,brand:r.brandName,kind:r.kind,installs:r.installs,"install%":pct4(r.installRate),"mention%":pct4(r.mentionRate),delta:delta(r.deltaRank)});var packageRow=r=>({package:r.name,ecosystem:r.ecosystem,brand:r.brandName,kind:r.kind,installs:r.installs,rate:pct4(r.rate)});var brandRow=r=>({brand:r.brandName,kind:r.kind,installs:r.installs,rate:pct4(r.rate)});var funnelRow=r=>({brand:r.brandName,kind:r.kind,"mention%":pct4(r.mentionRate),"install%":pct4(r.installRate)});var domainRow=r=>({domain:r.domain,brand:r.brandName,runs:r.runs,rate:pct4(r.rate),requests:r.requests});function register31(program2){const group=hostedOnly(program2.command("stats")).description("Read-only analytics over your runs");addFilterOptions(group.command("rankings").description("Brands ranked by install rate")).action((opts,command)=>runStats(command,"rankings",opts,rankingRow));addFilterOptions(group.command("installs").description("Package install rates (share of voice)")).action((opts,command)=>runStats(command,"installs",opts,packageRow));addFilterOptions(group.command("brands").description("Per-brand (company-level) install rates")).action((opts,command)=>runStats(command,"brands",opts,brandRow));addFilterOptions(group.command("funnel").description("Per-brand mention rate beside install rate")).action((opts,command)=>runStats(command,"funnel",opts,funnelRow));addFilterOptions(group.command("domains").description("Outbound fetch rates by registrable domain")).action((opts,command)=>runStats(command,"domains",opts,domainRow))}var status_exports={};__export(status_exports,{register:()=>register32});function pct5(rate){return rate==null?"\u2014":`${Math.round(rate*100)}%`}function delta2(pp){if(pp==null)return"";const rounded=Math.round(pp);return rounded===0?"":` (${rounded>0?"+":""}${rounded}pp)`}function register32(program2){hostedOnly(program2.command("status")).description("One-shot org summary: actions, Agent Preference, agent experience").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const s=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/summary`);if(resolveOutput(cmd)==="json"){printJson(s);return}if(!s.hasAnyRuns){console.log("No runs yet. Create an eval (`gauge eval create`) or run one (`gauge run --eval <id>`).");return}console.log(`last synthesis: ${s.lastSynthesisAt??"never"}`);console.log("\nagent experience:");if(!s.agentExperience.ok){console.log(` (unavailable: ${s.agentExperience.error})`)}else if(s.agentExperience.data.cards.length===0){console.log(" (no judged runs in the last 30 days)")}else{for(const card of s.agentExperience.data.cards){console.log(` ${card.name}: ${pct5(card.passRate)}${delta2(card.deltaPP)} over ${card.judgedRuns} judged run(s)`)}const extra=s.agentExperience.data.setCount-s.agentExperience.data.cards.length;if(extra>0)console.log(` (+${extra} more \u2014 see the dashboard)`)}console.log("\nAgent Preference:");if(!s.visibility.ok){console.log(` (unavailable: ${s.visibility.error})`)}else if(s.visibility.data.cards.length===0){console.log(" (no preference runs in the last 30 days)")}else{for(const card of s.visibility.data.cards){const owned=card.owned?`${card.owned.name} #${card.owned.rank}, chosen ${pct5(card.owned.chosenRate)}${delta2(card.chosenDeltaPP)}`:"no owned brand ranked";console.log(` ${card.name}: ${owned}`)}}console.log("\nopen actions:");if(!s.actions.ok){console.log(` (unavailable: ${s.actions.error})`)}else if(s.actions.data.length===0){console.log(" (none)")}else{for(const a of s.actions.data.slice(0,5)){console.log(` [${a.priority}] ${a.title} (${a.id})`)}if(s.actions.data.length>5)console.log(` (+${s.actions.data.length-5} more \u2014 \`gauge actions list\`)`)}})}var tags_exports={};__export(tags_exports,{register:()=>register33});function tagsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/tags`}function register33(program2){const group=hostedOnly(program2.command("tags")).description("Manage org tags");group.command("list").description("List tags").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const items=await fetchAllPages(createClient(),tagsPath(org));if(resolveOutput(cmd)==="json")printJson(items);else printItems(items,"table")});group.command("add <name>").description("Add a tag").option("--color <color>","palette color (e.g. blue, grape; default gray)").action(async(name,opts,cmd)=>{const org=await resolveOrg(cmd);const tag=await createClient().post(tagsPath(org),{name,color:opts.color});if(resolveOutput(cmd)==="json")printJson(tag);else console.log(`Created tag ${tag.id} (${tag.name}, ${tag.color})`)});group.command("rename <tag> <new-name>").description("Rename a tag (by id or current name)").option("--color <color>","also change the color").action(async(tag,newName,opts,cmd)=>{const org=await resolveOrg(cmd);const updated=await createClient().patch(`${tagsPath(org)}/${encodeURIComponent(tag)}`,{name:newName,color:opts.color});if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Renamed tag ${updated.id} to ${updated.name} (${updated.color})`)});group.command("rm <tag>").description("Delete a tag (by id or name)").action(async(tag,_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(`${tagsPath(org)}/${encodeURIComponent(tag)}`);if(resolveOutput(cmd)!=="json")console.log(`Deleted tag ${tag}`)})}var telemetry_exports={};__export(telemetry_exports,{register:()=>register34});async function printStatus(command){const status=await telemetryStatus();if(command.optsWithGlobals().output==="json"){printJson(status);return}console.log(`CLI telemetry is ${status.enabled?"enabled":"disabled"} (${status.source}).`);if(status.source==="environment")console.log("GAUGE_TELEMETRY_DISABLED=1 overrides the saved preference.");if(status.source==="unavailable")console.log("Preferences could not be read. Collection is disabled until repaired with gauge telemetry enable or disable.")}function register34(program2){const telemetry=program2.command("telemetry").description("Control anonymous CLI usage telemetry (enabled by default)");telemetry.command("status").description("Show whether CLI telemetry is enabled").action(async(_options,command)=>printStatus(command));for(const enabled of[true,false]){telemetry.command(enabled?"enable":"disable").description(`${enabled?"Enable":"Disable"} CLI telemetry for this user`).action(async(_options,command)=>{await setTelemetryEnabled(enabled);await printStatus(command)})}}var update_exports={};__export(update_exports,{register:()=>register35});import{spawnSync as spawnSync2}from"child_process";function register35(program2,{run=spawnSync2,platform=process.platform}={}){program2.command("update").description("Update the globally installed Gauge CLI to the latest version").action((_opts,command)=>{const output=resolveOutput(command);console.error("Updating Gauge CLI to the latest published version...");const result=run(platform==="win32"?"npm.cmd":"npm",["install","--global","@withgauge/cli@latest"],{shell:platform==="win32",stdio:["inherit",output==="json"?2:"inherit","inherit"]});if(result.error||result.status!==0){const reason=result.error?result.error.message:result.signal?`npm was interrupted by ${result.signal}`:`npm exited with code ${result.status}`;throw new CliError("internal",`CLI update failed: ${reason}.
|
|
973
|
+
gauge run compare --local --baseline batch=<id> --candidate batch=<id>`).action(async(opts,cmd)=>{const body={baseline:parseCompareSide("--baseline",opts.baseline),candidate:parseCompareSide("--candidate",opts.candidate)};const state=gaugeState(cmd);let result;let warnings;if(state.mode==="local"){const baseline=localCompareFilter("--baseline",body.baseline);const candidate=localCompareFilter("--candidate",body.candidate);const root=state.root;if(!root)throw new CliError("usage","Local runs need a Gauge workspace. Run `gauge init` in this directory first.");const gaugeDir=join11(root,".gauge");result=compareLocalRuns(listLocalRuns(gaugeDir,baseline),listLocalRuns(gaugeDir,candidate));warnings=localCompareWarnings}else{result=await createClient().post(`/api/v1/orgs/${encodeURIComponent(await resolveOrg(cmd))}/runs/compare`,body);warnings=compareWarnings}if(resolveOutput(cmd)==="json"){printJson(result);return}for(const warning of warnings(result))console.error(`Warning: ${warning}`);printComparison(result)});run.command("wait [request-id]").description("Wait for every session and judge in a launch request, then print the verdict. Exits 0 once settled, whatever the verdict. Local: wait for a batch with --batch").option("--batch <batch-id>","wait for this local batch").option("--check","exit 1 unless the verdict is PASSED or MEASURED (for CI gates); local: unless every session completed").action(async(id,opts,cmd)=>{const state=gaugeState(cmd);if(state.mode==="local"){if(!opts.batch||id)throw new CliError("usage","Local wait takes --batch <batch-id>, not a request ID");if(!state.root)throw new CliError("usage","Run this command inside the batch's Gauge workspace");return waitForLocalBatch(cmd,join11(state.root,".gauge"),opts.batch,opts)}if(!id||opts.batch)throw new CliError("usage","Hosted wait requires a request ID; --batch is local only");return waitForRequest(cmd,id,opts)})}var skills_exports={};__export(skills_exports,{buildSkillEditBody:()=>buildSkillEditBody,register:()=>register30});function skillsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/skills`}function skillPath(org,ref){return`${skillsPath(org)}/${encodeURIComponent(ref)}`}function sourceLabel(s){if(!s)return"";const path=s.subpath?`/${s.subpath}`:"";return`${s.repoFullName}${path}#${s.ref}${s.autoIngest?" (auto)":""}`}async function buildSkillEditBody(opts,fetchCurrentSource){if(opts.note!==void 0&&opts.clearNote)throw new CliError("usage","--note and --clear-note conflict");if(opts.source!==void 0&&opts.clearSource)throw new CliError("usage","--source and --clear-source conflict");if(opts.autoIngest!==void 0&&opts.clearSource)throw new CliError("usage","--auto-ingest and --clear-source conflict");let autoIngest;if(opts.autoIngest!==void 0){if(opts.autoIngest!=="true"&&opts.autoIngest!=="false")throw new CliError("usage","--auto-ingest expects true or false");autoIngest=opts.autoIngest==="true"}const body={};if(opts.rename!==void 0)body.name=opts.rename;if(opts.clearNote)body.note=null;else if(opts.note!==void 0)body.note=opts.note;if(opts.clearSource)body.source=null;else if(opts.source!==void 0){const m=opts.source.match(/^([^#]+\/[^#/]+)#([^/]+)(?:\/(.+))?$/);if(!m)throw new CliError("usage",'invalid --source (expected "owner/repo#branch" or "owner/repo#branch/subpath")');body.source={repoFullName:m[1],ref:m[2],subpath:m[3],autoIngest}}else if(autoIngest!==void 0){const current=await fetchCurrentSource();if(!current)throw new CliError("usage","this skill has no tracked source \u2014 set one first with --source owner/repo#branch");body.source={repoFullName:current.repoFullName,ref:current.ref,subpath:current.subpath||void 0,autoIngest}}if(Object.keys(body).length===0)throw new CliError("usage","nothing to change");return body}function register30(program2){const group=hostedOnly(program2.command("skills")).description("Manage agent skills used by eval and preference runs");group.command("list").description("List the org's skills with their newest version").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(skillsPath(org));if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(s=>({name:s.name,latest:s.latestVersion?.label??"",versions:s.versionCount,source:sourceLabel(s.source),note:s.note??""})),"table")});group.command("get <ref>").description("Show one skill: source, note, every version").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const skill=await createClient().get(skillPath(org,ref));if(resolveOutput(cmd)==="json"){printJson(skill);return}console.log(`name: ${skill.name}`);console.log(`id: ${skill.id}`);console.log(`note: ${skill.note??"(none)"}`);console.log(`source: ${sourceLabel(skill.source)||"(untracked)"}`);console.log("versions:");for(const v of skill.versions)console.log(` ${skill.name}@${v.label} ${v.contentDigest.slice(0,19)} ${v.createdAt}`)});group.command("add <reference>").description("Ingest a skill from owner/repo[/skill][#ref] or a GitHub URL").option("--as <handle>","the handle (default: the directory basename)").option("--label <label>","version label (a typed label always mints; omitted dedups by digest)").option("--skill <name>","select one skill from a source holding several").option("--note <text>","note shown on the index").option("--track","track the source branch for auto re-ingest (reference must name a branch)").action(async(reference,opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().post(skillsPath(org),{reference,handle:opts.as,label:opts.label,skill:opts.skill,note:opts.note,trackSource:opts.track});if(resolveOutput(cmd)==="json"){printJson(result);return}console.log(result.unchanged?`Already current: ${result.handle}@${result.label} carries this content`:`Ingested ${result.handle}@${result.label}`);console.error(`Attach it to an eval: gauge eval edit <id> --skill ${result.handle}@${result.label}`)});group.command("edit <ref>").description("Rename, note, or repoint a skill's tracked source").option("--rename <handle>","new handle (name@label refs must be updated)").option("--note <text>","replace the note").option("--clear-note","remove the note").option("--source <owner/repo#branch[/subpath]>","set the tracked source branch").option("--clear-source","stop tracking the source").option("--auto-ingest <bool>","true|false \u2014 auto re-ingest on new commits").action(async(ref,opts,cmd)=>{const org=await resolveOrg(cmd);const client=createClient();const body=await buildSkillEditBody(opts,async()=>{const skill=await client.get(skillPath(org,ref));return skill.source});const updated=await client.patch(skillPath(org,ref),body);if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Updated skill ${updated.name}`)});group.command("reingest <ref>").description("Pull the tracked source now (no-op when already current)").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().post(`${skillPath(org,ref)}/reingest`,{});if(resolveOutput(cmd)==="json")printJson(result);else console.log(result.unchanged?"Already current \u2014 nothing new upstream":`Ingested new version ${ref}@${result.label}`)});group.command("rm <ref>").description("Remove a skill; the handle stays reserved and finished runs keep their snapshot").action(async(ref,_opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().delete(skillPath(org,ref));if(resolveOutput(cmd)==="json")printJson(result);else console.log(`Deleted skill ${ref} (${result.referencedRuns} finished run(s) referenced it)`)});group.command("runs <ref>").description("Recent sessions that staged this skill").option("--limit <n>","max runs (default 10, cap 50)").action(async(ref,opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(`${skillPath(org,ref)}/runs`,{limit:opts.limit});if(resolveOutput(cmd)==="json"){printJson(res.items);return}printItems(res.items.map(r=>({id:r.runId,agent:r.agent,version:r.versionLabel??"",turns:r.turns??"",cost:r.costUsd==null?"":`$${r.costUsd.toFixed(2)}`,finished:r.finishedAt??"(running)"})),"table")})}var stats_exports={};__export(stats_exports,{register:()=>register31});function addFilterOptions(cmd){return cmd.option("--prompt <id>","filter to a single prompt").option("--tag <ids>","comma-separated tag ids").option("--repo <urls>","comma-separated base repo urls").option("--language <langs>","comma-separated repo languages").option("--framework <names>","comma-separated repo frameworks").option("--size <buckets>","comma-separated codebase-size buckets").option("--branded","only branded prompts").option("--unbranded","only non-branded prompts").option("--limit <n>","max rows (default 50)").option("--full","return all rows (up to 1000)")}function filterQuery(opts){if(opts.branded&&opts.unbranded){throw new CliError("usage","Pass at most one of --branded / --unbranded.")}return{prompt:opts.prompt,tag:opts.tag,repo:opts.repo,language:opts.language,framework:opts.framework,size:opts.size,branded:opts.branded?"true":opts.unbranded?"false":void 0,limit:opts.full?"1000":opts.limit}}var pct3=rate=>`${(rate*100).toFixed(1)}%`;var delta=d=>d===null||d===0?"\u2014":d>0?`+${d}`:String(d);function noteFreshness2(f){if(!f.available){console.error("warning: analytics store is temporarily unavailable; results may be incomplete.")}else if(f.pending>0){console.error(`note: ${f.pending} recent run(s) still ingesting; results may not reflect them yet.`)}}async function runStats(command,resource,opts,toRow5){const org=await resolveOrg(command);const output=resolveOutput(command);const res=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/stats/${resource}`,filterQuery(opts));if(output==="json"){printJson(res);return}printItems(res.items.map(toRow5),"table");noteFreshness2(res.freshness)}var rankingRow=r=>({rank:r.rank,brand:r.brandName,kind:r.kind,installs:r.installs,"install%":pct3(r.installRate),"mention%":pct3(r.mentionRate),delta:delta(r.deltaRank)});var packageRow=r=>({package:r.name,ecosystem:r.ecosystem,brand:r.brandName,kind:r.kind,installs:r.installs,rate:pct3(r.rate)});var brandRow=r=>({brand:r.brandName,kind:r.kind,installs:r.installs,rate:pct3(r.rate)});var funnelRow=r=>({brand:r.brandName,kind:r.kind,"mention%":pct3(r.mentionRate),"install%":pct3(r.installRate)});var domainRow=r=>({domain:r.domain,brand:r.brandName,runs:r.runs,rate:pct3(r.rate),requests:r.requests});function register31(program2){const group=hostedOnly(program2.command("stats")).description("Read-only analytics over your runs");addFilterOptions(group.command("rankings").description("Brands ranked by install rate")).action((opts,command)=>runStats(command,"rankings",opts,rankingRow));addFilterOptions(group.command("installs").description("Package install rates (share of voice)")).action((opts,command)=>runStats(command,"installs",opts,packageRow));addFilterOptions(group.command("brands").description("Per-brand (company-level) install rates")).action((opts,command)=>runStats(command,"brands",opts,brandRow));addFilterOptions(group.command("funnel").description("Per-brand mention rate beside install rate")).action((opts,command)=>runStats(command,"funnel",opts,funnelRow));addFilterOptions(group.command("domains").description("Outbound fetch rates by registrable domain")).action((opts,command)=>runStats(command,"domains",opts,domainRow))}var status_exports={};__export(status_exports,{register:()=>register32});function pct4(rate){return rate==null?"\u2014":`${Math.round(rate*100)}%`}function delta2(pp){if(pp==null)return"";const rounded=Math.round(pp);return rounded===0?"":` (${rounded>0?"+":""}${rounded}pp)`}function register32(program2){hostedOnly(program2.command("status")).description("One-shot org summary: Agent Preference and agent experience").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const s=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/summary`);if(resolveOutput(cmd)==="json"){printJson(s);return}if(!s.hasAnyRuns){console.log("No runs yet. Create an eval (`gauge eval create`) or run one (`gauge run --eval <id>`).");return}console.log(`last synthesis: ${s.lastSynthesisAt??"never"}`);console.log("\nagent experience:");if(!s.agentExperience.ok){console.log(` (unavailable: ${s.agentExperience.error})`)}else if(s.agentExperience.data.cards.length===0){console.log(" (no judged runs in the last 30 days)")}else{for(const card of s.agentExperience.data.cards){console.log(` ${card.name}: ${pct4(card.passRate)}${delta2(card.deltaPP)} over ${card.judgedRuns} judged run(s)`)}const extra=s.agentExperience.data.setCount-s.agentExperience.data.cards.length;if(extra>0)console.log(` (+${extra} more \u2014 see the dashboard)`)}console.log("\nAgent Preference:");if(!s.visibility.ok){console.log(` (unavailable: ${s.visibility.error})`)}else if(s.visibility.data.cards.length===0){console.log(" (no preference runs in the last 30 days)")}else{for(const card of s.visibility.data.cards){const owned=card.owned?`${card.owned.name} #${card.owned.rank}, chosen ${pct4(card.owned.chosenRate)}${delta2(card.chosenDeltaPP)}`:"no owned brand ranked";console.log(` ${card.name}: ${owned}`)}}})}var tags_exports={};__export(tags_exports,{register:()=>register33});function tagsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/tags`}function register33(program2){const group=hostedOnly(program2.command("tags")).description("Manage org tags");group.command("list").description("List tags").action(async(_opts,cmd)=>{const org=await resolveOrg(cmd);const items=await fetchAllPages(createClient(),tagsPath(org));if(resolveOutput(cmd)==="json")printJson(items);else printItems(items,"table")});group.command("add <name>").description("Add a tag").option("--color <color>","palette color (e.g. blue, grape; default gray)").action(async(name,opts,cmd)=>{const org=await resolveOrg(cmd);const tag=await createClient().post(tagsPath(org),{name,color:opts.color});if(resolveOutput(cmd)==="json")printJson(tag);else console.log(`Created tag ${tag.id} (${tag.name}, ${tag.color})`)});group.command("rename <tag> <new-name>").description("Rename a tag (by id or current name)").option("--color <color>","also change the color").action(async(tag,newName,opts,cmd)=>{const org=await resolveOrg(cmd);const updated=await createClient().patch(`${tagsPath(org)}/${encodeURIComponent(tag)}`,{name:newName,color:opts.color});if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Renamed tag ${updated.id} to ${updated.name} (${updated.color})`)});group.command("rm <tag>").description("Delete a tag (by id or name)").action(async(tag,_opts,cmd)=>{const org=await resolveOrg(cmd);await createClient().delete(`${tagsPath(org)}/${encodeURIComponent(tag)}`);if(resolveOutput(cmd)!=="json")console.log(`Deleted tag ${tag}`)})}var telemetry_exports={};__export(telemetry_exports,{register:()=>register34});async function printStatus(command){const status=await telemetryStatus();if(command.optsWithGlobals().output==="json"){printJson(status);return}console.log(`CLI telemetry is ${status.enabled?"enabled":"disabled"} (${status.source}).`);if(status.source==="environment")console.log("GAUGE_TELEMETRY_DISABLED=1 overrides the saved preference.");if(status.source==="unavailable")console.log("Preferences could not be read. Collection is disabled until repaired with gauge telemetry enable or disable.")}function register34(program2){const telemetry=program2.command("telemetry").description("Control anonymous CLI usage telemetry (enabled by default)");telemetry.command("status").description("Show whether CLI telemetry is enabled").action(async(_options,command)=>printStatus(command));for(const enabled of[true,false]){telemetry.command(enabled?"enable":"disable").description(`${enabled?"Enable":"Disable"} CLI telemetry for this user`).action(async(_options,command)=>{await setTelemetryEnabled(enabled);await printStatus(command)})}}var update_exports={};__export(update_exports,{register:()=>register35});import{spawnSync as spawnSync2}from"child_process";function register35(program2,{run=spawnSync2,platform=process.platform}={}){program2.command("update").description("Update the globally installed Gauge CLI to the latest version").action((_opts,command)=>{const output=resolveOutput(command);console.error("Updating Gauge CLI to the latest published version...");const result=run(platform==="win32"?"npm.cmd":"npm",["install","--global","@withgauge/cli@latest"],{shell:platform==="win32",stdio:["inherit",output==="json"?2:"inherit","inherit"]});if(result.error||result.status!==0){const reason=result.error?result.error.message:result.signal?`npm was interrupted by ${result.signal}`:`npm exited with code ${result.status}`;throw new CliError("internal",`CLI update failed: ${reason}.
|
|
801
974
|
Retry with: npm install --global @withgauge/cli@latest`)}if(output==="json"){printJson({updated:true,package:"@withgauge/cli",tag:"latest"})}else{console.log("Updated Gauge CLI. Run `gauge --version` to see the installed version.")}})}var usage_exports={};__export(usage_exports,{register:()=>register36,usageRow:()=>usageRow});function usageRow(e){return{createdAt:e.createdAt,kind:e.kind,delta:e.delta,"started by":triggeredByLabel(e.triggeredBy),runId:e.runId??""}}function register36(program2){hostedOnly(program2.command("usage")).description("List credit ledger entries").option("--since <window>",'e.g. "30d", "24h", or an ISO date').option("--limit <n>","page size (1-200)",parseLimit,50).option("--cursor <cursor>","resume from a previous page's nextCursor").action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/usage`,{since:opts.since,limit:opts.limit,cursor:opts.cursor});if(resolveOutput(cmd)==="json"){printJson(res);return}printItems(res.items.map(usageRow),"table");if(res.items.length===0&&!opts.cursor){try{const status=await createClient().get(`/api/v1/orgs/${encodeURIComponent(org)}/billing`);if(!status.enforced){process.stderr.write("Billing is not enforced for this org, so runs don't create ledger entries.\n")}}catch{}}if(res.nextCursor){process.stderr.write(`(more results \u2014 rerun with --cursor ${res.nextCursor})
|
|
802
|
-
`)}})}var EXAMPLES=[["gauge init --local","Initialize a workspace; creates no measurements"],["gauge instructions hillclimb","Choose inputs and prepare a first pilot"],["gauge eval create --help","Save a task and success criteria"],["gauge run --eval <id> --local","Run a saved eval; rerun after input changes"]];var GROUPS=[["Get started",{instructions:"Instructions for coding agents",init:"Initialize a Gauge workspace",local:"Configure local execution","install-skill":"Install the Gauge agent skill",onboard:"Sign in and set up your workspace",status:"One-shot org summary (Gauge account)"}],["Evals (local or hosted)",{eval:"Manage saved evals",preference:"Manage saved preference prompts",run:"Start runs and read them",query:"Query local runs and findings or hosted analytics",cleanup:"Clean interrupted local sessions",models:"List models, harnesses, and providers"}],["Hosted tools (Gauge account)",{personas:"Manage simulated-user personas",skills:"Manage agent skills",mcp:"Manage MCP servers",optimizations:"Find the best change to an owned surface"}],["Insights (Gauge account)",{stats:"Analytics over your runs",dashboards:"Saved dashboards",actions:"Recommended follow-ups"}],["Gauge Chat (AI visibility)",{chat:"Prompts, topics, answers, citations, queries"}],["Organization (Gauge account)",{orgs:"List and switch organizations",org:"This org's settings",members:"Team members and invites",repos:"Manage repositories",connections:"Stored credentials (no secrets)",keys:"Manage BYOK provider keys",brands:"Manage org brands",tags:"Manage org tags",billing:"Billing and credits",usage:"Credit ledger entries"}],["Account",{auth:"Sign in, sign out, and tokens",whoami:"Show the signed-in user"}],["CLI",{config:"Manage local CLI config",telemetry:"Enable or disable anonymous CLI usage telemetry",update:"Update to the latest version",feedback:"Send feedback to the Gauge team"}]];function formatIntro(lines,helper,width,code,spaced=false){return lines.flatMap((line,i)=>{const bullet=line.startsWith("- ");const indent=bullet?" ":"";const gap=spaced&&i>0&&!bullet&&!lines[i-1].startsWith("- ")?[""]:[];const text=(bullet?line.slice(2):line).replace(/`([^`]+)`/g,(span,inner)=>{const styled=code(inner);return(styled===inner?span:styled).replaceAll(" ","\0")});return[...gap,...helper.boxWrap(text,width-indent.length).replaceAll("\0"," ").split("\n").map((row2,j)=>(j===0&&bullet?"- ":indent)+row2)]})}function padFor(rows){return Math.max(...rows.map(([term])=>term.length))+2}function formatTable(rows,indent,helper,width,color,pad=padFor(rows)){return rows.flatMap(([term,text])=>{const[first,...rest]=helper.boxWrap(text,width-indent.length-pad).split("\n");const head=`${indent}${color(term.padEnd(pad))}${first}`;if(term.length>=pad)return[`${indent}${color(term)}`,...[first,...rest].map(row2=>indent+" ".repeat(pad)+row2)];return[head,...rest.map(row2=>indent+" ".repeat(pad)+row2)]})}function firstClause(text){return text.split(/[.:;(]| — /)[0].trim()}function formatRootHelp(program2,helper){const bold=style("1");const dim=style("2");const green=style("32");const width=helper.helpWidth??80;const state=readGaugeState(program2.opts());const visible=new Map(helper.visibleCommands(program2).filter(cmd=>cmd.name()!=="help").map(cmd=>[cmd.name(),cmd]));const sections=GROUPS.map(([title,summaries])=>[title,Object.entries(summaries).filter(([name])=>visible.delete(name))]);if(visible.size)sections.push(["Other",[...visible.values()].map(cmd=>[cmd.name(),firstClause(cmd.summary()||cmd.description())])]);const options=helper.visibleOptions(program2).map(option=>[helper.optionTerm(option),firstClause(option.description)]);const table=(rows,indent,pad)=>formatTable(rows,indent,helper,width,green,pad);const commandPad=padFor(sections.flatMap(([,rows])=>rows));const lines=["",`${bold("Gauge CLI")} ${dim(`v${CLI_VERSION}`)}`,"",...formatIntro(program2.description().split(/(?<=\.) /),helper,width,green),"",...formatIntro(["Gauge strongly recommends the gauge-agents skill for coding agents: `gauge install-skill`. It makes Gauge workflow guidance discoverable in future sessions.",...header(state,void 0,true)],helper,width,green,true),"",bold(" EXAMPLES"),...table(EXAMPLES," "),"",bold(" COMMANDS")];for(const[title,rows]of sections){if(!rows.length)continue;lines.push("",bold(` ${title}`),...table(rows," ",commandPad))}lines.push("",bold(" OPTIONS"),...table(options," "),"",` Run ${green("gauge <command> --help")} for details on any command.`,"");return lines.join("\n")}function formatCommandHelp(cmd,helper){const bold=style("1");const green=style("32");const width=helper.helpWidth??80;const path=[];for(let c=cmd;c;c=c.parent)path.unshift(c.name());const name=path.join(" ");const usage=[name,cmd.usage()].join(" ");let root=cmd;while(root.parent)root=root.parent;const rows={ARGUMENTS:helper.visibleArguments(cmd).map(arg=>[helper.argumentTerm(arg),helper.argumentDescription(arg)]),COMMANDS:helper.visibleCommands(cmd).filter(sub=>sub.name()!=="help").map(sub=>[helper.subcommandTerm(sub),helper.subcommandDescription(sub)]),OPTIONS:helper.visibleOptions(cmd).map(option=>[helper.optionTerm(option),helper.optionDescription(option)]),"GLOBAL OPTIONS":helper.visibleGlobalOptions(cmd).filter(option=>root.options.includes(option)&&option.long!=="--version").map(option=>[helper.optionTerm(option),helper.optionDescription(option)])};const pad=Math.min(padFor(Object.values(rows).flat().concat([["",""]])),Math.floor(width/3));const lines=["",bold(name)+(cmd.aliases().length?` (${cmd.aliases().join(", ")})`:"")];const description=helper.commandDescription(cmd);if(description)lines.push("",...formatIntro(description.split("\n"),helper,width-1,green).map(row2=>` ${row2}`));lines.push("",bold(" USAGE"),` ${usage}`);for(const[title,items]of Object.entries(rows)){if(!items.length)continue;lines.push("",bold(` ${title}`),...formatTable(items," ",helper,width,green,pad))}if(rows.COMMANDS.length)lines.push("",` Run ${green(`${name} <command> --help`)} for details on any command.`);const topic=path[1]&&topicForCommand(path[1]);if(topic)lines.push("",` Run ${green(`gauge instructions ${topic}`)} for explicit guidance.`);lines.push("");return lines.join("\n")}function rejectUnknownCommandHelp(program2){program2.on("beforeAllHelp",({command})=>{const[name]=command.args;if(name===void 0||name.startsWith("-")||command.commands.length===0||command.registeredArguments.length>0||command.commands.some(sub=>sub.name()===name||sub.aliases().includes(name)))return;command.unknownCommand()})}import childProcess from"child_process";function registerTelemetry(program2,workerPath){program2.hook("preAction",(_program,command)=>{try{if(process.env.GAUGE_TELEMETRY_DISABLED==="1")return;const names=[];for(let current=command;current?.parent;current=current.parent)names.unshift(current.name());if(!names.length||names[0]==="telemetry")return;const event={command:names.join(" "),timestamp:new Date().toISOString(),cliVersion:CLI_VERSION};const child=childProcess.spawn(process.execPath,[workerPath,JSON.stringify(event)],{detached:true,stdio:"ignore",windowsHide:true});child.on("error",()=>{});child.unref()}catch{}})}var program=new Command;program.name("gauge").description("Hillclimb Agent Experience: run coding agents against your product, analyze friction, improve, and retest. Run locally with Docker and your agent credentials, or use hosted Gauge. Agents: start with `gauge instructions`.").configureHelp({showGlobalOptions:true,formatHelp:(cmd,helper)=>cmd===program?formatRootHelp(cmd,helper):formatCommandHelp(cmd,helper)}).version(CLI_VERSION).option("--org <org>","organization slug or id, for Gauge Chat and Gauge Agents (default: gauge.json orgId, then your saved default)").addOption(new Option5("-o, --output <format>","output format").choices([...OUTPUT_FORMATS]).default("table")).addOption(new Option5("--local","run on this machine, no account needed").conflicts("hosted")).option("--hosted","run on hosted Gauge (needs an account). Default: workspace preference, then user preference, else local");program.exitOverride();rejectUnknownCommandHelp(program);program.configureOutput({outputError:()=>{}});for(const group of[instructions_exports,localRuntime_exports,cleanup_exports,update_exports,auth_exports,init_exports,installSkill_exports,onboard_exports,status_exports,orgs_exports,org_exports,members_exports,models_exports,localProfiles_exports,run_exports,eval_exports,preference_exports,personas_exports,skills_exports,mcp_exports,connections_exports,repos_exports,optimizations_exports,brands_exports2,tags_exports,actions_exports,dashboards_exports,keys_exports2,billing_exports,usage_exports,stats_exports,query_exports,chat_exports,config_exports,telemetry_exports,feedback_exports2]){group.register(program)}if(!import.meta.url.endsWith(".ts"))registerTelemetry(program,fileURLToPath(new URL("./telemetryWorker.js",import.meta.url)));function errorContext(){try{const state=readGaugeState(program.opts());return{local:state.mode==="local"?state.source:void 0}}catch{return{}}}try{await program.parseAsync(process.argv)}catch(error){if(error instanceof CommanderError2&&(error.exitCode===0||error.code==="commander.help"||error.code==="commander.helpDisplayed"))process.exit(error.exitCode===0?0:2);const argv=process.argv.slice(2);const report=errorReport(error,commandPath(program,argv),errorContext());if(report.code==="upgrade_required"&&!/npm install/i.test(report.message))report.hint="Run: npm install -g @withgauge/cli";if(report.code==="not_found"&&/organization/i.test(report.message)&&!program.opts().org)report.hint="That is your configured default organization. Run `gauge orgs list` to see yours, then `gauge orgs use <org>` to update it.";process.stderr.write(renderError(report,wantsJson(argv)));process.exit(report.exitCode)}
|
|
975
|
+
`)}})}var watch_exports={};__export(watch_exports,{ensureWatchServer:()=>ensureWatchServer,register:()=>register37,stopWatchServer:()=>stopWatchServer});import{spawn as spawn3}from"child_process";import{existsSync as existsSync5}from"fs";import{mkdir,open,readFile,rm as rm2,writeFile}from"fs/promises";import{join as join12}from"path";import{setTimeout as delay3}from"timers/promises";import{fileURLToPath}from"url";function alive(pid){if(!Number.isSafeInteger(pid)||pid<=0)return false;try{process.kill(pid,0);return true}catch(error){return error.code==="EPERM"}}var readWatchState=gaugeDir=>readJson(join12(gaugeDir,"watch.json"));async function health(state){try{const res=await fetch(`${state.url}/api/health`,{signal:AbortSignal.timeout(5e3)});return await res.json()}catch(error){return error.name==="TimeoutError"?"busy":void 0}}async function ours(state,gaugeDir){if(!alive(state.pid))return void 0;const body=await health(state);if(body==="busy")return{};return body?.gauge==="watch"&&body.gaugeDir===gaugeDir?body:void 0}async function withWatchLock(gaugeDir,fn){await mkdir(gaugeDir,{recursive:true});const path=join12(gaugeDir,"watch.lock");const deadline=Date.now()+3e4;while(true){try{await writeFile(path,String(process.pid),{flag:"wx",mode:384});break}catch(error){if(error.code!=="EEXIST")throw error;const holder=Number(await readFile(path,"utf8").catch(()=>""));if(holder&&!alive(holder))await rm2(path,{force:true});else if(Date.now()>=deadline)throw new CliError("internal",`Timed out waiting for ${path}. Remove it if no gauge watch is running.`);else await delay3(50)}}try{return await fn()}finally{await rm2(path,{force:true})}}async function stopLocked(gaugeDir){const state=await readWatchState(gaugeDir);if(!state)return false;const running=Boolean(await ours(state,gaugeDir));if(running){process.kill(state.pid,"SIGTERM");const deadline=Date.now()+5e3;while(alive(state.pid)&&Date.now()<deadline)await delay3(50);if(alive(state.pid))throw new CliError("internal",`The watch server (pid ${state.pid}) did not stop.`)}await rm2(join12(gaugeDir,"watch.json"),{force:true});return running}function stopWatchServer(gaugeDir){return withWatchLock(gaugeDir,()=>stopLocked(gaugeDir))}function ensureWatchServer(gaugeDir){return withWatchLock(gaugeDir,()=>startLocked(gaugeDir))}async function startLocked(gaugeDir){const existing=await readWatchState(gaugeDir);if(existing&&(await ours(existing,gaugeDir))?.version===CLI_VERSION)return{...existing,reused:true};await stopLocked(gaugeDir);const bundled=fileURLToPath(new URL("./watchServer.js",import.meta.url));const args=existsSync5(bundled)?[bundled]:[...process.execArgv,fileURLToPath(new URL("../watchServer.ts",import.meta.url))];const logPath=join12(gaugeDir,"watch.log");const log=await open(logPath,"a",384);try{const child=spawn3(process.execPath,[...args,gaugeDir,CLI_VERSION],{detached:true,stdio:["ignore",log.fd,log.fd]});await new Promise((resolve5,reject)=>{child.once("spawn",resolve5);child.once("error",reject)});let exited=false;child.once("exit",()=>{exited=true});const deadline=Date.now()+1e4;while(true){const state=await readWatchState(gaugeDir);if(state&&state.pid===child.pid){child.unref();return{...state,reused:false}}if(exited||Date.now()>=deadline){child.kill();throw new CliError("internal",`The watch server did not start. See ${logPath}`)}await delay3(50)}}finally{await log.close()}}async function hostedUrl(command){const selection=selectedOrg(command);if(!selection)throw new CliError("usage","No organization selected. Pass --org <org> or set a default with `gauge orgs use <org>`.",{needsAccount:true});const{slug}=await findOrganization(selection.value);const base=resolveBaseUrl().replace(/\/+$/,"");return`${base}/${encodeURIComponent(slug)}/dashboard?liveruns=true`}function register37(program2){const watch=program2.command("watch").description("Print a URL for a live, read-only view of past and running sessions (local: starts a background server)").action(async(_opts,command)=>{const json=resolveOutput(command)==="json";if(gaugeState(command).mode==="hosted"){const url=await hostedUrl(command);if(json)printJson({mode:"hosted",url});else console.log(url);return}const state=await ensureWatchServer(localDirectory2(command));if(json)printJson({mode:"local",...state});else{console.log(state.url);console.error("Stop: gauge watch stop")}});watch.command("stop").description("Stop this workspace's background watch server").action(async(_opts,command)=>{const stopped=await stopWatchServer(localDirectory2(command));if(resolveOutput(command)==="json")printJson({stopped});else console.log(stopped?"Watch server stopped.":"No watch server is running.")})}var EXAMPLES=[["gauge init --local","Initialize a workspace; creates no measurements"],['gauge run "<prompt>"',"Run one prompt now; saves no measurement"],["gauge instructions hillclimb","Choose inputs and prepare a first pilot"],["gauge eval create --help","Save a task and success criteria"],["gauge run --eval <id> --local","Run a saved eval; rerun after input changes"]];var GROUPS=[["Get started",{instructions:"Instructions for coding agents",init:"Initialize a Gauge workspace",local:"Configure local execution","install-skill":"Install the Gauge agent skill",onboard:"Sign in and set up your workspace",status:"One-shot org summary (Gauge account)"}],["Evals (local or hosted)",{eval:"Manage saved evals",preference:"Manage saved preference prompts",run:"Start runs and read them",watch:"Live view of past and running sessions",query:"Query local runs and findings or hosted analytics",cleanup:"Clean interrupted local sessions",models:"List models, harnesses, and providers"}],["Hosted tools (Gauge account)",{personas:"Manage simulated-user personas",skills:"Manage agent skills",mcp:"Manage MCP servers",optimizations:"Edit what agents read and rerun their sessions"}],["Insights (Gauge account)",{stats:"Analytics over your runs",dashboards:"Saved dashboards",issues:"The same finding across many sessions (also called findings)"}],["Gauge Chat (AI visibility)",{chat:"Prompts, topics, answers, citations, queries"}],["Organization (Gauge account)",{orgs:"List and switch organizations",org:"This org's settings",members:"Team members and invites",repos:"Manage repositories",connections:"Stored credentials (no secrets)",keys:"Manage BYOK provider keys",brands:"Manage org brands",tags:"Manage org tags",billing:"Billing and credits",usage:"Credit ledger entries"}],["Account",{auth:"Sign in, sign out, and tokens",whoami:"Show the signed-in user"}],["CLI",{config:"Manage local CLI config",telemetry:"Enable or disable anonymous CLI usage telemetry",update:"Update to the latest version",feedback:"Send feedback to the Gauge team"}]];function formatIntro(lines,helper,width,code,spaced=false){return lines.flatMap((line,i)=>{const bullet=line.startsWith("- ");const indent=bullet?" ":"";const gap=spaced&&i>0&&!bullet&&!lines[i-1].startsWith("- ")?[""]:[];const text=(bullet?line.slice(2):line).replace(/`([^`]+)`/g,(span,inner)=>{const styled=code(inner);return(styled===inner?span:styled).replaceAll(" ","\0")});return[...gap,...helper.boxWrap(text,width-indent.length).replaceAll("\0"," ").split("\n").map((row2,j)=>(j===0&&bullet?"- ":indent)+row2)]})}function padFor(rows){return Math.max(...rows.map(([term])=>term.length))+2}function formatTable(rows,indent,helper,width,color,pad=padFor(rows)){return rows.flatMap(([term,text])=>{const[first,...rest]=helper.boxWrap(text,width-indent.length-pad).split("\n");const head=`${indent}${color(term.padEnd(pad))}${first}`;if(term.length>=pad)return[`${indent}${color(term)}`,...[first,...rest].map(row2=>indent+" ".repeat(pad)+row2)];return[head,...rest.map(row2=>indent+" ".repeat(pad)+row2)]})}function firstClause(text){return text.split(/[.:;(]| — /)[0].trim()}function formatRootHelp(program2,helper){const bold=style("1");const dim=style("2");const green=style("32");const width=helper.helpWidth??80;const state=readGaugeState(program2.opts());const visible=new Map(helper.visibleCommands(program2).filter(cmd=>cmd.name()!=="help").map(cmd=>[cmd.name(),cmd]));const sections=GROUPS.map(([title,summaries])=>[title,Object.entries(summaries).filter(([name])=>visible.delete(name))]);if(visible.size)sections.push(["Other",[...visible.values()].map(cmd=>[cmd.name(),firstClause(cmd.summary()||cmd.description())])]);const options=helper.visibleOptions(program2).map(option=>[helper.optionTerm(option),firstClause(option.description)]);const table=(rows,indent,pad)=>formatTable(rows,indent,helper,width,green,pad);const commandPad=padFor(sections.flatMap(([,rows])=>rows));const lines=["",`${bold("Gauge CLI")} ${dim(`v${CLI_VERSION}`)}`,"",...formatIntro(program2.description().split(/(?<=\.) /),helper,width,green),"",...formatIntro(["Gauge strongly recommends the gauge-agents skill for coding agents: `gauge install-skill`. It makes Gauge workflow guidance discoverable in future sessions.",...header(state,void 0,true)],helper,width,green,true),"",bold(" EXAMPLES"),...table(EXAMPLES," "),"",bold(" COMMANDS")];for(const[title,rows]of sections){if(!rows.length)continue;lines.push("",bold(` ${title}`),...table(rows," ",commandPad))}lines.push("",bold(" OPTIONS"),...table(options," "),"",` Run ${green("gauge <command> --help")} for details on any command.`,"");return lines.join("\n")}function formatCommandHelp(cmd,helper){const bold=style("1");const green=style("32");const width=helper.helpWidth??80;const path=[];for(let c=cmd;c;c=c.parent)path.unshift(c.name());const name=path.join(" ");const usage=[name,cmd.usage()].join(" ");let root=cmd;while(root.parent)root=root.parent;const rows={ARGUMENTS:helper.visibleArguments(cmd).map(arg=>[helper.argumentTerm(arg),helper.argumentDescription(arg)]),COMMANDS:helper.visibleCommands(cmd).filter(sub=>sub.name()!=="help").map(sub=>[helper.subcommandTerm(sub),helper.subcommandDescription(sub)]),OPTIONS:helper.visibleOptions(cmd).map(option=>[helper.optionTerm(option),helper.optionDescription(option)]),"GLOBAL OPTIONS":helper.visibleGlobalOptions(cmd).filter(option=>root.options.includes(option)&&option.long!=="--version").map(option=>[helper.optionTerm(option),helper.optionDescription(option)])};const pad=Math.min(padFor(Object.values(rows).flat().concat([["",""]])),Math.floor(width/3));const lines=["",bold(name)+(cmd.aliases().length?` (${cmd.aliases().join(", ")})`:"")];const description=helper.commandDescription(cmd);if(description)lines.push("",...formatIntro(description.split("\n"),helper,width-1,green).map(row2=>` ${row2}`));lines.push("",bold(" USAGE"),` ${usage}`);for(const[title,items]of Object.entries(rows)){if(!items.length)continue;lines.push("",bold(` ${title}`),...formatTable(items," ",helper,width,green,pad))}if(rows.COMMANDS.length)lines.push("",` Run ${green(`${name} <command> --help`)} for details on any command.`);const topic=path[1]&&topicForCommand(path[1]);if(topic)lines.push("",` Run ${green(`gauge instructions ${topic}`)} for explicit guidance.`);lines.push("");return lines.join("\n")}function rejectUnknownCommandHelp(program2){program2.on("beforeAllHelp",({command})=>{const[name]=command.args;if(name===void 0||name.startsWith("-")||command.commands.length===0||command.registeredArguments.length>0||command.commands.some(sub=>sub.name()===name||sub.aliases().includes(name)))return;command.unknownCommand()})}import childProcess from"child_process";function registerTelemetry(program2,workerPath){program2.hook("preAction",(_program,command)=>{try{if(process.env.GAUGE_TELEMETRY_DISABLED==="1")return;const names=[];for(let current=command;current?.parent;current=current.parent)names.unshift(current.name());if(!names.length||names[0]==="telemetry")return;const event={command:names.join(" "),timestamp:new Date().toISOString(),cliVersion:CLI_VERSION};const child=childProcess.spawn(process.execPath,[workerPath,JSON.stringify(event)],{detached:true,stdio:"ignore",windowsHide:true});child.on("error",()=>{});child.unref()}catch{}})}var program=new Command;program.name("gauge").description("Hillclimb Agent Experience: run coding agents against your product, analyze friction, improve, and retest. Run locally with Docker and your agent credentials, or use hosted Gauge. Agents: start with `gauge instructions`.").configureHelp({showGlobalOptions:true,formatHelp:(cmd,helper)=>cmd===program?formatRootHelp(cmd,helper):formatCommandHelp(cmd,helper)}).version(CLI_VERSION).option("--org <org>","organization slug or id, for Gauge Chat and Gauge Agents (default: gauge.json orgId, then your saved default)").addOption(new Option6("-o, --output <format>","output format").choices([...OUTPUT_FORMATS]).default("table")).addOption(new Option6("--local","run on this machine, no account needed").conflicts("hosted")).option("--hosted","run on hosted Gauge (needs an account). Default: workspace preference, then user preference, else local");program.exitOverride();rejectUnknownCommandHelp(program);program.configureOutput({outputError:()=>{}});for(const group of[instructions_exports,localRuntime_exports,cleanup_exports,update_exports,auth_exports,init_exports,installSkill_exports,onboard_exports,status_exports,orgs_exports,org_exports,members_exports,models_exports,localProfiles_exports,run_exports,watch_exports,eval_exports,preference_exports,personas_exports,skills_exports,mcp_exports,connections_exports,repos_exports,optimizations_exports,brands_exports2,tags_exports,issues_exports,dashboards_exports,keys_exports2,billing_exports,usage_exports,stats_exports,query_exports,chat_exports,config_exports,telemetry_exports,feedback_exports2]){group.register(program)}if(!import.meta.url.endsWith(".ts"))registerTelemetry(program,fileURLToPath2(new URL("./telemetryWorker.js",import.meta.url)));function errorContext(){try{const state=readGaugeState(program.opts());return{local:state.mode==="local"?state.source:void 0}}catch{return{}}}try{await program.parseAsync(process.argv)}catch(error){if(error instanceof CommanderError2&&(error.exitCode===0||error.code==="commander.help"||error.code==="commander.helpDisplayed"))process.exit(error.exitCode===0?0:2);const argv=process.argv.slice(2);const report=errorReport(error,commandPath(program,argv),errorContext());if(report.code==="upgrade_required"&&!/npm install/i.test(report.message))report.hint="Run: npm install -g @withgauge/cli";if(report.code==="not_found"&&/organization/i.test(report.message)&&!program.opts().org)report.hint="That is your configured default organization. Run `gauge orgs list` to see yours, then `gauge orgs use <org>` to update it.";process.stderr.write(renderError(report,wantsJson(argv)));process.exit(report.exitCode)}
|