@withgauge/cli 0.25.0 → 0.25.1
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 +35 -3
- package/dist/batchWorker.js +1 -1
- package/dist/{chatApi-KHPW7ED5.js → chatApi-W6HW3HKD.js} +1 -1
- package/dist/chunk-E24F53BZ.js +1 -0
- package/dist/{chunk-XHPA33OA.js → chunk-RRLAI72I.js} +1 -1
- package/dist/{chunk-EVT234WT.js → chunk-UANK5LHE.js} +1 -1
- package/dist/{chunk-KRHTXL5O.js → chunk-WJTCHKG6.js} +1 -1
- package/dist/index.js +9 -7
- package/dist/{orgDirectory-V3RCKGLA.js → orgDirectory-BZW2ZAD3.js} +1 -1
- package/package.json +1 -1
- package/dist/chunk-DEX43SUK.js +0 -1
package/dist/index.js
CHANGED
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
#!/usr/bin/env node
|
|
2
|
-
import{LOCAL_AGENTS,LOCAL_MCP_HINT,LOCAL_TARGETS,analysisNext,assertLocalCapabilities,cancelLocal,cleanupLocal,commentSourceSchema,deploymentSourceSchema,detachLocal,freezeWebFixtures,inspectLocalCredential,inspectLocalGateway,inspectLocalRuntime,localBatchStatus,localStatus,mergeWebFixtures,openLocalStore,parseWebFixtureInput,planLocal,pullLocalGateway,pullLocalRuntime,readProfiles,readWebFixtureArchive,requireProfile,runLocal,setProfile,websiteSourceSchema,websiteUseSchema,writeProfiles}from"./chunk-
|
|
2
|
+
import{LOCAL_AGENTS,LOCAL_MCP_HINT,LOCAL_TARGETS,analysisNext,assertLocalCapabilities,cancelLocal,cleanupLocal,commentSourceSchema,deploymentSourceSchema,detachLocal,freezeWebFixtures,inspectLocalCredential,inspectLocalGateway,inspectLocalRuntime,localBatchStatus,localStatus,mergeWebFixtures,openLocalStore,parseWebFixtureInput,planLocal,pullLocalGateway,pullLocalRuntime,readProfiles,readWebFixtureArchive,requireProfile,runLocal,setProfile,websiteSourceSchema,websiteUseSchema,writeProfiles}from"./chunk-UANK5LHE.js";import{setTelemetryEnabled,telemetryStatus}from"./chunk-WYYRMPRX.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,style,truncate,withCategory}from"./chunk-WJTCHKG6.js";import{createdOrganization,credentialKind,fetchDirectory,findOrganization,isUuid,legacySelection,productsLabel,resolveOrgValue}from"./chunk-E24F53BZ.js";import{ANALYSIS_SCHEMA_VERSION,ApiClient,CliError,agentConfigInputSchema,agentConfigSchema,agentSchema,analysisSubmissionSchema,assetRefListSchema,batteries_exports,brands_exports,clearToken,createClient,createdBySchema,credentialsPath,errorEnvelopeSchema,exitCodeFor,feedback_exports,keys_exports,listResponseSchema,readCredentials,resolveAuthUrl,resolveBaseUrl,resolveToken,runConfig_exports,runCreditsChargedSchema,runEventSchema,runTriggeredBySchema,saveCredentials,saveToken,sharedTokenSchema,storedSharedCredentials,withCredentialLock}from"./chunk-RRLAI72I.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 Option5}from"commander";import{fileURLToPath}from"url";var actions_exports={};__export(actions_exports,{register:()=>register});function actionsPath(org){return`/api/v1/orgs/${encodeURIComponent(org)}/actions`}var STATUS_CHOICES=["open","resolved","dismissed"];function parseStatus(raw){const upper=raw.trim().toUpperCase();if(!["OPEN","RESOLVED","DISMISSED"].includes(upper))throw new CliError("usage",`invalid status "${raw}" (expected: ${STATUS_CHOICES.join(" | ")})`);return upper}function register(program2){const group=hostedOnly(program2.command("actions")).description("The action center: recommended follow-ups");group.command("list").description("List actions, highest-leverage first").option("--status <status>","filter: open | resolved | dismissed (csv)").action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const res=await createClient().get(actionsPath(org),{status:opts.status===void 0?void 0:opts.status.split(",").map(s=>parseStatus(s)).join(",")});if(resolveOutput(cmd)==="json")printJson(res.items);else printItems(res.items.map(a=>({id:a.id,priority:a.priority,status:a.status,title:a.title,evidence:a.evidenceCount,surface:a.sourceType})),"table")});group.command("status <id> <status>").description("Transition an action: open | resolved | dismissed").option("--note <text>","resolution note (recorded when closing)").action(async(id,status,opts,cmd)=>{const org=await resolveOrg(cmd);const updated=await createClient().patch(`${actionsPath(org)}/${encodeURIComponent(id)}`,{status:parseStatus(status),note:opts.note});if(resolveOutput(cmd)==="json")printJson(updated);else console.log(`Action ${updated.id} is now ${updated.status}`)})}var auth_exports={};__export(auth_exports,{browserLogin:()=>browserLogin,deviceLogin:()=>deviceLogin,logout:()=>logout,parseLoopbackCallback:()=>parseLoopbackCallback,readStdin:()=>readStdin,register:()=>register2,useOnlyOrg:()=>useOnlyOrg});import{spawn}from"child_process";import{createHash,randomBytes,randomUUID}from"crypto";import{createServer}from"http";import{hostname}from"os";import{createInterface}from"readline";import{setTimeout as sleep}from"timers/promises";import{z}from"zod";var deviceAuthorizeRequestSchema=z.object({deviceName:z.string().trim().max(64).optional(),email:z.string().trim().email().max(320).optional()});var deviceAuthorizeResponseSchema=z.object({deviceCode:z.string(),userCode:z.string(),verificationUri:z.string(),verificationUriComplete:z.string(),interval:z.number().int().positive(),expiresIn:z.number().int().positive(),emailSent:z.boolean()});var deviceTokenRequestSchema=z.object({deviceCode:z.string().min(1)});var deviceTokenApprovedSchema=z.object({status:z.literal("approved"),accessToken:z.string(),tokenType:z.literal("bearer"),expiresAt:z.string().nullable()});var deviceTokenResultSchema=z.discriminatedUnion("status",[z.object({status:z.literal("pending")}),z.object({status:z.literal("slow_down"),interval:z.number().int()}),z.object({status:z.literal("denied")}),z.object({status:z.literal("expired")}),deviceTokenApprovedSchema]);async function credentialIdentity(){if(credentialKind()==="legacy"){const{user}=await createClient().get("/api/v1/me");return{user,apiKey:null,token:{kind:"legacy",scopes:null,expiresAt:null}}}return createChatApiClient().me()}async function principalIdentity(){const{user,apiKey,token}=await credentialIdentity();const orgs=token.kind==="legacy"?null:apiKey?.organization?[apiKey.organization]:await fetchDirectory();return{user,apiKey,orgs}}function printPrincipal(identity){const{user,apiKey,orgs}=identity;if(apiKey)console.log(`Org API key "${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}
|
|
@@ -212,7 +212,7 @@ description: Write Gauge eval cases, choose models and judging criteria, and
|
|
|
212
212
|
- Automatic GitHub PR checks support root \`gauge.json\` \`checks.paths\` as a repo-wide gate and per-case frontmatter \`config.checks.paths\` as a narrower filter. Both accept positive repository-relative globs. Cases without filters run when the repo-wide gate passes; changes to \`gauge.json\` select all cases and edits to a case select that case. Explicit CLI launches and GitHub reruns ignore both filters. Failed/unfinished previous checks, changed target branches, and unavailable diffs run conservatively. \`config.inputs\` selects inputs, not triggers.
|
|
213
213
|
- Pilot before scaling; confirm before spending credits. Creating definitions does not authorize launching runs. Report sample counts and distinguish incorrect behavior, missing evidence, environment failures, and budget exhaustion.
|
|
214
214
|
- Gauge tests artifacts from the PR's successful build; preserve CI's existing checkout settings, including the default merge checkout. Fork PRs are disabled by default. In Gauge's GitHub settings, check **Allow fork pull requests**; a workspace member then follows the GitHub check link to authorize that revision using workspace credits and connections. New revisions and explicit reruns need fresh approval. Inputs still determine readiness.
|
|
215
|
-
- For authenticated MCP access,
|
|
215
|
+
- For authenticated MCP access, use \`gauge mcp add --hosted --name <handle> --url <https-url> --oauth\`. Complete consent in your browser; \`--no-browser\` prints the URL for remote terminals. To reuse a saved connection use \`--connection <id>\` instead (\`gauge connections list --hosted\`). Put the resulting versioned reference in case \`config.mcpRefs\`. Gauge stores and refreshes session credentials; keep tokens out of prompts and Git. See \`gauge mcp add --help\` for manual OAuth client registration options.
|
|
216
216
|
- To test local build output without the GitHub App, use \`gauge run --input name=path --yes -o json\`. Repeat \`--input\` for each input. This uses current evaluation files, including edits; other inputs keep their configured sources. The path replaces the input's configured files entirely and may be a glob (quote it): skills accept a directory or archive, npm packages a .tgz or a glob matching one, files a directory, file, or glob, and binaries a directory or glob with \`commands\` relative to it. Use \`gauge run wait <id> --check -o json\` for the verdict, criterion results, and session links; with \`--check\`, failed checks exit 1 and stdout is only the JSON.
|
|
217
217
|
- Before committing setup changes, run \`gauge run --verify -o json\` for a local working-tree checklist. Only attempted static checks are scored yes, no, or unknown; unknown means a specific declaration could not be resolved. Informational findings (such as zero cases) are separate. JSON includes a version, static status (passed/failed/incomplete), summary counts, global/per-input checks with stable IDs, facts, info, and excluded capabilities in notChecked. Exit 1 means a definite problem; exit 0 can still contain unknowns. Live delivery, account settings, and sandbox execution are outside this command's scope. It requires no login or PR and launches no sessions.
|
|
218
218
|
|
|
@@ -773,11 +773,13 @@ ${END}`;function scaffoldWorkspace(root,backend,cases=false){mkdirSync(root,{rec
|
|
|
773
773
|
`;var SMALL_LOGO=String.raw`█▀▀ ▄▀█ █ █ █▀▀ █▀▀
|
|
774
774
|
█▄█ █▀█ █▄█ █▄█ ██▄`;async function printLogo(){const out=process.stdout;if(detectAgent(process.env,out.isTTY)==="agent"){console.log(SMALL_LOGO);return}if(process.env.CI||process.env.TERM==="dumb"){console.log(LOGO.replace(/^\n+/,""));return}out.write("\x1B[?25l");try{out.write(LOGO);const up=`\x1B[${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}
|
|
775
775
|
${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
|
-
${result.gateway.problem.hint}`)}if(!result.ready||!result.gateway.ready)process.exitCode=1}function register17(program2){const local=program2.commands.find(command=>command.name()==="local")??program2.command("local").description("Configure local execution (no Gauge account needed)");const runtime=local.command("runtime").description("Inspect or download the supported Docker runtime");runtime.command("inspect").description("Report local runtime availability and compatibility without downloading").action((_options,command)=>print({...inspectLocalRuntime(),gateway:inspectLocalGateway()},command));runtime.command("pull").description("Download the CLI's supported public runtime image").action(async(_options,command)=>print({...await pullLocalRuntime(),gateway:await pullLocalGateway()},command))}var mcp_exports={};__export(mcp_exports,{register:()=>register18});import{readFileSync as readFileSync7}from"fs";function mcpPath(org){return`/api/v1/orgs/${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}`);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").action(async(opts,cmd)=>{const org=await resolveOrg(cmd);const result=await createClient().post(mcpPath(org),{name:opts.name,spec:specFromOpts(opts),label:opts.label,note:opts.note});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(`
|
|
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:
|
|
777
|
+
${url}
|
|
778
|
+
Waiting for consent\u2026`);if(options.browser)deps.openBrowser(url.toString());const deadline=Math.min(expiresAt,deps.now()+options.timeoutSeconds*1e3);while(deps.now()<deadline){const signal=AbortSignal.timeout(Math.max(1,deadline-deps.now()));let status;try{status=await client.get(`${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(`
|
|
777
779
|
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(`
|
|
778
|
-
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 sleep2(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 ${trials2.filter(t=>t.status==="RUNNING").map(t=>t.key).join(",")||"idle"}`;if(line!==last){last=line;console.error(line)}const needsInput=awaitingInput(detail);if(watchStops(detail,!!opts.follow)){if(resolveOutput(cmd)==="json")printJson(detail);else printDetail2(detail);if(needsInput)process.exitCode=NEEDS_INPUT_EXIT;return}await sleep2(intervalMs)}});group.command("authoring <id>").description("What the next round may change: exposed surfaces, CLI targets, baseline run evidence, revision").action(async(id,_opts,cmd)=>{printJson(await createClient().get(`${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)}
|
|
779
|
-
`)}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 z12}from"zod";import{join as join12,relative as relative6}from"path";import{z as z5}from"zod";var preferenceDefinitionSchema=z5.discriminatedUnion("kind",[z5.object({kind:z5.literal("open-ended"),branded:z5.boolean().default(false)}).strict(),z5.object({kind:z5.literal("head-to-head"),brandA:z5.string().trim().min(1),brandB:z5.string().trim().min(1)}).strict().refine(p=>p.brandA.toLowerCase()!==p.brandB.toLowerCase(),"A head-to-head needs two different brands")]);var preferenceSnapshotSchema=z5.object({version:z5.literal(1),preference:preferenceDefinitionSchema}).strict();var headToHeadSchema=z5.object({kind:z5.literal("HEAD_TO_HEAD"),assessability:z5.enum(["assessed","partial","unassessable"]).optional(),selectionStatus:z5.enum(["measured","not_measured","unassessable"]).optional(),winner:z5.enum(["A","B","both","neither"]),brandA:z5.string(),brandB:z5.string(),rationale:z5.string(),brandAReasoning:z5.string(),brandBReasoning:z5.string(),brandAWhyRaised:z5.string().optional(),brandBWhyRaised:z5.string().optional()});var recommendPickSchema=z5.object({brand:z5.string(),rank:z5.number().int(),adopted:z5.boolean(),reasoning:z5.string().optional(),whyRaised:z5.string().optional()});var recommendSchema=z5.object({kind:z5.literal("RECOMMEND"),selectedBrands:z5.array(z5.string()).optional(),assessability:z5.enum(["assessed","partial","unassessable"]).optional(),selectionStatus:z5.enum(["measured","not_measured","unassessable"]).optional(),picks:z5.array(recommendPickSchema).default([]),singlePick:z5.boolean(),complianceNote:z5.string()});var visibilityVerdictSchema=z5.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 ${verdict.brandB}`:"Neither";if(verdict.selectedBrands!==void 0)return verdict.selectedBrands.length?verdict.selectedBrands.join(", "):"No selection";if(!verdict.singlePick)return verdict.picks.length?"No clear selection":"No selection";return verdict.picks.find(pick=>pick.rank===1)?.brand??"No selection"}import{Option as Option4}from"commander";import{readFileSync as readFileSync9}from"fs";import{z as z6}from"zod";function assertRequiredExperiment(spec,required){if(required.length===0){throw new CliError("invalid_request","--require-experiment requires a nonempty label")}const mismatches=[];for(const[promptIndex,prompt]of spec.prompts.entries()){const promptExperiment=prompt.experiment??spec.experiment;const schedules=prompt.schedules??[{}];if(schedules.length===0){mismatches.push(`prompts.${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:
|
|
780
|
-
${details}`,{hint:"Run `gauge run --schema --hosted` for the full schema, or `gauge run --example --hosted` for a starter."})}return result.data}async function resolveApplyOrg(command,spec){return resolveOrg(command,spec.org)}function repoLabel2(repoUrl,repoRef){if(!repoUrl)return"greenfield";return repoRef?`${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=()=>z6.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/${encodeURIComponent(org)}/batteries`,spec,opts.dryRun?{dryRun:1}:void 0);if(resolveOutput(command)==="json")printJson(result);else if(opts.dryRun)printPlan(result);else printApplied(result)}import{createHash as createHash3}from"crypto";import{z as z10}from"zod";import{z as z7}from"zod";var runAddonsSchema=z7.object({browser:z7.object({session:z7.string().min(1),projectId:z7.number().int().positive(),region:z7.literal("us").default("us")}).strict().optional()}).strict();import{z as z9}from"zod";import{z as z8}from"zod";var inputPathSchema=z8.string().min(1).max(256).regex(/^(?![!/])(?!.*[\\\x00-\x1f\x7f])(?!.*(?:^|\/)\.{1,2}(?:\/|$))(?!.*\/\/)(?!.*\/$).+$/,"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=z8.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=z8.object({type:z8.literal("binary"),paths:z8.array(inputPathSchema).min(1).max(32),platform:z8.literal("linux-x64").describe("Linux x86-64 ELF executables compatible with the consumer sandbox. No system packages are installed."),commands:z8.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=z8.discriminatedUnion("type",[binaryUseSchema,websiteUseSchema,z8.object({type:z8.literal("web-fixture"),path:inputPathSchema.regex(/^[^*?{}[\]]+$/)}).strict(),z8.object({type:z8.literal("files"),paths:z8.array(inputPathSchema).min(1).max(32).optional()}).strict(),z8.object({type:z8.literal("npm-cli"),package:inputPathSchema,install:z8.enum(["environment","workspace"]).optional()}).strict(),z8.object({type:z8.literal("npm-package"),package:inputPathSchema,install:z8.enum(["environment","workspace"]).optional()}).strict(),z8.object({type:z8.literal("skill"),root:z8.union([z8.literal("."),inputPathSchema]).refine(path=>!/[*?{}[\]]/.test(path),"Skill root must be an explicit directory")}).strict()]);var inputSourceSchema=z8.discriminatedUnion("type",[z8.object({type:z8.literal("git"),paths:z8.array(inputPathSchema).min(1).max(32)}).strict(),z8.object({type:z8.literal("github-actions"),artifact:z8.string().min(1).max(256).optional()}).strict()]);var runPlanResourceSchema=z8.object({name:z8.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64),type:z8.literal("input"),source:z8.discriminatedUnion("type",[z8.object({type:z8.literal("upload"),sha256:z8.string().regex(/^sha256:[a-f0-9]{64}$/),bytes:z8.number().int().positive().max(512*1024*1024),repoUrl:z8.string().nullable(),commit:z8.string().regex(/^[a-f0-9]{40}$/i).nullable()}).strict(),commentSourceSchema.extend({repoUrl:z8.string(),commit:z8.string().regex(/^[a-f0-9]{40}$/i)}).strict(),deploymentSourceSchema.extend({repoUrl:z8.string(),commit:z8.string().regex(/^[a-f0-9]{40}$/i)}).strict(),inputSourceSchema.options[0].extend({repoUrl:z8.string(),commit:z8.string().regex(/^[a-f0-9]{40}$/i)}).strict(),inputSourceSchema.options[1].extend({repoUrl:z8.string(),commit:z8.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=z9.discriminatedUnion("type",[z9.object({type:z9.literal("git")}).strict(),z9.object({type:z9.literal("github-actions"),workflow:inputPathSchema.regex(/^\.github\/workflows\/[^/*?{}[\]]+\.ya?ml$/).describe("Workflow file in this repository, for example .github/workflows/build.yml."),artifact:z9.string().min(1).max(256).optional().describe("Upload name or pattern. Omit only when the workflow produces exactly one artifact.")}).strict()]);var skillInputPathSchema=z9.union([z9.literal("."),inputPathSchema.regex(/^[^*?{}[\]]+$/)]).describe("Explicit Skill directory or ZIP/tar.gz file containing SKILL.md at its root.");var runDefinitionInputSchema=z9.discriminatedUnion("type",[binaryUseSchema.extend({source:definitionInputSourceSchema.default({type:"git"})}).strict(),websiteUseSchema.extend({source:websiteSourceSchema}).strict(),z9.object({type:z9.literal("web-fixture"),source:definitionInputSourceSchema.default({type:"git"}),path:inputPathSchema.regex(/^[^*?{}[\]]+$/).describe("Explicit path to a version 1 web fixture JSON file.")}).strict(),z9.object({type:z9.literal("npm-package"),source:definitionInputSourceSchema.default({type:"git"}),path:inputPathSchema.describe("Path or glob selecting exactly one packed npm .tgz file."),install:z9.enum(["environment","workspace"]).default("environment").describe("environment exposes the package's commands on PATH; workspace adds a consumer application dependency.")}).strict(),z9.object({type:z9.literal("skill"),source:definitionInputSourceSchema.default({type:"git"}),path:skillInputPathSchema}).strict(),z9.object({type:z9.literal("files"),source:definitionInputSourceSchema.default({type:"git"}),paths:z9.array(inputPathSchema).min(1).max(32).describe("Files or globs selected from the committed Git tree or the uploaded artifact.")}).strict()]);function isSkillArchive(path){return/\.(?:zip|tar\.gz|tgz)$/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==="."?"**":`${input.path}/**`]:[input.path];return{name,type:"input",source:input.source.type==="git"?{type:"git",paths,repoUrl,commit}:{...input.source,repoUrl,commit},use:input.type==="binary"?{type:"binary",paths:input.paths,platform:input.platform,commands:input.commands}:input.type==="npm-package"?{type:"npm-package",package:input.path,install:input.install}:input.type==="web-fixture"?{type:"web-fixture",path:input.path}:input.type==="skill"?{type:"skill",root:input.path}:{type:"files",paths:input.paths}}}var inputUploadSpecSchema=z10.object({sha256:z10.string().regex(/^sha256:[a-f0-9]{64}$/),bytes:z10.number().int().positive().max(512*1024*1024)}).strict();var uncredentialedRepoUrlSchema=z10.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=z10.object({paths:z10.array(inputPathSchema).min(1).max(32)}).strict();var runDefinitionConfigSchema=z10.object({agents:z10.array(agentConfigInputSchema).min(1),sampleCount:z10.number().int().min(1).max(50).optional(),repoUrl:uncredentialedRepoUrlSchema.nullable().optional(),repoRef:z10.string().min(1).nullable().optional(),profileId:z10.string().min(1).nullable().optional(),skillRefs:assetRefListSchema.optional(),mcpRefs:assetRefListSchema.optional(),connectionIds:z10.array(z10.string().min(1)).max(32).optional(),addons:runAddonsSchema.optional(),inputs:z10.array(z10.string()).optional(),checks:runDefinitionChecksSchema.optional()}).strict();var runDefinitionFileConfigSchema=z10.object({version:z10.literal(2).default(2),org:z10.string().min(1).optional(),orgId:z10.string().uuid().optional(),checks:runDefinitionChecksSchema.optional(),cases:z10.array(inputPathSchema).min(1).max(32).default(["gauge-cases/**/*.md"]),inputs:z10.record(z10.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64),runDefinitionInputSchema).refine(inputs=>Object.keys(inputs).length<=32,"At most 32 inputs are allowed").meta({maxProperties:32}).default({})}).strict();var runDefinitionCriterionSchema=z10.object({name:z10.string().min(1),rubric:z10.string().min(1)}).strict();var caseMeasurement={criteria:z10.array(runDefinitionCriterionSchema).min(1).optional(),preference:preferenceDefinitionSchema.optional()};function exactlyOneMeasurement(item){return item.criteria!==void 0!==(item.preference!==void 0)}var runDefinitionFrontmatterSchema=z10.object({...caseMeasurement,config:runDefinitionConfigSchema}).strict().refine(exactlyOneMeasurement,"Specify exactly one of criteria or preference");var runDefinitionCaseSchema=z10.object({path:z10.string().min(1),prompt:z10.string().min(1),...caseMeasurement,config:runDefinitionConfigSchema}).strict().refine(exactlyOneMeasurement,"Specify exactly one of criteria or preference");var runDefinitionSubmissionSchema=z10.object({version:z10.literal(2),source:z10.object({repoUrl:uncredentialedRepoUrlSchema.nullable(),commit:z10.string().regex(/^[a-f0-9]{40}$/i).nullable(),digest:z10.string().regex(/^sha256:[a-f0-9]{64}$/),verification:z10.enum(["SUBMITTED_UNVERIFIED","GITHUB_VERIFIED","LOCAL"])}).strict(),resources:z10.array(runPlanResourceSchema).max(32).default([]),cases:z10.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: ${name}`});for(const[index,resource]of definition.resources.entries()){if(resource.source.repoUrl!==definition.source.repoUrl||resource.source.commit!==definition.source.commit)ctx.addIssue({code:"custom",path:["resources",index,"source"],message:"Project preparation must use the definition repository and commit"})}});var runDefinitionPublicSubmissionSchema=runDefinitionSubmissionSchema.refine(definition=>definition.source.verification!=="GITHUB_VERIFIED",{path:["source","verification"],message:"Only the GitHub App may verify committed bytes"});var runDefinitionPlanResponseSchema=z10.object({source:runDefinitionSubmissionSchema.shape.source,measurementMode:z10.enum(["AGENT_EXPERIENCE","AGENT_PREFERENCE","MIXED"]),cases:z10.array(z10.object({path:z10.string(),agents:z10.array(agentConfigSchema),samples:z10.number().int().nonnegative(),expectedRuns:z10.number().int().nonnegative(),workspace:z10.object({repoUrl:z10.string().nullable(),repoRef:z10.string().nullable()}),criteriaCount:z10.number().int().nonnegative(),preference:preferenceDefinitionSchema.optional(),inputs:z10.array(z10.string())})),resources:z10.array(z10.object({name:z10.string(),type:z10.literal("input"),source:z10.enum(["git","github-actions","github-deployment","github-pr-comment","upload"]),use:inputUseSchema,readiness:z10.enum(["READY","WAITING","FAILED"]),detail:z10.string().optional(),commands:z10.array(z10.string()).optional()})),expectedRuns:z10.number().int().nonnegative(),canRun:z10.boolean()});var runDefinitionRunResponseSchema=z10.object({runRequestId:z10.string(),batchIds:z10.array(z10.string()),runIds:z10.array(z10.string()),requestedRuns:z10.number().int().nonnegative(),createdRuns:z10.number().int().nonnegative(),aggregateUrl:z10.string()});import{z as z11}from"zod";var caseSpecSchema=z11.object({version:z11.literal(2),cases:z11.array(z11.object({key:z11.string().regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/).max(64).describe("Unique case id, lowercase-kebab"),prompt:z11.string().trim().min(1).describe("The user's request"),criteria:z11.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=()=>z11.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:
|
|
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:
|
|
781
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)}
|
|
782
784
|
`;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")}
|
|
783
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(`
|
|
@@ -789,7 +791,7 @@ Wait: ${detached.wait}
|
|
|
789
791
|
Stop: ${detached.stop}`)}finally{await local.cleanup()}return}let batch;try{batch=await runLocal(local.definition,local.uploads,join12(root,".gauge"),run=>{if(!json)console.error(`${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}
|
|
790
792
|
Status: gauge run status --batch ${batchId} --local
|
|
791
793
|
Wait: gauge run wait --batch ${batchId} --local
|
|
792
|
-
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(z12.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",`
|
|
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",`
|
|
793
795
|
${style("1")(" EXAMPLES")}
|
|
794
796
|
gauge run compare --baseline request=<id> --candidate request=<id>
|
|
795
797
|
gauge run compare --baseline eval=<id> --candidate trial=<trialId>
|