bostrat-node 1.17.0 → 1.17.1

package/bin/bostrat-replay

@@ -1,2 +1,14 @@
 #!/usr/bin/env node
-var p=process.env.BOSTRAT_AGENT_URL||`http://127.0.0.1:${process.env.BOSTRAT_AGENT_PORT||3362}`;function a(s){console.error(`[bostrat-replay] ${s}`),process.exit(1)}async function u(s){let t=await fetch(`${p}${s}`),r=await t.json().catch(()=>({}));if(!t.ok)throw new Error(r.error||`HTTP ${t.status}`);return r}function f(s){let t={_:[]};for(let r=0;r<s.length;r++){let e=s[r];e==="--rows"||e==="--json"||e==="--once"?t[e.slice(2)]=!0:e==="--budget"?t.budget=Number(s[++r]):e.startsWith("--")?a(`unknown flag ${e}`):t._.push(e)}return t}var h=s=>{let t=Math.round((Number(s)||0)/6e4);return t>=60?`${(t/60).toFixed(1)}h`:`${t}m`},m=(s,t)=>t?`${Math.round(100*s/t)}%`:"\u2013";async function y(s){let t=Number.isFinite(s.budget)&&s.budget>0?s.budget:8e3,r=s.rows?"1":"0",e;for(let o=0;o<120&&(e=await u(`/api/usage/replay-census?budgetMs=${t}&rows=${r}`),!(!e.partial||s.once));o++)process.stderr.write(`[bostrat-replay] census in progress \u2014 ${e.pending} still pending (${e.judged} judged)\r`);if(e||a("no answer from the agent"),s.json){console.log(JSON.stringify(e,null,2));return}let l=e.tracks,c=e.criteria,i=e.excluded;console.log(`[bostrat-replay] ${e.label}${e.partial?` \u2014 PARTIAL, ${e.pending} pending (run again)`:""}`),console.log(`  eligible: ${e.eligible} of ${e.merged} merged archived Tracks (${m(e.eligible,e.merged)}) \u2014 ${l.archived} archived on this node, ${l.open} open, ${l.loops} loop runs excluded, ${l.unreadable} unreadable`),console.log(`  funnel (each criterion counted on its own, over the ${e.judged} judged):`),console.log(`    merged PR            ${c.mergedPr}`),console.log(`    recoverable base     ${c.recoverableBase}   (not: ${i.baseNotRecoverable}${Object.keys(e.baseReasons).length?" \u2014 "+Object.entries(e.baseReasons).map(([o,n])=>`${o} ${n}`).join(", "):""})`),console.log(`    tests touched        ${c.testsTouched}   (not: ${i.noTestsTouched} of those with a base)`),console.log(`    no ops/deploy step   ${c.noOpsStep}   (not: ${i.opsStep}${Object.keys(e.opsRules).length?" \u2014 "+Object.entries(e.opsRules).map(([o,n])=>`${o} ${n}`).join(", "):""})`),console.log(`    not merged (archived without a merged PR): ${i.notMerged}`);let g=Object.entries(e.byRepo);if(g.length){console.log("  by repo:");for(let[o,n]of g)console.log(`    ${o.padEnd(32)} merged ${String(n.merged).padStart(4)}  eligible ${String(n.eligible).padStart(4)}  (base ${n.baseNotRecoverable} \xB7 tests ${n.noTestsTouched} \xB7 ops ${n.opsStep})`)}let d=Object.entries(e.byCategory);d.length&&console.log(`  eligible by category: ${d.map(([o,n])=>`${o} ${n}`).join(" \xB7 ")}`);let $=Object.entries(e.byModel);if($.length&&console.log(`  eligible by model: ${$.map(([o,n])=>`${o} ${n}`).join(" \xB7 ")}`),s.rows&&Array.isArray(e.rows)){console.log(`  eligible Tracks (${e.rows.length}):`);for(let o of e.rows)console.log(`    ${o.repo}#${o.pr}  ${String(o.name||o.trackId).slice(0,48).padEnd(48)}  ${o.category||"\u2013"}  ${o.model||o.provider||"\u2013"}  active ${h(o.activeMs)}  tests ${o.tests}/${o.files}  base ${String(o.base).slice(0,8)}`)}console.log(`  basis: ${e.basis.label}; ${e.basis.scope}`),console.log(`    no ops/deploy step: ${e.basis.noOpsStep}`)}var b=f(process.argv.slice(2)),w=b._[0];try{w==="census"?await y(b):a("usage: bostrat-replay census [--rows] [--json] [--budget <ms>] [--once]")}catch(s){a(s?.message||String(s))}
+var k=process.env.BOSTRAT_AGENT_URL||`http://127.0.0.1:${process.env.BOSTRAT_AGENT_PORT||3362}`,p=["usage: bostrat-replay census [--rows] [--json] [--budget <ms>] [--once]","       bostrat-replay spec frame <source> [--json]",'       bostrat-replay spec draft <source> (--statement "<text>" | --statement-file <path>) [--suite "<cmd>"] [--note "<why>"]','       bostrat-replay spec revise <spec> (--statement "<text>" | --statement-file <path>) [--suite "<cmd>"] [--note "<what changed>"]',"       bostrat-replay spec ls [--status drafted|approved|dismissed] [--json]","       bostrat-replay spec show <spec> [--json]"].join(`
+`);function c(e){console.error(`[bostrat-replay] ${e}`),process.exit(1)}async function m(e,o,s){let t=await fetch(`${k}${o}`,{method:e,headers:s?{"content-type":"application/json"}:void 0,body:s?JSON.stringify(s):void 0}),n=await t.json().catch(()=>({}));if(!t.ok){let r=new Error(n.error||`HTTP ${t.status}`);throw r.status=t.status,r.data=n,r}return n}var $=e=>m("GET",e),j=new Set(["rows","json","once"]),O=new Set(["budget","statement","statement-file","suite","note","status"]);function w(e){let o={_:[]};for(let s=0;s<e.length;s++){let t=e[s];if(t.startsWith("--")){let n=t.slice(2);j.has(n)?o[n]=!0:O.has(n)?(s+1>=e.length&&c(`${t} needs a value
+${p}`),o[n]=e[++s]):c(`unknown flag ${t}
+${p}`)}else o._.push(t)}return o}var g=e=>{let o=Math.round((Number(e)||0)/6e4);return o>=60?`${(o/60).toFixed(1)}h`:`${o}m`},N=(e,o)=>o?`${Math.round(100*e/o)}%`:"\u2013",h=e=>String(e||"").slice(0,8)||"\u2013",R=e=>e?new Date(e).toISOString().slice(0,10):"\u2013",A=(e,o=96)=>{let s=String(e||"").replace(/\s+/g," ").trim();return s.length>o?`${s.slice(0,o-1)}\u2026`:s};async function E(e){let o=Number.isFinite(Number(e.budget))&&Number(e.budget)>0?Number(e.budget):8e3,s=e.rows?"1":"0",t;for(let a=0;a<120&&(t=await $(`/api/usage/replay-census?budgetMs=${o}&rows=${s}`),!(!t.partial||e.once));a++)process.stderr.write(`[bostrat-replay] census in progress \u2014 ${t.pending} still pending (${t.judged} judged)\r`);if(t||c("no answer from the agent"),e.json){console.log(JSON.stringify(t,null,2));return}let n=t.tracks,r=t.criteria,l=t.excluded;console.log(`[bostrat-replay] ${t.label}${t.partial?` \u2014 PARTIAL, ${t.pending} pending (run again)`:""}`),console.log(`  eligible: ${t.eligible} of ${t.merged} merged archived Tracks (${N(t.eligible,t.merged)}) \u2014 ${n.archived} archived on this node, ${n.open} open, ${n.loops} loop runs excluded, ${n.unreadable} unreadable`),console.log(`  funnel (each criterion counted on its own, over the ${t.judged} judged):`),console.log(`    merged PR            ${r.mergedPr}`),console.log(`    recoverable base     ${r.recoverableBase}   (not: ${l.baseNotRecoverable}${Object.keys(t.baseReasons).length?" \u2014 "+Object.entries(t.baseReasons).map(([a,i])=>`${a} ${i}`).join(", "):""})`),console.log(`    tests touched        ${r.testsTouched}   (not: ${l.noTestsTouched} of those with a base)`),console.log(`    no ops/deploy step   ${r.noOpsStep}   (not: ${l.opsStep}${Object.keys(t.opsRules).length?" \u2014 "+Object.entries(t.opsRules).map(([a,i])=>`${a} ${i}`).join(", "):""})`),console.log(`    not merged (archived without a merged PR): ${l.notMerged}`);let d=Object.entries(t.byRepo);if(d.length){console.log("  by repo:");for(let[a,i]of d)console.log(`    ${a.padEnd(32)} merged ${String(i.merged).padStart(4)}  eligible ${String(i.eligible).padStart(4)}  (base ${i.baseNotRecoverable} \xB7 tests ${i.noTestsTouched} \xB7 ops ${i.opsStep})`)}let v=Object.entries(t.byCategory);v.length&&console.log(`  eligible by category: ${v.map(([a,i])=>`${a} ${i}`).join(" \xB7 ")}`);let y=Object.entries(t.byModel);if(y.length&&console.log(`  eligible by model: ${y.map(([a,i])=>`${a} ${i}`).join(" \xB7 ")}`),e.rows&&Array.isArray(t.rows)){console.log(`  eligible Tracks (${t.rows.length}):`);for(let a of t.rows)console.log(`    ${a.repo}#${a.pr}  ${String(a.name||a.trackId).slice(0,48).padEnd(48)}  ${a.category||"\u2013"}  ${a.model||a.provider||"\u2013"}  active ${g(a.activeMs)}  tests ${a.tests}/${a.files}  base ${h(a.base)}  ${a.trackId}`)}console.log(`  basis: ${t.basis.label}; ${t.basis.scope}`),console.log(`    no ops/deploy step: ${t.basis.noOpsStep}`)}async function I(e){let o;for(let s=0;s<40&&(o=await $(`/api/replay/frame?source=${encodeURIComponent(e)}&budgetMs=15000`),!!o.partial);s++)process.stderr.write(`[bostrat-replay] reading the archived logs \u2014 ${o.pending} still pending\r`);return(!o||o.partial)&&c("the archived logs are still being read \u2014 run again"),o}function f(e,o="  "){console.log(`${o}source: ${e.repo}#${e.pr} \u2014 Track \u201C${e.trackName||e.trackId}\u201D (${e.trackId})`),console.log(`${o}  merged ${R(e.mergedAt)} at ${h(e.mergeCommit)} \xB7 base ${h(e.base)} (the commit a sealed clone is cut at)`),console.log(`${o}  ${e.category||"uncategorized"} \xB7 ${e.model||e.provider||"\u2013"} \xB7 agent-active ${g(e.activeMs)} \xB7 wall ${g(e.wallMs)} \xB7 ${e.prompts} operator prompt${e.prompts===1?"":"s"} \xB7 $${(Number(e.usd)||0).toFixed(2)} notional`)}async function x(e){let o=e._[0];o||c(`spec frame needs a <source>
+${p}`);let s=await I(o);if(e.json){console.log(JSON.stringify(s,null,2));return}if(console.log(`[bostrat-replay] ${s.label||"replay task spec"} \u2014 frame for ${o}`),!s.eligible){console.log(`  NOT replay-eligible: ${s.reasons.join(", ")}`),s.source&&f(s.source),console.log("  a spec cannot be drafted from this Track (R11: no acceptance tests, or an ops step a sealed clone cannot replay)");return}f(s.source),console.log(`  acceptance: ${s.acceptance.tests.length} test path${s.acceptance.tests.length===1?"":"s"} the PR touched (of ${s.acceptance.files} files):`);for(let n of s.acceptance.tests)console.log(`    ${n}`);let t=s.promptNotes||{};console.log(`  operator prompts, verbatim and in order (${s.prompts.length}${t.echoesDropped?`; ${t.echoesDropped} transcript echo${t.echoesDropped===1?"":"es"} dropped`:""}${t.truncated?`; ${t.truncated} cut at 16 KB`:""}${t.omitted?`; ${t.omitted} beyond the first 200 omitted`:""}):`),s.prompts.forEach((n,r)=>{console.log(`  \u2500\u2500 ${r+1}${r===0?" (the opening ask)":" (follow-up \u2014 NOT replayed)"} \xB7 ${new Date(n.at).toISOString()} \xB7 ${n.box||"\u2013"}${n.truncated?" \xB7 truncated":""}`),console.log(n.text.split(`
+`).map(l=>`  \u2502 ${l}`).join(`
+`))}),console.log("  next: write the STATEMENT \u2014 the standalone task a replay box gets instead of this conversation (fold what the follow-ups settled into it; name the outcome the merged PR delivered, never the conversation) \u2014 and hand it in:"),console.log(`    bostrat-replay spec draft ${o} --statement-file <path> [--suite "<the repo's suite command>"]`)}async function T(e){if(e.statement&&e["statement-file"]&&c("give --statement OR --statement-file, not both"),e["statement-file"]){let{readFile:o}=await import("node:fs/promises");try{return await o(e["statement-file"],"utf8")}catch(s){c(`cannot read ${e["statement-file"]}: ${s?.message||s}`)}}return e.statement||c(`a statement is required: --statement "<text>" or --statement-file <path>
+${p}`),e.statement}function b(e){let o=e.versions[e.versions.length-1],s=e.approvedVersion!=null?`approved v${e.approvedVersion}`:"not approved";return console.log(`[bostrat-replay] spec ${e.id} v${e.version} \u2014 ${e.status} (${s}) \u2014 ${e.source.repo}#${e.source.pr} \u201C${e.source.trackName||e.source.trackId}\u201D`),o}async function M(e){let o=e._[0];o||c(`spec draft needs a <source>
+${p}`);let s=await T(e),t;try{t=await m("POST","/api/replay/specs",{cwd:process.cwd(),source:o,statement:s,suite:e.suite??null,note:e.note??null})}catch(r){r.status===503&&r.data?.partial&&c(`${r.message}`),c(r.message)}let n=t.spec;b(n),console.log(`  v1 drafted \u2014 the approval card is in this Track's chat (${n.acceptance.tests.length} acceptance test path${n.acceptance.tests.length===1?"":"s"}, ${n.prompts.length} original prompt${n.prompts.length===1?"":"s"} stored verbatim).`),console.log("  This is a BLOCKING prompt: stop and end your turn. The spec enters the bench only when the operator approves this version; to change it, `bostrat-replay spec revise "+n.id+" \u2026` (a new version, a fresh card).")}async function _(e){let o=e._[0];o||c(`spec revise needs a <spec>
+${p}`);let s=await $(`/api/replay/specs/${encodeURIComponent(o)}`).catch(d=>c(d.message)),t=await T(e),n={cwd:process.cwd(),statement:t,note:e.note??null};e.suite!==void 0&&(n.suite=e.suite);let r;try{r=await m("POST",`/api/replay/specs/${encodeURIComponent(s.spec.id)}/revise`,n)}catch(d){c(d.message)}let l=r.spec;b(l),console.log(`  v${l.version} drafted \u2014 a fresh approval card is in this Track's chat${l.approvedVersion!=null?`; v${l.approvedVersion} stays the bench entry until this one is approved`:""}.`),console.log("  This is a BLOCKING prompt: stop and end your turn.")}async function P(e){let{specs:o}=await $(`/api/replay/specs${e.status?`?status=${encodeURIComponent(e.status)}`:""}`);if(e.json){console.log(JSON.stringify(o,null,2));return}if(!o.length){console.log("[bostrat-replay] no task specs on this node yet \u2014 `bostrat-replay census --rows` lists the eligible sources");return}let s={drafted:0,approved:0,dismissed:0};for(let t of o)s[t.status]=(s[t.status]||0)+1;console.log(`[bostrat-replay] ${o.length} task spec${o.length===1?"":"s"} on this node \u2014 ${s.approved} approved (the bench) \xB7 ${s.drafted} awaiting approval \xB7 ${s.dismissed} dismissed`);for(let t of o){let n=t.versions[t.versions.length-1],r=t.status==="approved"?"+":t.status==="drafted"?"?":"-";console.log(`${r} ${t.id}  v${t.version}${t.approvedVersion!=null&&t.approvedVersion!==t.version?` (approved v${t.approvedVersion})`:""}  ${`${t.source.repo}#${t.source.pr}`.padEnd(28)}  ${(t.source.category||"\u2013").padEnd(14)}  ${(t.source.model||t.source.provider||"\u2013").padEnd(20)}  active ${g(t.source.activeMs).padStart(5)}  tests ${String(t.acceptance.tests.length).padStart(2)}  ${A(n.statement,60)}`)}}async function C(e){let o=e._[0];o||c(`spec show needs a <spec>
+${p}`);let{spec:s}=await $(`/api/replay/specs/${encodeURIComponent(o)}`).catch(t=>c(t.message));if(e.json){console.log(JSON.stringify(s,null,2));return}b(s),f(s.source),console.log(`  acceptance: ${s.acceptance.tests.length} test path${s.acceptance.tests.length===1?"":"s"} (of ${s.acceptance.files} files the PR changed)${s.versions[s.versions.length-1].suite?` \xB7 suite: ${s.versions[s.versions.length-1].suite}`:""}`);for(let t of s.acceptance.tests)console.log(`    ${t}`);console.log(`  versions (${s.versions.length}):`);for(let t of s.versions)console.log(`  \u2500\u2500 v${t.v} \xB7 ${t.outcome}${t.v===s.approvedVersion?" \xB7 THE BENCH ENTRY":""} \xB7 ${new Date(t.createdAt).toISOString()} \xB7 by ${t.by?.agent||"the lead"} on Track \u201C${t.by?.trackName||t.by?.trackId||"\u2013"}\u201D${t.note?` \xB7 ${t.note}`:""}${t.suite?` \xB7 suite: ${t.suite}`:""}`),console.log(t.statement.split(`
+`).map(n=>`  \u2502 ${n}`).join(`
+`));console.log(`  original prompts stored verbatim: ${s.prompts.length} (the opening ask is what the statement replaces; follow-ups are not replayed) \u2014 \`spec show ${s.id} --json\` prints them`)}var u=process.argv.slice(2),S=u.shift();try{if(S==="census")await E(w(u));else if(S==="spec"){let e=u.shift(),o=w(u);e==="frame"?await x(o):e==="draft"?await M(o):e==="revise"?await _(o):e==="ls"?await P(o):e==="show"?await C(o):c(p)}else c(p)}catch(e){c(e?.message||String(e))}