@tangle-network/agent-eval 0.23.1 → 0.24.0

package/dist/{chunk-EXGR4XEM.js → chunk-SESZDQPX.js}

@@ -1,23 +1,23 @@
 import {
   assertLlmRoute
-} from "./chunk-KAO3Q65R.js";
+} from "./chunk-4S4BM3QQ.js";
 import {
   researchReport
-} from "./chunk-IOXMGMHQ.js";
+} from "./chunk-2A5XJB43.js";
 import {
   RunIntegrityError,
   assertRunCaptured
-} from "./chunk-QUKKGHTZ.js";
+} from "./chunk-KTGTIOFD.js";
 import {
   FileSystemRawProviderSink
-} from "./chunk-SQQLHODJ.js";
+} from "./chunk-PC4UYEBM.js";
 import {
   TraceEmitter
-} from "./chunk-5IIQKMD5.js";
+} from "./chunk-TVVP3ZZQ.js";
 import {
   canonicalize,
   hashJson
-} from "./chunk-6M774GY6.js";
+} from "./chunk-4F5DQN55.js";
 
 // src/eval-campaign.ts
 var DEFAULT_INTEGRITY = {
@@ -52,7 +52,9 @@ async function runEvalCampaign(opts) {
     scenarioIds.add(s.scenarioId);
   }
   if (opts.report?.comparator && !variantIds.has(opts.report.comparator)) {
-    throw new Error(`runEvalCampaign: report.comparator "${opts.report.comparator}" is not a configured variantId.`);
+    throw new Error(
+      `runEvalCampaign: report.comparator "${opts.report.comparator}" is not a configured variantId.`
+    );
   }
   if (!opts.commitSha) {
     throw new Error("runEvalCampaign: commitSha is required (every RunRecord needs it).");
@@ -67,17 +69,19 @@ async function runEvalCampaign(opts) {
   const provider = opts.llmOpts.provider ?? null;
   const preregistrationHash = opts.preregistrationHash ?? null;
   const rawSinkFactory = opts.rawSinkFactory ?? defaultRawSinkFactory(opts.workDir);
-  const campaignFingerprint = await hashJson(canonicalize({
-    campaignId: opts.campaignId,
-    variants: opts.variants.map((v) => v.id).sort(),
-    scenarios: opts.scenarios.map((s) => s.scenarioId).sort(),
-    seeds: [...seeds].sort((a, b) => a - b),
-    splitTag,
-    comparator: opts.report?.comparator ?? null,
-    baseUrl,
-    provider,
-    preregistrationHash
-  }));
+  const campaignFingerprint = await hashJson(
+    canonicalize({
+      campaignId: opts.campaignId,
+      variants: opts.variants.map((v) => v.id).sort(),
+      scenarios: opts.scenarios.map((s) => s.scenarioId).sort(),
+      seeds: [...seeds].sort((a, b) => a - b),
+      splitTag,
+      comparator: opts.report?.comparator ?? null,
+      baseUrl,
+      provider,
+      preregistrationHash
+    })
+  );
   const cells = [];
   for (const variant of opts.variants) {
     for (const scenario of opts.scenarios) {
@@ -280,4 +284,4 @@ function defaultRunId(params) {
 export {
   runEvalCampaign
 };
-//# sourceMappingURL=chunk-EXGR4XEM.js.map
+//# sourceMappingURL=chunk-SESZDQPX.js.map

package/dist/chunk-SESZDQPX.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/eval-campaign.ts"],"sourcesContent":["/**\n * EvalCampaign — opinionated matrix runner that wires the four\n * capture-integrity directives by construction.\n *\n * Every consumer that ran a launch-grade benchmark before 0.22 reinvented\n * the same shape: matrix runner → for each (variant, scenario, seed) →\n * start a TraceEmitter → call LLMs → end the run → maybe analyze.\n * The bug class blueprint-agent reported (raw events not captured, route\n * silently wrong, integrity not asserted, analyst never ran) lives at the\n * integration boundary — not the agent-eval API surface. The four\n * directives in `SKILL.md § Capture integrity` are mitigations.\n *\n * `EvalCampaign` is the structural fix. Consumers don't wire the integrity\n * surface anymore; the campaign owns it. Specifically, the campaign:\n *\n *   - calls `assertLlmRoute` once at preflight before any work runs\n *   - constructs a per-run `TraceStore` and `RawProviderSink` via factories\n *   - constructs the `TraceEmitter` with `onRunComplete: [analyst hook]`\n *   - hands the runner an `LlmClientOptions` pre-wired with the sink and\n *     trace context — the runner can't accidentally call an LLM without\n *     capturing the raw HTTP envelope\n *   - calls `assertRunCaptured` after every `endRun` and routes failures\n *     through a configurable policy (`throw` / `mark_failed` / `log`)\n *   - assembles per-run `RunRecord`s and runs `researchReport` at the end\n *     so the campaign artifact is launch-decision-grade by default\n *   - embeds the campaign fingerprint (a SHA-256 over the canonicalised\n *     run set) and optional `preregistrationHash` in the report\n *\n * The runner contract is intentionally narrow: produce a `CampaignRunOutcome`\n * given a fully-wired `CampaignRunContext`. Everything orchestration-shaped\n * lives in the campaign. This is the inversion-of-control point — consumers\n * stop writing matrix runners and start writing scenario-runners.\n *\n * Out of scope for v1 (tracked in `docs/research-report-methodology.md`):\n *\n *   - Distributed/cluster execution (concurrency is local async)\n *   - Adaptive sampling / sequential interim looks\n *   - Resume from partial state across crashes\n *   - LLM-call retry beyond what `LlmClient` already does\n */\n\nimport { assertLlmRoute, type LlmClientOptions, type LlmRouteRequirements } from './llm-client'\nimport { canonicalize, hashJson } from './pre-registration'\nimport type {\n  RunJudgeMetadata,\n  RunOutcome,\n  RunRecord,\n  RunSplitTag,\n  RunTokenUsage,\n} from './run-record'\nimport { type ResearchReport, type ResearchReportOptions, researchReport } from './summary-report'\nimport type { RunCompleteHook } from './trace/emitter'\nimport { TraceEmitter } from './trace/emitter'\nimport {\n  assertRunCaptured,\n  RunIntegrityError,\n  type RunIntegrityExpectations,\n  type RunIntegrityReport,\n} from './trace/integrity'\nimport { FileSystemRawProviderSink, type RawProviderSink } from './trace/raw-provider-sink'\nimport type { TraceStore } from './trace/store'\n\n// ── Public types ─────────────────────────────────────────────────────────\n\nexport interface CampaignVariant<V> {\n  id: string\n  payload: V\n}\n\nexport interface CampaignScenario {\n  scenarioId: string\n  /** Free-form metadata propagated to runs and reports. */\n  tags?: Record<string, string>\n}\n\nexport interface CampaignRunContext<V> {\n  /** Stable run id. The campaign generates this; the runner does not. */\n  runId: string\n  /** Logical experiment id (campaignId by default; overridable per-run via opts). */\n  experimentId: string\n  variant: V\n  variantId: string\n  scenarioId: string\n  scenarioTags: Record<string, string>\n  seed: number\n  splitTag: RunSplitTag\n  /**\n   * The TraceEmitter for this run, with `onRunComplete` hooks pre-wired\n   * (analyst auto-execution if configured, plus integrity check). The\n   * runner MUST call `emitter.startRun` before doing any work and either\n   * `emitter.endRun` or `emitter.abortRun` before returning.\n   */\n  emitter: TraceEmitter\n  store: TraceStore\n  rawSink: RawProviderSink\n  /**\n   * Pre-wired LLM client options — `rawSink` and `traceContext` are populated\n   * so any `callLlm(req, ctx.llmOpts)` automatically captures raw HTTP. The\n   * runner can spread additional fields if needed.\n   */\n  llmOpts: LlmClientOptions\n}\n\nexport interface CampaignRunOutcome {\n  /** Did the run pass? Mirrors `RunOutcome.pass` semantics. */\n  pass: boolean\n  /** Score for the run on its split. Maps to `searchScore` or `holdoutScore`. */\n  score: number\n  /** Mandatory cost in USD. Use 0 + raw.cost_unknown=1 only if truly unknown. */\n  costUsd: number\n  tokenUsage: RunTokenUsage\n  /** Snapshot model id (e.g. `claude-sonnet-4-6@2025-04-15`). */\n  model: string\n  /** sha256 of the effective prompt sent to the model. */\n  promptHash: string\n  /** sha256 of the effective config (model, temperature, tools, judges, splits). */\n  configHash: string\n  /** Optional extra numeric metrics to land in `outcome.raw`. */\n  raw?: Record<string, number>\n  /** Optional failure-taxonomy tag if the run failed. */\n  failureMode?: string\n  /** Optional judge metadata when a judge was used. */\n  judgeMetadata?: RunJudgeMetadata\n}\n\nexport type CampaignRunner<V> = (ctx: CampaignRunContext<V>) => Promise<CampaignRunOutcome>\n\nexport type CampaignIntegrityPolicy = 'throw' | 'mark_failed' | 'log'\n\nexport interface EvalCampaignOptions<V> {\n  /**\n   * Stable id for the campaign. Used as the default `experimentId` on\n   * every run, and folded into the campaign fingerprint.\n   */\n  campaignId: string\n  variants: CampaignVariant<V>[]\n  scenarios: CampaignScenario[]\n  /** Default `[0, 1, 2]`. */\n  seeds?: number[]\n  /** Default `'holdout'` — the split that anchors a launch decision. */\n  splitTag?: RunSplitTag\n  /** Git SHA the campaign is run against. Mandatory; `RunRecord` rejects unset. */\n  commitSha: string\n  /**\n   * LLM client config. Augmented per-run with `rawSink` and `traceContext`\n   * before being passed to the runner. The campaign asserts this config\n   * matches `routeRequirements` once at preflight.\n   */\n  llmOpts: LlmClientOptions\n  /**\n   * Default `{ requireExplicitBaseUrl: true, requireAuth: true }` — fail\n   * loud if the campaign would silently fall back to the public router or\n   * run unauthenticated. Override with an empty object to disable.\n   */\n  routeRequirements?: LlmRouteRequirements\n  /**\n   * Per-run TraceStore factory. Common shape: a fresh store per run keyed\n   * on `runId`. Implementations that share a store across the campaign\n   * are valid — the campaign only writes through `emitter`.\n   */\n  storeFactory: (params: CampaignFactoryParams) => TraceStore\n  /**\n   * Per-run RawProviderSink factory. Defaults to `FileSystemRawProviderSink`\n   * rooted at `${workDir}/raw-events/${runId}` if `workDir` is supplied;\n   * otherwise required. Forensic capture is non-negotiable in a campaign\n   * run — pass `NoopRawProviderSink` explicitly if you want to opt out.\n   */\n  rawSinkFactory?: (params: CampaignFactoryParams) => RawProviderSink\n  /**\n   * Filesystem root for default `rawSinkFactory`. Ignored if\n   * `rawSinkFactory` is supplied.\n   */\n  workDir?: string\n  /**\n   * Extra `onRunComplete` hooks the campaign appends (after its own\n   * integrity-check hook). Pass `traceAnalystOnRunComplete(...)` here.\n   */\n  onRunComplete?: RunCompleteHook[]\n  /**\n   * Per-run integrity expectations. Defaults to:\n   *   `{ llmSpansMin: 1, requireRawCoverageOfLlmSpans: true, requireOutcome: true }`.\n   * Override (e.g. `{ llmSpansMin: 0 }`) for runs that don't call LLMs.\n   */\n  integrity?: RunIntegrityExpectations\n  /** Behaviour when integrity fails. Default `'mark_failed'`. */\n  onIntegrityFailure?: CampaignIntegrityPolicy\n  /**\n   * Per-run runner. Receives a fully-wired context; produces an outcome\n   * the campaign converts into a `RunRecord`.\n   */\n  runner: CampaignRunner<V>\n  /**\n   * If set, the campaign computes `researchReport` at the end. `comparator`\n   * is a `variantId`. Other fields are forwarded verbatim.\n   */\n  report?: { comparator?: string } & Omit<\n    ResearchReportOptions,\n    'comparator' | 'preregistrationHash' | 'generatedAt'\n  >\n  /**\n   * Hash of a signed `HypothesisManifest` (see `pre-registration.ts`).\n   * Embedded in the campaign fingerprint and the research report.\n   */\n  preregistrationHash?: string\n  /** Local concurrency. Default `1` (sequential). */\n  concurrency?: number\n  /**\n   * Override the time source. Tests pass a mock to make wallMs deterministic.\n   */\n  now?: () => number\n  /** Override the runId generator. Tests pin this. */\n  runId?: (params: CampaignFactoryParams) => string\n}\n\nexport interface CampaignFactoryParams {\n  campaignId: string\n  runId: string\n  variantId: string\n  scenarioId: string\n  seed: number\n}\n\nexport interface FailedRun {\n  runId: string\n  variantId: string\n  scenarioId: string\n  seed: number\n  reason: string\n  error?: string\n}\n\nexport interface EvalCampaignResult {\n  campaignId: string\n  /** SHA-256 over canonicalised `(variantIds, scenarioIds, seeds, comparator, splitTag, baseUrl, provider, preregistrationHash)`. */\n  campaignFingerprint: string\n  preregistrationHash: string | null\n  /** Successful runs only. Failed runs land in `failedRuns`. */\n  runs: RunRecord[]\n  /** Integrity reports for every successful run. */\n  integrityReports: RunIntegrityReport[]\n  failedRuns: FailedRun[]\n  /** Computed when `report` is set on options. */\n  report?: ResearchReport\n  startedAt: string\n  endedAt: string\n}\n\n// ── Implementation ───────────────────────────────────────────────────────\n\nconst DEFAULT_INTEGRITY: RunIntegrityExpectations = {\n  llmSpansMin: 1,\n  requireRawCoverageOfLlmSpans: true,\n  requireOutcome: true,\n}\n\nconst DEFAULT_ROUTE: LlmRouteRequirements = {\n  requireExplicitBaseUrl: true,\n  requireAuth: true,\n}\n\nexport async function runEvalCampaign<V>(\n  opts: EvalCampaignOptions<V>,\n): Promise<EvalCampaignResult> {\n  // ── Preflight ──────────────────────────────────────────────────────\n  assertLlmRoute(opts.llmOpts, opts.routeRequirements ?? DEFAULT_ROUTE)\n\n  if (opts.variants.length === 0) {\n    throw new Error('runEvalCampaign: variants must be non-empty.')\n  }\n  if (opts.scenarios.length === 0) {\n    throw new Error('runEvalCampaign: scenarios must be non-empty.')\n  }\n  const variantIds = new Set<string>()\n  for (const v of opts.variants) {\n    if (variantIds.has(v.id)) {\n      throw new Error(`runEvalCampaign: duplicate variant id \"${v.id}\".`)\n    }\n    variantIds.add(v.id)\n  }\n  const scenarioIds = new Set<string>()\n  for (const s of opts.scenarios) {\n    if (scenarioIds.has(s.scenarioId)) {\n      throw new Error(`runEvalCampaign: duplicate scenarioId \"${s.scenarioId}\".`)\n    }\n    scenarioIds.add(s.scenarioId)\n  }\n  if (opts.report?.comparator && !variantIds.has(opts.report.comparator)) {\n    throw new Error(\n      `runEvalCampaign: report.comparator \"${opts.report.comparator}\" is not a configured variantId.`,\n    )\n  }\n  if (!opts.commitSha) {\n    throw new Error('runEvalCampaign: commitSha is required (every RunRecord needs it).')\n  }\n\n  const seeds = opts.seeds ?? [0, 1, 2]\n  const splitTag: RunSplitTag = opts.splitTag ?? 'holdout'\n  const concurrency = Math.max(1, opts.concurrency ?? 1)\n  const integrity = { ...DEFAULT_INTEGRITY, ...(opts.integrity ?? {}) }\n  const onIntegrityFailure: CampaignIntegrityPolicy = opts.onIntegrityFailure ?? 'mark_failed'\n  const now = opts.now ?? (() => Date.now())\n  const baseUrl = (opts.llmOpts.baseUrl ?? '').replace(/\\/+$/, '')\n  const provider = opts.llmOpts.provider ?? null\n  const preregistrationHash = opts.preregistrationHash ?? null\n\n  const rawSinkFactory = opts.rawSinkFactory ?? defaultRawSinkFactory(opts.workDir)\n\n  // ── Fingerprint ────────────────────────────────────────────────────\n  const campaignFingerprint = await hashJson(\n    canonicalize({\n      campaignId: opts.campaignId,\n      variants: opts.variants.map((v) => v.id).sort(),\n      scenarios: opts.scenarios.map((s) => s.scenarioId).sort(),\n      seeds: [...seeds].sort((a, b) => a - b),\n      splitTag,\n      comparator: opts.report?.comparator ?? null,\n      baseUrl,\n      provider,\n      preregistrationHash,\n    }),\n  )\n\n  // ── Plan the matrix ────────────────────────────────────────────────\n  type Cell = { variant: CampaignVariant<V>; scenario: CampaignScenario; seed: number }\n  const cells: Cell[] = []\n  for (const variant of opts.variants) {\n    for (const scenario of opts.scenarios) {\n      for (const seed of seeds) {\n        cells.push({ variant, scenario, seed })\n      }\n    }\n  }\n\n  const startedAt = new Date(now()).toISOString()\n  const runs: RunRecord[] = []\n  const integrityReports: RunIntegrityReport[] = []\n  const failedRuns: FailedRun[] = []\n\n  // ── Execute (bounded-concurrency worker pool) ──────────────────────\n  let cursor = 0\n  async function worker(): Promise<void> {\n    while (true) {\n      const i = cursor++\n      if (i >= cells.length) return\n      const cell = cells[i]!\n      try {\n        const result = await runOneCell(cell)\n        runs.push(result.record)\n        integrityReports.push(result.integrity)\n      } catch (err) {\n        if (err instanceof CellExecutionError) {\n          failedRuns.push(err.failed)\n          if (err.integrity) integrityReports.push(err.integrity)\n        } else {\n          // Genuine bug — not a runner failure, not an integrity failure.\n          // Surface it; don't silently mask.\n          throw err\n        }\n      }\n    }\n  }\n\n  async function runOneCell(\n    cell: Cell,\n  ): Promise<{ record: RunRecord; integrity: RunIntegrityReport }> {\n    const runId = (opts.runId ?? defaultRunId)({\n      campaignId: opts.campaignId,\n      runId: '', // unused by default generator\n      variantId: cell.variant.id,\n      scenarioId: cell.scenario.scenarioId,\n      seed: cell.seed,\n    })\n    const factoryParams: CampaignFactoryParams = {\n      campaignId: opts.campaignId,\n      runId,\n      variantId: cell.variant.id,\n      scenarioId: cell.scenario.scenarioId,\n      seed: cell.seed,\n    }\n    const store = opts.storeFactory(factoryParams)\n    const rawSink = rawSinkFactory(factoryParams)\n\n    const emitter = new TraceEmitter(store, {\n      runId,\n      now: opts.now,\n      onRunComplete: opts.onRunComplete,\n    })\n\n    const llmOpts: LlmClientOptions = {\n      ...opts.llmOpts,\n      rawSink,\n      traceContext: { runId },\n    }\n\n    const ctx: CampaignRunContext<V> = {\n      runId,\n      experimentId: opts.campaignId,\n      variant: cell.variant.payload,\n      variantId: cell.variant.id,\n      scenarioId: cell.scenario.scenarioId,\n      scenarioTags: cell.scenario.tags ?? {},\n      seed: cell.seed,\n      splitTag,\n      emitter,\n      store,\n      rawSink,\n      llmOpts,\n    }\n\n    const wallStart = now()\n    let outcome: CampaignRunOutcome\n    try {\n      outcome = await opts.runner(ctx)\n    } catch (err) {\n      const message = err instanceof Error ? err.message : String(err)\n      // The runner threw mid-execution; give it a chance to have aborted.\n      try {\n        await emitter.abortRun(message)\n      } catch {\n        // Already aborted/ended; ignore.\n      }\n      throw new CellExecutionError({\n        runId,\n        variantId: cell.variant.id,\n        scenarioId: cell.scenario.scenarioId,\n        seed: cell.seed,\n        reason: 'runner_threw',\n        error: message,\n      })\n    }\n    const wallMs = now() - wallStart\n\n    const integrityReport = await assertRunCaptured(store, runId, { ...integrity, rawSink })\n    if (!integrityReport.ok) {\n      switch (onIntegrityFailure) {\n        case 'throw':\n          throw new RunIntegrityError(integrityReport)\n        case 'mark_failed':\n          throw new CellExecutionError(\n            {\n              runId,\n              variantId: cell.variant.id,\n              scenarioId: cell.scenario.scenarioId,\n              seed: cell.seed,\n              reason: 'integrity_failed',\n              error: integrityReport.issues.map((i) => i.code).join(', '),\n            },\n            integrityReport,\n          )\n        case 'log':\n          // Caller wants the run admitted with a flagged report; fall through.\n          break\n      }\n    }\n\n    const recordOutcome: RunOutcome = {\n      raw: outcome.raw ?? {},\n    }\n    if (splitTag === 'holdout') recordOutcome.holdoutScore = outcome.score\n    else recordOutcome.searchScore = outcome.score\n\n    const record: RunRecord = {\n      runId,\n      experimentId: opts.campaignId,\n      candidateId: cell.variant.id,\n      seed: cell.seed,\n      model: outcome.model,\n      promptHash: outcome.promptHash,\n      configHash: outcome.configHash,\n      commitSha: opts.commitSha,\n      wallMs,\n      costUsd: outcome.costUsd,\n      tokenUsage: outcome.tokenUsage,\n      judgeMetadata: outcome.judgeMetadata,\n      outcome: recordOutcome,\n      failureMode: outcome.failureMode,\n      splitTag,\n      scenarioId: cell.scenario.scenarioId,\n    }\n    return { record, integrity: integrityReport }\n  }\n\n  const workers = Array.from({ length: Math.min(concurrency, cells.length) }, () => worker())\n  await Promise.all(workers)\n\n  // ── Optional research report ───────────────────────────────────────\n  let report: ResearchReport | undefined\n  if (opts.report) {\n    const reportOpts: ResearchReportOptions = {\n      ...opts.report,\n      comparator: opts.report.comparator,\n      split: splitTag === 'dev' ? 'search' : splitTag,\n      generatedAt: new Date(now()).toISOString(),\n      preregistrationHash: preregistrationHash ?? undefined,\n    }\n    report = await researchReport(runs, reportOpts)\n  }\n\n  const endedAt = new Date(now()).toISOString()\n\n  return {\n    campaignId: opts.campaignId,\n    campaignFingerprint,\n    preregistrationHash,\n    runs,\n    integrityReports,\n    failedRuns,\n    report,\n    startedAt,\n    endedAt,\n  }\n}\n\n// ── Internal ─────────────────────────────────────────────────────────────\n\nclass CellExecutionError extends Error {\n  readonly failed: FailedRun\n  readonly integrity?: RunIntegrityReport\n  constructor(failed: FailedRun, integrity?: RunIntegrityReport) {\n    super(`cell ${failed.variantId}/${failed.scenarioId}@${failed.seed} failed: ${failed.reason}`)\n    this.failed = failed\n    this.integrity = integrity\n  }\n}\n\nfunction defaultRawSinkFactory(workDir: string | undefined) {\n  return (params: CampaignFactoryParams): RawProviderSink => {\n    if (!workDir) {\n      throw new Error(\n        'runEvalCampaign: rawSinkFactory not supplied and workDir not set. Pass either to enable raw provider capture, or pass `new NoopRawProviderSink()` via rawSinkFactory to opt out explicitly.',\n      )\n    }\n    return new FileSystemRawProviderSink({\n      dir: `${workDir}/raw-events/${params.runId}`,\n    })\n  }\n}\n\nfunction defaultRunId(params: CampaignFactoryParams): string {\n  // Stable across re-runs: fingerprint of (campaignId, variantId, scenarioId, seed).\n  // Caller can override via opts.runId for non-deterministic IDs.\n  const base = `${params.campaignId}::${params.variantId}::${params.scenarioId}::${params.seed}`\n  // Lightweight hex: we don't need crypto-grade here, just stability + uniqueness.\n  let h1 = 0x811c9dc5\n  let h2 = 0x12345678\n  for (let i = 0; i < base.length; i++) {\n    const c = base.charCodeAt(i)\n    h1 = Math.imul(h1 ^ c, 0x01000193) >>> 0\n    h2 = Math.imul(h2 ^ c, 0x9e3779b1) >>> 0\n  }\n  return `run-${h1.toString(16).padStart(8, '0')}${h2.toString(16).padStart(8, '0')}`\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;AAyPA,IAAM,oBAA8C;AAAA,EAClD,aAAa;AAAA,EACb,8BAA8B;AAAA,EAC9B,gBAAgB;AAClB;AAEA,IAAM,gBAAsC;AAAA,EAC1C,wBAAwB;AAAA,EACxB,aAAa;AACf;AAEA,eAAsB,gBACpB,MAC6B;AAE7B,iBAAe,KAAK,SAAS,KAAK,qBAAqB,aAAa;AAEpE,MAAI,KAAK,SAAS,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE;AACA,MAAI,KAAK,UAAU,WAAW,GAAG;AAC/B,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,aAAa,oBAAI,IAAY;AACnC,aAAW,KAAK,KAAK,UAAU;AAC7B,QAAI,WAAW,IAAI,EAAE,EAAE,GAAG;AACxB,YAAM,IAAI,MAAM,0CAA0C,EAAE,EAAE,IAAI;AAAA,IACpE;AACA,eAAW,IAAI,EAAE,EAAE;AAAA,EACrB;AACA,QAAM,cAAc,oBAAI,IAAY;AACpC,aAAW,KAAK,KAAK,WAAW;AAC9B,QAAI,YAAY,IAAI,EAAE,UAAU,GAAG;AACjC,YAAM,IAAI,MAAM,0CAA0C,EAAE,UAAU,IAAI;AAAA,IAC5E;AACA,gBAAY,IAAI,EAAE,UAAU;AAAA,EAC9B;AACA,MAAI,KAAK,QAAQ,cAAc,CAAC,WAAW,IAAI,KAAK,OAAO,UAAU,GAAG;AACtE,UAAM,IAAI;AAAA,MACR,uCAAuC,KAAK,OAAO,UAAU;AAAA,IAC/D;AAAA,EACF;AACA,MAAI,CAAC,KAAK,WAAW;AACnB,UAAM,IAAI,MAAM,oEAAoE;AAAA,EACtF;AAEA,QAAM,QAAQ,KAAK,SAAS,CAAC,GAAG,GAAG,CAAC;AACpC,QAAM,WAAwB,KAAK,YAAY;AAC/C,QAAM,cAAc,KAAK,IAAI,GAAG,KAAK,eAAe,CAAC;AACrD,QAAM,YAAY,EAAE,GAAG,mBAAmB,GAAI,KAAK,aAAa,CAAC,EAAG;AACpE,QAAM,qBAA8C,KAAK,sBAAsB;AAC/E,QAAM,MAAM,KAAK,QAAQ,MAAM,KAAK,IAAI;AACxC,QAAM,WAAW,KAAK,QAAQ,WAAW,IAAI,QAAQ,QAAQ,EAAE;AAC/D,QAAM,WAAW,KAAK,QAAQ,YAAY;AAC1C,QAAM,sBAAsB,KAAK,uBAAuB;AAExD,QAAM,iBAAiB,KAAK,kBAAkB,sBAAsB,KAAK,OAAO;AAGhF,QAAM,sBAAsB,MAAM;AAAA,IAChC,aAAa;AAAA,MACX,YAAY,KAAK;AAAA,MACjB,UAAU,KAAK,SAAS,IAAI,CAAC,MAAM,EAAE,EAAE,EAAE,KAAK;AAAA,MAC9C,WAAW,KAAK,UAAU,IAAI,CAAC,MAAM,EAAE,UAAU,EAAE,KAAK;AAAA,MACxD,OAAO,CAAC,GAAG,KAAK,EAAE,KAAK,CAAC,GAAG,MAAM,IAAI,CAAC;AAAA,MACtC;AAAA,MACA,YAAY,KAAK,QAAQ,cAAc;AAAA,MACvC;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAIA,QAAM,QAAgB,CAAC;AACvB,aAAW,WAAW,KAAK,UAAU;AACnC,eAAW,YAAY,KAAK,WAAW;AACrC,iBAAW,QAAQ,OAAO;AACxB,cAAM,KAAK,EAAE,SAAS,UAAU,KAAK,CAAC;AAAA,MACxC;AAAA,IACF;AAAA,EACF;AAEA,QAAM,YAAY,IAAI,KAAK,IAAI,CAAC,EAAE,YAAY;AAC9C,QAAM,OAAoB,CAAC;AAC3B,QAAM,mBAAyC,CAAC;AAChD,QAAM,aAA0B,CAAC;AAGjC,MAAI,SAAS;AACb,iBAAe,SAAwB;AACrC,WAAO,MAAM;AACX,YAAM,IAAI;AACV,UAAI,KAAK,MAAM,OAAQ;AACvB,YAAM,OAAO,MAAM,CAAC;AACpB,UAAI;AACF,cAAM,SAAS,MAAM,WAAW,IAAI;AACpC,aAAK,KAAK,OAAO,MAAM;AACvB,yBAAiB,KAAK,OAAO,SAAS;AAAA,MACxC,SAAS,KAAK;AACZ,YAAI,eAAe,oBAAoB;AACrC,qBAAW,KAAK,IAAI,MAAM;AAC1B,cAAI,IAAI,UAAW,kBAAiB,KAAK,IAAI,SAAS;AAAA,QACxD,OAAO;AAGL,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,iBAAe,WACb,MAC+D;AAC/D,UAAM,SAAS,KAAK,SAAS,cAAc;AAAA,MACzC,YAAY,KAAK;AAAA,MACjB,OAAO;AAAA;AAAA,MACP,WAAW,KAAK,QAAQ;AAAA,MACxB,YAAY,KAAK,SAAS;AAAA,MAC1B,MAAM,KAAK;AAAA,IACb,CAAC;AACD,UAAM,gBAAuC;AAAA,MAC3C,YAAY,KAAK;AAAA,MACjB;AAAA,MACA,WAAW,KAAK,QAAQ;AAAA,MACxB,YAAY,KAAK,SAAS;AAAA,MAC1B,MAAM,KAAK;AAAA,IACb;AACA,UAAM,QAAQ,KAAK,aAAa,aAAa;AAC7C,UAAM,UAAU,eAAe,aAAa;AAE5C,UAAM,UAAU,IAAI,aAAa,OAAO;AAAA,MACtC;AAAA,MACA,KAAK,KAAK;AAAA,MACV,eAAe,KAAK;AAAA,IACtB,CAAC;AAED,UAAM,UAA4B;AAAA,MAChC,GAAG,KAAK;AAAA,MACR;AAAA,MACA,cAAc,EAAE,MAAM;AAAA,IACxB;AAEA,UAAM,MAA6B;AAAA,MACjC;AAAA,MACA,cAAc,KAAK;AAAA,MACnB,SAAS,KAAK,QAAQ;AAAA,MACtB,WAAW,KAAK,QAAQ;AAAA,MACxB,YAAY,KAAK,SAAS;AAAA,MAC1B,cAAc,KAAK,SAAS,QAAQ,CAAC;AAAA,MACrC,MAAM,KAAK;AAAA,MACX;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,UAAM,YAAY,IAAI;AACtB,QAAI;AACJ,QAAI;AACF,gBAAU,MAAM,KAAK,OAAO,GAAG;AAAA,IACjC,SAAS,KAAK;AACZ,YAAM,UAAU,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAE/D,UAAI;AACF,cAAM,QAAQ,SAAS,OAAO;AAAA,MAChC,QAAQ;AAAA,MAER;AACA,YAAM,IAAI,mBAAmB;AAAA,QAC3B;AAAA,QACA,WAAW,KAAK,QAAQ;AAAA,QACxB,YAAY,KAAK,SAAS;AAAA,QAC1B,MAAM,KAAK;AAAA,QACX,QAAQ;AAAA,QACR,OAAO;AAAA,MACT,CAAC;AAAA,IACH;AACA,UAAM,SAAS,IAAI,IAAI;AAEvB,UAAM,kBAAkB,MAAM,kBAAkB,OAAO,OAAO,EAAE,GAAG,WAAW,QAAQ,CAAC;AACvF,QAAI,CAAC,gBAAgB,IAAI;AACvB,cAAQ,oBAAoB;AAAA,QAC1B,KAAK;AACH,gBAAM,IAAI,kBAAkB,eAAe;AAAA,QAC7C,KAAK;AACH,gBAAM,IAAI;AAAA,YACR;AAAA,cACE;AAAA,cACA,WAAW,KAAK,QAAQ;AAAA,cACxB,YAAY,KAAK,SAAS;AAAA,cAC1B,MAAM,KAAK;AAAA,cACX,QAAQ;AAAA,cACR,OAAO,gBAAgB,OAAO,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,IAAI;AAAA,YAC5D;AAAA,YACA;AAAA,UACF;AAAA,QACF,KAAK;AAEH;AAAA,MACJ;AAAA,IACF;AAEA,UAAM,gBAA4B;AAAA,MAChC,KAAK,QAAQ,OAAO,CAAC;AAAA,IACvB;AACA,QAAI,aAAa,UAAW,eAAc,eAAe,QAAQ;AAAA,QAC5D,eAAc,cAAc,QAAQ;AAEzC,UAAM,SAAoB;AAAA,MACxB;AAAA,MACA,cAAc,KAAK;AAAA,MACnB,aAAa,KAAK,QAAQ;AAAA,MAC1B,MAAM,KAAK;AAAA,MACX,OAAO,QAAQ;AAAA,MACf,YAAY,QAAQ;AAAA,MACpB,YAAY,QAAQ;AAAA,MACpB,WAAW,KAAK;AAAA,MAChB;AAAA,MACA,SAAS,QAAQ;AAAA,MACjB,YAAY,QAAQ;AAAA,MACpB,eAAe,QAAQ;AAAA,MACvB,SAAS;AAAA,MACT,aAAa,QAAQ;AAAA,MACrB;AAAA,MACA,YAAY,KAAK,SAAS;AAAA,IAC5B;AACA,WAAO,EAAE,QAAQ,WAAW,gBAAgB;AAAA,EAC9C;AAEA,QAAM,UAAU,MAAM,KAAK,EAAE,QAAQ,KAAK,IAAI,aAAa,MAAM,MAAM,EAAE,GAAG,MAAM,OAAO,CAAC;AAC1F,QAAM,QAAQ,IAAI,OAAO;AAGzB,MAAI;AACJ,MAAI,KAAK,QAAQ;AACf,UAAM,aAAoC;AAAA,MACxC,GAAG,KAAK;AAAA,MACR,YAAY,KAAK,OAAO;AAAA,MACxB,OAAO,aAAa,QAAQ,WAAW;AAAA,MACvC,aAAa,IAAI,KAAK,IAAI,CAAC,EAAE,YAAY;AAAA,MACzC,qBAAqB,uBAAuB;AAAA,IAC9C;AACA,aAAS,MAAM,eAAe,MAAM,UAAU;AAAA,EAChD;AAEA,QAAM,UAAU,IAAI,KAAK,IAAI,CAAC,EAAE,YAAY;AAE5C,SAAO;AAAA,IACL,YAAY,KAAK;AAAA,IACjB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAIA,IAAM,qBAAN,cAAiC,MAAM;AAAA,EAC5B;AAAA,EACA;AAAA,EACT,YAAY,QAAmB,WAAgC;AAC7D,UAAM,QAAQ,OAAO,SAAS,IAAI,OAAO,UAAU,IAAI,OAAO,IAAI,YAAY,OAAO,MAAM,EAAE;AAC7F,SAAK,SAAS;AACd,SAAK,YAAY;AAAA,EACnB;AACF;AAEA,SAAS,sBAAsB,SAA6B;AAC1D,SAAO,CAAC,WAAmD;AACzD,QAAI,CAAC,SAAS;AACZ,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,WAAO,IAAI,0BAA0B;AAAA,MACnC,KAAK,GAAG,OAAO,eAAe,OAAO,KAAK;AAAA,IAC5C,CAAC;AAAA,EACH;AACF;AAEA,SAAS,aAAa,QAAuC;AAG3D,QAAM,OAAO,GAAG,OAAO,UAAU,KAAK,OAAO,SAAS,KAAK,OAAO,UAAU,KAAK,OAAO,IAAI;AAE5F,MAAI,KAAK;AACT,MAAI,KAAK;AACT,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAM,IAAI,KAAK,WAAW,CAAC;AAC3B,SAAK,KAAK,KAAK,KAAK,GAAG,QAAU,MAAM;AACvC,SAAK,KAAK,KAAK,KAAK,GAAG,UAAU,MAAM;AAAA,EACzC;AACA,SAAO,OAAO,GAAG,SAAS,EAAE,EAAE,SAAS,GAAG,GAAG,CAAC,GAAG,GAAG,SAAS,EAAE,EAAE,SAAS,GAAG,GAAG,CAAC;AACnF;","names":[]}

package/dist/{chunk-6KQG5HAH.js → chunk-SY6WAAAD.js}

@@ -1,6 +1,6 @@
 import {
   callLlmJson
-} from "./chunk-KAO3Q65R.js";
+} from "./chunk-4S4BM3QQ.js";
 
 // src/wire/schemas.ts
 import { extendZodWithOpenApi } from "@asteasolutions/zod-to-openapi";
@@ -76,7 +76,9 @@ var HealthResponseSchema = z.object({
 }).openapi("HealthResponse");
 var ErrorResponseSchema = z.object({
   error: z.object({
-    code: z.string().describe('Machine-readable code: "validation_error", "rubric_not_found", "judge_error".'),
+    code: z.string().describe(
+      'Machine-readable code: "validation_error", "rubric_not_found", "judge_error".'
+    ),
     message: z.string().describe("Human-readable message."),
     details: z.unknown().optional().describe("Optional structured detail.")
   }).describe("Errors are always wrapped in this shape across all endpoints.")
@@ -247,7 +249,12 @@ function validateJudgeOutput(value, rubric) {
   for (const dim of rubric.dimensions) {
     const score = dimensionRecord[dim.id];
     if (typeof score !== "number" || !Number.isFinite(score) || score < dim.min || score > dim.max) {
-      throw new WireError("judge_error", `Judge returned invalid score for dimension "${dim.id}".`, 500, value);
+      throw new WireError(
+        "judge_error",
+        `Judge returned invalid score for dimension "${dim.id}".`,
+        500,
+        value
+      );
     }
     dimensions[dim.id] = score;
   }
@@ -268,7 +275,12 @@ function validateIdArray(raw, allowed, field, original) {
   const out = [];
   for (const item of raw) {
     if (typeof item !== "string" || !allowed.has(item)) {
-      throw new WireError("judge_error", `Judge returned unknown ${field} id "${String(item)}".`, 500, original);
+      throw new WireError(
+        "judge_error",
+        `Judge returned unknown ${field} id "${String(item)}".`,
+        500,
+        original
+      );
     }
     out.push(item);
   }
@@ -371,7 +383,7 @@ function handleVersion() {
 }
 
 // src/wire/openapi.ts
-import { OpenApiGeneratorV31, OpenAPIRegistry } from "@asteasolutions/zod-to-openapi";
+import { OpenAPIRegistry, OpenApiGeneratorV31 } from "@asteasolutions/zod-to-openapi";
 function buildOpenApi(packageVersion) {
   const registry = new OpenAPIRegistry();
   registry.register("JudgeRequest", JudgeRequestSchema);
@@ -494,62 +506,6 @@ Wire-protocol version: ${WIRE_VERSION}. Bumps on breaking changes to request/res
   return doc;
 }
 
-// src/wire/server.ts
-import { serve } from "@hono/node-server";
-import { Hono } from "hono";
-import { cors } from "hono/cors";
-var STARTED_AT = Date.now();
-function createApp() {
-  const app = new Hono();
-  app.use("*", cors());
-  app.onError((err, c) => {
-    if (err instanceof WireError) {
-      return c.json(
-        { error: { code: err.code, message: err.message, details: err.details } },
-        err.status
-      );
-    }
-    console.error("[agent-eval] unhandled error:", err);
-    return c.json(
-      { error: { code: "internal_error", message: "Internal server error." } },
-      500
-    );
-  });
-  app.get(
-    "/healthz",
-    (c) => c.json({ status: "ok", uptimeSec: (Date.now() - STARTED_AT) / 1e3 })
-  );
-  app.get("/v1/version", (c) => c.json(handleVersion()));
-  app.get("/v1/rubrics", (c) => c.json(handleListRubrics()));
-  app.post("/v1/judge", async (c) => {
-    const raw = await c.req.json().catch(() => null);
-    if (raw == null) {
-      throw new WireError("validation_error", "Request body must be JSON.", 400);
-    }
-    const parsed = JudgeRequestSchema.safeParse(raw);
-    if (!parsed.success) {
-      throw new WireError(
-        "validation_error",
-        "Request did not match JudgeRequest schema.",
-        400,
-        parsed.error.issues
-      );
-    }
-    const result = await handleJudge(parsed.data);
-    return c.json(result);
-  });
-  app.get("/openapi.json", (c) => c.json(buildOpenApi(handleVersion().version)));
-  return app;
-}
-function startServer(opts = {}) {
-  const app = createApp();
-  const port = opts.port ?? 5005;
-  const host = opts.host ?? "127.0.0.1";
-  return serve({ fetch: app.fetch, port, hostname: host }, ({ address, port: actualPort }) => {
-    console.log(`[agent-eval] serving on http://${address}:${actualPort}`);
-  });
-}
-
 // src/wire/rpc.ts
 async function dispatchRpc(req) {
   try {
@@ -602,17 +558,19 @@ async function runRpcOnce(method) {
     req = method ? { method, params: body } : body;
   } catch (err) {
     process.stdout.write(
-      JSON.stringify({
+      `${JSON.stringify({
         error: {
           code: "parse_error",
           message: `stdin was not valid JSON: ${err instanceof Error ? err.message : String(err)}`
         }
-      }) + "\n"
+      })}
+`
     );
     return 1;
   }
   const out = await dispatchRpc(req);
-  process.stdout.write(JSON.stringify(out) + "\n");
+  process.stdout.write(`${JSON.stringify(out)}
+`);
   return "error" in out ? 1 : 0;
 }
 async function runRpcBatch(method) {
@@ -626,23 +584,78 @@ async function runRpcBatch(method) {
       req = method ? { method, params: body } : body;
     } catch (err) {
       process.stdout.write(
-        JSON.stringify({
+        `${JSON.stringify({
           error: {
             code: "parse_error",
             message: `line was not valid JSON: ${err instanceof Error ? err.message : String(err)}`
           }
-        }) + "\n"
+        })}
+`
       );
       exitCode = 1;
       continue;
     }
     const out = await dispatchRpc(req);
-    process.stdout.write(JSON.stringify(out) + "\n");
+    process.stdout.write(`${JSON.stringify(out)}
+`);
     if ("error" in out) exitCode = 1;
   }
   return exitCode;
 }
 
+// src/wire/server.ts
+import { serve } from "@hono/node-server";
+import { Hono } from "hono";
+import { cors } from "hono/cors";
+var STARTED_AT = Date.now();
+function createApp() {
+  const app = new Hono();
+  app.use("*", cors());
+  app.onError((err, c) => {
+    if (err instanceof WireError) {
+      return c.json(
+        { error: { code: err.code, message: err.message, details: err.details } },
+        err.status
+      );
+    }
+    console.error("[agent-eval] unhandled error:", err);
+    return c.json({ error: { code: "internal_error", message: "Internal server error." } }, 500);
+  });
+  app.get(
+    "/healthz",
+    (c) => c.json({ status: "ok", uptimeSec: (Date.now() - STARTED_AT) / 1e3 })
+  );
+  app.get("/v1/version", (c) => c.json(handleVersion()));
+  app.get("/v1/rubrics", (c) => c.json(handleListRubrics()));
+  app.post("/v1/judge", async (c) => {
+    const raw = await c.req.json().catch(() => null);
+    if (raw == null) {
+      throw new WireError("validation_error", "Request body must be JSON.", 400);
+    }
+    const parsed = JudgeRequestSchema.safeParse(raw);
+    if (!parsed.success) {
+      throw new WireError(
+        "validation_error",
+        "Request did not match JudgeRequest schema.",
+        400,
+        parsed.error.issues
+      );
+    }
+    const result = await handleJudge(parsed.data);
+    return c.json(result);
+  });
+  app.get("/openapi.json", (c) => c.json(buildOpenApi(handleVersion().version)));
+  return app;
+}
+function startServer(opts = {}) {
+  const app = createApp();
+  const port = opts.port ?? 5005;
+  const host = opts.host ?? "127.0.0.1";
+  return serve({ fetch: app.fetch, port, hostname: host }, ({ address, port: actualPort }) => {
+    console.log(`[agent-eval] serving on http://${address}:${actualPort}`);
+  });
+}
+
 export {
   RubricDimensionSchema,
   FailureModeSchema,
@@ -664,10 +677,10 @@ export {
   handleListRubrics,
   handleVersion,
   buildOpenApi,
-  createApp,
-  startServer,
   dispatchRpc,
   runRpcOnce,
-  runRpcBatch
+  runRpcBatch,
+  createApp,
+  startServer
 };
-//# sourceMappingURL=chunk-6KQG5HAH.js.map
+//# sourceMappingURL=chunk-SY6WAAAD.js.map

package/dist/chunk-SY6WAAAD.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/wire/schemas.ts","../src/wire/rubrics.ts","../src/wire/handlers.ts","../src/wire/openapi.ts","../src/wire/rpc.ts","../src/wire/server.ts"],"sourcesContent":["/**\n * Wire-protocol schemas.\n *\n * These Zod schemas are the contract between the agent-eval runtime and\n * any non-TypeScript client (Python, Rust, Go, …). They get rendered to\n * OpenAPI by `wire/openapi.ts` and code-generators consume that spec to\n * produce typed clients in other languages.\n *\n * Rule: if it's not in this file, it isn't on the wire. Keep names and\n * shapes self-explanatory — every field has a `.describe()` so the\n * generated docs are useful without reading the source.\n */\nimport { extendZodWithOpenApi } from '@asteasolutions/zod-to-openapi'\nimport { z } from 'zod'\n\nextendZodWithOpenApi(z)\n\n// ── Building blocks ─────────────────────────────────────────────────\n\nexport const RubricDimensionSchema = z\n  .object({\n    id: z\n      .string()\n      .min(1)\n      .describe('Short stable id like \"buyer_quality\" — used as the key in scoring output.'),\n    description: z\n      .string()\n      .min(1)\n      .describe('One-line plain-English meaning. Read by humans reviewing low scores.'),\n    weight: z\n      .number()\n      .min(0)\n      .default(1)\n      .describe('Relative weight in the composite score. Default 1; 0 disables.'),\n    min: z.number().default(0).describe('Lower bound of valid score for this dimension.'),\n    max: z.number().default(1).describe('Upper bound of valid score for this dimension.'),\n  })\n  .openapi('RubricDimension')\n\nexport const FailureModeSchema = z\n  .object({\n    id: z.string().min(1).describe('Short stable id like \"ai-cadence\" — used in detection lists.'),\n    description: z.string().min(1).describe('Plain-English description of the failure pattern.'),\n  })\n  .openapi('FailureMode')\n\n// ── Rubric ──────────────────────────────────────────────────────────\n\nexport const RubricSchema = z\n  .object({\n    name: z\n      .string()\n      .min(1)\n      .describe('Stable name like \"anti-slop\" — used by clients to invoke this rubric.'),\n    description: z\n      .string()\n      .min(1)\n      .describe('What this rubric measures. Shown in /v1/rubrics listing.'),\n    systemPrompt: z\n      .string()\n      .min(1)\n      .describe(\n        'Instructs the judging LLM. Should explain the persona (e.g. \"senior engineer reviewing voice\"), what to score on, and what to return.',\n      ),\n    dimensions: z\n      .array(RubricDimensionSchema)\n      .min(1)\n      .describe('Scoring axes. The composite score is a weighted sum of these.'),\n    failureModes: z\n      .array(FailureModeSchema)\n      .default([])\n      .describe('Patterns to detect; each detected mode appears in the result.failureModes list.'),\n    wins: z\n      .array(FailureModeSchema)\n      .default([])\n      .describe('Positive patterns; each detected one appears in the result.wins list.'),\n  })\n  .openapi('Rubric')\n\n// ── Judge call ──────────────────────────────────────────────────────\n\nexport const JudgeRequestSchema = z\n  .object({\n    rubricName: z\n      .string()\n      .optional()\n      .describe('Use a built-in rubric by name. Mutually exclusive with `rubric`.'),\n    rubric: RubricSchema.optional().describe(\n      'Inline rubric definition. Mutually exclusive with `rubricName`.',\n    ),\n    content: z\n      .string()\n      .min(1)\n      .describe('The text being judged — a tweet, a blog post, a code snippet, anything stringly.'),\n    context: z\n      .record(z.string(), z.unknown())\n      .optional()\n      .describe(\n        'Free-form metadata for the rubric to use — analytics, source URL, author, etc. Surfaced to the LLM.',\n      ),\n    model: z\n      .string()\n      .optional()\n      .describe('Override the judge model (default routes via tcloud). e.g. \"claude-opus-4-7\".'),\n  })\n  .refine((v) => Boolean(v.rubricName) !== Boolean(v.rubric), {\n    message: 'Provide exactly one of `rubricName` or `rubric`.',\n  })\n  .openapi('JudgeRequest')\n\nexport const JudgeResultSchema = z\n  .object({\n    composite: z\n      .number()\n      .min(0)\n      .max(1)\n      .describe('Weighted combination of dimension scores in 0..1. The single number to gate on.'),\n    dimensions: z\n      .record(z.string(), z.number())\n      .describe('Per-dimension score, keyed by RubricDimension.id.'),\n    failureModes: z\n      .array(z.string())\n      .default([])\n      .describe('Failure-mode ids detected in the content (subset of rubric.failureModes ids).'),\n    wins: z\n      .array(z.string())\n      .default([])\n      .describe('Win ids detected in the content (subset of rubric.wins ids).'),\n    rationale: z\n      .string()\n      .describe('Plain-English explanation of the score. Surfaced to the human reviewer.'),\n    rubricVersion: z\n      .string()\n      .describe(\n        'Stable hash of the rubric used. Scores are only comparable across runs when this matches.',\n      ),\n    model: z.string().describe('Model that produced the judgement, for reproducibility.'),\n    durationMs: z.number().int().nonnegative().describe('End-to-end wall time for this call.'),\n  })\n  .openapi('JudgeResult')\n\n// ── Rubric listing ──────────────────────────────────────────────────\n\nexport const RubricInfoSchema = z\n  .object({\n    name: z.string().describe('Pass this to /v1/judge as `rubricName`.'),\n    description: z.string().describe('What this rubric measures.'),\n    dimensions: z\n      .array(z.object({ id: z.string(), description: z.string(), weight: z.number() }))\n      .describe('The scoring axes this rubric uses, with weights.'),\n    failureModes: z.array(z.string()).default([]).describe('Failure-mode ids this rubric detects.'),\n    rubricVersion: z.string().describe('Stable hash — match this to compare scores across runs.'),\n  })\n  .openapi('RubricInfo')\n\nexport const ListRubricsResponseSchema = z\n  .object({\n    rubrics: z.array(RubricInfoSchema),\n  })\n  .openapi('ListRubricsResponse')\n\n// ── Version / health ────────────────────────────────────────────────\n\nexport const VersionResponseSchema = z\n  .object({\n    package: z.string().describe('Package name (always \"@tangle-network/agent-eval\").'),\n    version: z.string().describe('Semver of the running server. Match your client to this.'),\n    wireVersion: z\n      .string()\n      .describe(\n        'Wire-protocol semver. Bumps separately from package version when the schema changes.',\n      ),\n    apiSurface: z.array(z.string()).describe('List of supported method names.'),\n  })\n  .openapi('VersionResponse')\n\nexport const HealthResponseSchema = z\n  .object({\n    status: z.literal('ok'),\n    uptimeSec: z.number(),\n  })\n  .openapi('HealthResponse')\n\n// ── Errors ──────────────────────────────────────────────────────────\n\nexport const ErrorResponseSchema = z\n  .object({\n    error: z\n      .object({\n        code: z\n          .string()\n          .describe(\n            'Machine-readable code: \"validation_error\", \"rubric_not_found\", \"judge_error\".',\n          ),\n        message: z.string().describe('Human-readable message.'),\n        details: z.unknown().optional().describe('Optional structured detail.'),\n      })\n      .describe('Errors are always wrapped in this shape across all endpoints.'),\n  })\n  .openapi('ErrorResponse')\n\n// ── Type exports for callers in the same package ────────────────────\n\nexport type RubricDimension = z.infer<typeof RubricDimensionSchema>\nexport type FailureMode = z.infer<typeof FailureModeSchema>\nexport type Rubric = z.infer<typeof RubricSchema>\nexport type JudgeRequest = z.infer<typeof JudgeRequestSchema>\nexport type JudgeResult = z.infer<typeof JudgeResultSchema>\nexport type RubricInfo = z.infer<typeof RubricInfoSchema>\nexport type ListRubricsResponse = z.infer<typeof ListRubricsResponseSchema>\nexport type VersionResponse = z.infer<typeof VersionResponseSchema>\nexport type ErrorResponse = z.infer<typeof ErrorResponseSchema>\n\n// ── Wire-protocol version ───────────────────────────────────────────\n\n/**\n * Bump on any breaking change to a request/response schema.\n * Non-breaking (additive) changes don't require a bump.\n */\nexport const WIRE_VERSION = '1.0.0'\n\n/**\n * Stable hash of a rubric. Used to make scores comparable across runs:\n * if the rubricVersion matches, the rubric was identical.\n */\nexport function hashRubric(rubric: Rubric): string {\n  const stable = stableStringify(rubric)\n  let h = 5381\n  for (let i = 0; i < stable.length; i++) {\n    h = (h * 33) ^ stable.charCodeAt(i)\n  }\n  // Unsigned 32-bit hex, prefixed with rubric name + version slot\n  return `${rubric.name}@${(h >>> 0).toString(16).padStart(8, '0')}`\n}\n\nfunction stableStringify(value: unknown): string {\n  if (Array.isArray(value)) return `[${value.map((item) => stableStringify(item)).join(',')}]`\n  if (value && typeof value === 'object') {\n    const entries = Object.entries(value as Record<string, unknown>)\n      .sort(([a], [b]) => a.localeCompare(b))\n      .map(([key, item]) => `${JSON.stringify(key)}:${stableStringify(item)}`)\n    return `{${entries.join(',')}}`\n  }\n  return JSON.stringify(value)\n}\n","/**\n * Built-in rubrics shipped with agent-eval.\n *\n * A rubric is a set of scoring axes plus a system prompt that tells the\n * judging LLM how to grade against those axes. Built-in rubrics are\n * curated for use cases that recur across Tangle projects — call them\n * by name from any client.\n *\n * Adding a rubric:\n *   1. Define the Rubric object below with a clear `description` and\n *      named `dimensions`.\n *   2. Register it in `BUILTIN_RUBRICS` at the bottom.\n *   3. Add a test in `tests/wire/rubrics.test.ts`.\n *\n * Custom rubrics: callers pass `rubric` inline to /v1/judge instead of\n * `rubricName` — see schemas.ts.\n */\nimport type { Rubric } from './schemas'\nimport { hashRubric } from './schemas'\n\n// ── anti-slop ───────────────────────────────────────────────────────\n// Voice/style judge tuned for technical-buyer audiences. Used by the\n// Postiz autoresearch loop and any content-quality gate.\n\nconst ANTI_SLOP: Rubric = {\n  name: 'anti-slop',\n  description:\n    'Voice and signal quality for content aimed at senior engineers. Catches AI cadence, marketing tone, and engagement-bait shapes.',\n  systemPrompt: `You are evaluating a piece of content written for senior engineers and technical founders.\n\nYou score three things:\n- buyer_quality (0..1): would a senior engineer in the target ICP find this worth their attention? High = specific, earned, technically interesting. Low = generic, hyped, off-target.\n- voice (0..1): does it read like a person who built the thing, or like AI/marketing copy?\n- signal (0..1): does it contain a non-obvious detail, constraint, or claim a reader couldn't get from the public docs?\n\nDetect failure modes (return ids matching):\n- ai-cadence: rule-of-three openings, em-dash flourish, \"Let me explain\", \"Here's the thing\", AI rhythm\n- marketing-tone: \"We're excited to announce\", \"thrilled\", \"delighted\", \"game-changer\", buzzword stack\n- vague-claim: technical claim without a specific component, file, or measurement\n- no-hook: opening doesn't earn attention from the target reader\n- engagement-bait: \"agree?\", \"thoughts?\", listicles, controversy-fishing, hook-detail-pitch\n- off-icp: content shape would attract motivational/grift/hype audiences instead of buyers\n- stale-claim: repeats a positioning line we've used many times this month\n\nDetect wins (return ids matching):\n- specific-component: names a real file, component, or measurement\n- earned-detail: shares a non-obvious detail not derivable from public docs\n- constraint-articulated: names a real tradeoff and the side chosen\n- honest-failure: describes a real failure mode and what was done about it\n\nReturn ONLY JSON matching the response schema. Be conservative — most content has 0-1 wins and 1-2 failure modes, not many of each.`,\n  dimensions: [\n    {\n      id: 'buyer_quality',\n      description: 'Would the target buyer find this worth attention?',\n      weight: 0.5,\n      min: 0,\n      max: 1,\n    },\n    {\n      id: 'voice',\n      description: 'Does it sound like a builder, not AI or marketing?',\n      weight: 0.3,\n      min: 0,\n      max: 1,\n    },\n    {\n      id: 'signal',\n      description: 'Non-obvious detail, constraint, or claim?',\n      weight: 0.2,\n      min: 0,\n      max: 1,\n    },\n  ],\n  failureModes: [\n    { id: 'ai-cadence', description: 'AI-rhythm openings and transitions' },\n    { id: 'marketing-tone', description: 'Buzzwords, hype, corporate-PR voice' },\n    { id: 'vague-claim', description: 'Technical claim without specifics' },\n    { id: 'no-hook', description: 'Opening fails to earn attention' },\n    { id: 'engagement-bait', description: 'Listicle/controversy/agree-pattern' },\n    { id: 'off-icp', description: 'Voice attracts the wrong audience' },\n    { id: 'stale-claim', description: 'Reuses an over-used positioning line' },\n  ],\n  wins: [\n    { id: 'specific-component', description: 'Names a real file/component/number' },\n    { id: 'earned-detail', description: 'Detail not in public docs' },\n    { id: 'constraint-articulated', description: 'Names a real tradeoff' },\n    { id: 'honest-failure', description: 'Describes a real failure honestly' },\n  ],\n}\n\n// ── Registry ────────────────────────────────────────────────────────\n\nexport const BUILTIN_RUBRICS: Record<string, Rubric> = {\n  'anti-slop': ANTI_SLOP,\n}\n\n/** Get a built-in rubric by name, or undefined. */\nexport function getBuiltinRubric(name: string): Rubric | undefined {\n  return BUILTIN_RUBRICS[name]\n}\n\n/** List built-in rubrics with their stable versions. */\nexport function listBuiltinRubrics() {\n  return Object.values(BUILTIN_RUBRICS).map((r) => ({\n    name: r.name,\n    description: r.description,\n    dimensions: r.dimensions.map((d) => ({\n      id: d.id,\n      description: d.description,\n      weight: d.weight,\n    })),\n    failureModes: r.failureModes.map((f) => f.id),\n    rubricVersion: hashRubric(r),\n  }))\n}\n","/**\n * Pure handler functions — the \"business logic\" behind every wire-protocol\n * method. The HTTP server (`server.ts`) and the stdio RPC (`rpc.ts`) both\n * call these. Tests call these directly without spinning a server.\n *\n * Each handler:\n *   - Takes a parsed request (already Zod-validated by the transport).\n *   - Returns a result that matches the response schema.\n *   - Throws `WireError` for caller-fixable errors (404, 400, 422).\n *   - Lets unexpected errors bubble — the transport maps them to 500.\n */\nimport { callLlmJson } from '../llm-client'\nimport { getBuiltinRubric, listBuiltinRubrics } from './rubrics'\nimport {\n  hashRubric,\n  type JudgeRequest,\n  type JudgeResult,\n  type ListRubricsResponse,\n  type Rubric,\n  type VersionResponse,\n  WIRE_VERSION,\n} from './schemas'\n\n/** Caller-fixable error. The transport renders this to 4xx + ErrorResponse. */\nexport class WireError extends Error {\n  constructor(\n    public readonly code: string,\n    message: string,\n    public readonly status: number = 400,\n    public readonly details?: unknown,\n  ) {\n    super(message)\n    this.name = 'WireError'\n  }\n}\n\n// ── judge ───────────────────────────────────────────────────────────\n\n/** The JSON schema we ask the judging LLM to fill in. */\nfunction judgeOutputSchema(rubric: Rubric) {\n  return {\n    name: 'JudgeOutput',\n    schema: {\n      type: 'object',\n      additionalProperties: false,\n      properties: {\n        dimensions: {\n          type: 'object',\n          additionalProperties: false,\n          properties: Object.fromEntries(\n            rubric.dimensions.map((d) => [\n              d.id,\n              { type: 'number', minimum: d.min, maximum: d.max },\n            ]),\n          ),\n          required: rubric.dimensions.map((d) => d.id),\n        },\n        failureModes: {\n          type: 'array',\n          items: { type: 'string', enum: rubric.failureModes.map((f) => f.id) },\n        },\n        wins: {\n          type: 'array',\n          items: { type: 'string', enum: rubric.wins.map((w) => w.id) },\n        },\n        rationale: { type: 'string' },\n      },\n      required: ['dimensions', 'rationale'],\n    } as Record<string, unknown>,\n  }\n}\n\ninterface JudgeOutput {\n  dimensions: Record<string, number>\n  failureModes?: string[]\n  wins?: string[]\n  rationale: string\n}\n\nfunction validateJudgeOutput(value: unknown, rubric: Rubric): JudgeOutput {\n  if (!value || typeof value !== 'object') {\n    throw new WireError('judge_error', 'Judge returned malformed output.', 500, value)\n  }\n  const raw = value as Record<string, unknown>\n  const rawDimensions = raw.dimensions\n  if (!rawDimensions || typeof rawDimensions !== 'object' || Array.isArray(rawDimensions)) {\n    throw new WireError('judge_error', 'Judge returned malformed dimensions.', 500, value)\n  }\n\n  const dimensions: Record<string, number> = {}\n  const dimensionRecord = rawDimensions as Record<string, unknown>\n  for (const dim of rubric.dimensions) {\n    const score = dimensionRecord[dim.id]\n    if (\n      typeof score !== 'number' ||\n      !Number.isFinite(score) ||\n      score < dim.min ||\n      score > dim.max\n    ) {\n      throw new WireError(\n        'judge_error',\n        `Judge returned invalid score for dimension \"${dim.id}\".`,\n        500,\n        value,\n      )\n    }\n    dimensions[dim.id] = score\n  }\n\n  const allowedFailures = new Set(rubric.failureModes.map((mode) => mode.id))\n  const allowedWins = new Set(rubric.wins.map((win) => win.id))\n  const failureModes = validateIdArray(raw.failureModes, allowedFailures, 'failureModes', value)\n  const wins = validateIdArray(raw.wins, allowedWins, 'wins', value)\n  if (typeof raw.rationale !== 'string' || raw.rationale.trim().length === 0) {\n    throw new WireError('judge_error', 'Judge returned missing rationale.', 500, value)\n  }\n\n  return { dimensions, failureModes, wins, rationale: raw.rationale }\n}\n\nfunction validateIdArray(\n  raw: unknown,\n  allowed: Set<string>,\n  field: 'failureModes' | 'wins',\n  original: unknown,\n): string[] {\n  if (raw === undefined) return []\n  if (!Array.isArray(raw)) {\n    throw new WireError('judge_error', `Judge returned non-array ${field}.`, 500, original)\n  }\n  const out: string[] = []\n  for (const item of raw) {\n    if (typeof item !== 'string' || !allowed.has(item)) {\n      throw new WireError(\n        'judge_error',\n        `Judge returned unknown ${field} id \"${String(item)}\".`,\n        500,\n        original,\n      )\n    }\n    out.push(item)\n  }\n  return out\n}\n\nfunction compositeScore(dimensions: Record<string, number>, rubric: Rubric): number {\n  let weighted = 0\n  let totalWeight = 0\n  for (const dim of rubric.dimensions) {\n    const raw = dimensions[dim.id] ?? 0\n    const range = dim.max - dim.min || 1\n    const normalized = Math.max(0, Math.min(1, (raw - dim.min) / range))\n    weighted += normalized * dim.weight\n    totalWeight += dim.weight\n  }\n  return totalWeight > 0 ? weighted / totalWeight : 0\n}\n\nfunction buildJudgePrompt(content: string, context: unknown): string {\n  const ctx = context && Object.keys(context as object).length ? JSON.stringify(context) : ''\n  return [\n    `CONTENT TO JUDGE:`,\n    content,\n    '',\n    ctx ? `CONTEXT (metadata, analytics, etc.):` : '',\n    ctx ? ctx : '',\n  ]\n    .filter(Boolean)\n    .join('\\n')\n}\n\nconst DEFAULT_JUDGE_MODEL = 'claude-sonnet-4-6'\n\nexport async function handleJudge(req: JudgeRequest): Promise<JudgeResult> {\n  // Resolve rubric\n  let rubric: Rubric\n  if (req.rubricName) {\n    const found = getBuiltinRubric(req.rubricName)\n    if (!found) {\n      throw new WireError('rubric_not_found', `No built-in rubric named \"${req.rubricName}\".`, 404)\n    }\n    rubric = found\n  } else if (req.rubric) {\n    rubric = req.rubric\n  } else {\n    // refine() in the schema should already have caught this — defense in depth\n    throw new WireError('validation_error', 'Provide either `rubricName` or `rubric`.', 422)\n  }\n\n  const startedAt = Date.now()\n  const model = req.model ?? DEFAULT_JUDGE_MODEL\n\n  const { value, result } = await callLlmJson<JudgeOutput>({\n    model,\n    messages: [\n      { role: 'system', content: rubric.systemPrompt },\n      { role: 'user', content: buildJudgePrompt(req.content, req.context) },\n    ],\n    jsonSchema: judgeOutputSchema(rubric),\n    temperature: 0.0,\n    timeoutMs: 60_000,\n  })\n\n  const output = validateJudgeOutput(value, rubric)\n\n  const composite = compositeScore(output.dimensions, rubric)\n  const durationMs = Date.now() - startedAt\n\n  return {\n    composite,\n    dimensions: output.dimensions,\n    failureModes: output.failureModes ?? [],\n    wins: output.wins ?? [],\n    rationale: output.rationale,\n    rubricVersion: hashRubric(rubric),\n    model: result.model,\n    durationMs,\n  }\n}\n\n// ── listRubrics ─────────────────────────────────────────────────────\n\nexport function handleListRubrics(): ListRubricsResponse {\n  return { rubrics: listBuiltinRubrics() }\n}\n\n// ── version ─────────────────────────────────────────────────────────\n\nimport { readFileSync } from 'node:fs'\nimport { dirname, resolve } from 'node:path'\nimport { fileURLToPath } from 'node:url'\n\nlet CACHED_VERSION: string | undefined\n\nfunction readPackageVersion(): string {\n  if (CACHED_VERSION) return CACHED_VERSION\n  // Walk up from this file looking for the nearest package.json.\n  // In dist/ this is dist/.., in src/wire/ this is ../../package.json.\n  const here = dirname(fileURLToPath(import.meta.url))\n  const candidates = [\n    resolve(here, '..', '..', 'package.json'), // src/wire → repo root\n    resolve(here, '..', 'package.json'), // dist → repo root\n  ]\n  for (const path of candidates) {\n    try {\n      const pkg = JSON.parse(readFileSync(path, 'utf-8')) as { version?: string }\n      if (pkg.version) {\n        CACHED_VERSION = pkg.version\n        return pkg.version\n      }\n    } catch {\n      // try next\n    }\n  }\n  return '0.0.0-unknown'\n}\n\nexport function handleVersion(): VersionResponse {\n  return {\n    package: '@tangle-network/agent-eval',\n    version: readPackageVersion(),\n    wireVersion: WIRE_VERSION,\n    apiSurface: ['judge', 'listRubrics', 'version'],\n  }\n}\n","/**\n * Build an OpenAPI spec from the wire schemas.\n *\n * The spec is the contract that other-language clients (Python, Rust,\n * Go, …) generate from. There is no hand-written client — clients are\n * derived artifacts of this file plus `schemas.ts`.\n *\n * Run `pnpm openapi` (defined in package.json) to write the spec to\n * `dist/openapi.json`. CI uses that file to regenerate the Python\n * client and gate the dual-publish workflow.\n */\nimport { OpenAPIRegistry, OpenApiGeneratorV31 } from '@asteasolutions/zod-to-openapi'\nimport type { OpenAPIObject } from 'openapi3-ts/oas31'\n\nimport {\n  ErrorResponseSchema,\n  HealthResponseSchema,\n  JudgeRequestSchema,\n  JudgeResultSchema,\n  ListRubricsResponseSchema,\n  VersionResponseSchema,\n  WIRE_VERSION,\n} from './schemas'\n\nexport function buildOpenApi(packageVersion: string): OpenAPIObject {\n  const registry = new OpenAPIRegistry()\n\n  // Components — each schema becomes a $ref-able component\n  registry.register('JudgeRequest', JudgeRequestSchema)\n  registry.register('JudgeResult', JudgeResultSchema)\n  registry.register('ListRubricsResponse', ListRubricsResponseSchema)\n  registry.register('VersionResponse', VersionResponseSchema)\n  registry.register('HealthResponse', HealthResponseSchema)\n  registry.register('ErrorResponse', ErrorResponseSchema)\n\n  // Routes\n  registry.registerPath({\n    method: 'post',\n    path: '/v1/judge',\n    summary: 'Score a piece of content against a rubric',\n    description:\n      'Runs the judging LLM with the named (or inline) rubric and returns dimension scores, detected failure modes, wins, and a composite score in 0..1.',\n    request: {\n      body: {\n        content: {\n          'application/json': { schema: JudgeRequestSchema },\n        },\n      },\n    },\n    responses: {\n      200: {\n        description: 'Successful judgement',\n        content: { 'application/json': { schema: JudgeResultSchema } },\n      },\n      400: {\n        description: 'Validation error',\n        content: { 'application/json': { schema: ErrorResponseSchema } },\n      },\n      404: {\n        description: 'Rubric not found',\n        content: { 'application/json': { schema: ErrorResponseSchema } },\n      },\n      500: {\n        description: 'Judge error',\n        content: { 'application/json': { schema: ErrorResponseSchema } },\n      },\n    },\n  })\n\n  registry.registerPath({\n    method: 'get',\n    path: '/v1/rubrics',\n    summary: 'List built-in rubrics',\n    description:\n      'Returns every rubric registered server-side, with their dimensions and stable rubricVersion hash.',\n    responses: {\n      200: {\n        description: 'Listing',\n        content: { 'application/json': { schema: ListRubricsResponseSchema } },\n      },\n    },\n  })\n\n  registry.registerPath({\n    method: 'get',\n    path: '/v1/version',\n    summary: 'Server and wire-protocol version',\n    description: 'Match your client version to `version`; check `wireVersion` for compatibility.',\n    responses: {\n      200: {\n        description: 'Version info',\n        content: { 'application/json': { schema: VersionResponseSchema } },\n      },\n    },\n  })\n\n  registry.registerPath({\n    method: 'get',\n    path: '/healthz',\n    summary: 'Liveness check',\n    responses: {\n      200: {\n        description: 'OK',\n        content: { 'application/json': { schema: HealthResponseSchema } },\n      },\n    },\n  })\n\n  const generator = new OpenApiGeneratorV31(registry.definitions)\n  const doc = generator.generateDocument({\n    openapi: '3.1.0',\n    info: {\n      title: '@tangle-network/agent-eval — wire protocol',\n      version: packageVersion,\n      description: `HTTP and stdio RPC interface to agent-eval. The TypeScript runtime is the source of truth; this spec is the contract that cross-language clients (Python, Rust, Go) generate from.\n\nWire-protocol version: ${WIRE_VERSION}. Bumps on breaking changes to request/response schemas.`,\n      contact: { name: 'Tangle Network', url: 'https://github.com/tangle-network/agent-eval' },\n      license: { name: 'MIT' },\n    },\n    servers: [{ url: 'http://localhost:5005', description: 'Local agent-eval serve' }],\n  })\n  const rubricRef = { $ref: '#/components/schemas/Rubric' } as const\n  const commonJudgeFields = {\n    content: { type: 'string', minLength: 1 },\n    context: { type: 'object', additionalProperties: true },\n    model: { type: 'string' },\n  } as const\n  doc.components ??= {}\n  doc.components.schemas ??= {}\n  doc.components.schemas.JudgeRequest = {\n    oneOf: [\n      {\n        type: 'object',\n        additionalProperties: false,\n        required: ['rubricName', 'content'],\n        properties: {\n          rubricName: { type: 'string', minLength: 1 },\n          ...commonJudgeFields,\n        },\n      },\n      {\n        type: 'object',\n        additionalProperties: false,\n        required: ['rubric', 'content'],\n        properties: {\n          rubric: rubricRef,\n          ...commonJudgeFields,\n        },\n      },\n    ],\n    description: 'Judge request. Provide exactly one of rubricName or rubric.',\n  }\n  return doc\n}\n","/**\n * stdio RPC transport.\n *\n * For batch / cron use without a running server. The Python client falls\n * back to this when no server is reachable.\n *\n * Protocol (line-delimited JSON over stdin/stdout):\n *   IN:  one JSON object on stdin: {\"method\":\"judge\",\"params\":{...}}\n *   OUT: one JSON object on stdout: {\"result\":{...}} or {\"error\":{...}}\n *\n * One request per process invocation. To pipeline many calls, the client\n * writes JSONL to stdin and reads JSONL from stdout — see batch mode below.\n */\nimport { handleJudge, handleListRubrics, handleVersion, WireError } from './handlers'\nimport { JudgeRequestSchema } from './schemas'\n\ninterface RpcRequest {\n  method: 'judge' | 'listRubrics' | 'version'\n  params?: unknown\n}\n\ninterface RpcSuccess {\n  result: unknown\n}\n\ninterface RpcError {\n  error: { code: string; message: string; details?: unknown }\n}\n\nexport async function dispatchRpc(req: RpcRequest): Promise<RpcSuccess | RpcError> {\n  try {\n    switch (req.method) {\n      case 'judge': {\n        const parsed = JudgeRequestSchema.safeParse(req.params)\n        if (!parsed.success) {\n          return {\n            error: {\n              code: 'validation_error',\n              message: 'params did not match JudgeRequest schema.',\n              details: parsed.error.issues,\n            },\n          }\n        }\n        return { result: await handleJudge(parsed.data) }\n      }\n      case 'listRubrics':\n        return { result: handleListRubrics() }\n      case 'version':\n        return { result: handleVersion() }\n      default:\n        return {\n          error: {\n            code: 'unknown_method',\n            message: `No such method: ${(req as { method: string }).method}`,\n          },\n        }\n    }\n  } catch (err) {\n    if (err instanceof WireError) {\n      return { error: { code: err.code, message: err.message, details: err.details } }\n    }\n    const message = err instanceof Error ? err.message : String(err)\n    return { error: { code: 'internal_error', message } }\n  }\n}\n\n// ── stdin/stdout driver ─────────────────────────────────────────────\n\nasync function readAll(stream: NodeJS.ReadableStream): Promise<string> {\n  const chunks: Buffer[] = []\n  for await (const chunk of stream) {\n    chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk as string))\n  }\n  return Buffer.concat(chunks).toString('utf-8')\n}\n\n/** Read one JSON request from stdin, write one JSON response to stdout. */\nexport async function runRpcOnce(method?: string): Promise<number> {\n  const raw = await readAll(process.stdin)\n  let req: RpcRequest\n  try {\n    const body = JSON.parse(raw)\n    req = method ? { method: method as RpcRequest['method'], params: body } : (body as RpcRequest)\n  } catch (err) {\n    process.stdout.write(\n      `${JSON.stringify({\n        error: {\n          code: 'parse_error',\n          message: `stdin was not valid JSON: ${err instanceof Error ? err.message : String(err)}`,\n        },\n      })}\\n`,\n    )\n    return 1\n  }\n  const out = await dispatchRpc(req)\n  process.stdout.write(`${JSON.stringify(out)}\\n`)\n  return 'error' in out ? 1 : 0\n}\n\n/** Read JSONL requests from stdin, write JSONL responses to stdout. */\nexport async function runRpcBatch(method?: string): Promise<number> {\n  const raw = await readAll(process.stdin)\n  const lines = raw.split('\\n').filter((l) => l.trim().length > 0)\n  let exitCode = 0\n  for (const line of lines) {\n    let req: RpcRequest\n    try {\n      const body = JSON.parse(line)\n      req = method ? { method: method as RpcRequest['method'], params: body } : (body as RpcRequest)\n    } catch (err) {\n      process.stdout.write(\n        `${JSON.stringify({\n          error: {\n            code: 'parse_error',\n            message: `line was not valid JSON: ${err instanceof Error ? err.message : String(err)}`,\n          },\n        })}\\n`,\n      )\n      exitCode = 1\n      continue\n    }\n    const out = await dispatchRpc(req)\n    process.stdout.write(`${JSON.stringify(out)}\\n`)\n    if ('error' in out) exitCode = 1\n  }\n  return exitCode\n}\n","/**\n * HTTP transport for the wire protocol.\n *\n * Hono + @hono/node-server. Every endpoint:\n *   1. Validates the request against its Zod schema.\n *   2. Calls the matching handler in `handlers.ts`.\n *   3. Renders 4xx for `WireError` with structured body, 500 for unexpected.\n *\n * The server has no internal state besides the handler imports — restart\n * costs nothing. Run via `agent-eval serve --port 5005`.\n */\nimport { type ServerType, serve } from '@hono/node-server'\nimport { Hono } from 'hono'\nimport { cors } from 'hono/cors'\n\nimport { handleJudge, handleListRubrics, handleVersion, WireError } from './handlers'\nimport { buildOpenApi } from './openapi'\nimport { JudgeRequestSchema } from './schemas'\n\nconst STARTED_AT = Date.now()\n\nexport function createApp() {\n  const app = new Hono()\n\n  app.use('*', cors())\n\n  app.onError((err, c) => {\n    if (err instanceof WireError) {\n      return c.json(\n        { error: { code: err.code, message: err.message, details: err.details } },\n        err.status as 400 | 404 | 422 | 500,\n      )\n    }\n    // Unexpected — log and return generic 500 without leaking internals.\n    console.error('[agent-eval] unhandled error:', err)\n    return c.json({ error: { code: 'internal_error', message: 'Internal server error.' } }, 500)\n  })\n\n  // ── Health ──\n  app.get('/healthz', (c) =>\n    c.json({ status: 'ok' as const, uptimeSec: (Date.now() - STARTED_AT) / 1000 }),\n  )\n\n  // ── Version ──\n  app.get('/v1/version', (c) => c.json(handleVersion()))\n\n  // ── Rubrics ──\n  app.get('/v1/rubrics', (c) => c.json(handleListRubrics()))\n\n  // ── Judge ──\n  app.post('/v1/judge', async (c) => {\n    const raw = await c.req.json().catch(() => null)\n    if (raw == null) {\n      throw new WireError('validation_error', 'Request body must be JSON.', 400)\n    }\n    const parsed = JudgeRequestSchema.safeParse(raw)\n    if (!parsed.success) {\n      throw new WireError(\n        'validation_error',\n        'Request did not match JudgeRequest schema.',\n        400,\n        parsed.error.issues,\n      )\n    }\n    const result = await handleJudge(parsed.data)\n    return c.json(result)\n  })\n\n  // ── OpenAPI spec ──\n  app.get('/openapi.json', (c) => c.json(buildOpenApi(handleVersion().version)))\n\n  return app\n}\n\nexport interface ServeOptions {\n  /** Default 5005. */\n  port?: number\n  /** Default '127.0.0.1'. Set to '0.0.0.0' to listen on all interfaces. */\n  host?: string\n}\n\nexport function startServer(opts: ServeOptions = {}): ServerType {\n  const app = createApp()\n  const port = opts.port ?? 5005\n  const host = opts.host ?? '127.0.0.1'\n  return serve({ fetch: app.fetch, port, hostname: host }, ({ address, port: actualPort }) => {\n    // eslint-disable-next-line no-console\n    console.log(`[agent-eval] serving on http://${address}:${actualPort}`)\n  })\n}\n"],"mappings":";;;;;AAYA,SAAS,4BAA4B;AACrC,SAAS,SAAS;AAElB,qBAAqB,CAAC;AAIf,IAAM,wBAAwB,EAClC,OAAO;AAAA,EACN,IAAI,EACD,OAAO,EACP,IAAI,CAAC,EACL,SAAS,gFAA2E;AAAA,EACvF,aAAa,EACV,OAAO,EACP,IAAI,CAAC,EACL,SAAS,sEAAsE;AAAA,EAClF,QAAQ,EACL,OAAO,EACP,IAAI,CAAC,EACL,QAAQ,CAAC,EACT,SAAS,gEAAgE;AAAA,EAC5E,KAAK,EAAE,OAAO,EAAE,QAAQ,CAAC,EAAE,SAAS,gDAAgD;AAAA,EACpF,KAAK,EAAE,OAAO,EAAE,QAAQ,CAAC,EAAE,SAAS,gDAAgD;AACtF,CAAC,EACA,QAAQ,iBAAiB;AAErB,IAAM,oBAAoB,EAC9B,OAAO;AAAA,EACN,IAAI,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,mEAA8D;AAAA,EAC7F,aAAa,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,mDAAmD;AAC7F,CAAC,EACA,QAAQ,aAAa;AAIjB,IAAM,eAAe,EACzB,OAAO;AAAA,EACN,MAAM,EACH,OAAO,EACP,IAAI,CAAC,EACL,SAAS,4EAAuE;AAAA,EACnF,aAAa,EACV,OAAO,EACP,IAAI,CAAC,EACL,SAAS,0DAA0D;AAAA,EACtE,cAAc,EACX,OAAO,EACP,IAAI,CAAC,EACL;AAAA,IACC;AAAA,EACF;AAAA,EACF,YAAY,EACT,MAAM,qBAAqB,EAC3B,IAAI,CAAC,EACL,SAAS,+DAA+D;AAAA,EAC3E,cAAc,EACX,MAAM,iBAAiB,EACvB,QAAQ,CAAC,CAAC,EACV,SAAS,iFAAiF;AAAA,EAC7F,MAAM,EACH,MAAM,iBAAiB,EACvB,QAAQ,CAAC,CAAC,EACV,SAAS,uEAAuE;AACrF,CAAC,EACA,QAAQ,QAAQ;AAIZ,IAAM,qBAAqB,EAC/B,OAAO;AAAA,EACN,YAAY,EACT,OAAO,EACP,SAAS,EACT,SAAS,kEAAkE;AAAA,EAC9E,QAAQ,aAAa,SAAS,EAAE;AAAA,IAC9B;AAAA,EACF;AAAA,EACA,SAAS,EACN,OAAO,EACP,IAAI,CAAC,EACL,SAAS,uFAAkF;AAAA,EAC9F,SAAS,EACN,OAAO,EAAE,OAAO,GAAG,EAAE,QAAQ,CAAC,EAC9B,SAAS,EACT;AAAA,IACC;AAAA,EACF;AAAA,EACF,OAAO,EACJ,OAAO,EACP,SAAS,EACT,SAAS,+EAA+E;AAC7F,CAAC,EACA,OAAO,CAAC,MAAM,QAAQ,EAAE,UAAU,MAAM,QAAQ,EAAE,MAAM,GAAG;AAAA,EAC1D,SAAS;AACX,CAAC,EACA,QAAQ,cAAc;AAElB,IAAM,oBAAoB,EAC9B,OAAO;AAAA,EACN,WAAW,EACR,OAAO,EACP,IAAI,CAAC,EACL,IAAI,CAAC,EACL,SAAS,iFAAiF;AAAA,EAC7F,YAAY,EACT,OAAO,EAAE,OAAO,GAAG,EAAE,OAAO,CAAC,EAC7B,SAAS,mDAAmD;AAAA,EAC/D,cAAc,EACX,MAAM,EAAE,OAAO,CAAC,EAChB,QAAQ,CAAC,CAAC,EACV,SAAS,+EAA+E;AAAA,EAC3F,MAAM,EACH,MAAM,EAAE,OAAO,CAAC,EAChB,QAAQ,CAAC,CAAC,EACV,SAAS,8DAA8D;AAAA,EAC1E,WAAW,EACR,OAAO,EACP,SAAS,yEAAyE;AAAA,EACrF,eAAe,EACZ,OAAO,EACP;AAAA,IACC;AAAA,EACF;AAAA,EACF,OAAO,EAAE,OAAO,EAAE,SAAS,yDAAyD;AAAA,EACpF,YAAY,EAAE,OAAO,EAAE,IAAI,EAAE,YAAY,EAAE,SAAS,qCAAqC;AAC3F,CAAC,EACA,QAAQ,aAAa;AAIjB,IAAM,mBAAmB,EAC7B,OAAO;AAAA,EACN,MAAM,EAAE,OAAO,EAAE,SAAS,yCAAyC;AAAA,EACnE,aAAa,EAAE,OAAO,EAAE,SAAS,4BAA4B;AAAA,EAC7D,YAAY,EACT,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,OAAO,GAAG,aAAa,EAAE,OAAO,GAAG,QAAQ,EAAE,OAAO,EAAE,CAAC,CAAC,EAC/E,SAAS,kDAAkD;AAAA,EAC9D,cAAc,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,QAAQ,CAAC,CAAC,EAAE,SAAS,uCAAuC;AAAA,EAC9F,eAAe,EAAE,OAAO,EAAE,SAAS,8DAAyD;AAC9F,CAAC,EACA,QAAQ,YAAY;AAEhB,IAAM,4BAA4B,EACtC,OAAO;AAAA,EACN,SAAS,EAAE,MAAM,gBAAgB;AACnC,CAAC,EACA,QAAQ,qBAAqB;AAIzB,IAAM,wBAAwB,EAClC,OAAO;AAAA,EACN,SAAS,EAAE,OAAO,EAAE,SAAS,qDAAqD;AAAA,EAClF,SAAS,EAAE,OAAO,EAAE,SAAS,0DAA0D;AAAA,EACvF,aAAa,EACV,OAAO,EACP;AAAA,IACC;AAAA,EACF;AAAA,EACF,YAAY,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,SAAS,iCAAiC;AAC5E,CAAC,EACA,QAAQ,iBAAiB;AAErB,IAAM,uBAAuB,EACjC,OAAO;AAAA,EACN,QAAQ,EAAE,QAAQ,IAAI;AAAA,EACtB,WAAW,EAAE,OAAO;AACtB,CAAC,EACA,QAAQ,gBAAgB;AAIpB,IAAM,sBAAsB,EAChC,OAAO;AAAA,EACN,OAAO,EACJ,OAAO;AAAA,IACN,MAAM,EACH,OAAO,EACP;AAAA,MACC;AAAA,IACF;AAAA,IACF,SAAS,EAAE,OAAO,EAAE,SAAS,yBAAyB;AAAA,IACtD,SAAS,EAAE,QAAQ,EAAE,SAAS,EAAE,SAAS,6BAA6B;AAAA,EACxE,CAAC,EACA,SAAS,+DAA+D;AAC7E,CAAC,EACA,QAAQ,eAAe;AAoBnB,IAAM,eAAe;AAMrB,SAAS,WAAW,QAAwB;AACjD,QAAM,SAAS,gBAAgB,MAAM;AACrC,MAAI,IAAI;AACR,WAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK;AACtC,QAAK,IAAI,KAAM,OAAO,WAAW,CAAC;AAAA,EACpC;AAEA,SAAO,GAAG,OAAO,IAAI,KAAK,MAAM,GAAG,SAAS,EAAE,EAAE,SAAS,GAAG,GAAG,CAAC;AAClE;AAEA,SAAS,gBAAgB,OAAwB;AAC/C,MAAI,MAAM,QAAQ,KAAK,EAAG,QAAO,IAAI,MAAM,IAAI,CAAC,SAAS,gBAAgB,IAAI,CAAC,EAAE,KAAK,GAAG,CAAC;AACzF,MAAI,SAAS,OAAO,UAAU,UAAU;AACtC,UAAM,UAAU,OAAO,QAAQ,KAAgC,EAC5D,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,cAAc,CAAC,CAAC,EACrC,IAAI,CAAC,CAAC,KAAK,IAAI,MAAM,GAAG,KAAK,UAAU,GAAG,CAAC,IAAI,gBAAgB,IAAI,CAAC,EAAE;AACzE,WAAO,IAAI,QAAQ,KAAK,GAAG,CAAC;AAAA,EAC9B;AACA,SAAO,KAAK,UAAU,KAAK;AAC7B;;;AC5NA,IAAM,YAAoB;AAAA,EACxB,MAAM;AAAA,EACN,aACE;AAAA,EACF,cAAc;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAuBd,YAAY;AAAA,IACV;AAAA,MACE,IAAI;AAAA,MACJ,aAAa;AAAA,MACb,QAAQ;AAAA,MACR,KAAK;AAAA,MACL,KAAK;AAAA,IACP;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,aAAa;AAAA,MACb,QAAQ;AAAA,MACR,KAAK;AAAA,MACL,KAAK;AAAA,IACP;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,aAAa;AAAA,MACb,QAAQ;AAAA,MACR,KAAK;AAAA,MACL,KAAK;AAAA,IACP;AAAA,EACF;AAAA,EACA,cAAc;AAAA,IACZ,EAAE,IAAI,cAAc,aAAa,qCAAqC;AAAA,IACtE,EAAE,IAAI,kBAAkB,aAAa,sCAAsC;AAAA,IAC3E,EAAE,IAAI,eAAe,aAAa,oCAAoC;AAAA,IACtE,EAAE,IAAI,WAAW,aAAa,kCAAkC;AAAA,IAChE,EAAE,IAAI,mBAAmB,aAAa,qCAAqC;AAAA,IAC3E,EAAE,IAAI,WAAW,aAAa,oCAAoC;AAAA,IAClE,EAAE,IAAI,eAAe,aAAa,uCAAuC;AAAA,EAC3E;AAAA,EACA,MAAM;AAAA,IACJ,EAAE,IAAI,sBAAsB,aAAa,qCAAqC;AAAA,IAC9E,EAAE,IAAI,iBAAiB,aAAa,4BAA4B;AAAA,IAChE,EAAE,IAAI,0BAA0B,aAAa,wBAAwB;AAAA,IACrE,EAAE,IAAI,kBAAkB,aAAa,oCAAoC;AAAA,EAC3E;AACF;AAIO,IAAM,kBAA0C;AAAA,EACrD,aAAa;AACf;AAGO,SAAS,iBAAiB,MAAkC;AACjE,SAAO,gBAAgB,IAAI;AAC7B;AAGO,SAAS,qBAAqB;AACnC,SAAO,OAAO,OAAO,eAAe,EAAE,IAAI,CAAC,OAAO;AAAA,IAChD,MAAM,EAAE;AAAA,IACR,aAAa,EAAE;AAAA,IACf,YAAY,EAAE,WAAW,IAAI,CAAC,OAAO;AAAA,MACnC,IAAI,EAAE;AAAA,MACN,aAAa,EAAE;AAAA,MACf,QAAQ,EAAE;AAAA,IACZ,EAAE;AAAA,IACF,cAAc,EAAE,aAAa,IAAI,CAAC,MAAM,EAAE,EAAE;AAAA,IAC5C,eAAe,WAAW,CAAC;AAAA,EAC7B,EAAE;AACJ;;;ACiHA,SAAS,oBAAoB;AAC7B,SAAS,SAAS,eAAe;AACjC,SAAS,qBAAqB;AA9MvB,IAAM,YAAN,cAAwB,MAAM;AAAA,EACnC,YACkB,MAChB,SACgB,SAAiB,KACjB,SAChB;AACA,UAAM,OAAO;AALG;AAEA;AACA;AAGhB,SAAK,OAAO;AAAA,EACd;AAAA,EAPkB;AAAA,EAEA;AAAA,EACA;AAKpB;AAKA,SAAS,kBAAkB,QAAgB;AACzC,SAAO;AAAA,IACL,MAAM;AAAA,IACN,QAAQ;AAAA,MACN,MAAM;AAAA,MACN,sBAAsB;AAAA,MACtB,YAAY;AAAA,QACV,YAAY;AAAA,UACV,MAAM;AAAA,UACN,sBAAsB;AAAA,UACtB,YAAY,OAAO;AAAA,YACjB,OAAO,WAAW,IAAI,CAAC,MAAM;AAAA,cAC3B,EAAE;AAAA,cACF,EAAE,MAAM,UAAU,SAAS,EAAE,KAAK,SAAS,EAAE,IAAI;AAAA,YACnD,CAAC;AAAA,UACH;AAAA,UACA,UAAU,OAAO,WAAW,IAAI,CAAC,MAAM,EAAE,EAAE;AAAA,QAC7C;AAAA,QACA,cAAc;AAAA,UACZ,MAAM;AAAA,UACN,OAAO,EAAE,MAAM,UAAU,MAAM,OAAO,aAAa,IAAI,CAAC,MAAM,EAAE,EAAE,EAAE;AAAA,QACtE;AAAA,QACA,MAAM;AAAA,UACJ,MAAM;AAAA,UACN,OAAO,EAAE,MAAM,UAAU,MAAM,OAAO,KAAK,IAAI,CAAC,MAAM,EAAE,EAAE,EAAE;AAAA,QAC9D;AAAA,QACA,WAAW,EAAE,MAAM,SAAS;AAAA,MAC9B;AAAA,MACA,UAAU,CAAC,cAAc,WAAW;AAAA,IACtC;AAAA,EACF;AACF;AASA,SAAS,oBAAoB,OAAgB,QAA6B;AACxE,MAAI,CAAC,SAAS,OAAO,UAAU,UAAU;AACvC,UAAM,IAAI,UAAU,eAAe,oCAAoC,KAAK,KAAK;AAAA,EACnF;AACA,QAAM,MAAM;AACZ,QAAM,gBAAgB,IAAI;AAC1B,MAAI,CAAC,iBAAiB,OAAO,kBAAkB,YAAY,MAAM,QAAQ,aAAa,GAAG;AACvF,UAAM,IAAI,UAAU,eAAe,wCAAwC,KAAK,KAAK;AAAA,EACvF;AAEA,QAAM,aAAqC,CAAC;AAC5C,QAAM,kBAAkB;AACxB,aAAW,OAAO,OAAO,YAAY;AACnC,UAAM,QAAQ,gBAAgB,IAAI,EAAE;AACpC,QACE,OAAO,UAAU,YACjB,CAAC,OAAO,SAAS,KAAK,KACtB,QAAQ,IAAI,OACZ,QAAQ,IAAI,KACZ;AACA,YAAM,IAAI;AAAA,QACR;AAAA,QACA,+CAA+C,IAAI,EAAE;AAAA,QACrD;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,eAAW,IAAI,EAAE,IAAI;AAAA,EACvB;AAEA,QAAM,kBAAkB,IAAI,IAAI,OAAO,aAAa,IAAI,CAAC,SAAS,KAAK,EAAE,CAAC;AAC1E,QAAM,cAAc,IAAI,IAAI,OAAO,KAAK,IAAI,CAAC,QAAQ,IAAI,EAAE,CAAC;AAC5D,QAAM,eAAe,gBAAgB,IAAI,cAAc,iBAAiB,gBAAgB,KAAK;AAC7F,QAAM,OAAO,gBAAgB,IAAI,MAAM,aAAa,QAAQ,KAAK;AACjE,MAAI,OAAO,IAAI,cAAc,YAAY,IAAI,UAAU,KAAK,EAAE,WAAW,GAAG;AAC1E,UAAM,IAAI,UAAU,eAAe,qCAAqC,KAAK,KAAK;AAAA,EACpF;AAEA,SAAO,EAAE,YAAY,cAAc,MAAM,WAAW,IAAI,UAAU;AACpE;AAEA,SAAS,gBACP,KACA,SACA,OACA,UACU;AACV,MAAI,QAAQ,OAAW,QAAO,CAAC;AAC/B,MAAI,CAAC,MAAM,QAAQ,GAAG,GAAG;AACvB,UAAM,IAAI,UAAU,eAAe,4BAA4B,KAAK,KAAK,KAAK,QAAQ;AAAA,EACxF;AACA,QAAM,MAAgB,CAAC;AACvB,aAAW,QAAQ,KAAK;AACtB,QAAI,OAAO,SAAS,YAAY,CAAC,QAAQ,IAAI,IAAI,GAAG;AAClD,YAAM,IAAI;AAAA,QACR;AAAA,QACA,0BAA0B,KAAK,QAAQ,OAAO,IAAI,CAAC;AAAA,QACnD;AAAA,QACA;AAAA,MACF;AAAA,IACF;AACA,QAAI,KAAK,IAAI;AAAA,EACf;AACA,SAAO;AACT;AAEA,SAAS,eAAe,YAAoC,QAAwB;AAClF,MAAI,WAAW;AACf,MAAI,cAAc;AAClB,aAAW,OAAO,OAAO,YAAY;AACnC,UAAM,MAAM,WAAW,IAAI,EAAE,KAAK;AAClC,UAAM,QAAQ,IAAI,MAAM,IAAI,OAAO;AACnC,UAAM,aAAa,KAAK,IAAI,GAAG,KAAK,IAAI,IAAI,MAAM,IAAI,OAAO,KAAK,CAAC;AACnE,gBAAY,aAAa,IAAI;AAC7B,mBAAe,IAAI;AAAA,EACrB;AACA,SAAO,cAAc,IAAI,WAAW,cAAc;AACpD;AAEA,SAAS,iBAAiB,SAAiB,SAA0B;AACnE,QAAM,MAAM,WAAW,OAAO,KAAK,OAAiB,EAAE,SAAS,KAAK,UAAU,OAAO,IAAI;AACzF,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA;AAAA,IACA,MAAM,yCAAyC;AAAA,IAC/C,MAAM,MAAM;AAAA,EACd,EACG,OAAO,OAAO,EACd,KAAK,IAAI;AACd;AAEA,IAAM,sBAAsB;AAE5B,eAAsB,YAAY,KAAyC;AAEzE,MAAI;AACJ,MAAI,IAAI,YAAY;AAClB,UAAM,QAAQ,iBAAiB,IAAI,UAAU;AAC7C,QAAI,CAAC,OAAO;AACV,YAAM,IAAI,UAAU,oBAAoB,6BAA6B,IAAI,UAAU,MAAM,GAAG;AAAA,IAC9F;AACA,aAAS;AAAA,EACX,WAAW,IAAI,QAAQ;AACrB,aAAS,IAAI;AAAA,EACf,OAAO;AAEL,UAAM,IAAI,UAAU,oBAAoB,4CAA4C,GAAG;AAAA,EACzF;AAEA,QAAM,YAAY,KAAK,IAAI;AAC3B,QAAM,QAAQ,IAAI,SAAS;AAE3B,QAAM,EAAE,OAAO,OAAO,IAAI,MAAM,YAAyB;AAAA,IACvD;AAAA,IACA,UAAU;AAAA,MACR,EAAE,MAAM,UAAU,SAAS,OAAO,aAAa;AAAA,MAC/C,EAAE,MAAM,QAAQ,SAAS,iBAAiB,IAAI,SAAS,IAAI,OAAO,EAAE;AAAA,IACtE;AAAA,IACA,YAAY,kBAAkB,MAAM;AAAA,IACpC,aAAa;AAAA,IACb,WAAW;AAAA,EACb,CAAC;AAED,QAAM,SAAS,oBAAoB,OAAO,MAAM;AAEhD,QAAM,YAAY,eAAe,OAAO,YAAY,MAAM;AAC1D,QAAM,aAAa,KAAK,IAAI,IAAI;AAEhC,SAAO;AAAA,IACL;AAAA,IACA,YAAY,OAAO;AAAA,IACnB,cAAc,OAAO,gBAAgB,CAAC;AAAA,IACtC,MAAM,OAAO,QAAQ,CAAC;AAAA,IACtB,WAAW,OAAO;AAAA,IAClB,eAAe,WAAW,MAAM;AAAA,IAChC,OAAO,OAAO;AAAA,IACd;AAAA,EACF;AACF;AAIO,SAAS,oBAAyC;AACvD,SAAO,EAAE,SAAS,mBAAmB,EAAE;AACzC;AAQA,IAAI;AAEJ,SAAS,qBAA6B;AACpC,MAAI,eAAgB,QAAO;AAG3B,QAAM,OAAO,QAAQ,cAAc,YAAY,GAAG,CAAC;AACnD,QAAM,aAAa;AAAA,IACjB,QAAQ,MAAM,MAAM,MAAM,cAAc;AAAA;AAAA,IACxC,QAAQ,MAAM,MAAM,cAAc;AAAA;AAAA,EACpC;AACA,aAAW,QAAQ,YAAY;AAC7B,QAAI;AACF,YAAM,MAAM,KAAK,MAAM,aAAa,MAAM,OAAO,CAAC;AAClD,UAAI,IAAI,SAAS;AACf,yBAAiB,IAAI;AACrB,eAAO,IAAI;AAAA,MACb;AAAA,IACF,QAAQ;AAAA,IAER;AAAA,EACF;AACA,SAAO;AACT;AAEO,SAAS,gBAAiC;AAC/C,SAAO;AAAA,IACL,SAAS;AAAA,IACT,SAAS,mBAAmB;AAAA,IAC5B,aAAa;AAAA,IACb,YAAY,CAAC,SAAS,eAAe,SAAS;AAAA,EAChD;AACF;;;AC7PA,SAAS,iBAAiB,2BAA2B;AAa9C,SAAS,aAAa,gBAAuC;AAClE,QAAM,WAAW,IAAI,gBAAgB;AAGrC,WAAS,SAAS,gBAAgB,kBAAkB;AACpD,WAAS,SAAS,eAAe,iBAAiB;AAClD,WAAS,SAAS,uBAAuB,yBAAyB;AAClE,WAAS,SAAS,mBAAmB,qBAAqB;AAC1D,WAAS,SAAS,kBAAkB,oBAAoB;AACxD,WAAS,SAAS,iBAAiB,mBAAmB;AAGtD,WAAS,aAAa;AAAA,IACpB,QAAQ;AAAA,IACR,MAAM;AAAA,IACN,SAAS;AAAA,IACT,aACE;AAAA,IACF,SAAS;AAAA,MACP,MAAM;AAAA,QACJ,SAAS;AAAA,UACP,oBAAoB,EAAE,QAAQ,mBAAmB;AAAA,QACnD;AAAA,MACF;AAAA,IACF;AAAA,IACA,WAAW;AAAA,MACT,KAAK;AAAA,QACH,aAAa;AAAA,QACb,SAAS,EAAE,oBAAoB,EAAE,QAAQ,kBAAkB,EAAE;AAAA,MAC/D;AAAA,MACA,KAAK;AAAA,QACH,aAAa;AAAA,QACb,SAAS,EAAE,oBAAoB,EAAE,QAAQ,oBAAoB,EAAE;AAAA,MACjE;AAAA,MACA,KAAK;AAAA,QACH,aAAa;AAAA,QACb,SAAS,EAAE,oBAAoB,EAAE,QAAQ,oBAAoB,EAAE;AAAA,MACjE;AAAA,MACA,KAAK;AAAA,QACH,aAAa;AAAA,QACb,SAAS,EAAE,oBAAoB,EAAE,QAAQ,oBAAoB,EAAE;AAAA,MACjE;AAAA,IACF;AAAA,EACF,CAAC;AAED,WAAS,aAAa;AAAA,IACpB,QAAQ;AAAA,IACR,MAAM;AAAA,IACN,SAAS;AAAA,IACT,aACE;AAAA,IACF,WAAW;AAAA,MACT,KAAK;AAAA,QACH,aAAa;AAAA,QACb,SAAS,EAAE,oBAAoB,EAAE,QAAQ,0BAA0B,EAAE;AAAA,MACvE;AAAA,IACF;AAAA,EACF,CAAC;AAED,WAAS,aAAa;AAAA,IACpB,QAAQ;AAAA,IACR,MAAM;AAAA,IACN,SAAS;AAAA,IACT,aAAa;AAAA,IACb,WAAW;AAAA,MACT,KAAK;AAAA,QACH,aAAa;AAAA,QACb,SAAS,EAAE,oBAAoB,EAAE,QAAQ,sBAAsB,EAAE;AAAA,MACnE;AAAA,IACF;AAAA,EACF,CAAC;AAED,WAAS,aAAa;AAAA,IACpB,QAAQ;AAAA,IACR,MAAM;AAAA,IACN,SAAS;AAAA,IACT,WAAW;AAAA,MACT,KAAK;AAAA,QACH,aAAa;AAAA,QACb,SAAS,EAAE,oBAAoB,EAAE,QAAQ,qBAAqB,EAAE;AAAA,MAClE;AAAA,IACF;AAAA,EACF,CAAC;AAED,QAAM,YAAY,IAAI,oBAAoB,SAAS,WAAW;AAC9D,QAAM,MAAM,UAAU,iBAAiB;AAAA,IACrC,SAAS;AAAA,IACT,MAAM;AAAA,MACJ,OAAO;AAAA,MACP,SAAS;AAAA,MACT,aAAa;AAAA;AAAA,yBAEM,YAAY;AAAA,MAC/B,SAAS,EAAE,MAAM,kBAAkB,KAAK,+CAA+C;AAAA,MACvF,SAAS,EAAE,MAAM,MAAM;AAAA,IACzB;AAAA,IACA,SAAS,CAAC,EAAE,KAAK,yBAAyB,aAAa,yBAAyB,CAAC;AAAA,EACnF,CAAC;AACD,QAAM,YAAY,EAAE,MAAM,8BAA8B;AACxD,QAAM,oBAAoB;AAAA,IACxB,SAAS,EAAE,MAAM,UAAU,WAAW,EAAE;AAAA,IACxC,SAAS,EAAE,MAAM,UAAU,sBAAsB,KAAK;AAAA,IACtD,OAAO,EAAE,MAAM,SAAS;AAAA,EAC1B;AACA,MAAI,eAAe,CAAC;AACpB,MAAI,WAAW,YAAY,CAAC;AAC5B,MAAI,WAAW,QAAQ,eAAe;AAAA,IACpC,OAAO;AAAA,MACL;AAAA,QACE,MAAM;AAAA,QACN,sBAAsB;AAAA,QACtB,UAAU,CAAC,cAAc,SAAS;AAAA,QAClC,YAAY;AAAA,UACV,YAAY,EAAE,MAAM,UAAU,WAAW,EAAE;AAAA,UAC3C,GAAG;AAAA,QACL;AAAA,MACF;AAAA,MACA;AAAA,QACE,MAAM;AAAA,QACN,sBAAsB;AAAA,QACtB,UAAU,CAAC,UAAU,SAAS;AAAA,QAC9B,YAAY;AAAA,UACV,QAAQ;AAAA,UACR,GAAG;AAAA,QACL;AAAA,MACF;AAAA,IACF;AAAA,IACA,aAAa;AAAA,EACf;AACA,SAAO;AACT;;;AC7HA,eAAsB,YAAY,KAAiD;AACjF,MAAI;AACF,YAAQ,IAAI,QAAQ;AAAA,MAClB,KAAK,SAAS;AACZ,cAAM,SAAS,mBAAmB,UAAU,IAAI,MAAM;AACtD,YAAI,CAAC,OAAO,SAAS;AACnB,iBAAO;AAAA,YACL,OAAO;AAAA,cACL,MAAM;AAAA,cACN,SAAS;AAAA,cACT,SAAS,OAAO,MAAM;AAAA,YACxB;AAAA,UACF;AAAA,QACF;AACA,eAAO,EAAE,QAAQ,MAAM,YAAY,OAAO,IAAI,EAAE;AAAA,MAClD;AAAA,MACA,KAAK;AACH,eAAO,EAAE,QAAQ,kBAAkB,EAAE;AAAA,MACvC,KAAK;AACH,eAAO,EAAE,QAAQ,cAAc,EAAE;AAAA,MACnC;AACE,eAAO;AAAA,UACL,OAAO;AAAA,YACL,MAAM;AAAA,YACN,SAAS,mBAAoB,IAA2B,MAAM;AAAA,UAChE;AAAA,QACF;AAAA,IACJ;AAAA,EACF,SAAS,KAAK;AACZ,QAAI,eAAe,WAAW;AAC5B,aAAO,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,SAAS,IAAI,SAAS,SAAS,IAAI,QAAQ,EAAE;AAAA,IACjF;AACA,UAAM,UAAU,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAC/D,WAAO,EAAE,OAAO,EAAE,MAAM,kBAAkB,QAAQ,EAAE;AAAA,EACtD;AACF;AAIA,eAAe,QAAQ,QAAgD;AACrE,QAAM,SAAmB,CAAC;AAC1B,mBAAiB,SAAS,QAAQ;AAChC,WAAO,KAAK,OAAO,SAAS,KAAK,IAAI,QAAQ,OAAO,KAAK,KAAe,CAAC;AAAA,EAC3E;AACA,SAAO,OAAO,OAAO,MAAM,EAAE,SAAS,OAAO;AAC/C;AAGA,eAAsB,WAAW,QAAkC;AACjE,QAAM,MAAM,MAAM,QAAQ,QAAQ,KAAK;AACvC,MAAI;AACJ,MAAI;AACF,UAAM,OAAO,KAAK,MAAM,GAAG;AAC3B,UAAM,SAAS,EAAE,QAAwC,QAAQ,KAAK,IAAK;AAAA,EAC7E,SAAS,KAAK;AACZ,YAAQ,OAAO;AAAA,MACb,GAAG,KAAK,UAAU;AAAA,QAChB,OAAO;AAAA,UACL,MAAM;AAAA,UACN,SAAS,6BAA6B,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG,CAAC;AAAA,QACxF;AAAA,MACF,CAAC,CAAC;AAAA;AAAA,IACJ;AACA,WAAO;AAAA,EACT;AACA,QAAM,MAAM,MAAM,YAAY,GAAG;AACjC,UAAQ,OAAO,MAAM,GAAG,KAAK,UAAU,GAAG,CAAC;AAAA,CAAI;AAC/C,SAAO,WAAW,MAAM,IAAI;AAC9B;AAGA,eAAsB,YAAY,QAAkC;AAClE,QAAM,MAAM,MAAM,QAAQ,QAAQ,KAAK;AACvC,QAAM,QAAQ,IAAI,MAAM,IAAI,EAAE,OAAO,CAAC,MAAM,EAAE,KAAK,EAAE,SAAS,CAAC;AAC/D,MAAI,WAAW;AACf,aAAW,QAAQ,OAAO;AACxB,QAAI;AACJ,QAAI;AACF,YAAM,OAAO,KAAK,MAAM,IAAI;AAC5B,YAAM,SAAS,EAAE,QAAwC,QAAQ,KAAK,IAAK;AAAA,IAC7E,SAAS,KAAK;AACZ,cAAQ,OAAO;AAAA,QACb,GAAG,KAAK,UAAU;AAAA,UAChB,OAAO;AAAA,YACL,MAAM;AAAA,YACN,SAAS,4BAA4B,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG,CAAC;AAAA,UACvF;AAAA,QACF,CAAC,CAAC;AAAA;AAAA,MACJ;AACA,iBAAW;AACX;AAAA,IACF;AACA,UAAM,MAAM,MAAM,YAAY,GAAG;AACjC,YAAQ,OAAO,MAAM,GAAG,KAAK,UAAU,GAAG,CAAC;AAAA,CAAI;AAC/C,QAAI,WAAW,IAAK,YAAW;AAAA,EACjC;AACA,SAAO;AACT;;;ACnHA,SAA0B,aAAa;AACvC,SAAS,YAAY;AACrB,SAAS,YAAY;AAMrB,IAAM,aAAa,KAAK,IAAI;AAErB,SAAS,YAAY;AAC1B,QAAM,MAAM,IAAI,KAAK;AAErB,MAAI,IAAI,KAAK,KAAK,CAAC;AAEnB,MAAI,QAAQ,CAAC,KAAK,MAAM;AACtB,QAAI,eAAe,WAAW;AAC5B,aAAO,EAAE;AAAA,QACP,EAAE,OAAO,EAAE,MAAM,IAAI,MAAM,SAAS,IAAI,SAAS,SAAS,IAAI,QAAQ,EAAE;AAAA,QACxE,IAAI;AAAA,MACN;AAAA,IACF;AAEA,YAAQ,MAAM,iCAAiC,GAAG;AAClD,WAAO,EAAE,KAAK,EAAE,OAAO,EAAE,MAAM,kBAAkB,SAAS,yBAAyB,EAAE,GAAG,GAAG;AAAA,EAC7F,CAAC;AAGD,MAAI;AAAA,IAAI;AAAA,IAAY,CAAC,MACnB,EAAE,KAAK,EAAE,QAAQ,MAAe,YAAY,KAAK,IAAI,IAAI,cAAc,IAAK,CAAC;AAAA,EAC/E;AAGA,MAAI,IAAI,eAAe,CAAC,MAAM,EAAE,KAAK,cAAc,CAAC,CAAC;AAGrD,MAAI,IAAI,eAAe,CAAC,MAAM,EAAE,KAAK,kBAAkB,CAAC,CAAC;AAGzD,MAAI,KAAK,aAAa,OAAO,MAAM;AACjC,UAAM,MAAM,MAAM,EAAE,IAAI,KAAK,EAAE,MAAM,MAAM,IAAI;AAC/C,QAAI,OAAO,MAAM;AACf,YAAM,IAAI,UAAU,oBAAoB,8BAA8B,GAAG;AAAA,IAC3E;AACA,UAAM,SAAS,mBAAmB,UAAU,GAAG;AAC/C,QAAI,CAAC,OAAO,SAAS;AACnB,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,QACA;AAAA,QACA,OAAO,MAAM;AAAA,MACf;AAAA,IACF;AACA,UAAM,SAAS,MAAM,YAAY,OAAO,IAAI;AAC5C,WAAO,EAAE,KAAK,MAAM;AAAA,EACtB,CAAC;AAGD,MAAI,IAAI,iBAAiB,CAAC,MAAM,EAAE,KAAK,aAAa,cAAc,EAAE,OAAO,CAAC,CAAC;AAE7E,SAAO;AACT;AASO,SAAS,YAAY,OAAqB,CAAC,GAAe;AAC/D,QAAM,MAAM,UAAU;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO,MAAM,EAAE,OAAO,IAAI,OAAO,MAAM,UAAU,KAAK,GAAG,CAAC,EAAE,SAAS,MAAM,WAAW,MAAM;AAE1F,YAAQ,IAAI,kCAAkC,OAAO,IAAI,UAAU,EAAE;AAAA,EACvE,CAAC;AACH;","names":[]}

package/dist/{chunk-5IIQKMD5.js → chunk-TVVP3ZZQ.js}

@@ -37,7 +37,7 @@ var TraceEmitter = class {
    * to anchor to don't have to invent placeholder strings at the call site.
    */
   async startRun(run) {
-    const scenarioId = run.scenarioId ?? run.layer ?? run.tags?.["kind"] ?? "runtime";
+    const scenarioId = run.scenarioId ?? run.layer ?? run.tags?.kind ?? "runtime";
     const full = {
       ...run,
       scenarioId,
@@ -60,7 +60,13 @@ var TraceEmitter = class {
       status: "aborted",
       outcome
     });
-    await this.runHooks({ runId: this._runId, emitter: this, store: this.store, outcome, status: "aborted" });
+    await this.runHooks({
+      runId: this._runId,
+      emitter: this,
+      store: this.store,
+      outcome,
+      status: "aborted"
+    });
   }
   async runHooks(ctx) {
     for (const hook of this.hooks) {
@@ -104,7 +110,11 @@ var TraceEmitter = class {
       span,
       end: async (patch) => {
         const endedAt = this.now();
-        await this.store.updateSpan(span.spanId, { endedAt, status: "ok", ...patch });
+        await this.store.updateSpan(span.spanId, {
+          endedAt,
+          status: "ok",
+          ...patch
+        });
         this.pop(span.spanId);
       },
       fail: async (error, patch) => {
@@ -233,4 +243,4 @@ export {
   TraceEmitter,
   llmSpanFromProvider
 };
-//# sourceMappingURL=chunk-5IIQKMD5.js.map
+//# sourceMappingURL=chunk-TVVP3ZZQ.js.map

package/dist/chunk-TVVP3ZZQ.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/trace/emitter.ts"],"sourcesContent":["/**\n * TraceEmitter — hierarchical span builder that auto-parents using an\n * internal stack. One emitter per Run; emitters do NOT share state.\n *\n * Convenience methods (`llm`, `tool`, `retrieval`, `judge`, `sandbox`)\n * return a `SpanHandle` with `.end()` / `.fail()` so callers don't\n * have to thread spanIds manually. For async workflows that can't use\n * the stack (e.g. fan-out parallel calls), pass `parentSpanId`\n * explicitly.\n */\n\nimport type {\n  Artifact,\n  BudgetLedgerEntry,\n  EventKind,\n  JudgeSpan,\n  LlmSpan,\n  Message,\n  RetrievalSpan,\n  Run,\n  RunOutcome,\n  SandboxSpan,\n  Span,\n  SpanKind,\n  ToolSpan,\n  TraceEvent,\n} from './schema'\nimport type { TraceStore } from './store'\n\nexport interface SpanHandle<S extends Span = Span> {\n  span: S\n  end(patch?: Partial<S>): Promise<void>\n  fail(error: string | Error, patch?: Partial<S>): Promise<void>\n}\n\nexport interface RunCompleteHookContext {\n  runId: string\n  emitter: TraceEmitter\n  store: TraceStore\n  /** Outcome the caller passed to `endRun` (undefined for `abortRun`). */\n  outcome?: RunOutcome\n  /** Final run status. */\n  status: 'completed' | 'failed' | 'aborted'\n}\n\nexport type RunCompleteHook = (ctx: RunCompleteHookContext) => Promise<void> | void\n\nexport interface TraceEmitterOptions {\n  runId?: string\n  /** Inject a clock for deterministic tests. */\n  now?: () => number\n  /** Inject an id generator for deterministic tests. */\n  id?: () => string\n  /**\n   * Hooks fired after `endRun` / `abortRun` writes the final run state.\n   * Designed for trace-analyst auto-execution, integrity assertions, and\n   * outbound notifications. Hooks run sequentially in the order supplied.\n   *\n   * By default a hook that throws is swallowed and logged as a `note` event\n   * on the run — auto-orchestration must not crash the underlying flow.\n   * Set `hookErrors: 'throw'` to propagate.\n   */\n  onRunComplete?: RunCompleteHook[]\n  /** `'swallow'` (default) | `'throw'`. */\n  hookErrors?: 'swallow' | 'throw'\n}\n\nexport class TraceEmitter {\n  private store: TraceStore\n  private stack: string[] = []\n  private _runId: string\n  private now: () => number\n  private id: () => string\n  private hooks: RunCompleteHook[]\n  private hookErrors: 'swallow' | 'throw'\n\n  constructor(store: TraceStore, options: TraceEmitterOptions = {}) {\n    this.store = store\n    this.now = options.now ?? (() => Date.now())\n    this.id = options.id ?? (() => cryptoRandomId())\n    this._runId = options.runId ?? this.id()\n    this.hooks = options.onRunComplete ?? []\n    this.hookErrors = options.hookErrors ?? 'swallow'\n  }\n\n  get runId(): string {\n    return this._runId\n  }\n\n  get traceStore(): TraceStore {\n    return this.store\n  }\n\n  /** Append a hook after construction (e.g. attach the trace analyst). */\n  addRunCompleteHook(hook: RunCompleteHook): void {\n    this.hooks.push(hook)\n  }\n\n  // ── Run lifecycle ──────────────────────────────────────────────────\n\n  /**\n   * Begin a Run.\n   *\n   * `scenarioId` is required on the persisted Run shape — every Run downstream\n   * gets a non-empty scenarioId so filters and aggregations stay simple — but\n   * the INPUT here accepts it as optional. When omitted, startRun substitutes\n   * a sensible default (`run.layer ?? run.tags?.['kind'] ?? 'runtime'`) so\n   * runtime / operator / meta-eval runs that have no curated-scenario corpus\n   * to anchor to don't have to invent placeholder strings at the call site.\n   */\n  async startRun(\n    run: Omit<Run, 'runId' | 'scenarioId' | 'startedAt' | 'status'> & { scenarioId?: string },\n  ): Promise<Run> {\n    const scenarioId = run.scenarioId ?? run.layer ?? run.tags?.kind ?? 'runtime'\n    const full: Run = {\n      ...run,\n      scenarioId,\n      runId: this._runId,\n      startedAt: this.now(),\n      status: 'running',\n    }\n    await this.store.appendRun(full)\n    return full\n  }\n\n  async endRun(outcome?: RunOutcome): Promise<void> {\n    const status: 'completed' | 'failed' = outcome?.pass === false ? 'failed' : 'completed'\n    await this.store.updateRun(this._runId, { endedAt: this.now(), status, outcome })\n    await this.runHooks({ runId: this._runId, emitter: this, store: this.store, outcome, status })\n  }\n\n  async abortRun(reason: string): Promise<void> {\n    const outcome = { pass: false, notes: reason }\n    await this.store.updateRun(this._runId, {\n      endedAt: this.now(),\n      status: 'aborted',\n      outcome,\n    })\n    await this.runHooks({\n      runId: this._runId,\n      emitter: this,\n      store: this.store,\n      outcome,\n      status: 'aborted',\n    })\n  }\n\n  private async runHooks(ctx: RunCompleteHookContext): Promise<void> {\n    for (const hook of this.hooks) {\n      try {\n        await hook(ctx)\n      } catch (err) {\n        if (this.hookErrors === 'throw') throw err\n        try {\n          await this.store.appendEvent({\n            eventId: this.id(),\n            runId: this._runId,\n            kind: 'log',\n            timestamp: this.now(),\n            payload: {\n              source: 'run_complete_hook',\n              error: err instanceof Error ? err.message : String(err),\n            },\n          })\n        } catch {\n          // best-effort\n        }\n      }\n    }\n  }\n\n  // ── Generic span ───────────────────────────────────────────────────\n\n  async span<S extends Span = Span>(\n    init: {\n      kind: SpanKind\n      name: string\n      parentSpanId?: string\n      attributes?: Record<string, unknown>\n    } & Partial<Omit<S, 'spanId' | 'runId' | 'startedAt' | 'kind' | 'name'>>,\n  ): Promise<SpanHandle<S>> {\n    const spanId = this.id()\n    const parent = init.parentSpanId ?? this.stack[this.stack.length - 1]\n    const span = {\n      spanId,\n      parentSpanId: parent,\n      runId: this._runId,\n      startedAt: this.now(),\n      ...init,\n    } as unknown as S\n    await this.store.appendSpan(span)\n    this.stack.push(spanId)\n    return this.handle<S>(span)\n  }\n\n  private handle<S extends Span>(span: S): SpanHandle<S> {\n    return {\n      span,\n      end: async (patch?: Partial<S>) => {\n        const endedAt = this.now()\n        await this.store.updateSpan(span.spanId, {\n          endedAt,\n          status: 'ok',\n          ...patch,\n        } as Partial<Span>)\n        this.pop(span.spanId)\n      },\n      fail: async (error: string | Error, patch?: Partial<S>) => {\n        const endedAt = this.now()\n        const errStr = error instanceof Error ? error.message : error\n        await this.store.updateSpan(span.spanId, {\n          endedAt,\n          status: 'error',\n          error: errStr,\n          ...patch,\n        } as Partial<Span>)\n        this.pop(span.spanId)\n      },\n    }\n  }\n\n  private pop(spanId: string): void {\n    const idx = this.stack.lastIndexOf(spanId)\n    if (idx >= 0) this.stack.splice(idx, 1)\n  }\n\n  // ── Typed span conveniences ────────────────────────────────────────\n\n  llm(\n    init: Omit<LlmSpan, 'spanId' | 'runId' | 'kind' | 'startedAt'>,\n  ): Promise<SpanHandle<LlmSpan>> {\n    return this.span<LlmSpan>({ kind: 'llm', ...init })\n  }\n\n  tool(\n    init: Omit<ToolSpan, 'spanId' | 'runId' | 'kind' | 'startedAt'>,\n  ): Promise<SpanHandle<ToolSpan>> {\n    return this.span<ToolSpan>({ kind: 'tool', ...init })\n  }\n\n  retrieval(\n    init: Omit<RetrievalSpan, 'spanId' | 'runId' | 'kind' | 'startedAt'>,\n  ): Promise<SpanHandle<RetrievalSpan>> {\n    return this.span<RetrievalSpan>({ kind: 'retrieval', ...init })\n  }\n\n  async recordJudge(\n    verdict: Omit<JudgeSpan, 'spanId' | 'runId' | 'kind' | 'startedAt' | 'endedAt'>,\n  ): Promise<JudgeSpan> {\n    const spanId = this.id()\n    const now = this.now()\n    const full: JudgeSpan = {\n      spanId,\n      runId: this._runId,\n      kind: 'judge',\n      startedAt: now,\n      endedAt: now,\n      status: 'ok',\n      ...verdict,\n    }\n    await this.store.appendSpan(full)\n    return full\n  }\n\n  sandbox(\n    init: Omit<SandboxSpan, 'spanId' | 'runId' | 'kind' | 'startedAt'>,\n  ): Promise<SpanHandle<SandboxSpan>> {\n    return this.span<SandboxSpan>({ kind: 'sandbox', ...init })\n  }\n\n  // ── Events ─────────────────────────────────────────────────────────\n\n  async emit(event: {\n    kind: EventKind\n    spanId?: string\n    payload?: Record<string, unknown>\n  }): Promise<TraceEvent> {\n    const full: TraceEvent = {\n      eventId: this.id(),\n      runId: this._runId,\n      spanId: event.spanId ?? this.stack[this.stack.length - 1],\n      kind: event.kind,\n      timestamp: this.now(),\n      payload: event.payload ?? {},\n    }\n    await this.store.appendEvent(full)\n    return full\n  }\n\n  // ── Budget ledger ──────────────────────────────────────────────────\n\n  async recordBudget(\n    entry: Omit<BudgetLedgerEntry, 'runId' | 'timestamp'> & { timestamp?: number },\n  ): Promise<BudgetLedgerEntry> {\n    const full: BudgetLedgerEntry = {\n      runId: this._runId,\n      timestamp: entry.timestamp ?? this.now(),\n      dimension: entry.dimension,\n      limit: entry.limit,\n      consumed: entry.consumed,\n      remaining: entry.remaining,\n      breached: entry.breached,\n      spanId: entry.spanId ?? this.stack[this.stack.length - 1],\n    }\n    await this.store.appendBudgetEntry(full)\n    if (full.breached) {\n      await this.emit({\n        kind: 'budget_breach',\n        spanId: full.spanId,\n        payload: { dimension: full.dimension, limit: full.limit, consumed: full.consumed },\n      })\n    }\n    return full\n  }\n\n  // ── Artifacts ──────────────────────────────────────────────────────\n\n  async recordArtifact(artifact: Omit<Artifact, 'artifactId' | 'runId'>): Promise<Artifact> {\n    const full: Artifact = { artifactId: this.id(), runId: this._runId, ...artifact }\n    await this.store.appendArtifact(full)\n    return full\n  }\n\n  // ── Nested composition ─────────────────────────────────────────────\n\n  /**\n   * Runs `fn` inside a span; auto-ends on success, auto-fails on throw.\n   * Returns the fn's return value. Use this for the 95% case.\n   */\n  async within<T>(\n    init: Parameters<TraceEmitter['span']>[0],\n    fn: (handle: SpanHandle) => Promise<T>,\n  ): Promise<T> {\n    const handle = await this.span(init)\n    try {\n      const result = await fn(handle)\n      await handle.end()\n      return result\n    } catch (err) {\n      await handle.fail(err instanceof Error ? err : String(err))\n      throw err\n    }\n  }\n}\n\n// Helpers -------------------------------------------------------------\n\nfunction cryptoRandomId(): string {\n  if (typeof globalThis.crypto?.randomUUID === 'function') return globalThis.crypto.randomUUID()\n  return `${Date.now().toString(36)}-${Math.random().toString(36).slice(2, 10)}`\n}\n\n/** Helper to build an LLM span handle args object from a provider-shaped response. */\nexport function llmSpanFromProvider(args: {\n  name?: string\n  model: string\n  messages: Message[]\n  output: string\n  usage?: {\n    inputTokens?: number\n    outputTokens?: number\n    cachedTokens?: number\n    reasoningTokens?: number\n  }\n  costUsd?: number\n  finishReason?: string\n}): Omit<LlmSpan, 'spanId' | 'runId' | 'kind' | 'startedAt'> {\n  return {\n    name: args.name ?? args.model,\n    model: args.model,\n    messages: args.messages,\n    output: args.output,\n    inputTokens: args.usage?.inputTokens,\n    outputTokens: args.usage?.outputTokens,\n    cachedTokens: args.usage?.cachedTokens,\n    reasoningTokens: args.usage?.reasoningTokens,\n    costUsd: args.costUsd,\n    finishReason: args.finishReason,\n  }\n}\n"],"mappings":";AAmEO,IAAM,eAAN,MAAmB;AAAA,EAChB;AAAA,EACA,QAAkB,CAAC;AAAA,EACnB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,OAAmB,UAA+B,CAAC,GAAG;AAChE,SAAK,QAAQ;AACb,SAAK,MAAM,QAAQ,QAAQ,MAAM,KAAK,IAAI;AAC1C,SAAK,KAAK,QAAQ,OAAO,MAAM,eAAe;AAC9C,SAAK,SAAS,QAAQ,SAAS,KAAK,GAAG;AACvC,SAAK,QAAQ,QAAQ,iBAAiB,CAAC;AACvC,SAAK,aAAa,QAAQ,cAAc;AAAA,EAC1C;AAAA,EAEA,IAAI,QAAgB;AAClB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,aAAyB;AAC3B,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,mBAAmB,MAA6B;AAC9C,SAAK,MAAM,KAAK,IAAI;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAcA,MAAM,SACJ,KACc;AACd,UAAM,aAAa,IAAI,cAAc,IAAI,SAAS,IAAI,MAAM,QAAQ;AACpE,UAAM,OAAY;AAAA,MAChB,GAAG;AAAA,MACH;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,WAAW,KAAK,IAAI;AAAA,MACpB,QAAQ;AAAA,IACV;AACA,UAAM,KAAK,MAAM,UAAU,IAAI;AAC/B,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,OAAO,SAAqC;AAChD,UAAM,SAAiC,SAAS,SAAS,QAAQ,WAAW;AAC5E,UAAM,KAAK,MAAM,UAAU,KAAK,QAAQ,EAAE,SAAS,KAAK,IAAI,GAAG,QAAQ,QAAQ,CAAC;AAChF,UAAM,KAAK,SAAS,EAAE,OAAO,KAAK,QAAQ,SAAS,MAAM,OAAO,KAAK,OAAO,SAAS,OAAO,CAAC;AAAA,EAC/F;AAAA,EAEA,MAAM,SAAS,QAA+B;AAC5C,UAAM,UAAU,EAAE,MAAM,OAAO,OAAO,OAAO;AAC7C,UAAM,KAAK,MAAM,UAAU,KAAK,QAAQ;AAAA,MACtC,SAAS,KAAK,IAAI;AAAA,MAClB,QAAQ;AAAA,MACR;AAAA,IACF,CAAC;AACD,UAAM,KAAK,SAAS;AAAA,MAClB,OAAO,KAAK;AAAA,MACZ,SAAS;AAAA,MACT,OAAO,KAAK;AAAA,MACZ;AAAA,MACA,QAAQ;AAAA,IACV,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,SAAS,KAA4C;AACjE,eAAW,QAAQ,KAAK,OAAO;AAC7B,UAAI;AACF,cAAM,KAAK,GAAG;AAAA,MAChB,SAAS,KAAK;AACZ,YAAI,KAAK,eAAe,QAAS,OAAM;AACvC,YAAI;AACF,gBAAM,KAAK,MAAM,YAAY;AAAA,YAC3B,SAAS,KAAK,GAAG;AAAA,YACjB,OAAO,KAAK;AAAA,YACZ,MAAM;AAAA,YACN,WAAW,KAAK,IAAI;AAAA,YACpB,SAAS;AAAA,cACP,QAAQ;AAAA,cACR,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAAA,YACxD;AAAA,UACF,CAAC;AAAA,QACH,QAAQ;AAAA,QAER;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAIA,MAAM,KACJ,MAMwB;AACxB,UAAM,SAAS,KAAK,GAAG;AACvB,UAAM,SAAS,KAAK,gBAAgB,KAAK,MAAM,KAAK,MAAM,SAAS,CAAC;AACpE,UAAM,OAAO;AAAA,MACX;AAAA,MACA,cAAc;AAAA,MACd,OAAO,KAAK;AAAA,MACZ,WAAW,KAAK,IAAI;AAAA,MACpB,GAAG;AAAA,IACL;AACA,UAAM,KAAK,MAAM,WAAW,IAAI;AAChC,SAAK,MAAM,KAAK,MAAM;AACtB,WAAO,KAAK,OAAU,IAAI;AAAA,EAC5B;AAAA,EAEQ,OAAuB,MAAwB;AACrD,WAAO;AAAA,MACL;AAAA,MACA,KAAK,OAAO,UAAuB;AACjC,cAAM,UAAU,KAAK,IAAI;AACzB,cAAM,KAAK,MAAM,WAAW,KAAK,QAAQ;AAAA,UACvC;AAAA,UACA,QAAQ;AAAA,UACR,GAAG;AAAA,QACL,CAAkB;AAClB,aAAK,IAAI,KAAK,MAAM;AAAA,MACtB;AAAA,MACA,MAAM,OAAO,OAAuB,UAAuB;AACzD,cAAM,UAAU,KAAK,IAAI;AACzB,cAAM,SAAS,iBAAiB,QAAQ,MAAM,UAAU;AACxD,cAAM,KAAK,MAAM,WAAW,KAAK,QAAQ;AAAA,UACvC;AAAA,UACA,QAAQ;AAAA,UACR,OAAO;AAAA,UACP,GAAG;AAAA,QACL,CAAkB;AAClB,aAAK,IAAI,KAAK,MAAM;AAAA,MACtB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,IAAI,QAAsB;AAChC,UAAM,MAAM,KAAK,MAAM,YAAY,MAAM;AACzC,QAAI,OAAO,EAAG,MAAK,MAAM,OAAO,KAAK,CAAC;AAAA,EACxC;AAAA;AAAA,EAIA,IACE,MAC8B;AAC9B,WAAO,KAAK,KAAc,EAAE,MAAM,OAAO,GAAG,KAAK,CAAC;AAAA,EACpD;AAAA,EAEA,KACE,MAC+B;AAC/B,WAAO,KAAK,KAAe,EAAE,MAAM,QAAQ,GAAG,KAAK,CAAC;AAAA,EACtD;AAAA,EAEA,UACE,MACoC;AACpC,WAAO,KAAK,KAAoB,EAAE,MAAM,aAAa,GAAG,KAAK,CAAC;AAAA,EAChE;AAAA,EAEA,MAAM,YACJ,SACoB;AACpB,UAAM,SAAS,KAAK,GAAG;AACvB,UAAM,MAAM,KAAK,IAAI;AACrB,UAAM,OAAkB;AAAA,MACtB;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,MAAM;AAAA,MACN,WAAW;AAAA,MACX,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,GAAG;AAAA,IACL;AACA,UAAM,KAAK,MAAM,WAAW,IAAI;AAChC,WAAO;AAAA,EACT;AAAA,EAEA,QACE,MACkC;AAClC,WAAO,KAAK,KAAkB,EAAE,MAAM,WAAW,GAAG,KAAK,CAAC;AAAA,EAC5D;AAAA;AAAA,EAIA,MAAM,KAAK,OAIa;AACtB,UAAM,OAAmB;AAAA,MACvB,SAAS,KAAK,GAAG;AAAA,MACjB,OAAO,KAAK;AAAA,MACZ,QAAQ,MAAM,UAAU,KAAK,MAAM,KAAK,MAAM,SAAS,CAAC;AAAA,MACxD,MAAM,MAAM;AAAA,MACZ,WAAW,KAAK,IAAI;AAAA,MACpB,SAAS,MAAM,WAAW,CAAC;AAAA,IAC7B;AACA,UAAM,KAAK,MAAM,YAAY,IAAI;AACjC,WAAO;AAAA,EACT;AAAA;AAAA,EAIA,MAAM,aACJ,OAC4B;AAC5B,UAAM,OAA0B;AAAA,MAC9B,OAAO,KAAK;AAAA,MACZ,WAAW,MAAM,aAAa,KAAK,IAAI;AAAA,MACvC,WAAW,MAAM;AAAA,MACjB,OAAO,MAAM;AAAA,MACb,UAAU,MAAM;AAAA,MAChB,WAAW,MAAM;AAAA,MACjB,UAAU,MAAM;AAAA,MAChB,QAAQ,MAAM,UAAU,KAAK,MAAM,KAAK,MAAM,SAAS,CAAC;AAAA,IAC1D;AACA,UAAM,KAAK,MAAM,kBAAkB,IAAI;AACvC,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,KAAK;AAAA,QACd,MAAM;AAAA,QACN,QAAQ,KAAK;AAAA,QACb,SAAS,EAAE,WAAW,KAAK,WAAW,OAAO,KAAK,OAAO,UAAU,KAAK,SAAS;AAAA,MACnF,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AAAA;AAAA,EAIA,MAAM,eAAe,UAAqE;AACxF,UAAM,OAAiB,EAAE,YAAY,KAAK,GAAG,GAAG,OAAO,KAAK,QAAQ,GAAG,SAAS;AAChF,UAAM,KAAK,MAAM,eAAe,IAAI;AACpC,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,MAAM,OACJ,MACA,IACY;AACZ,UAAM,SAAS,MAAM,KAAK,KAAK,IAAI;AACnC,QAAI;AACF,YAAM,SAAS,MAAM,GAAG,MAAM;AAC9B,YAAM,OAAO,IAAI;AACjB,aAAO;AAAA,IACT,SAAS,KAAK;AACZ,YAAM,OAAO,KAAK,eAAe,QAAQ,MAAM,OAAO,GAAG,CAAC;AAC1D,YAAM;AAAA,IACR;AAAA,EACF;AACF;AAIA,SAAS,iBAAyB;AAChC,MAAI,OAAO,WAAW,QAAQ,eAAe,WAAY,QAAO,WAAW,OAAO,WAAW;AAC7F,SAAO,GAAG,KAAK,IAAI,EAAE,SAAS,EAAE,CAAC,IAAI,KAAK,OAAO,EAAE,SAAS,EAAE,EAAE,MAAM,GAAG,EAAE,CAAC;AAC9E;AAGO,SAAS,oBAAoB,MAayB;AAC3D,SAAO;AAAA,IACL,MAAM,KAAK,QAAQ,KAAK;AAAA,IACxB,OAAO,KAAK;AAAA,IACZ,UAAU,KAAK;AAAA,IACf,QAAQ,KAAK;AAAA,IACb,aAAa,KAAK,OAAO;AAAA,IACzB,cAAc,KAAK,OAAO;AAAA,IAC1B,cAAc,KAAK,OAAO;AAAA,IAC1B,iBAAiB,KAAK,OAAO;AAAA,IAC7B,SAAS,KAAK;AAAA,IACd,cAAc,KAAK;AAAA,EACrB;AACF;","names":[]}