@tangle-network/agent-eval 0.77.0 → 0.80.0

package/dist/chunk-KWRRMR3J.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/agent-profile-cell.ts","../src/run-record.ts"],"sourcesContent":["import { ValidationError } from './errors'\nimport { hashJson } from './pre-registration'\n\nexport type AgentProfileCellSchemaVersion = 'agent-profile-cell/v1'\n\nexport type AgentProfileJson =\n  | string\n  | number\n  | boolean\n  | null\n  | AgentProfileJson[]\n  | { [key: string]: AgentProfileJson }\n\nexport type AgentProfileDimensionValue = string | number | boolean | null\n\nexport interface AgentProfileSource {\n  /** Runtime/profile contract being fingerprinted, e.g. `sandbox-agent-profile`. */\n  kind: string\n  /** sha256 over the canonical source profile object. */\n  hash: string\n}\n\nexport interface AgentProfileSourceInput {\n  kind: string\n  /** Precomputed sha256 for callers that already sign their profile artifact. */\n  hash?: string\n  /** Full canonical runtime profile; hashed and then discarded from the cell. */\n  profile?: AgentProfileJson\n}\n\nexport interface AgentProfileHarness {\n  id: string\n  version?: string\n  hash?: string\n}\n\nexport interface AgentProfileCellInput {\n  profileId: string\n  sourceProfile: AgentProfileSourceInput\n  harness?: AgentProfileHarness\n  model?: string\n  promptHash?: string\n  dimensions?: Record<string, AgentProfileDimensionValue>\n}\n\nexport interface AgentProfileCell {\n  schemaVersion: AgentProfileCellSchemaVersion\n  cellId: string\n  profileId: string\n  sourceProfile: AgentProfileSource\n  harness?: AgentProfileHarness\n  model?: string\n  promptHash?: string\n  dimensions?: Record<string, AgentProfileDimensionValue>\n}\n\nexport class AgentProfileCellValidationError extends ValidationError {\n  readonly path: string\n  constructor(message: string, path = '') {\n    super(path ? `${message} (at ${path})` : message)\n    this.path = path\n  }\n}\n\nconst SHA256_HEX = /^[0-9a-f]{64}$/\nconst CELL_ID = /^agent-profile-cell:sha256:[0-9a-f]{64}$/\n\nexport async function buildAgentProfileCell(\n  input: AgentProfileCellInput,\n): Promise<AgentProfileCell> {\n  const material = await normalizeAgentProfileCellInput(input)\n  const cellId = `agent-profile-cell:sha256:${await hashJson(material)}`\n  return { ...material, cellId }\n}\n\nexport function agentProfileCellHashMaterial(\n  cell: AgentProfileCell,\n): Omit<AgentProfileCell, 'cellId'> {\n  const { cellId: _cellId, ...material } = cell\n  void _cellId\n  return normalizeAgentProfileCell(material)\n}\n\nexport async function verifyAgentProfileCell(cell: AgentProfileCell): Promise<boolean> {\n  validateAgentProfileCell(cell)\n  return (\n    cell.cellId ===\n    `agent-profile-cell:sha256:${await hashJson(agentProfileCellHashMaterial(cell))}`\n  )\n}\n\nexport function validateAgentProfileCell(input: unknown): AgentProfileCell {\n  if (input === null || typeof input !== 'object') {\n    throw new AgentProfileCellValidationError('expected object')\n  }\n  const obj = input as Record<string, unknown>\n  expectLiteral(obj.schemaVersion, 'agent-profile-cell/v1', 'schemaVersion')\n  if (typeof obj.cellId !== 'string' || !CELL_ID.test(obj.cellId)) {\n    throw new AgentProfileCellValidationError(\n      'cellId must match agent-profile-cell:sha256:<64 lowercase hex chars>',\n      'cellId',\n    )\n  }\n  expectString(obj.profileId, 'profileId')\n  validateSource(obj.sourceProfile, 'sourceProfile')\n  if (obj.harness !== undefined) validateHarness(obj.harness, 'harness')\n  if (obj.model !== undefined) expectString(obj.model, 'model')\n  if (obj.promptHash !== undefined) expectString(obj.promptHash, 'promptHash')\n  if (obj.dimensions !== undefined) validateDimensions(obj.dimensions, 'dimensions')\n  return input as AgentProfileCell\n}\n\nexport function requireAgentProfileCell(record: {\n  runId: string\n  agentProfile?: AgentProfileCell\n}): AgentProfileCell {\n  if (!record.agentProfile) {\n    throw new AgentProfileCellValidationError(\n      `run \"${record.runId}\" is missing agentProfile; profile-cell grouping requires explicit profile identity`,\n      'agentProfile',\n    )\n  }\n  return validateAgentProfileCell(record.agentProfile)\n}\n\nexport function agentProfileCellKey(record: {\n  runId: string\n  agentProfile?: AgentProfileCell\n}): string {\n  return requireAgentProfileCell(record).cellId\n}\n\nexport async function assertRunAgentProfileCell(record: {\n  runId: string\n  model: string\n  promptHash: string\n  agentProfile?: AgentProfileCell\n}): Promise<AgentProfileCell> {\n  const profile = requireAgentProfileCell(record)\n  if (!(await verifyAgentProfileCell(profile))) {\n    throw new AgentProfileCellValidationError(\n      `run \"${record.runId}\" has an agentProfile.cellId that does not match its content`,\n      'agentProfile.cellId',\n    )\n  }\n  if (profile.model !== undefined && profile.model !== record.model) {\n    throw new AgentProfileCellValidationError(\n      `run \"${record.runId}\" agentProfile.model \"${profile.model}\" does not match model \"${record.model}\"`,\n      'agentProfile.model',\n    )\n  }\n  if (profile.promptHash !== undefined && profile.promptHash !== record.promptHash) {\n    throw new AgentProfileCellValidationError(\n      `run \"${record.runId}\" agentProfile.promptHash \"${profile.promptHash}\" does not match promptHash \"${record.promptHash}\"`,\n      'agentProfile.promptHash',\n    )\n  }\n  return profile\n}\n\nexport function groupRunsByAgentProfileCell<\n  T extends { runId: string; agentProfile?: AgentProfileCell },\n>(records: readonly T[]): Map<string, T[]> {\n  const groups = new Map<string, T[]>()\n  for (const record of records) {\n    const key = agentProfileCellKey(record)\n    const bucket = groups.get(key)\n    if (bucket) bucket.push(record)\n    else groups.set(key, [record])\n  }\n  return groups\n}\n\nasync function normalizeAgentProfileCellInput(\n  input: AgentProfileCellInput,\n): Promise<Omit<AgentProfileCell, 'cellId'>> {\n  return normalizeAgentProfileCell({\n    schemaVersion: 'agent-profile-cell/v1',\n    profileId: input.profileId,\n    sourceProfile: await normalizeSourceInput(input.sourceProfile),\n    harness: input.harness,\n    model: input.model,\n    promptHash: input.promptHash,\n    dimensions: input.dimensions,\n  })\n}\n\nfunction normalizeAgentProfileCell(\n  input: Omit<AgentProfileCell, 'cellId'>,\n): Omit<AgentProfileCell, 'cellId'> {\n  return compactObject({\n    schemaVersion: 'agent-profile-cell/v1' as const,\n    profileId: requireNonEmpty(input.profileId, 'profileId'),\n    sourceProfile: normalizeSource(input.sourceProfile),\n    harness: input.harness ? normalizeHarness(input.harness, 'harness') : undefined,\n    model: optionalNonEmpty(input.model, 'model'),\n    promptHash: optionalNonEmpty(input.promptHash, 'promptHash'),\n    dimensions: input.dimensions\n      ? nonEmptyRecord(normalizeDimensions(input.dimensions))\n      : undefined,\n  })\n}\n\nasync function normalizeSourceInput(input: AgentProfileSourceInput): Promise<AgentProfileSource> {\n  const kind = requireNonEmpty(input.kind, 'sourceProfile.kind')\n  if (input.hash !== undefined && input.profile !== undefined) {\n    throw new AgentProfileCellValidationError(\n      'sourceProfile must provide either hash or profile, not both',\n      'sourceProfile',\n    )\n  }\n  if (input.hash !== undefined) {\n    return { kind, hash: requireSha256Hex(input.hash, 'sourceProfile.hash') }\n  }\n  if (input.profile === undefined) {\n    throw new AgentProfileCellValidationError(\n      'sourceProfile must provide hash or profile',\n      'sourceProfile',\n    )\n  }\n  assertJson(input.profile, 'sourceProfile.profile')\n  return { kind, hash: await hashJson(input.profile) }\n}\n\nfunction normalizeSource(input: AgentProfileSource): AgentProfileSource {\n  return {\n    kind: requireNonEmpty(input.kind, 'sourceProfile.kind'),\n    hash: requireSha256Hex(input.hash, 'sourceProfile.hash'),\n  }\n}\n\nfunction normalizeHarness(input: AgentProfileHarness, path: string): AgentProfileHarness {\n  return compactObject({\n    id: requireNonEmpty(input.id, `${path}.id`),\n    version: optionalNonEmpty(input.version, `${path}.version`),\n    hash: optionalNonEmpty(input.hash, `${path}.hash`),\n  })\n}\n\nfunction normalizeDimensions(\n  input: Record<string, AgentProfileDimensionValue>,\n): Record<string, AgentProfileDimensionValue> {\n  const out: Record<string, AgentProfileDimensionValue> = {}\n  for (const key of Object.keys(input).sort()) {\n    const value = input[key]\n    requireNonEmpty(key, 'dimensions.<key>')\n    if (\n      value !== null &&\n      typeof value !== 'string' &&\n      typeof value !== 'number' &&\n      typeof value !== 'boolean'\n    ) {\n      throw new AgentProfileCellValidationError(\n        'expected primitive dimension value',\n        `dimensions.${key}`,\n      )\n    }\n    if (typeof value === 'number' && !Number.isFinite(value)) {\n      throw new AgentProfileCellValidationError('expected finite number', `dimensions.${key}`)\n    }\n    out[key] = value\n  }\n  return out\n}\n\nfunction compactObject<T extends Record<string, unknown>>(input: T): T {\n  const out: Record<string, unknown> = {}\n  for (const [key, value] of Object.entries(input)) {\n    if (value !== undefined) out[key] = value\n  }\n  return out as T\n}\n\nfunction nonEmptyRecord<T extends Record<string, unknown>>(input: T): T | undefined {\n  return Object.keys(input).length > 0 ? input : undefined\n}\n\nfunction validateSource(value: unknown, path: string): void {\n  if (value === null || typeof value !== 'object' || Array.isArray(value)) {\n    throw new AgentProfileCellValidationError('expected object', path)\n  }\n  const rec = value as Record<string, unknown>\n  expectString(rec.kind, `${path}.kind`)\n  requireSha256Hex(rec.hash, `${path}.hash`)\n}\n\nfunction validateHarness(value: unknown, path: string): void {\n  if (value === null || typeof value !== 'object' || Array.isArray(value)) {\n    throw new AgentProfileCellValidationError('expected object', path)\n  }\n  const rec = value as Record<string, unknown>\n  expectString(rec.id, `${path}.id`)\n  if (rec.version !== undefined) expectString(rec.version, `${path}.version`)\n  if (rec.hash !== undefined) expectString(rec.hash, `${path}.hash`)\n}\n\nfunction validateDimensions(value: unknown, path: string): void {\n  if (value === null || typeof value !== 'object' || Array.isArray(value)) {\n    throw new AgentProfileCellValidationError('expected object', path)\n  }\n  normalizeDimensions(value as Record<string, AgentProfileDimensionValue>)\n}\n\nfunction assertJson(value: AgentProfileJson, path: string): void {\n  if (value === null) return\n  const type = typeof value\n  if (type === 'string' || type === 'boolean') return\n  if (type === 'number') {\n    if (!Number.isFinite(value)) {\n      throw new AgentProfileCellValidationError('expected finite number', path)\n    }\n    return\n  }\n  if (Array.isArray(value)) {\n    value.forEach((item, index) => {\n      assertJson(item, `${path}[${index}]`)\n    })\n    return\n  }\n  if (type === 'object') {\n    for (const [key, nested] of Object.entries(value)) {\n      requireNonEmpty(key, `${path}.<key>`)\n      assertJson(nested, `${path}.${key}`)\n    }\n    return\n  }\n  throw new AgentProfileCellValidationError('expected JSON-compatible value', path)\n}\n\nfunction expectLiteral(value: unknown, expected: string, path: string): void {\n  if (value !== expected) {\n    throw new AgentProfileCellValidationError(`expected ${expected}`, path)\n  }\n}\n\nfunction expectString(value: unknown, path: string): void {\n  if (typeof value !== 'string' || value.length === 0) {\n    throw new AgentProfileCellValidationError('expected non-empty string', path)\n  }\n}\n\nfunction requireNonEmpty(value: string, path: string): string {\n  if (typeof value !== 'string' || value.length === 0) {\n    throw new AgentProfileCellValidationError('expected non-empty string', path)\n  }\n  return value\n}\n\nfunction optionalNonEmpty(value: string | undefined, path: string): string | undefined {\n  if (value === undefined) return undefined\n  return requireNonEmpty(value, path)\n}\n\nfunction requireSha256Hex(value: unknown, path: string): string {\n  if (typeof value !== 'string' || !SHA256_HEX.test(value)) {\n    throw new AgentProfileCellValidationError('expected 64 lowercase sha256 hex chars', path)\n  }\n  return value\n}\n\n// ── Consumer helpers ─────────────────────────────────────────────────\n//\n// Two pieces of boilerplate every product consuming `buildAgentProfileCell`\n// has been duplicating (gtm-agent #137, blueprint-agent #1756/#1757):\n//\n//   1. A `JSON.parse(JSON.stringify(value))` helper that canonicalizes an\n//      arbitrary sandbox-SDK `AgentProfile` into the recursive\n//      `AgentProfileJson` shape, with a fail-loud error when the profile\n//      is not JSON-serializable.\n//\n//   2. The magic string `'sandbox-agent-profile'` for `sourceProfile.kind`.\n//\n// Both belong here so the cross-product cell join (same canonical profile\n// hashes to the same `sourceProfile.hash` across products) is enforced by\n// the type system, not by every consumer remembering to do it right.\n// See blueprint-agent issue tangle-network/agent-eval#82.\n\n/** Canonical `sourceProfile.kind` values. Two products fingerprinting the\n *  same canonical profile MUST use the same kind for their cells to share\n *  `sourceProfile.hash`. Extend rather than create new strings — adding a\n *  new kind is a deliberate cross-product schema change. */\nexport const AGENT_PROFILE_KINDS = {\n  /** A profile declared via `defineAgentProfile(...)` from\n   *  `@tangle-network/sandbox`. The default kind for sandbox-hosted\n   *  products (gtm-agent, blueprint-agent, sandbox, evals). */\n  SANDBOX_AGENT_PROFILE: 'sandbox-agent-profile',\n} as const\n\nexport type AgentProfileKind = (typeof AGENT_PROFILE_KINDS)[keyof typeof AGENT_PROFILE_KINDS]\n\n/** Canonicalize an arbitrary value into `AgentProfileJson` by JSON\n *  round-trip. Throws when the value contains anything not representable\n *  as JSON (functions, BigInt, cycles) — non-portable profiles fail loud\n *  rather than silently dropping fields. */\nexport function toAgentProfileJson(value: unknown): AgentProfileJson {\n  let serialized: string | undefined\n  try {\n    serialized = JSON.stringify(value)\n  } catch (err) {\n    throw new AgentProfileCellValidationError(\n      `agent profile must be JSON-serializable: ${err instanceof Error ? err.message : String(err)}`,\n      'sourceProfile.profile',\n    )\n  }\n  if (serialized === undefined) {\n    throw new AgentProfileCellValidationError(\n      'agent profile must be JSON-serializable (got undefined after JSON.stringify)',\n      'sourceProfile.profile',\n    )\n  }\n  return JSON.parse(serialized) as AgentProfileJson\n}\n\n/** Minimal shape required of any sandbox-SDK `AgentProfile` — anything\n *  with a non-empty `name` and `version` plus JSON-serializable contents.\n *  Compatible with `defineAgentProfile(...)` output from\n *  `@tangle-network/sandbox`; products that have not yet declared a real\n *  profile can pass a `{ name, version, ...metadata }` stub. */\nexport interface SandboxAgentProfileLike {\n  name: string\n  version: string\n  [key: string]: unknown\n}\n\n/** Higher-level helper that hard-codes the canonical\n *  `sandbox-agent-profile` kind plus the JSON canonicalization. Equivalent\n *  to calling `buildAgentProfileCell` with `profileId = \\`${name}@${version}\\``\n *  and `sourceProfile = { kind: SANDBOX_AGENT_PROFILE, profile: <round-tripped> }`.\n *\n *  Use this from any product consuming a sandbox-SDK `AgentProfile`; the\n *  manual `buildAgentProfileCell` call is reserved for advanced cases\n *  (custom kinds, pre-computed source hashes, alternate profileId\n *  conventions). */\nexport async function buildSandboxAgentProfileCell(\n  profile: SandboxAgentProfileLike,\n  input: Omit<AgentProfileCellInput, 'profileId' | 'sourceProfile'>,\n): Promise<AgentProfileCell> {\n  if (!profile || typeof profile !== 'object') {\n    throw new AgentProfileCellValidationError('sandbox AgentProfile must be an object', 'profile')\n  }\n  if (typeof profile.name !== 'string' || profile.name.length === 0) {\n    throw new AgentProfileCellValidationError(\n      'sandbox AgentProfile must have a non-empty `name`',\n      'profile.name',\n    )\n  }\n  if (typeof profile.version !== 'string' || profile.version.length === 0) {\n    throw new AgentProfileCellValidationError(\n      'sandbox AgentProfile must have a non-empty `version`',\n      'profile.version',\n    )\n  }\n  return buildAgentProfileCell({\n    ...input,\n    profileId: `${profile.name}@${profile.version}`,\n    sourceProfile: {\n      kind: AGENT_PROFILE_KINDS.SANDBOX_AGENT_PROFILE,\n      profile: toAgentProfileJson(profile),\n    },\n  })\n}\n","/**\n * Paper-grade RunRecord schema + runtime validator.\n *\n * Every run that participates in a promotion gate, paper table, or\n * researcher loop SHOULD be recorded as a `RunRecord`. The mandatory\n * fields are exactly those the paper \"Two Loops, Three Roles\" requires\n * for reproducibility: who/what/when/cost/seed/hash, plus the search vs\n * holdout split tag and either a `searchScore` or a `holdoutScore`.\n *\n * This is intentionally NOT a replacement for the rich `Run` /\n * `ProposeReviewReport` / `ScenarioResult` types already in the\n * package. Those are runtime structures with full provenance. A\n * `RunRecord` is the analysis-time projection — the JSON-friendly\n * row you'd put in a parquet file or paste into a notebook.\n *\n * Validate at the boundary:\n *\n *   const rec = validateRunRecord(rawJson)         // throws on missing\n *   const ok  = isRunRecord(rawJson)               // boolean check\n *   const rec = parseRunRecordSafe(rawJson)        // { ok, value | error }\n *\n * The validator runs in pure TS — zod is intentionally NOT a\n * dependency. Round-trip tested in `tests/run-record.test.ts`.\n */\n\nimport type { AgentProfileCell } from './agent-profile-cell'\nimport { validateAgentProfileCell } from './agent-profile-cell'\nimport { ValidationError } from './errors'\nimport type { FailureClass } from './trace/schema'\n\n/** Search/dev/holdout split tag. 'search' is the paper-grade alias for the\n *  combined train+test pool that the optimizer is allowed to read. */\nexport type RunSplitTag = 'search' | 'dev' | 'holdout'\n\nexport interface RunTokenUsage {\n  input: number\n  output: number\n  cached?: number\n}\n\nexport interface RunJudgeMetadata {\n  model: string\n  promptVersion: string\n  /** [0,1] confidence the judge declared. Constant judge confidence\n   *  across many runs is a fallback signal (see `canary.ts`). */\n  confidence: number\n  /** True if the judge degraded to a fallback path (rules-only,\n   *  prior-call cache, etc.). The canary uses this to alert. */\n  fallback: boolean\n}\n\n/**\n * Per-judge / per-dimension breakdown for runs scored by an ensemble of\n * judges over a multi-dimensional rubric.\n *\n * The collapsed `outcome.searchScore` / `holdoutScore` carries the\n * composite the gate uses. The full breakdown belongs here so consumers\n * can answer \"which judge disagreed?\", \"which dimension dragged the\n * composite down?\", and \"did half the panel fail?\" without re-running.\n *\n * `perJudge[judgeId][dim]` is the canonical source; `perDimMean` and\n * `composite` are convenience projections — derivable but precomputed so\n * downstream IRR primitives (`interRaterReliability`,\n * `corpusInterRaterAgreement`) and reporters don't pay the same\n * aggregation twice.\n *\n * Fail-loud discipline: judges that errored out land in `failedJudges`\n * by id. A missing key in `perJudge` is ambiguous (silent zero vs not\n * run); the explicit list makes a partial-failure recorded as such.\n */\nexport interface JudgeScoresRecord {\n  /** Per-judge per-dimension scores. `{ \"kimi-k2.6\": { helpfulness: 0.8, clarity: 0.7 }, ... }`. */\n  perJudge: Record<string, Record<string, number>>\n  /** Per-dim mean across judges. Convenience — derivable from `perJudge`. */\n  perDimMean: Record<string, number>\n  /** Composite mean across all dims and judges. Mirrors the score\n   *  the gate sees on `outcome.searchScore` / `holdoutScore`. */\n  composite: number\n  /** Judges that errored or returned an unparseable verdict. Recorded\n   *  by id (e.g. `['glm-5.1']`) so a partial-failure case is explicit,\n   *  not inferred from missing keys in `perJudge`. */\n  failedJudges?: string[]\n  /** Free-form notes the judges emitted (joined across judges or\n   *  first-judge only — consumer's choice). */\n  notes?: string\n}\n\nexport interface RunOutcome {\n  /** Score on the search/optimization split. Optional because a\n   *  holdout-only evaluation only fills `holdoutScore`. */\n  searchScore?: number\n  /** Score on the held-out split. Optional because a search-only run\n   *  only fills `searchScore`. At least one must be present. */\n  holdoutScore?: number\n  /** Bag of any other metric the run produced — judge dimensions,\n   *  pass/fail counters, latency stats, etc. Numeric only — keeps\n   *  reporters honest. */\n  raw: Record<string, number>\n  /** Per-judge / per-dim breakdown. Consumers writing ensemble\n   *  judgements populate this; substrate primitives like\n   *  `interRaterReliability` and `corpusInterRaterAgreement` accept\n   *  these records as input. Optional — single-judge or scalar-only\n   *  runs leave it unset. */\n  judgeScores?: JudgeScoresRecord\n  /** Authenticity / realness verdict — did the run build the REAL thing on the\n   *  intended infra, or fake it (see `./authenticity`)? Optional: only domains\n   *  with an authenticity config populate it. Carried in the corpus so the\n   *  flywheel / off-policy learning can optimize for real completion, not gamed\n   *  pass-rate. `score` is 0-1; `gated` is the anti-Goodhart flag — a gated run\n   *  must not count as a real success regardless of `score`. */\n  realness?: { score: number; gated: boolean; reason?: string }\n}\n\n/**\n * Mandatory paper-grade fields for a single evaluation run. Optional\n * fields are extension points; mandatory fields throw if missing.\n *\n * Hash discipline:\n *   - `promptHash` is the sha256 of the EFFECTIVE prompt sent to the\n *     model (after any steering bundle merge).\n *   - `configHash` is the sha256 of the effective run config (model,\n *     temperature, tools, judges, splits). The pair (promptHash,\n *     configHash) uniquely identifies an experimental cell.\n *\n * Model snapshot discipline:\n *   - `model` MUST encode a snapshot version. Bare aliases like\n *     `claude-sonnet-4` or `gpt-4o` are banned — they remap silently.\n *     Use `claude-sonnet-4-6@2025-04-15` or `gpt-4o-2024-11-20`.\n */\nexport interface RunRecord {\n  /** UUID for the run. */\n  runId: string\n  /** Logical experiment grouping (a treatment vs a baseline within\n   *  the same sweep should share `experimentId`). */\n  experimentId: string\n  /** Stable identifier for the candidate (variant) being run. The\n   *  promotion gate compares two `candidateId`s on matched items. */\n  candidateId: string\n  /** RNG seed for the run. Always recorded — silent re-seeding is\n   *  the most common cause of non-reproducible numbers. */\n  seed: number\n  /** Model identifier WITH snapshot version. */\n  model: string\n  /** sha256 of the effective prompt (post-steering). */\n  promptHash: string\n  /** sha256 of the effective config. */\n  configHash: string\n  /** Git SHA the harness was run from. */\n  commitSha: string\n  /** End-to-end wall-clock duration in milliseconds. */\n  wallMs: number\n  /** Time spent queued before execution started, if known. */\n  queueMs?: number\n  /** Total USD cost. Mandatory — runs without a cost number are\n   *  unbounded by definition and must not be admitted into the gate. */\n  costUsd: number\n  /** Token usage breakdown. */\n  tokenUsage: RunTokenUsage\n  /** Judge-side metadata, if a judge was used. */\n  judgeMetadata?: RunJudgeMetadata\n  /** Per-split scores + raw bag. */\n  outcome: RunOutcome\n  /** Canonical, cross-agent failure class drawn from the shared\n   *  `FAILURE_CLASSES` taxonomy. This is the aggregation key that makes\n   *  \"which failure dominates across the whole fleet\" answerable in ONE\n   *  vocabulary — every agent classifies against the same enum. Producers\n   *  set it via the substrate classifier; leave unset only when the failure\n   *  genuinely can't be classified. */\n  failureClass?: FailureClass\n  /** Free-form domain-specific failure detail, scoped UNDER `failureClass`\n   *  (e.g. failureClass='tool_recovery_failure', failureMode='forge_build_unsatisfied').\n   *  The within-agent drill-down; `failureClass` is the cross-agent key. */\n  failureMode?: string\n  /** Which split this run was drawn from. */\n  splitTag: RunSplitTag\n  /**\n   * Stable scenario identifier the run was scored against. Optional for\n   * backwards compatibility, but **strongly recommended**: every primitive\n   * that pairs runs by scenario (preferences, paired stats, BT tournament)\n   * keys on this. The campaign artifact populates it canonically; legacy\n   * runs without it fall back to inference from `outcome.raw.scenario_id`\n   * or `experimentId`.\n   */\n  scenarioId?: string\n  /**\n   * Canonical identity for the agent profile cell that produced this row:\n   * profile artifact hash plus optional harness/model/prompt/reporting\n   * dimensions. Use `agentProfile.cellId` to group persona sweeps and\n   * longitudinal reports by the complete source profile, not by a loose\n   * candidate label or opaque config hash.\n   */\n  agentProfile?: AgentProfileCell\n}\n\n// ── Validation ───────────────────────────────────────────────────────\n\nconst MANDATORY_TOP_LEVEL = [\n  'runId',\n  'experimentId',\n  'candidateId',\n  'seed',\n  'model',\n  'promptHash',\n  'configHash',\n  'commitSha',\n  'wallMs',\n  'costUsd',\n  'tokenUsage',\n  'outcome',\n  'splitTag',\n] as const\n\nconst SPLIT_TAGS: ReadonlyArray<RunSplitTag> = ['search', 'dev', 'holdout']\n\nexport class RunRecordValidationError extends ValidationError {\n  readonly path: string\n  constructor(message: string, path = '') {\n    super(path ? `${message} (at ${path})` : message)\n    this.path = path\n  }\n}\n\n/**\n * Strict validator. Throws `RunRecordValidationError` on the first\n * missing or wrongly-typed field. Returns the input cast to\n * `RunRecord` on success — the validator does not coerce.\n */\nexport function validateRunRecord(input: unknown): RunRecord {\n  if (input === null || typeof input !== 'object') {\n    throw new RunRecordValidationError('expected object')\n  }\n  const obj = input as Record<string, unknown>\n\n  for (const key of MANDATORY_TOP_LEVEL) {\n    if (!(key in obj)) {\n      throw new RunRecordValidationError(`missing mandatory field \"${key}\"`)\n    }\n  }\n\n  expectString(obj.runId, 'runId')\n  expectString(obj.experimentId, 'experimentId')\n  expectString(obj.candidateId, 'candidateId')\n  expectFiniteNumber(obj.seed, 'seed')\n  expectString(obj.model, 'model')\n  expectString(obj.promptHash, 'promptHash')\n  expectString(obj.configHash, 'configHash')\n  expectString(obj.commitSha, 'commitSha')\n  expectFiniteNumber(obj.wallMs, 'wallMs')\n  if (obj.queueMs !== undefined) expectFiniteNumber(obj.queueMs, 'queueMs')\n  expectFiniteNumber(obj.costUsd, 'costUsd')\n\n  // Snapshot discipline: bare model aliases are not paper-grade.\n  if (!modelHasSnapshot(obj.model as string)) {\n    throw new RunRecordValidationError(\n      `model \"${obj.model}\" lacks a snapshot version (use 'name@YYYY-MM-DD' or 'name-YYYYMMDD')`,\n      'model',\n    )\n  }\n\n  // Token usage.\n  const tu = obj.tokenUsage\n  if (tu === null || typeof tu !== 'object') {\n    throw new RunRecordValidationError('tokenUsage must be an object', 'tokenUsage')\n  }\n  const tuRec = tu as Record<string, unknown>\n  expectFiniteNumber(tuRec.input, 'tokenUsage.input')\n  expectFiniteNumber(tuRec.output, 'tokenUsage.output')\n  if (tuRec.cached !== undefined) expectFiniteNumber(tuRec.cached, 'tokenUsage.cached')\n\n  // Judge metadata, optional.\n  if (obj.judgeMetadata !== undefined) {\n    const jm = obj.judgeMetadata\n    if (jm === null || typeof jm !== 'object') {\n      throw new RunRecordValidationError('judgeMetadata must be an object', 'judgeMetadata')\n    }\n    const jmRec = jm as Record<string, unknown>\n    expectString(jmRec.model, 'judgeMetadata.model')\n    expectString(jmRec.promptVersion, 'judgeMetadata.promptVersion')\n    expectFiniteNumber(jmRec.confidence, 'judgeMetadata.confidence')\n    if (typeof jmRec.fallback !== 'boolean') {\n      throw new RunRecordValidationError(\n        'judgeMetadata.fallback must be boolean',\n        'judgeMetadata.fallback',\n      )\n    }\n  }\n\n  // Outcome.\n  const out = obj.outcome\n  if (out === null || typeof out !== 'object') {\n    throw new RunRecordValidationError('outcome must be an object', 'outcome')\n  }\n  const outRec = out as Record<string, unknown>\n  if (outRec.searchScore !== undefined)\n    expectFiniteNumber(outRec.searchScore, 'outcome.searchScore')\n  if (outRec.holdoutScore !== undefined)\n    expectFiniteNumber(outRec.holdoutScore, 'outcome.holdoutScore')\n  if (outRec.searchScore === undefined && outRec.holdoutScore === undefined) {\n    throw new RunRecordValidationError(\n      'outcome must define searchScore or holdoutScore (or both)',\n      'outcome',\n    )\n  }\n  const raw = outRec.raw\n  if (raw === null || typeof raw !== 'object') {\n    throw new RunRecordValidationError('outcome.raw must be an object', 'outcome.raw')\n  }\n  for (const [k, v] of Object.entries(raw as Record<string, unknown>)) {\n    expectFiniteNumber(v, `outcome.raw.${k}`)\n  }\n  // Realness verdict, optional.\n  if (outRec.realness !== undefined) {\n    const r = outRec.realness\n    if (r === null || typeof r !== 'object') {\n      throw new RunRecordValidationError('outcome.realness must be an object', 'outcome.realness')\n    }\n    const rr = r as Record<string, unknown>\n    expectFiniteNumber(rr.score, 'outcome.realness.score')\n    if (typeof rr.gated !== 'boolean') {\n      throw new RunRecordValidationError(\n        'outcome.realness.gated must be a boolean',\n        'outcome.realness.gated',\n      )\n    }\n  }\n\n  // Per-judge / per-dim breakdown, optional.\n  if (outRec.judgeScores !== undefined) {\n    validateJudgeScores(outRec.judgeScores, 'outcome.judgeScores')\n  }\n\n  // Failure mode optional.\n  if (obj.failureMode !== undefined) expectString(obj.failureMode, 'failureMode')\n\n  if (obj.agentProfile !== undefined) {\n    try {\n      const profile = validateAgentProfileCell(obj.agentProfile)\n      if (profile.model !== undefined && profile.model !== obj.model) {\n        throw new RunRecordValidationError(\n          `agentProfile.model \"${profile.model}\" does not match model \"${obj.model}\"`,\n          'agentProfile.model',\n        )\n      }\n      if (profile.promptHash !== undefined && profile.promptHash !== obj.promptHash) {\n        throw new RunRecordValidationError(\n          `agentProfile.promptHash \"${profile.promptHash}\" does not match promptHash \"${obj.promptHash}\"`,\n          'agentProfile.promptHash',\n        )\n      }\n    } catch (error) {\n      if (error instanceof RunRecordValidationError) throw error\n      if (error instanceof Error) {\n        throw new RunRecordValidationError(error.message, 'agentProfile')\n      }\n      throw error\n    }\n  }\n\n  // Split tag.\n  if (typeof obj.splitTag !== 'string' || !SPLIT_TAGS.includes(obj.splitTag as RunSplitTag)) {\n    throw new RunRecordValidationError(\n      `splitTag must be one of ${SPLIT_TAGS.join(', ')}, got ${String(obj.splitTag)}`,\n      'splitTag',\n    )\n  }\n\n  return input as RunRecord\n}\n\n/** Boolean validator — convenience for filtering arrays. */\nexport function isRunRecord(input: unknown): input is RunRecord {\n  try {\n    validateRunRecord(input)\n    return true\n  } catch {\n    return false\n  }\n}\n\n/** Non-throwing validator — returns a discriminated union. */\nexport function parseRunRecordSafe(\n  input: unknown,\n): { ok: true; value: RunRecord } | { ok: false; error: RunRecordValidationError } {\n  try {\n    return { ok: true, value: validateRunRecord(input) }\n  } catch (e) {\n    if (e instanceof RunRecordValidationError) return { ok: false, error: e }\n    throw e\n  }\n}\n\n/** Round-trip helper — `JSON.parse(JSON.stringify(record))` then validate. */\nexport function roundTripRunRecord(record: RunRecord): RunRecord {\n  const json = JSON.stringify(record)\n  return validateRunRecord(JSON.parse(json))\n}\n\n// ── Internals ────────────────────────────────────────────────────────\n\nfunction expectString(value: unknown, path: string): void {\n  if (typeof value !== 'string' || value.length === 0) {\n    throw new RunRecordValidationError(`expected non-empty string`, path)\n  }\n}\n\nfunction expectFiniteNumber(value: unknown, path: string): void {\n  if (typeof value !== 'number' || !Number.isFinite(value)) {\n    throw new RunRecordValidationError(`expected finite number`, path)\n  }\n}\n\nfunction validateJudgeScores(value: unknown, path: string): void {\n  if (value === null || typeof value !== 'object') {\n    throw new RunRecordValidationError('judgeScores must be an object', path)\n  }\n  const rec = value as Record<string, unknown>\n\n  const perJudge = rec.perJudge\n  if (perJudge === null || typeof perJudge !== 'object') {\n    throw new RunRecordValidationError('perJudge must be an object', `${path}.perJudge`)\n  }\n  for (const [judgeId, dims] of Object.entries(perJudge as Record<string, unknown>)) {\n    if (dims === null || typeof dims !== 'object') {\n      throw new RunRecordValidationError(\n        'per-judge entry must be an object of dimension scores',\n        `${path}.perJudge.${judgeId}`,\n      )\n    }\n    for (const [dim, score] of Object.entries(dims as Record<string, unknown>)) {\n      expectFiniteNumber(score, `${path}.perJudge.${judgeId}.${dim}`)\n    }\n  }\n\n  const perDimMean = rec.perDimMean\n  if (perDimMean === null || typeof perDimMean !== 'object') {\n    throw new RunRecordValidationError('perDimMean must be an object', `${path}.perDimMean`)\n  }\n  for (const [dim, mean] of Object.entries(perDimMean as Record<string, unknown>)) {\n    expectFiniteNumber(mean, `${path}.perDimMean.${dim}`)\n  }\n\n  expectFiniteNumber(rec.composite, `${path}.composite`)\n\n  if (rec.failedJudges !== undefined) {\n    if (!Array.isArray(rec.failedJudges)) {\n      throw new RunRecordValidationError(\n        'failedJudges must be an array of strings',\n        `${path}.failedJudges`,\n      )\n    }\n    for (let i = 0; i < rec.failedJudges.length; i++) {\n      const id = rec.failedJudges[i]\n      if (typeof id !== 'string' || id.length === 0) {\n        throw new RunRecordValidationError(\n          'failedJudges entry must be a non-empty string',\n          `${path}.failedJudges[${i}]`,\n        )\n      }\n    }\n  }\n\n  if (rec.notes !== undefined && typeof rec.notes !== 'string') {\n    throw new RunRecordValidationError('notes must be a string', `${path}.notes`)\n  }\n}\n\n/**\n * Heuristic snapshot check. Accepts:\n *   - `name@YYYY-MM-DD` (Anthropic style: `claude-sonnet-4-6@2025-04-15`)\n *   - `name-YYYYMMDD`   (OpenAI style: `gpt-4o-2024-11-20`)\n *   - `name@<arbitrary-token>` (allow opaque snapshots like `@v3`)\n *   - explicit `:date-...` Vertex-style tags\n *\n * Rejects bare aliases like `claude-sonnet-4` or `gpt-4o` that remap\n * silently as providers ship new snapshots.\n */\nfunction modelHasSnapshot(model: string): boolean {\n  if (model.includes('@')) return true\n  if (/-\\d{8}$/.test(model)) return true\n  if (/-\\d{4}-\\d{2}-\\d{2}$/.test(model)) return true\n  if (/:date-/.test(model)) return true\n  return false\n}\n"],"mappings":";;;;;;;;AAwDO,IAAM,kCAAN,cAA8C,gBAAgB;AAAA,EAC1D;AAAA,EACT,YAAY,SAAiB,OAAO,IAAI;AACtC,UAAM,OAAO,GAAG,OAAO,QAAQ,IAAI,MAAM,OAAO;AAChD,SAAK,OAAO;AAAA,EACd;AACF;AAEA,IAAM,aAAa;AACnB,IAAM,UAAU;AAEhB,eAAsB,sBACpB,OAC2B;AAC3B,QAAM,WAAW,MAAM,+BAA+B,KAAK;AAC3D,QAAM,SAAS,6BAA6B,MAAM,SAAS,QAAQ,CAAC;AACpE,SAAO,EAAE,GAAG,UAAU,OAAO;AAC/B;AAEO,SAAS,6BACd,MACkC;AAClC,QAAM,EAAE,QAAQ,SAAS,GAAG,SAAS,IAAI;AACzC,OAAK;AACL,SAAO,0BAA0B,QAAQ;AAC3C;AAEA,eAAsB,uBAAuB,MAA0C;AACrF,2BAAyB,IAAI;AAC7B,SACE,KAAK,WACL,6BAA6B,MAAM,SAAS,6BAA6B,IAAI,CAAC,CAAC;AAEnF;AAEO,SAAS,yBAAyB,OAAkC;AACzE,MAAI,UAAU,QAAQ,OAAO,UAAU,UAAU;AAC/C,UAAM,IAAI,gCAAgC,iBAAiB;AAAA,EAC7D;AACA,QAAM,MAAM;AACZ,gBAAc,IAAI,eAAe,yBAAyB,eAAe;AACzE,MAAI,OAAO,IAAI,WAAW,YAAY,CAAC,QAAQ,KAAK,IAAI,MAAM,GAAG;AAC/D,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,eAAa,IAAI,WAAW,WAAW;AACvC,iBAAe,IAAI,eAAe,eAAe;AACjD,MAAI,IAAI,YAAY,OAAW,iBAAgB,IAAI,SAAS,SAAS;AACrE,MAAI,IAAI,UAAU,OAAW,cAAa,IAAI,OAAO,OAAO;AAC5D,MAAI,IAAI,eAAe,OAAW,cAAa,IAAI,YAAY,YAAY;AAC3E,MAAI,IAAI,eAAe,OAAW,oBAAmB,IAAI,YAAY,YAAY;AACjF,SAAO;AACT;AAEO,SAAS,wBAAwB,QAGnB;AACnB,MAAI,CAAC,OAAO,cAAc;AACxB,UAAM,IAAI;AAAA,MACR,QAAQ,OAAO,KAAK;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AACA,SAAO,yBAAyB,OAAO,YAAY;AACrD;AAEO,SAAS,oBAAoB,QAGzB;AACT,SAAO,wBAAwB,MAAM,EAAE;AACzC;AAEA,eAAsB,0BAA0B,QAKlB;AAC5B,QAAM,UAAU,wBAAwB,MAAM;AAC9C,MAAI,CAAE,MAAM,uBAAuB,OAAO,GAAI;AAC5C,UAAM,IAAI;AAAA,MACR,QAAQ,OAAO,KAAK;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AACA,MAAI,QAAQ,UAAU,UAAa,QAAQ,UAAU,OAAO,OAAO;AACjE,UAAM,IAAI;AAAA,MACR,QAAQ,OAAO,KAAK,yBAAyB,QAAQ,KAAK,2BAA2B,OAAO,KAAK;AAAA,MACjG;AAAA,IACF;AAAA,EACF;AACA,MAAI,QAAQ,eAAe,UAAa,QAAQ,eAAe,OAAO,YAAY;AAChF,UAAM,IAAI;AAAA,MACR,QAAQ,OAAO,KAAK,8BAA8B,QAAQ,UAAU,gCAAgC,OAAO,UAAU;AAAA,MACrH;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AAEO,SAAS,4BAEd,SAAyC;AACzC,QAAM,SAAS,oBAAI,IAAiB;AACpC,aAAW,UAAU,SAAS;AAC5B,UAAM,MAAM,oBAAoB,MAAM;AACtC,UAAM,SAAS,OAAO,IAAI,GAAG;AAC7B,QAAI,OAAQ,QAAO,KAAK,MAAM;AAAA,QACzB,QAAO,IAAI,KAAK,CAAC,MAAM,CAAC;AAAA,EAC/B;AACA,SAAO;AACT;AAEA,eAAe,+BACb,OAC2C;AAC3C,SAAO,0BAA0B;AAAA,IAC/B,eAAe;AAAA,IACf,WAAW,MAAM;AAAA,IACjB,eAAe,MAAM,qBAAqB,MAAM,aAAa;AAAA,IAC7D,SAAS,MAAM;AAAA,IACf,OAAO,MAAM;AAAA,IACb,YAAY,MAAM;AAAA,IAClB,YAAY,MAAM;AAAA,EACpB,CAAC;AACH;AAEA,SAAS,0BACP,OACkC;AAClC,SAAO,cAAc;AAAA,IACnB,eAAe;AAAA,IACf,WAAW,gBAAgB,MAAM,WAAW,WAAW;AAAA,IACvD,eAAe,gBAAgB,MAAM,aAAa;AAAA,IAClD,SAAS,MAAM,UAAU,iBAAiB,MAAM,SAAS,SAAS,IAAI;AAAA,IACtE,OAAO,iBAAiB,MAAM,OAAO,OAAO;AAAA,IAC5C,YAAY,iBAAiB,MAAM,YAAY,YAAY;AAAA,IAC3D,YAAY,MAAM,aACd,eAAe,oBAAoB,MAAM,UAAU,CAAC,IACpD;AAAA,EACN,CAAC;AACH;AAEA,eAAe,qBAAqB,OAA6D;AAC/F,QAAM,OAAO,gBAAgB,MAAM,MAAM,oBAAoB;AAC7D,MAAI,MAAM,SAAS,UAAa,MAAM,YAAY,QAAW;AAC3D,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,MAAI,MAAM,SAAS,QAAW;AAC5B,WAAO,EAAE,MAAM,MAAM,iBAAiB,MAAM,MAAM,oBAAoB,EAAE;AAAA,EAC1E;AACA,MAAI,MAAM,YAAY,QAAW;AAC/B,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,aAAW,MAAM,SAAS,uBAAuB;AACjD,SAAO,EAAE,MAAM,MAAM,MAAM,SAAS,MAAM,OAAO,EAAE;AACrD;AAEA,SAAS,gBAAgB,OAA+C;AACtE,SAAO;AAAA,IACL,MAAM,gBAAgB,MAAM,MAAM,oBAAoB;AAAA,IACtD,MAAM,iBAAiB,MAAM,MAAM,oBAAoB;AAAA,EACzD;AACF;AAEA,SAAS,iBAAiB,OAA4B,MAAmC;AACvF,SAAO,cAAc;AAAA,IACnB,IAAI,gBAAgB,MAAM,IAAI,GAAG,IAAI,KAAK;AAAA,IAC1C,SAAS,iBAAiB,MAAM,SAAS,GAAG,IAAI,UAAU;AAAA,IAC1D,MAAM,iBAAiB,MAAM,MAAM,GAAG,IAAI,OAAO;AAAA,EACnD,CAAC;AACH;AAEA,SAAS,oBACP,OAC4C;AAC5C,QAAM,MAAkD,CAAC;AACzD,aAAW,OAAO,OAAO,KAAK,KAAK,EAAE,KAAK,GAAG;AAC3C,UAAM,QAAQ,MAAM,GAAG;AACvB,oBAAgB,KAAK,kBAAkB;AACvC,QACE,UAAU,QACV,OAAO,UAAU,YACjB,OAAO,UAAU,YACjB,OAAO,UAAU,WACjB;AACA,YAAM,IAAI;AAAA,QACR;AAAA,QACA,cAAc,GAAG;AAAA,MACnB;AAAA,IACF;AACA,QAAI,OAAO,UAAU,YAAY,CAAC,OAAO,SAAS,KAAK,GAAG;AACxD,YAAM,IAAI,gCAAgC,0BAA0B,cAAc,GAAG,EAAE;AAAA,IACzF;AACA,QAAI,GAAG,IAAI;AAAA,EACb;AACA,SAAO;AACT;AAEA,SAAS,cAAiD,OAAa;AACrE,QAAM,MAA+B,CAAC;AACtC,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,KAAK,GAAG;AAChD,QAAI,UAAU,OAAW,KAAI,GAAG,IAAI;AAAA,EACtC;AACA,SAAO;AACT;AAEA,SAAS,eAAkD,OAAyB;AAClF,SAAO,OAAO,KAAK,KAAK,EAAE,SAAS,IAAI,QAAQ;AACjD;AAEA,SAAS,eAAe,OAAgB,MAAoB;AAC1D,MAAI,UAAU,QAAQ,OAAO,UAAU,YAAY,MAAM,QAAQ,KAAK,GAAG;AACvE,UAAM,IAAI,gCAAgC,mBAAmB,IAAI;AAAA,EACnE;AACA,QAAM,MAAM;AACZ,eAAa,IAAI,MAAM,GAAG,IAAI,OAAO;AACrC,mBAAiB,IAAI,MAAM,GAAG,IAAI,OAAO;AAC3C;AAEA,SAAS,gBAAgB,OAAgB,MAAoB;AAC3D,MAAI,UAAU,QAAQ,OAAO,UAAU,YAAY,MAAM,QAAQ,KAAK,GAAG;AACvE,UAAM,IAAI,gCAAgC,mBAAmB,IAAI;AAAA,EACnE;AACA,QAAM,MAAM;AACZ,eAAa,IAAI,IAAI,GAAG,IAAI,KAAK;AACjC,MAAI,IAAI,YAAY,OAAW,cAAa,IAAI,SAAS,GAAG,IAAI,UAAU;AAC1E,MAAI,IAAI,SAAS,OAAW,cAAa,IAAI,MAAM,GAAG,IAAI,OAAO;AACnE;AAEA,SAAS,mBAAmB,OAAgB,MAAoB;AAC9D,MAAI,UAAU,QAAQ,OAAO,UAAU,YAAY,MAAM,QAAQ,KAAK,GAAG;AACvE,UAAM,IAAI,gCAAgC,mBAAmB,IAAI;AAAA,EACnE;AACA,sBAAoB,KAAmD;AACzE;AAEA,SAAS,WAAW,OAAyB,MAAoB;AAC/D,MAAI,UAAU,KAAM;AACpB,QAAM,OAAO,OAAO;AACpB,MAAI,SAAS,YAAY,SAAS,UAAW;AAC7C,MAAI,SAAS,UAAU;AACrB,QAAI,CAAC,OAAO,SAAS,KAAK,GAAG;AAC3B,YAAM,IAAI,gCAAgC,0BAA0B,IAAI;AAAA,IAC1E;AACA;AAAA,EACF;AACA,MAAI,MAAM,QAAQ,KAAK,GAAG;AACxB,UAAM,QAAQ,CAAC,MAAM,UAAU;AAC7B,iBAAW,MAAM,GAAG,IAAI,IAAI,KAAK,GAAG;AAAA,IACtC,CAAC;AACD;AAAA,EACF;AACA,MAAI,SAAS,UAAU;AACrB,eAAW,CAAC,KAAK,MAAM,KAAK,OAAO,QAAQ,KAAK,GAAG;AACjD,sBAAgB,KAAK,GAAG,IAAI,QAAQ;AACpC,iBAAW,QAAQ,GAAG,IAAI,IAAI,GAAG,EAAE;AAAA,IACrC;AACA;AAAA,EACF;AACA,QAAM,IAAI,gCAAgC,kCAAkC,IAAI;AAClF;AAEA,SAAS,cAAc,OAAgB,UAAkB,MAAoB;AAC3E,MAAI,UAAU,UAAU;AACtB,UAAM,IAAI,gCAAgC,YAAY,QAAQ,IAAI,IAAI;AAAA,EACxE;AACF;AAEA,SAAS,aAAa,OAAgB,MAAoB;AACxD,MAAI,OAAO,UAAU,YAAY,MAAM,WAAW,GAAG;AACnD,UAAM,IAAI,gCAAgC,6BAA6B,IAAI;AAAA,EAC7E;AACF;AAEA,SAAS,gBAAgB,OAAe,MAAsB;AAC5D,MAAI,OAAO,UAAU,YAAY,MAAM,WAAW,GAAG;AACnD,UAAM,IAAI,gCAAgC,6BAA6B,IAAI;AAAA,EAC7E;AACA,SAAO;AACT;AAEA,SAAS,iBAAiB,OAA2B,MAAkC;AACrF,MAAI,UAAU,OAAW,QAAO;AAChC,SAAO,gBAAgB,OAAO,IAAI;AACpC;AAEA,SAAS,iBAAiB,OAAgB,MAAsB;AAC9D,MAAI,OAAO,UAAU,YAAY,CAAC,WAAW,KAAK,KAAK,GAAG;AACxD,UAAM,IAAI,gCAAgC,0CAA0C,IAAI;AAAA,EAC1F;AACA,SAAO;AACT;AAuBO,IAAM,sBAAsB;AAAA;AAAA;AAAA;AAAA,EAIjC,uBAAuB;AACzB;AAQO,SAAS,mBAAmB,OAAkC;AACnE,MAAI;AACJ,MAAI;AACF,iBAAa,KAAK,UAAU,KAAK;AAAA,EACnC,SAAS,KAAK;AACZ,UAAM,IAAI;AAAA,MACR,4CAA4C,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG,CAAC;AAAA,MAC5F;AAAA,IACF;AAAA,EACF;AACA,MAAI,eAAe,QAAW;AAC5B,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,SAAO,KAAK,MAAM,UAAU;AAC9B;AAsBA,eAAsB,6BACpB,SACA,OAC2B;AAC3B,MAAI,CAAC,WAAW,OAAO,YAAY,UAAU;AAC3C,UAAM,IAAI,gCAAgC,0CAA0C,SAAS;AAAA,EAC/F;AACA,MAAI,OAAO,QAAQ,SAAS,YAAY,QAAQ,KAAK,WAAW,GAAG;AACjE,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,MAAI,OAAO,QAAQ,YAAY,YAAY,QAAQ,QAAQ,WAAW,GAAG;AACvE,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,SAAO,sBAAsB;AAAA,IAC3B,GAAG;AAAA,IACH,WAAW,GAAG,QAAQ,IAAI,IAAI,QAAQ,OAAO;AAAA,IAC7C,eAAe;AAAA,MACb,MAAM,oBAAoB;AAAA,MAC1B,SAAS,mBAAmB,OAAO;AAAA,IACrC;AAAA,EACF,CAAC;AACH;;;ACxQA,IAAM,sBAAsB;AAAA,EAC1B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEA,IAAM,aAAyC,CAAC,UAAU,OAAO,SAAS;AAEnE,IAAM,2BAAN,cAAuC,gBAAgB;AAAA,EACnD;AAAA,EACT,YAAY,SAAiB,OAAO,IAAI;AACtC,UAAM,OAAO,GAAG,OAAO,QAAQ,IAAI,MAAM,OAAO;AAChD,SAAK,OAAO;AAAA,EACd;AACF;AAOO,SAAS,kBAAkB,OAA2B;AAC3D,MAAI,UAAU,QAAQ,OAAO,UAAU,UAAU;AAC/C,UAAM,IAAI,yBAAyB,iBAAiB;AAAA,EACtD;AACA,QAAM,MAAM;AAEZ,aAAW,OAAO,qBAAqB;AACrC,QAAI,EAAE,OAAO,MAAM;AACjB,YAAM,IAAI,yBAAyB,4BAA4B,GAAG,GAAG;AAAA,IACvE;AAAA,EACF;AAEA,EAAAA,cAAa,IAAI,OAAO,OAAO;AAC/B,EAAAA,cAAa,IAAI,cAAc,cAAc;AAC7C,EAAAA,cAAa,IAAI,aAAa,aAAa;AAC3C,qBAAmB,IAAI,MAAM,MAAM;AACnC,EAAAA,cAAa,IAAI,OAAO,OAAO;AAC/B,EAAAA,cAAa,IAAI,YAAY,YAAY;AACzC,EAAAA,cAAa,IAAI,YAAY,YAAY;AACzC,EAAAA,cAAa,IAAI,WAAW,WAAW;AACvC,qBAAmB,IAAI,QAAQ,QAAQ;AACvC,MAAI,IAAI,YAAY,OAAW,oBAAmB,IAAI,SAAS,SAAS;AACxE,qBAAmB,IAAI,SAAS,SAAS;AAGzC,MAAI,CAAC,iBAAiB,IAAI,KAAe,GAAG;AAC1C,UAAM,IAAI;AAAA,MACR,UAAU,IAAI,KAAK;AAAA,MACnB;AAAA,IACF;AAAA,EACF;AAGA,QAAM,KAAK,IAAI;AACf,MAAI,OAAO,QAAQ,OAAO,OAAO,UAAU;AACzC,UAAM,IAAI,yBAAyB,gCAAgC,YAAY;AAAA,EACjF;AACA,QAAM,QAAQ;AACd,qBAAmB,MAAM,OAAO,kBAAkB;AAClD,qBAAmB,MAAM,QAAQ,mBAAmB;AACpD,MAAI,MAAM,WAAW,OAAW,oBAAmB,MAAM,QAAQ,mBAAmB;AAGpF,MAAI,IAAI,kBAAkB,QAAW;AACnC,UAAM,KAAK,IAAI;AACf,QAAI,OAAO,QAAQ,OAAO,OAAO,UAAU;AACzC,YAAM,IAAI,yBAAyB,mCAAmC,eAAe;AAAA,IACvF;AACA,UAAM,QAAQ;AACd,IAAAA,cAAa,MAAM,OAAO,qBAAqB;AAC/C,IAAAA,cAAa,MAAM,eAAe,6BAA6B;AAC/D,uBAAmB,MAAM,YAAY,0BAA0B;AAC/D,QAAI,OAAO,MAAM,aAAa,WAAW;AACvC,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,QAAM,MAAM,IAAI;AAChB,MAAI,QAAQ,QAAQ,OAAO,QAAQ,UAAU;AAC3C,UAAM,IAAI,yBAAyB,6BAA6B,SAAS;AAAA,EAC3E;AACA,QAAM,SAAS;AACf,MAAI,OAAO,gBAAgB;AACzB,uBAAmB,OAAO,aAAa,qBAAqB;AAC9D,MAAI,OAAO,iBAAiB;AAC1B,uBAAmB,OAAO,cAAc,sBAAsB;AAChE,MAAI,OAAO,gBAAgB,UAAa,OAAO,iBAAiB,QAAW;AACzE,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,QAAM,MAAM,OAAO;AACnB,MAAI,QAAQ,QAAQ,OAAO,QAAQ,UAAU;AAC3C,UAAM,IAAI,yBAAyB,iCAAiC,aAAa;AAAA,EACnF;AACA,aAAW,CAAC,GAAG,CAAC,KAAK,OAAO,QAAQ,GAA8B,GAAG;AACnE,uBAAmB,GAAG,eAAe,CAAC,EAAE;AAAA,EAC1C;AAEA,MAAI,OAAO,aAAa,QAAW;AACjC,UAAM,IAAI,OAAO;AACjB,QAAI,MAAM,QAAQ,OAAO,MAAM,UAAU;AACvC,YAAM,IAAI,yBAAyB,sCAAsC,kBAAkB;AAAA,IAC7F;AACA,UAAM,KAAK;AACX,uBAAmB,GAAG,OAAO,wBAAwB;AACrD,QAAI,OAAO,GAAG,UAAU,WAAW;AACjC,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,MAAI,OAAO,gBAAgB,QAAW;AACpC,wBAAoB,OAAO,aAAa,qBAAqB;AAAA,EAC/D;AAGA,MAAI,IAAI,gBAAgB,OAAW,CAAAA,cAAa,IAAI,aAAa,aAAa;AAE9E,MAAI,IAAI,iBAAiB,QAAW;AAClC,QAAI;AACF,YAAM,UAAU,yBAAyB,IAAI,YAAY;AACzD,UAAI,QAAQ,UAAU,UAAa,QAAQ,UAAU,IAAI,OAAO;AAC9D,cAAM,IAAI;AAAA,UACR,uBAAuB,QAAQ,KAAK,2BAA2B,IAAI,KAAK;AAAA,UACxE;AAAA,QACF;AAAA,MACF;AACA,UAAI,QAAQ,eAAe,UAAa,QAAQ,eAAe,IAAI,YAAY;AAC7E,cAAM,IAAI;AAAA,UACR,4BAA4B,QAAQ,UAAU,gCAAgC,IAAI,UAAU;AAAA,UAC5F;AAAA,QACF;AAAA,MACF;AAAA,IACF,SAAS,OAAO;AACd,UAAI,iBAAiB,yBAA0B,OAAM;AACrD,UAAI,iBAAiB,OAAO;AAC1B,cAAM,IAAI,yBAAyB,MAAM,SAAS,cAAc;AAAA,MAClE;AACA,YAAM;AAAA,IACR;AAAA,EACF;AAGA,MAAI,OAAO,IAAI,aAAa,YAAY,CAAC,WAAW,SAAS,IAAI,QAAuB,GAAG;AACzF,UAAM,IAAI;AAAA,MACR,2BAA2B,WAAW,KAAK,IAAI,CAAC,SAAS,OAAO,IAAI,QAAQ,CAAC;AAAA,MAC7E;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AACT;AAGO,SAAS,YAAY,OAAoC;AAC9D,MAAI;AACF,sBAAkB,KAAK;AACvB,WAAO;AAAA,EACT,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAGO,SAAS,mBACd,OACiF;AACjF,MAAI;AACF,WAAO,EAAE,IAAI,MAAM,OAAO,kBAAkB,KAAK,EAAE;AAAA,EACrD,SAAS,GAAG;AACV,QAAI,aAAa,yBAA0B,QAAO,EAAE,IAAI,OAAO,OAAO,EAAE;AACxE,UAAM;AAAA,EACR;AACF;AAGO,SAAS,mBAAmB,QAA8B;AAC/D,QAAM,OAAO,KAAK,UAAU,MAAM;AAClC,SAAO,kBAAkB,KAAK,MAAM,IAAI,CAAC;AAC3C;AAIA,SAASA,cAAa,OAAgB,MAAoB;AACxD,MAAI,OAAO,UAAU,YAAY,MAAM,WAAW,GAAG;AACnD,UAAM,IAAI,yBAAyB,6BAA6B,IAAI;AAAA,EACtE;AACF;AAEA,SAAS,mBAAmB,OAAgB,MAAoB;AAC9D,MAAI,OAAO,UAAU,YAAY,CAAC,OAAO,SAAS,KAAK,GAAG;AACxD,UAAM,IAAI,yBAAyB,0BAA0B,IAAI;AAAA,EACnE;AACF;AAEA,SAAS,oBAAoB,OAAgB,MAAoB;AAC/D,MAAI,UAAU,QAAQ,OAAO,UAAU,UAAU;AAC/C,UAAM,IAAI,yBAAyB,iCAAiC,IAAI;AAAA,EAC1E;AACA,QAAM,MAAM;AAEZ,QAAM,WAAW,IAAI;AACrB,MAAI,aAAa,QAAQ,OAAO,aAAa,UAAU;AACrD,UAAM,IAAI,yBAAyB,8BAA8B,GAAG,IAAI,WAAW;AAAA,EACrF;AACA,aAAW,CAAC,SAAS,IAAI,KAAK,OAAO,QAAQ,QAAmC,GAAG;AACjF,QAAI,SAAS,QAAQ,OAAO,SAAS,UAAU;AAC7C,YAAM,IAAI;AAAA,QACR;AAAA,QACA,GAAG,IAAI,aAAa,OAAO;AAAA,MAC7B;AAAA,IACF;AACA,eAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,IAA+B,GAAG;AAC1E,yBAAmB,OAAO,GAAG,IAAI,aAAa,OAAO,IAAI,GAAG,EAAE;AAAA,IAChE;AAAA,EACF;AAEA,QAAM,aAAa,IAAI;AACvB,MAAI,eAAe,QAAQ,OAAO,eAAe,UAAU;AACzD,UAAM,IAAI,yBAAyB,gCAAgC,GAAG,IAAI,aAAa;AAAA,EACzF;AACA,aAAW,CAAC,KAAK,IAAI,KAAK,OAAO,QAAQ,UAAqC,GAAG;AAC/E,uBAAmB,MAAM,GAAG,IAAI,eAAe,GAAG,EAAE;AAAA,EACtD;AAEA,qBAAmB,IAAI,WAAW,GAAG,IAAI,YAAY;AAErD,MAAI,IAAI,iBAAiB,QAAW;AAClC,QAAI,CAAC,MAAM,QAAQ,IAAI,YAAY,GAAG;AACpC,YAAM,IAAI;AAAA,QACR;AAAA,QACA,GAAG,IAAI;AAAA,MACT;AAAA,IACF;AACA,aAAS,IAAI,GAAG,IAAI,IAAI,aAAa,QAAQ,KAAK;AAChD,YAAM,KAAK,IAAI,aAAa,CAAC;AAC7B,UAAI,OAAO,OAAO,YAAY,GAAG,WAAW,GAAG;AAC7C,cAAM,IAAI;AAAA,UACR;AAAA,UACA,GAAG,IAAI,iBAAiB,CAAC;AAAA,QAC3B;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,MAAI,IAAI,UAAU,UAAa,OAAO,IAAI,UAAU,UAAU;AAC5D,UAAM,IAAI,yBAAyB,0BAA0B,GAAG,IAAI,QAAQ;AAAA,EAC9E;AACF;AAYA,SAAS,iBAAiB,OAAwB;AAChD,MAAI,MAAM,SAAS,GAAG,EAAG,QAAO;AAChC,MAAI,UAAU,KAAK,KAAK,EAAG,QAAO;AAClC,MAAI,sBAAsB,KAAK,KAAK,EAAG,QAAO;AAC9C,MAAI,SAAS,KAAK,KAAK,EAAG,QAAO;AACjC,SAAO;AACT;","names":["expectString"]}

package/dist/chunk-NPCTHQIO.js

@@ -0,0 +1,91 @@
+// src/meta-eval/calibration.ts
+async function calibrationCurve(traceStore, outcomeStore, evalMetric, outcomeMetric, options = {}) {
+  const runs = await traceStore.listRuns();
+  const outcomes = await outcomeStore.list();
+  const byRun = /* @__PURE__ */ new Map();
+  for (const o of outcomes) {
+    const arr = byRun.get(o.runId) ?? [];
+    arr.push(o);
+    byRun.set(o.runId, arr);
+  }
+  const extract = evalMetric.extract ?? defaultExtract(evalMetric.id);
+  const pairs = [];
+  for (const run of runs) {
+    const os = byRun.get(run.runId);
+    if (!os?.length) continue;
+    const x = await extract(run, traceStore);
+    if (x === null || !Number.isFinite(x)) continue;
+    const latest = [...os].sort((a, b) => b.capturedAt - a.capturedAt)[0];
+    const y = latest.metrics[outcomeMetric];
+    if (typeof y !== "number" || !Number.isFinite(y)) continue;
+    pairs.push({ x, y });
+  }
+  if (pairs.length < 2) return null;
+  return calibrationFromPairs(
+    pairs.map((p) => ({ evalScore: p.x, outcome: p.y })),
+    evalMetric.id,
+    outcomeMetric,
+    options
+  );
+}
+function calibrationFromPairs(inputPairs, evalMetric, outcomeMetric, options = {}) {
+  const pairs = inputPairs.filter(
+    (pair) => Number.isFinite(pair.evalScore) && Number.isFinite(pair.outcome)
+  );
+  if (pairs.length < 2) return null;
+  const numBins = options.bins ?? 10;
+  const binning = options.binning ?? "equal-width";
+  const xs = pairs.map((p) => p.evalScore);
+  const lo = options.range?.lo ?? Math.min(...xs);
+  const hi = options.range?.hi ?? Math.max(...xs);
+  const bins = [];
+  if (binning === "equal-frequency") {
+    const sorted = [...pairs].sort((a, b) => a.evalScore - b.evalScore);
+    const perBin = Math.max(1, Math.floor(sorted.length / numBins));
+    for (let i = 0; i < sorted.length; i += perBin) {
+      const chunk = sorted.slice(i, i + perBin);
+      if (chunk.length === 0) continue;
+      bins.push(toBin(chunk));
+    }
+  } else {
+    const width = (hi - lo) / numBins;
+    if (width === 0) return null;
+    for (let i = 0; i < numBins; i++) {
+      const binLo = lo + i * width;
+      const binHi = i === numBins - 1 ? hi + 1e-9 : lo + (i + 1) * width;
+      const chunk = pairs.filter((p) => p.evalScore >= binLo && p.evalScore < binHi);
+      if (chunk.length === 0) continue;
+      bins.push(toBin(chunk, binLo, binHi));
+    }
+  }
+  const total = bins.reduce((a, b) => a + b.n, 0);
+  const ece = bins.reduce((a, b) => a + b.n / total * b.gap, 0);
+  const maxGap = bins.reduce((a, b) => Math.max(a, b.gap), 0);
+  return { evalMetric, outcomeMetric, n: pairs.length, bins, ece, maxGap };
+}
+function toBin(chunk, lower, upper) {
+  const xs = chunk.map((c) => c.evalScore);
+  const ys = chunk.map((c) => c.outcome);
+  const evalMean = mean(xs);
+  const outcomeMean = mean(ys);
+  return {
+    lower: lower ?? Math.min(...xs),
+    upper: upper ?? Math.max(...xs),
+    n: chunk.length,
+    evalMean,
+    outcomeMean,
+    gap: Math.abs(outcomeMean - evalMean)
+  };
+}
+function mean(xs) {
+  return xs.reduce((a, b) => a + b, 0) / xs.length;
+}
+function defaultExtract(metric) {
+  return async (run) => run.outcome?.score ?? (metric === "pass" ? run.outcome?.pass === true ? 1 : 0 : null);
+}
+
+export {
+  calibrationCurve,
+  calibrationFromPairs
+};
+//# sourceMappingURL=chunk-NPCTHQIO.js.map

package/dist/chunk-NPCTHQIO.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/meta-eval/calibration.ts"],"sourcesContent":["/**\n * Calibration curve — binned \"if eval says X, what does reality show?\"\n *\n * Companion to correlationStudy. Raw correlation is a single number;\n * the calibration curve shows *where* the eval is well-calibrated vs\n * overconfident / underconfident. Buckets the eval metric, computes\n * mean outcome per bucket, reports expected-calibration-error (ECE).\n */\n\nimport type { Run } from '../trace/schema'\nimport type { TraceStore } from '../trace/store'\nimport type { EvalMetricSpec } from './correlation-study'\nimport type { DeploymentOutcome, OutcomeStore } from './outcome-store'\n\nexport interface CalibrationBin {\n  lower: number\n  upper: number\n  n: number\n  evalMean: number\n  outcomeMean: number\n  /** |outcomeMean − evalMean|; contributes to ECE weighted by n/total. */\n  gap: number\n}\n\nexport interface CalibrationReport {\n  evalMetric: string\n  outcomeMetric: string\n  n: number\n  bins: CalibrationBin[]\n  /** Expected Calibration Error — Σ (n_i/N) × |outcomeMean_i − evalMean_i|. */\n  ece: number\n  /** Max bin gap — upper bound on miscalibration. */\n  maxGap: number\n}\n\nexport interface CalibrationOptions {\n  bins?: number\n  /** Equal-width (fixed bin edges) or equal-frequency (quantile bins). */\n  binning?: 'equal-width' | 'equal-frequency'\n  /** Clip eval values to [lo, hi] before binning. */\n  range?: { lo: number; hi: number }\n}\n\nexport interface CalibrationPair {\n  evalScore: number\n  outcome: number\n}\n\nexport async function calibrationCurve(\n  traceStore: TraceStore,\n  outcomeStore: OutcomeStore,\n  evalMetric: EvalMetricSpec,\n  outcomeMetric: string,\n  options: CalibrationOptions = {},\n): Promise<CalibrationReport | null> {\n  const runs = await traceStore.listRuns()\n  const outcomes = await outcomeStore.list()\n  const byRun = new Map<string, DeploymentOutcome[]>()\n  for (const o of outcomes) {\n    const arr = byRun.get(o.runId) ?? []\n    arr.push(o)\n    byRun.set(o.runId, arr)\n  }\n\n  const extract = evalMetric.extract ?? defaultExtract(evalMetric.id)\n  const pairs: Array<{ x: number; y: number }> = []\n  for (const run of runs) {\n    const os = byRun.get(run.runId)\n    if (!os?.length) continue\n    const x = await extract(run, traceStore)\n    if (x === null || !Number.isFinite(x)) continue\n    const latest = [...os].sort((a, b) => b.capturedAt - a.capturedAt)[0]!\n    const y = latest.metrics[outcomeMetric]\n    if (typeof y !== 'number' || !Number.isFinite(y)) continue\n    pairs.push({ x, y })\n  }\n  if (pairs.length < 2) return null\n\n  return calibrationFromPairs(\n    pairs.map((p) => ({ evalScore: p.x, outcome: p.y })),\n    evalMetric.id,\n    outcomeMetric,\n    options,\n  )\n}\n\nexport function calibrationFromPairs(\n  inputPairs: CalibrationPair[],\n  evalMetric: string,\n  outcomeMetric: string,\n  options: CalibrationOptions = {},\n): CalibrationReport | null {\n  const pairs = inputPairs.filter(\n    (pair) => Number.isFinite(pair.evalScore) && Number.isFinite(pair.outcome),\n  )\n  if (pairs.length < 2) return null\n\n  const numBins = options.bins ?? 10\n  const binning = options.binning ?? 'equal-width'\n  const xs = pairs.map((p) => p.evalScore)\n  const lo = options.range?.lo ?? Math.min(...xs)\n  const hi = options.range?.hi ?? Math.max(...xs)\n\n  const bins: CalibrationBin[] = []\n  if (binning === 'equal-frequency') {\n    const sorted = [...pairs].sort((a, b) => a.evalScore - b.evalScore)\n    const perBin = Math.max(1, Math.floor(sorted.length / numBins))\n    for (let i = 0; i < sorted.length; i += perBin) {\n      const chunk = sorted.slice(i, i + perBin)\n      if (chunk.length === 0) continue\n      bins.push(toBin(chunk))\n    }\n  } else {\n    const width = (hi - lo) / numBins\n    if (width === 0) return null\n    for (let i = 0; i < numBins; i++) {\n      const binLo = lo + i * width\n      const binHi = i === numBins - 1 ? hi + 1e-9 : lo + (i + 1) * width\n      const chunk = pairs.filter((p) => p.evalScore >= binLo && p.evalScore < binHi)\n      if (chunk.length === 0) continue\n      bins.push(toBin(chunk, binLo, binHi))\n    }\n  }\n\n  const total = bins.reduce((a, b) => a + b.n, 0)\n  const ece = bins.reduce((a, b) => a + (b.n / total) * b.gap, 0)\n  const maxGap = bins.reduce((a, b) => Math.max(a, b.gap), 0)\n\n  return { evalMetric, outcomeMetric, n: pairs.length, bins, ece, maxGap }\n}\n\nfunction toBin(chunk: CalibrationPair[], lower?: number, upper?: number): CalibrationBin {\n  const xs = chunk.map((c) => c.evalScore)\n  const ys = chunk.map((c) => c.outcome)\n  const evalMean = mean(xs)\n  const outcomeMean = mean(ys)\n  return {\n    lower: lower ?? Math.min(...xs),\n    upper: upper ?? Math.max(...xs),\n    n: chunk.length,\n    evalMean,\n    outcomeMean,\n    gap: Math.abs(outcomeMean - evalMean),\n  }\n}\n\nfunction mean(xs: number[]): number {\n  return xs.reduce((a, b) => a + b, 0) / xs.length\n}\n\nfunction defaultExtract(metric: string): (run: Run, store: TraceStore) => Promise<number | null> {\n  return async (run) =>\n    run.outcome?.score ?? (metric === 'pass' ? (run.outcome?.pass === true ? 1 : 0) : null)\n}\n"],"mappings":";AAgDA,eAAsB,iBACpB,YACA,cACA,YACA,eACA,UAA8B,CAAC,GACI;AACnC,QAAM,OAAO,MAAM,WAAW,SAAS;AACvC,QAAM,WAAW,MAAM,aAAa,KAAK;AACzC,QAAM,QAAQ,oBAAI,IAAiC;AACnD,aAAW,KAAK,UAAU;AACxB,UAAM,MAAM,MAAM,IAAI,EAAE,KAAK,KAAK,CAAC;AACnC,QAAI,KAAK,CAAC;AACV,UAAM,IAAI,EAAE,OAAO,GAAG;AAAA,EACxB;AAEA,QAAM,UAAU,WAAW,WAAW,eAAe,WAAW,EAAE;AAClE,QAAM,QAAyC,CAAC;AAChD,aAAW,OAAO,MAAM;AACtB,UAAM,KAAK,MAAM,IAAI,IAAI,KAAK;AAC9B,QAAI,CAAC,IAAI,OAAQ;AACjB,UAAM,IAAI,MAAM,QAAQ,KAAK,UAAU;AACvC,QAAI,MAAM,QAAQ,CAAC,OAAO,SAAS,CAAC,EAAG;AACvC,UAAM,SAAS,CAAC,GAAG,EAAE,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,aAAa,EAAE,UAAU,EAAE,CAAC;AACpE,UAAM,IAAI,OAAO,QAAQ,aAAa;AACtC,QAAI,OAAO,MAAM,YAAY,CAAC,OAAO,SAAS,CAAC,EAAG;AAClD,UAAM,KAAK,EAAE,GAAG,EAAE,CAAC;AAAA,EACrB;AACA,MAAI,MAAM,SAAS,EAAG,QAAO;AAE7B,SAAO;AAAA,IACL,MAAM,IAAI,CAAC,OAAO,EAAE,WAAW,EAAE,GAAG,SAAS,EAAE,EAAE,EAAE;AAAA,IACnD,WAAW;AAAA,IACX;AAAA,IACA;AAAA,EACF;AACF;AAEO,SAAS,qBACd,YACA,YACA,eACA,UAA8B,CAAC,GACL;AAC1B,QAAM,QAAQ,WAAW;AAAA,IACvB,CAAC,SAAS,OAAO,SAAS,KAAK,SAAS,KAAK,OAAO,SAAS,KAAK,OAAO;AAAA,EAC3E;AACA,MAAI,MAAM,SAAS,EAAG,QAAO;AAE7B,QAAM,UAAU,QAAQ,QAAQ;AAChC,QAAM,UAAU,QAAQ,WAAW;AACnC,QAAM,KAAK,MAAM,IAAI,CAAC,MAAM,EAAE,SAAS;AACvC,QAAM,KAAK,QAAQ,OAAO,MAAM,KAAK,IAAI,GAAG,EAAE;AAC9C,QAAM,KAAK,QAAQ,OAAO,MAAM,KAAK,IAAI,GAAG,EAAE;AAE9C,QAAM,OAAyB,CAAC;AAChC,MAAI,YAAY,mBAAmB;AACjC,UAAM,SAAS,CAAC,GAAG,KAAK,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAClE,UAAM,SAAS,KAAK,IAAI,GAAG,KAAK,MAAM,OAAO,SAAS,OAAO,CAAC;AAC9D,aAAS,IAAI,GAAG,IAAI,OAAO,QAAQ,KAAK,QAAQ;AAC9C,YAAM,QAAQ,OAAO,MAAM,GAAG,IAAI,MAAM;AACxC,UAAI,MAAM,WAAW,EAAG;AACxB,WAAK,KAAK,MAAM,KAAK,CAAC;AAAA,IACxB;AAAA,EACF,OAAO;AACL,UAAM,SAAS,KAAK,MAAM;AAC1B,QAAI,UAAU,EAAG,QAAO;AACxB,aAAS,IAAI,GAAG,IAAI,SAAS,KAAK;AAChC,YAAM,QAAQ,KAAK,IAAI;AACvB,YAAM,QAAQ,MAAM,UAAU,IAAI,KAAK,OAAO,MAAM,IAAI,KAAK;AAC7D,YAAM,QAAQ,MAAM,OAAO,CAAC,MAAM,EAAE,aAAa,SAAS,EAAE,YAAY,KAAK;AAC7E,UAAI,MAAM,WAAW,EAAG;AACxB,WAAK,KAAK,MAAM,OAAO,OAAO,KAAK,CAAC;AAAA,IACtC;AAAA,EACF;AAEA,QAAM,QAAQ,KAAK,OAAO,CAAC,GAAG,MAAM,IAAI,EAAE,GAAG,CAAC;AAC9C,QAAM,MAAM,KAAK,OAAO,CAAC,GAAG,MAAM,IAAK,EAAE,IAAI,QAAS,EAAE,KAAK,CAAC;AAC9D,QAAM,SAAS,KAAK,OAAO,CAAC,GAAG,MAAM,KAAK,IAAI,GAAG,EAAE,GAAG,GAAG,CAAC;AAE1D,SAAO,EAAE,YAAY,eAAe,GAAG,MAAM,QAAQ,MAAM,KAAK,OAAO;AACzE;AAEA,SAAS,MAAM,OAA0B,OAAgB,OAAgC;AACvF,QAAM,KAAK,MAAM,IAAI,CAAC,MAAM,EAAE,SAAS;AACvC,QAAM,KAAK,MAAM,IAAI,CAAC,MAAM,EAAE,OAAO;AACrC,QAAM,WAAW,KAAK,EAAE;AACxB,QAAM,cAAc,KAAK,EAAE;AAC3B,SAAO;AAAA,IACL,OAAO,SAAS,KAAK,IAAI,GAAG,EAAE;AAAA,IAC9B,OAAO,SAAS,KAAK,IAAI,GAAG,EAAE;AAAA,IAC9B,GAAG,MAAM;AAAA,IACT;AAAA,IACA;AAAA,IACA,KAAK,KAAK,IAAI,cAAc,QAAQ;AAAA,EACtC;AACF;AAEA,SAAS,KAAK,IAAsB;AAClC,SAAO,GAAG,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI,GAAG;AAC5C;AAEA,SAAS,eAAe,QAAyE;AAC/F,SAAO,OAAO,QACZ,IAAI,SAAS,UAAU,WAAW,SAAU,IAAI,SAAS,SAAS,OAAO,IAAI,IAAK;AACtF;","names":[]}

package/dist/{chunk-JYE3WOTE.js → chunk-RPLZ4OIB.js}

@@ -1428,6 +1428,15 @@ async function runImprovementLoop(opts) {
   if (opts.autoOnPromote === "pr" && (!opts.ghOwner || !opts.ghRepo)) {
     throw new Error("runImprovementLoop: autoOnPromote='pr' requires ghOwner + ghRepo.");
   }
+  const holdoutIds = new Set(opts.holdoutScenarios.map((s) => s.id));
+  const leaked = opts.scenarios.filter((s) => holdoutIds.has(s.id)).map((s) => s.id);
+  if (leaked.length > 0) {
+    throw new Error(
+      `runImprovementLoop: training scenarios and holdoutScenarios must be disjoint (overlap: [${leaked.join(
+        ", "
+      )}]) \u2014 a shared scenario leaks the held-out gate axis into the optimization, inflating reported lift.`
+    );
+  }
   const dispatchTimeoutMs = opts.dispatchTimeoutMs ?? DEFAULT_DISPATCH_TIMEOUT_MS;
   const optimization = await runOptimization({ ...opts, dispatchTimeoutMs });
   const winnerIsBaseline = optimization.winnerSurfaceHash === surfaceHash(opts.baselineSurface);
@@ -1832,4 +1841,4 @@ export {
   provenanceSpansPath,
   emitLoopProvenance
 };
-//# sourceMappingURL=chunk-JYE3WOTE.js.map
+//# sourceMappingURL=chunk-RPLZ4OIB.js.map