@tangle-network/agent-eval 0.108.1 → 0.110.0

package/dist/analyst/index.d.ts

@@ -1,22 +1,20 @@
-import { M as MultiLayerVerifier, V as VerifyOptions, S as Severity } from '../multi-layer-verifier-CI4jdX-q.js';
-import { c as RunCritic, a as RunTrace } from '../run-critic-CmMf05uV.js';
-import { S as SemanticConceptJudgeOptions, a as SemanticConceptJudgeInput, B as BehavioralMetrics } from '../semantic-concept-judge-BmNZPB_j.js';
-export { C as CreateAnalystAiConfig, D as DEFAULT_TRACE_ANALYST_KINDS, b as DiffPolicy, F as FAILURE_MODE_KIND_SPEC, c as FINDING_SUBJECT_GRAMMAR_PROMPT, d as FINDING_SUBJECT_KINDS, e as FindingSubject, f as FindingSubjectKind, g as FindingSubjectStringSchema, h as FindingsDiff, i as FindingsStore, I as IMPROVEMENT_KIND_SPEC, K as KIND_EXPECTED_SUBJECTS, j as KNOWLEDGE_GAP_KIND_SPEC, k as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, l as SKILL_USAGE_ANALYST, m as SkillUsageAnalyst, n as SkillUsageRecord, o as SkillUsageReport, p as SkillUsageScanConfig, q as buildSkillUsageReport, r as createAnalystAi, s as defaultIsMaterial, t as diffFindings, u as emitSkillUsageFindings, v as parseFindingSubject, w as renderFindingSubject } from '../semantic-concept-judge-BmNZPB_j.js';
+import { M as MultiLayerVerifier, V as VerifyOptions, S as Severity } from '../multi-layer-verifier-BsqKuLyN.js';
+import { R as RunCritic, S as SemanticConceptJudgeOptions, a as RunTrace, b as SemanticConceptJudgeInput, B as BehavioralMetrics } from '../semantic-concept-judge-D7z6JCLZ.js';
+export { C as CreateAnalystAiConfig, D as DEFAULT_TRACE_ANALYST_KINDS, c as DiffPolicy, F as FAILURE_MODE_KIND_SPEC, d as FINDING_SUBJECT_GRAMMAR_PROMPT, e as FINDING_SUBJECT_KINDS, f as FindingSubject, g as FindingSubjectKind, h as FindingSubjectStringSchema, i as FindingsDiff, j as FindingsStore, I as IMPROVEMENT_KIND_SPEC, K as KIND_EXPECTED_SUBJECTS, k as KNOWLEDGE_GAP_KIND_SPEC, l as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, m as SKILL_USAGE_ANALYST, n as SkillUsageAnalyst, o as SkillUsageRecord, p as SkillUsageReport, q as SkillUsageScanConfig, r as buildSkillUsageReport, s as createAnalystAi, t as defaultIsMaterial, u as diffFindings, v as emitSkillUsageFindings, w as parseFindingSubject, x as renderFindingSubject } from '../semantic-concept-judge-D7z6JCLZ.js';
 import { b as JudgeFn, a as JudgeInput } from '../types-C7DGg5ex.js';
-import { a as Analyst, h as AnalystSeverity, A as AnalystFinding } from '../types-D1ytG0Yg.js';
-export { b as AnalystContext, g as AnalystCost, i as AnalystInputKind, j as AnalystRequirements, f as AnalystRunEvent, e as AnalystRunInputs, d as AnalystRunResult, c as AnalystRunSummary, k as ChatCallOpts, C as ChatClient, l as ChatRequest, m as ChatResponse, n as ChatTransport, o as CliBridgeTransportOpts, p as CreateChatClientOpts, D as DirectProviderTransportOpts, E as EvidenceRef, M as MockTransportOpts, R as RouterTransportOpts, S as SandboxSdkTransportOpts, q as computeFindingId, r as createChatClient, s as makeFinding } from '../types-D1ytG0Yg.js';
+import { a as Analyst, g as AnalystSeverity, A as AnalystFinding } from '../kind-factory-20hcaYpf.js';
+export { h as ANALYST_SEVERITIES, b as AnalystContext, i as AnalystCost, j as AnalystInputKind, k as AnalystRequirements, f as AnalystRunEvent, e as AnalystRunInputs, d as AnalystRunResult, c as AnalystRunSummary, l as ChatCallOpts, C as ChatClient, m as ChatRequest, n as ChatResponse, o as ChatTransport, p as CliBridgeTransportOpts, q as CreateChatClientOpts, r as CreateTraceAnalystKindOpts, D as DirectProviderTransportOpts, E as EvidenceRef, M as MockTransportOpts, R as RAW_FINDING_SCHEMA_PROMPT, s as RawAnalystFinding, t as RawAnalystFindingSchema, u as RouterTransportOpts, S as SandboxSdkTransportOpts, v as TraceAnalystGolden, T as TraceAnalystKindSpec, w as computeFindingId, x as createChatClient, y as createTraceAnalystKind, z as makeFinding, B as parseRawFinding, F as renderPriorFindings } from '../kind-factory-20hcaYpf.js';
 import { TCloud } from '@tangle-network/tcloud';
 import { T as TraceAnalysisStore } from '../store-C1YxJDEK.js';
-export { a as AnalystHooks, A as AnalystRegistry, b as AnalystRegistryOptions, B as BudgetPolicy, D as DefaultAnalystRegistryOptions, R as RegistryRunOpts, c as buildDefaultAnalystRegistry } from '../default-registry-BswHCXnU.js';
-export { A as ANALYST_SEVERITIES, C as CreateTraceAnalystKindOpts, R as RAW_FINDING_SCHEMA_PROMPT, a as RawAnalystFinding, b as RawAnalystFindingSchema, c as TraceAnalystGolden, T as TraceAnalystKindSpec, d as createTraceAnalystKind, p as parseRawFinding, r as renderPriorFindings } from '../kind-factory-DvIGo_cP.js';
-export { F as FindingToPolicyEditOptions, P as POLICY_EDIT_AXES, a as POLICY_EDIT_TARGET_SURFACES, b as PolicyEdit, c as PolicyEditAdmission, d as PolicyEditAdmissionOptions, e as PolicyEditAxis, f as PolicyEditChange, g as PolicyEditExpectedGain, h as PolicyEditGainDirection, i as PolicyEditGainUnit, j as PolicyEditInit, k as PolicyEditRisk, l as PolicyEditSchemaVersion, m as PolicyEditSource, n as PolicyEditTarget, o as PolicyEditTargetSurface, p as PolicyEditValidationError, q as admitPolicyEdit, r as applyPolicyEditToSurface, s as computePolicyEditId, t as isPolicyEdit, u as makePolicyEdit, v as policyEditFromFinding, w as policyEditsFromFindings, x as scorePolicyEditReadiness, y as validatePolicyEdit } from '../policy-edit-DQUXYMDm.js';
+export { a as AnalystHooks, A as AnalystRegistry, b as AnalystRegistryOptions, B as BudgetPolicy, D as DefaultAnalystRegistryOptions, R as RegistryRunOpts, c as buildDefaultAnalystRegistry } from '../default-registry-Bcf1uKVI.js';
+export { F as FindingToPolicyEditOptions, P as POLICY_EDIT_AXES, a as POLICY_EDIT_TARGET_SURFACES, b as PolicyEdit, c as PolicyEditAdmission, d as PolicyEditAdmissionOptions, e as PolicyEditAxis, f as PolicyEditChange, g as PolicyEditExpectedGain, h as PolicyEditGainDirection, i as PolicyEditGainUnit, j as PolicyEditInit, k as PolicyEditRisk, l as PolicyEditSchemaVersion, m as PolicyEditSource, n as PolicyEditTarget, o as PolicyEditTargetSurface, p as PolicyEditValidationError, q as admitPolicyEdit, r as applyPolicyEditToSurface, s as computePolicyEditId, t as isPolicyEdit, u as makePolicyEdit, v as policyEditFromFinding, w as policyEditsFromFindings, x as scorePolicyEditReadiness, y as validatePolicyEdit } from '../policy-edit-D2bBDZDf.js';
 import { L as LlmClientOptions } from '../llm-client-DyqEH4jH.js';
 import { AxFunction } from '@ax-llm/ax';
 import '../verdict-C9MlYujm.js';
-import '../schema-m0gsnbt3.js';
-import '../store-BcFXE6LG.js';
 import 'zod';
-import '../run-record-I-Z3JNvO.js';
+import '../schema-SGWcK9wa.js';
+import '../store-BsVi7ncX.js';
+import '../run-record-DksGsfgv.js';
 import '@tangle-network/agent-interface';
 import '../errors-oeQrLqXC.js';
 import '../raw-provider-sink-C46HDghv.js';

package/dist/analyst/index.js

@@ -11,13 +11,12 @@ import {
   diffFindings,
   emitSkillUsageFindings,
   runSemanticConceptJudge
-} from "../chunk-5PK3626Q.js";
-import "../chunk-DJWX3GVS.js";
+} from "../chunk-XRGOKCMO.js";
 import {
   behavioralAnalyst,
   buildDefaultAnalystRegistry,
   deriveEfficiencyFindings
-} from "../chunk-QRVS7MX4.js";
+} from "../chunk-OW47B5WA.js";
 import {
   POLICY_EDIT_AXES,
   POLICY_EDIT_TARGET_SURFACES,
@@ -34,7 +33,7 @@ import {
   policyEditsFromFindings,
   scorePolicyEditReadiness,
   validatePolicyEdit
-} from "../chunk-LOJ2QVCE.js";
+} from "../chunk-2IY4ILP4.js";
 import {
   ANALYST_SEVERITIES,
   AnalystRegistry,
@@ -52,23 +51,21 @@ import {
   buildTraceToolsForGroup,
   coerceJson,
   coerceToFindingRows,
+  computeFindingId,
   createTraceAnalystKind,
+  makeFinding,
   parseFindingSubject,
   parseRawFinding,
   renderFindingSubject,
   renderPriorFindings,
   stripCodeFences,
   structureFindings
-} from "../chunk-2OGPXHOB.js";
-import "../chunk-FUCQVFMU.js";
-import {
-  computeFindingId,
-  makeFinding
-} from "../chunk-45EEMHTC.js";
+} from "../chunk-7NX6ZSBG.js";
 import "../chunk-LNQEP766.js";
-import "../chunk-PC4UYEBM.js";
 import "../chunk-XJYR7XFV.js";
 import "../chunk-VSMTAMNK.js";
+import "../chunk-FUCQVFMU.js";
+import "../chunk-PC4UYEBM.js";
 import "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";
 

package/dist/analyst/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/analyst/adapters.ts"],"sourcesContent":["/**\n * Adapter factories — lift each existing agent-eval primitive into the\n * Analyst contract without re-implementing it.\n *\n * Five primitives, five factories. Each one:\n *   - Builds an Analyst with a stable id (caller chooses; defaults\n *     given), a sensible default `inputKind`, a version derived from\n *     the wrapped primitive's version + an adapter revision, and an\n *     `analyze()` that calls the primitive and lifts its output to\n *     AnalystFinding[] using `makeFinding()`.\n *   - Maps severities: the existing `Severity` ('critical' | 'major' |\n *     'minor' | 'info') projects onto AnalystSeverity ('critical' |\n *     'high' | 'medium' | 'low' | 'info'); 'major' → 'high', 'minor' →\n *     'medium'. Domain analysts that want finer-grained mapping override.\n *\n * Adapters never own state. Calling the same factory twice with the\n * same primitive instance is safe.\n */\n\nimport type {\n  Finding as LayerFinding,\n  Severity as LayerSeverity,\n  MultiLayerVerifier,\n  VerifyOptions,\n} from '../multi-layer-verifier'\nimport { RunCritic, type RunTrace } from '../run-critic'\nimport {\n  runSemanticConceptJudge,\n  SEMANTIC_CONCEPT_JUDGE_VERSION,\n  type SemanticConceptJudgeInput,\n  type SemanticConceptJudgeOptions,\n} from '../semantic-concept-judge'\nimport type { JudgeFn, JudgeInput, JudgeScore, TCloud } from '../types'\nimport type { Analyst, AnalystFinding, AnalystSeverity } from './types'\nimport { makeFinding } from './types'\n\nconst ADAPTER_REV = '1'\n\n// ── Severity bridges ───────────────────────────────────────────────\n\nexport function liftSeverity(s: LayerSeverity): AnalystSeverity {\n  switch (s) {\n    case 'critical':\n      return 'critical'\n    case 'major':\n      return 'high'\n    case 'minor':\n      return 'medium'\n    case 'info':\n      return 'info'\n  }\n}\n\n// ── 1. MultiLayerVerifier → Analyst ─────────────────────────────────\n\nexport interface VerifierAdapterOpts<Env> {\n  id?: string\n  area?: string\n  verifier: MultiLayerVerifier<Env>\n  /**\n   * The verifier expects an `env` per run. Adapters take it from\n   * `AnalystRunInputs.custom[<id>]` via the registry's 'custom' routing.\n   */\n  options?: Omit<VerifyOptions<Env>, 'env'>\n}\n\nexport function createVerifierAdapter<Env>(opts: VerifierAdapterOpts<Env>): Analyst<Env> {\n  const id = opts.id ?? 'multi-layer-verifier'\n  const area = opts.area ?? 'verification'\n  return {\n    id,\n    description:\n      \"Runs a MultiLayerVerifier and lifts each layer's findings into the analyst envelope.\",\n    inputKind: 'custom',\n    cost: { kind: 'deterministic' },\n    version: `verifier-${ADAPTER_REV}`,\n    async analyze(env, ctx) {\n      const report = await opts.verifier.run({ env, ...opts.options })\n      const out: AnalystFinding[] = []\n      for (const layer of report.layers) {\n        for (const finding of layer.findings) {\n          out.push(liftLayerFinding(id, area, layer.layer, finding))\n        }\n        // Layer-level signal: a failed/error layer is itself a finding\n        // even if it didn't emit per-finding rows.\n        if (layer.status === 'fail' || layer.status === 'error' || layer.status === 'timeout') {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject: layer.layer,\n              claim: `layer \"${layer.layer}\" ${layer.status}: ${layer.reason ?? 'no reason given'}`,\n              severity:\n                layer.status === 'error' ? 'high' : layer.status === 'timeout' ? 'medium' : 'high',\n              confidence: 1,\n              evidence_refs: [],\n              metadata: {\n                layer_status: layer.status,\n                duration_ms: layer.durationMs,\n                score: layer.score,\n                diagnostics: layer.diagnostics,\n              },\n            }),\n          )\n        }\n      }\n      ctx.log?.('verifier complete', {\n        layers: report.layers.length,\n        blended: report.blendedScore,\n        all_pass: report.allPass,\n      })\n      return out\n    },\n  }\n}\n\nfunction liftLayerFinding(\n  analyst_id: string,\n  area: string,\n  layer: string,\n  f: LayerFinding,\n): AnalystFinding {\n  return makeFinding({\n    analyst_id,\n    area,\n    subject: f.layer ?? layer,\n    claim: f.message,\n    severity: liftSeverity(f.severity),\n    confidence: 0.85,\n    evidence_refs: f.evidence\n      ? [{ kind: 'artifact', uri: 'inline:evidence', excerpt: f.evidence }]\n      : [],\n    metadata: f.detail,\n  })\n}\n\n// ── 2. RunCritic → Analyst ──────────────────────────────────────────\n\nexport interface RunCriticAdapterOpts {\n  id?: string\n  area?: string\n  critic?: RunCritic\n  /** Optional threshold below which a dimension is reported as a finding. Default 0.5. */\n  threshold?: number\n}\n\nexport function createRunCriticAdapter(opts: RunCriticAdapterOpts = {}): Analyst<RunTrace> {\n  const id = opts.id ?? 'run-critic'\n  const area = opts.area ?? 'run-quality'\n  const critic = opts.critic ?? new RunCritic()\n  const threshold = opts.threshold ?? 0.5\n  return {\n    id,\n    description:\n      'Scores a single run across success / grounding / drift / tool-quality and surfaces below-threshold dimensions.',\n    inputKind: 'custom',\n    cost: { kind: 'deterministic' },\n    version: `run-critic-${ADAPTER_REV}`,\n    async analyze(trace) {\n      const score = critic.scoreTrace(trace)\n      const out: AnalystFinding[] = []\n      const dims: Array<[keyof typeof score, AnalystSeverity, string]> = [\n        ['success', 'critical', 'run did not complete successfully'],\n        ['goalProgress', 'high', 'goal progress is low'],\n        ['repoGroundedness', 'high', 'output is poorly grounded in the repository'],\n        ['toolUseQuality', 'medium', 'tool use quality is low'],\n        ['patchQuality', 'medium', 'no real patch/edit evidence'],\n        ['testReality', 'high', 'no real test/build evidence'],\n        ['finalGate', 'critical', 'final gate is blocking'],\n      ]\n      for (const [dim, sev, msg] of dims) {\n        const value = score[dim] as number\n        if (typeof value === 'number' && value < threshold) {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject: dim,\n              claim: msg,\n              rationale: `${dim}=${value.toFixed(2)} below threshold ${threshold}`,\n              severity: sev,\n              confidence: 1,\n              evidence_refs: [],\n              metadata: { dimension: dim, value, threshold, run_id: trace.run.runId },\n            }),\n          )\n        }\n      }\n      // Drift penalty is high → surface as a finding (inverse threshold).\n      if (score.driftPenalty > 1 - threshold) {\n        out.push(\n          makeFinding({\n            analyst_id: id,\n            area,\n            subject: 'drift',\n            claim: 'agent output drifted from repository signal',\n            rationale: `driftPenalty=${score.driftPenalty.toFixed(2)}`,\n            severity: 'medium',\n            confidence: 0.9,\n            evidence_refs: [],\n            metadata: { drift_penalty: score.driftPenalty, notes: score.notes },\n          }),\n        )\n      }\n      return out\n    },\n  }\n}\n\n// ── 3. JudgeFn → Analyst ────────────────────────────────────────────\n\nexport interface JudgeAdapterOpts {\n  id?: string\n  area?: string\n  judge: JudgeFn\n  /** TCloud handle the JudgeFn calls. */\n  tcloud: TCloud\n  /** Optional cost classification — most judges call an LLM. */\n  cost?: Analyst['cost']\n  /** Optional threshold below which a JudgeScore becomes a finding. Default 6 (on 0-10 scale). */\n  threshold?: number\n}\n\nexport function createJudgeAdapter(opts: JudgeAdapterOpts): Analyst<JudgeInput> {\n  const id = opts.id ?? 'judge'\n  const area = opts.area ?? 'judge'\n  const threshold = opts.threshold ?? 6\n  return {\n    id,\n    description:\n      'Wraps an agent-eval JudgeFn into an analyst; below-threshold dimensions surface as findings.',\n    inputKind: 'judge-input',\n    cost: opts.cost ?? { kind: 'llm' },\n    version: `judge-${ADAPTER_REV}`,\n    async analyze(input) {\n      const scores = await opts.judge(opts.tcloud, input)\n      return scores\n        .filter((s) => normalize10(s.score) < threshold)\n        .map((s) => liftJudgeScore(id, area, s))\n    },\n  }\n}\n\nfunction normalize10(s: number): number {\n  // JudgeScore convention is 0-10 but some judges emit 0-1. Coerce to 0-10.\n  return s <= 1 ? s * 10 : s\n}\n\nfunction liftJudgeScore(analyst_id: string, area: string, s: JudgeScore): AnalystFinding {\n  const score10 = normalize10(s.score)\n  const severity: AnalystSeverity =\n    score10 < 3 ? 'critical' : score10 < 5 ? 'high' : score10 < 7 ? 'medium' : 'low'\n  return makeFinding({\n    analyst_id,\n    area,\n    subject: s.dimension,\n    claim: `${s.judgeName}/${s.dimension} scored ${score10.toFixed(1)}/10`,\n    rationale: s.reasoning,\n    severity,\n    confidence: 0.8,\n    evidence_refs: s.evidence\n      ? [{ kind: 'artifact', uri: 'inline:evidence', excerpt: s.evidence }]\n      : [],\n    // Provenance: this finding IS a judge verdict (an acceptance score), not an\n    // observation of behavior. The steer firewall (assertNoJudgeVerdict) rejects\n    // it from steering — even when it cites an artifact above — because letting a\n    // verdict steer the next attempt is the held-out judge leaking into the loop.\n    derived_from_judge: true,\n    metadata: { judge_name: s.judgeName, dimension: s.dimension, score_10: score10 },\n  })\n}\n\n// ── 4. SemanticConceptJudge → Analyst ──────────────────────────────\n\nexport interface SemanticConceptJudgeAdapterOpts {\n  id?: string\n  area?: string\n  options?: SemanticConceptJudgeOptions\n}\n\nexport function createSemanticConceptJudgeAdapter(\n  opts: SemanticConceptJudgeAdapterOpts = {},\n): Analyst<SemanticConceptJudgeInput> {\n  const id = opts.id ?? 'semantic-concept-judge'\n  const area = opts.area ?? 'concept-coverage'\n  return {\n    id,\n    description:\n      'Runs the semantic-concept judge and surfaces missing / weak concepts as findings.',\n    inputKind: 'custom',\n    cost: { kind: 'llm', models: opts.options?.model ? [opts.options.model] : undefined },\n    version: `${SEMANTIC_CONCEPT_JUDGE_VERSION}-adapter-${ADAPTER_REV}`,\n    async analyze(input) {\n      const result = await runSemanticConceptJudge(input, opts.options)\n      if (!result.available) {\n        return [\n          makeFinding({\n            analyst_id: id,\n            area,\n            claim: 'semantic-concept judge unavailable',\n            rationale: result.error,\n            severity: 'info',\n            confidence: 1,\n            evidence_refs: [],\n            metadata: { reason: result.error },\n          }),\n        ]\n      }\n      const out: AnalystFinding[] = []\n      for (const f of result.findings) {\n        // Only surface gaps: missing concepts or low scores. Concepts at\n        // 7+/10 with present=true are not findings — they're successes.\n        if (f.present && f.score >= 7) continue\n        out.push(\n          makeFinding({\n            analyst_id: id,\n            area,\n            subject: f.concept,\n            claim: f.present\n              ? `concept \"${f.concept}\" is weak (${f.score}/10)`\n              : `concept \"${f.concept}\" is missing`,\n            rationale: f.evidence,\n            severity: liftSeverity(f.severity),\n            confidence: 0.85,\n            evidence_refs: [{ kind: 'artifact', uri: 'inline:evidence', excerpt: f.evidence }],\n            metadata: {\n              concept: f.concept,\n              present: f.present,\n              score_10: f.score,\n              cost_usd: result.costUsd ?? undefined,\n            },\n          }),\n        )\n      }\n      return out\n    },\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAoCA,IAAM,cAAc;AAIb,SAAS,aAAa,GAAmC;AAC9D,UAAQ,GAAG;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,EACX;AACF;AAeO,SAAS,sBAA2B,MAA8C;AACvF,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,gBAAgB;AAAA,IAC9B,SAAS,YAAY,WAAW;AAAA,IAChC,MAAM,QAAQ,KAAK,KAAK;AACtB,YAAM,SAAS,MAAM,KAAK,SAAS,IAAI,EAAE,KAAK,GAAG,KAAK,QAAQ,CAAC;AAC/D,YAAM,MAAwB,CAAC;AAC/B,iBAAW,SAAS,OAAO,QAAQ;AACjC,mBAAW,WAAW,MAAM,UAAU;AACpC,cAAI,KAAK,iBAAiB,IAAI,MAAM,MAAM,OAAO,OAAO,CAAC;AAAA,QAC3D;AAGA,YAAI,MAAM,WAAW,UAAU,MAAM,WAAW,WAAW,MAAM,WAAW,WAAW;AACrF,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA,SAAS,MAAM;AAAA,cACf,OAAO,UAAU,MAAM,KAAK,KAAK,MAAM,MAAM,KAAK,MAAM,UAAU,iBAAiB;AAAA,cACnF,UACE,MAAM,WAAW,UAAU,SAAS,MAAM,WAAW,YAAY,WAAW;AAAA,cAC9E,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU;AAAA,gBACR,cAAc,MAAM;AAAA,gBACpB,aAAa,MAAM;AAAA,gBACnB,OAAO,MAAM;AAAA,gBACb,aAAa,MAAM;AAAA,cACrB;AAAA,YACF,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,qBAAqB;AAAA,QAC7B,QAAQ,OAAO,OAAO;AAAA,QACtB,SAAS,OAAO;AAAA,QAChB,UAAU,OAAO;AAAA,MACnB,CAAC;AACD,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAEA,SAAS,iBACP,YACA,MACA,OACA,GACgB;AAChB,SAAO,YAAY;AAAA,IACjB;AAAA,IACA;AAAA,IACA,SAAS,EAAE,SAAS;AAAA,IACpB,OAAO,EAAE;AAAA,IACT,UAAU,aAAa,EAAE,QAAQ;AAAA,IACjC,YAAY;AAAA,IACZ,eAAe,EAAE,WACb,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC,IAClE,CAAC;AAAA,IACL,UAAU,EAAE;AAAA,EACd,CAAC;AACH;AAYO,SAAS,uBAAuB,OAA6B,CAAC,GAAsB;AACzF,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,QAAM,SAAS,KAAK,UAAU,IAAI,UAAU;AAC5C,QAAM,YAAY,KAAK,aAAa;AACpC,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,gBAAgB;AAAA,IAC9B,SAAS,cAAc,WAAW;AAAA,IAClC,MAAM,QAAQ,OAAO;AACnB,YAAM,QAAQ,OAAO,WAAW,KAAK;AACrC,YAAM,MAAwB,CAAC;AAC/B,YAAM,OAA6D;AAAA,QACjE,CAAC,WAAW,YAAY,mCAAmC;AAAA,QAC3D,CAAC,gBAAgB,QAAQ,sBAAsB;AAAA,QAC/C,CAAC,oBAAoB,QAAQ,6CAA6C;AAAA,QAC1E,CAAC,kBAAkB,UAAU,yBAAyB;AAAA,QACtD,CAAC,gBAAgB,UAAU,6BAA6B;AAAA,QACxD,CAAC,eAAe,QAAQ,6BAA6B;AAAA,QACrD,CAAC,aAAa,YAAY,wBAAwB;AAAA,MACpD;AACA,iBAAW,CAAC,KAAK,KAAK,GAAG,KAAK,MAAM;AAClC,cAAM,QAAQ,MAAM,GAAG;AACvB,YAAI,OAAO,UAAU,YAAY,QAAQ,WAAW;AAClD,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA,SAAS;AAAA,cACT,OAAO;AAAA,cACP,WAAW,GAAG,GAAG,IAAI,MAAM,QAAQ,CAAC,CAAC,oBAAoB,SAAS;AAAA,cAClE,UAAU;AAAA,cACV,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU,EAAE,WAAW,KAAK,OAAO,WAAW,QAAQ,MAAM,IAAI,MAAM;AAAA,YACxE,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,UAAI,MAAM,eAAe,IAAI,WAAW;AACtC,YAAI;AAAA,UACF,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,SAAS;AAAA,YACT,OAAO;AAAA,YACP,WAAW,gBAAgB,MAAM,aAAa,QAAQ,CAAC,CAAC;AAAA,YACxD,UAAU;AAAA,YACV,YAAY;AAAA,YACZ,eAAe,CAAC;AAAA,YAChB,UAAU,EAAE,eAAe,MAAM,cAAc,OAAO,MAAM,MAAM;AAAA,UACpE,CAAC;AAAA,QACH;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAgBO,SAAS,mBAAmB,MAA6C;AAC9E,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,QAAM,YAAY,KAAK,aAAa;AACpC,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,KAAK,QAAQ,EAAE,MAAM,MAAM;AAAA,IACjC,SAAS,SAAS,WAAW;AAAA,IAC7B,MAAM,QAAQ,OAAO;AACnB,YAAM,SAAS,MAAM,KAAK,MAAM,KAAK,QAAQ,KAAK;AAClD,aAAO,OACJ,OAAO,CAAC,MAAM,YAAY,EAAE,KAAK,IAAI,SAAS,EAC9C,IAAI,CAAC,MAAM,eAAe,IAAI,MAAM,CAAC,CAAC;AAAA,IAC3C;AAAA,EACF;AACF;AAEA,SAAS,YAAY,GAAmB;AAEtC,SAAO,KAAK,IAAI,IAAI,KAAK;AAC3B;AAEA,SAAS,eAAe,YAAoB,MAAc,GAA+B;AACvF,QAAM,UAAU,YAAY,EAAE,KAAK;AACnC,QAAM,WACJ,UAAU,IAAI,aAAa,UAAU,IAAI,SAAS,UAAU,IAAI,WAAW;AAC7E,SAAO,YAAY;AAAA,IACjB;AAAA,IACA;AAAA,IACA,SAAS,EAAE;AAAA,IACX,OAAO,GAAG,EAAE,SAAS,IAAI,EAAE,SAAS,WAAW,QAAQ,QAAQ,CAAC,CAAC;AAAA,IACjE,WAAW,EAAE;AAAA,IACb;AAAA,IACA,YAAY;AAAA,IACZ,eAAe,EAAE,WACb,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC,IAClE,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA,IAKL,oBAAoB;AAAA,IACpB,UAAU,EAAE,YAAY,EAAE,WAAW,WAAW,EAAE,WAAW,UAAU,QAAQ;AAAA,EACjF,CAAC;AACH;AAUO,SAAS,kCACd,OAAwC,CAAC,GACL;AACpC,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,OAAO,QAAQ,KAAK,SAAS,QAAQ,CAAC,KAAK,QAAQ,KAAK,IAAI,OAAU;AAAA,IACpF,SAAS,GAAG,8BAA8B,YAAY,WAAW;AAAA,IACjE,MAAM,QAAQ,OAAO;AACnB,YAAM,SAAS,MAAM,wBAAwB,OAAO,KAAK,OAAO;AAChE,UAAI,CAAC,OAAO,WAAW;AACrB,eAAO;AAAA,UACL,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,OAAO;AAAA,YACP,WAAW,OAAO;AAAA,YAClB,UAAU;AAAA,YACV,YAAY;AAAA,YACZ,eAAe,CAAC;AAAA,YAChB,UAAU,EAAE,QAAQ,OAAO,MAAM;AAAA,UACnC,CAAC;AAAA,QACH;AAAA,MACF;AACA,YAAM,MAAwB,CAAC;AAC/B,iBAAW,KAAK,OAAO,UAAU;AAG/B,YAAI,EAAE,WAAW,EAAE,SAAS,EAAG;AAC/B,YAAI;AAAA,UACF,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,SAAS,EAAE;AAAA,YACX,OAAO,EAAE,UACL,YAAY,EAAE,OAAO,cAAc,EAAE,KAAK,SAC1C,YAAY,EAAE,OAAO;AAAA,YACzB,WAAW,EAAE;AAAA,YACb,UAAU,aAAa,EAAE,QAAQ;AAAA,YACjC,YAAY;AAAA,YACZ,eAAe,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC;AAAA,YACjF,UAAU;AAAA,cACR,SAAS,EAAE;AAAA,cACX,SAAS,EAAE;AAAA,cACX,UAAU,EAAE;AAAA,cACZ,UAAU,OAAO,WAAW;AAAA,YAC9B;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}
+{"version":3,"sources":["../../src/analyst/adapters.ts"],"sourcesContent":["/**\n * Adapter factories — lift each existing agent-eval primitive into the\n * Analyst contract without re-implementing it.\n *\n * Five primitives, five factories. Each one:\n *   - Builds an Analyst with a stable id (caller chooses; defaults\n *     given), a sensible default `inputKind`, a version derived from\n *     the wrapped primitive's version + an adapter revision, and an\n *     `analyze()` that calls the primitive and lifts its output to\n *     AnalystFinding[] using `makeFinding()`.\n *   - Maps severities: the existing `Severity` ('critical' | 'major' |\n *     'minor' | 'info') projects onto AnalystSeverity ('critical' |\n *     'high' | 'medium' | 'low' | 'info'); 'major' → 'high', 'minor' →\n *     'medium'. Domain analysts that want finer-grained mapping override.\n *\n * Adapters never own state. Calling the same factory twice with the\n * same primitive instance is safe.\n */\n\nimport type {\n  Finding as LayerFinding,\n  Severity as LayerSeverity,\n  MultiLayerVerifier,\n  VerifyOptions,\n} from '../multi-layer-verifier'\nimport { RunCritic, type RunTrace } from '../run-critic'\nimport {\n  runSemanticConceptJudge,\n  SEMANTIC_CONCEPT_JUDGE_VERSION,\n  type SemanticConceptJudgeInput,\n  type SemanticConceptJudgeOptions,\n} from '../semantic-concept-judge'\nimport type { JudgeFn, JudgeInput, JudgeScore, TCloud } from '../types'\nimport type { Analyst, AnalystFinding, AnalystSeverity } from './types'\nimport { makeFinding } from './types'\n\nconst ADAPTER_REV = '1'\n\n// ── Severity bridges ───────────────────────────────────────────────\n\nexport function liftSeverity(s: LayerSeverity): AnalystSeverity {\n  switch (s) {\n    case 'critical':\n      return 'critical'\n    case 'major':\n      return 'high'\n    case 'minor':\n      return 'medium'\n    case 'info':\n      return 'info'\n  }\n}\n\n// ── 1. MultiLayerVerifier → Analyst ─────────────────────────────────\n\nexport interface VerifierAdapterOpts<Env> {\n  id?: string\n  area?: string\n  verifier: MultiLayerVerifier<Env>\n  /**\n   * The verifier expects an `env` per run. Adapters take it from\n   * `AnalystRunInputs.custom[<id>]` via the registry's 'custom' routing.\n   */\n  options?: Omit<VerifyOptions<Env>, 'env'>\n}\n\nexport function createVerifierAdapter<Env>(opts: VerifierAdapterOpts<Env>): Analyst<Env> {\n  const id = opts.id ?? 'multi-layer-verifier'\n  const area = opts.area ?? 'verification'\n  return {\n    id,\n    description:\n      \"Runs a MultiLayerVerifier and lifts each layer's findings into the analyst envelope.\",\n    inputKind: 'custom',\n    cost: { kind: 'deterministic' },\n    version: `verifier-${ADAPTER_REV}`,\n    async analyze(env, ctx) {\n      const report = await opts.verifier.run({ env, ...opts.options })\n      const out: AnalystFinding[] = []\n      for (const layer of report.layers) {\n        for (const finding of layer.findings) {\n          out.push(liftLayerFinding(id, area, layer.layer, finding))\n        }\n        // Layer-level signal: a failed/error layer is itself a finding\n        // even if it didn't emit per-finding rows.\n        if (layer.status === 'fail' || layer.status === 'error' || layer.status === 'timeout') {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject: layer.layer,\n              claim: `layer \"${layer.layer}\" ${layer.status}: ${layer.reason ?? 'no reason given'}`,\n              severity:\n                layer.status === 'error' ? 'high' : layer.status === 'timeout' ? 'medium' : 'high',\n              confidence: 1,\n              evidence_refs: [],\n              metadata: {\n                layer_status: layer.status,\n                duration_ms: layer.durationMs,\n                score: layer.score,\n                diagnostics: layer.diagnostics,\n              },\n            }),\n          )\n        }\n      }\n      ctx.log?.('verifier complete', {\n        layers: report.layers.length,\n        blended: report.blendedScore,\n        all_pass: report.allPass,\n      })\n      return out\n    },\n  }\n}\n\nfunction liftLayerFinding(\n  analyst_id: string,\n  area: string,\n  layer: string,\n  f: LayerFinding,\n): AnalystFinding {\n  return makeFinding({\n    analyst_id,\n    area,\n    subject: f.layer ?? layer,\n    claim: f.message,\n    severity: liftSeverity(f.severity),\n    confidence: 0.85,\n    evidence_refs: f.evidence\n      ? [{ kind: 'artifact', uri: 'inline:evidence', excerpt: f.evidence }]\n      : [],\n    metadata: f.detail,\n  })\n}\n\n// ── 2. RunCritic → Analyst ──────────────────────────────────────────\n\nexport interface RunCriticAdapterOpts {\n  id?: string\n  area?: string\n  critic?: RunCritic\n  /** Optional threshold below which a dimension is reported as a finding. Default 0.5. */\n  threshold?: number\n}\n\nexport function createRunCriticAdapter(opts: RunCriticAdapterOpts = {}): Analyst<RunTrace> {\n  const id = opts.id ?? 'run-critic'\n  const area = opts.area ?? 'run-quality'\n  const critic = opts.critic ?? new RunCritic()\n  const threshold = opts.threshold ?? 0.5\n  return {\n    id,\n    description:\n      'Scores a single run across success / grounding / drift / tool-quality and surfaces below-threshold dimensions.',\n    inputKind: 'custom',\n    cost: { kind: 'deterministic' },\n    version: `run-critic-${ADAPTER_REV}`,\n    async analyze(trace) {\n      const score = critic.scoreTrace(trace)\n      const out: AnalystFinding[] = []\n      const dims: Array<[keyof typeof score, AnalystSeverity, string]> = [\n        ['success', 'critical', 'run did not complete successfully'],\n        ['goalProgress', 'high', 'goal progress is low'],\n        ['repoGroundedness', 'high', 'output is poorly grounded in the repository'],\n        ['toolUseQuality', 'medium', 'tool use quality is low'],\n        ['patchQuality', 'medium', 'no real patch/edit evidence'],\n        ['testReality', 'high', 'no real test/build evidence'],\n        ['finalGate', 'critical', 'final gate is blocking'],\n      ]\n      for (const [dim, sev, msg] of dims) {\n        const value = score[dim] as number\n        if (typeof value === 'number' && value < threshold) {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject: dim,\n              claim: msg,\n              rationale: `${dim}=${value.toFixed(2)} below threshold ${threshold}`,\n              severity: sev,\n              confidence: 1,\n              evidence_refs: [],\n              metadata: { dimension: dim, value, threshold, run_id: trace.run.runId },\n            }),\n          )\n        }\n      }\n      // Drift penalty is high → surface as a finding (inverse threshold).\n      if (score.driftPenalty > 1 - threshold) {\n        out.push(\n          makeFinding({\n            analyst_id: id,\n            area,\n            subject: 'drift',\n            claim: 'agent output drifted from repository signal',\n            rationale: `driftPenalty=${score.driftPenalty.toFixed(2)}`,\n            severity: 'medium',\n            confidence: 0.9,\n            evidence_refs: [],\n            metadata: { drift_penalty: score.driftPenalty, notes: score.notes },\n          }),\n        )\n      }\n      return out\n    },\n  }\n}\n\n// ── 3. JudgeFn → Analyst ────────────────────────────────────────────\n\nexport interface JudgeAdapterOpts {\n  id?: string\n  area?: string\n  judge: JudgeFn\n  /** TCloud handle the JudgeFn calls. */\n  tcloud: TCloud\n  /** Optional cost classification — most judges call an LLM. */\n  cost?: Analyst['cost']\n  /** Optional threshold below which a JudgeScore becomes a finding. Default 6 (on 0-10 scale). */\n  threshold?: number\n}\n\nexport function createJudgeAdapter(opts: JudgeAdapterOpts): Analyst<JudgeInput> {\n  const id = opts.id ?? 'judge'\n  const area = opts.area ?? 'judge'\n  const threshold = opts.threshold ?? 6\n  return {\n    id,\n    description:\n      'Wraps an agent-eval JudgeFn into an analyst; below-threshold dimensions surface as findings.',\n    inputKind: 'judge-input',\n    cost: opts.cost ?? { kind: 'llm' },\n    version: `judge-${ADAPTER_REV}`,\n    async analyze(input) {\n      const scores = await opts.judge(opts.tcloud, input)\n      return scores\n        .filter((s) => normalize10(s.score) < threshold)\n        .map((s) => liftJudgeScore(id, area, s))\n    },\n  }\n}\n\nfunction normalize10(s: number): number {\n  // JudgeScore convention is 0-10 but some judges emit 0-1. Coerce to 0-10.\n  return s <= 1 ? s * 10 : s\n}\n\nfunction liftJudgeScore(analyst_id: string, area: string, s: JudgeScore): AnalystFinding {\n  const score10 = normalize10(s.score)\n  const severity: AnalystSeverity =\n    score10 < 3 ? 'critical' : score10 < 5 ? 'high' : score10 < 7 ? 'medium' : 'low'\n  return makeFinding({\n    analyst_id,\n    area,\n    subject: s.dimension,\n    claim: `${s.judgeName}/${s.dimension} scored ${score10.toFixed(1)}/10`,\n    rationale: s.reasoning,\n    severity,\n    confidence: 0.8,\n    evidence_refs: s.evidence\n      ? [{ kind: 'artifact', uri: 'inline:evidence', excerpt: s.evidence }]\n      : [],\n    // Provenance: this finding IS a judge verdict (an acceptance score), not an\n    // observation of behavior. The steer firewall (assertNoJudgeVerdict) rejects\n    // it from steering — even when it cites an artifact above — because letting a\n    // verdict steer the next attempt is the held-out judge leaking into the loop.\n    derived_from_judge: true,\n    metadata: { judge_name: s.judgeName, dimension: s.dimension, score_10: score10 },\n  })\n}\n\n// ── 4. SemanticConceptJudge → Analyst ──────────────────────────────\n\nexport interface SemanticConceptJudgeAdapterOpts {\n  id?: string\n  area?: string\n  options?: SemanticConceptJudgeOptions\n}\n\nexport function createSemanticConceptJudgeAdapter(\n  opts: SemanticConceptJudgeAdapterOpts = {},\n): Analyst<SemanticConceptJudgeInput> {\n  const id = opts.id ?? 'semantic-concept-judge'\n  const area = opts.area ?? 'concept-coverage'\n  return {\n    id,\n    description:\n      'Runs the semantic-concept judge and surfaces missing / weak concepts as findings.',\n    inputKind: 'custom',\n    cost: { kind: 'llm', models: opts.options?.model ? [opts.options.model] : undefined },\n    version: `${SEMANTIC_CONCEPT_JUDGE_VERSION}-adapter-${ADAPTER_REV}`,\n    async analyze(input) {\n      const result = await runSemanticConceptJudge(input, opts.options)\n      if (!result.available) {\n        return [\n          makeFinding({\n            analyst_id: id,\n            area,\n            claim: 'semantic-concept judge unavailable',\n            rationale: result.error,\n            severity: 'info',\n            confidence: 1,\n            evidence_refs: [],\n            metadata: { reason: result.error },\n          }),\n        ]\n      }\n      const out: AnalystFinding[] = []\n      for (const f of result.findings) {\n        // Only surface gaps: missing concepts or low scores. Concepts at\n        // 7+/10 with present=true are not findings — they're successes.\n        if (f.present && f.score >= 7) continue\n        out.push(\n          makeFinding({\n            analyst_id: id,\n            area,\n            subject: f.concept,\n            claim: f.present\n              ? `concept \"${f.concept}\" is weak (${f.score}/10)`\n              : `concept \"${f.concept}\" is missing`,\n            rationale: f.evidence,\n            severity: liftSeverity(f.severity),\n            confidence: 0.85,\n            evidence_refs: [{ kind: 'artifact', uri: 'inline:evidence', excerpt: f.evidence }],\n            metadata: {\n              concept: f.concept,\n              present: f.present,\n              score_10: f.score,\n              cost_usd: result.costUsd ?? undefined,\n            },\n          }),\n        )\n      }\n      return out\n    },\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAoCA,IAAM,cAAc;AAIb,SAAS,aAAa,GAAmC;AAC9D,UAAQ,GAAG;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,EACX;AACF;AAeO,SAAS,sBAA2B,MAA8C;AACvF,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,gBAAgB;AAAA,IAC9B,SAAS,YAAY,WAAW;AAAA,IAChC,MAAM,QAAQ,KAAK,KAAK;AACtB,YAAM,SAAS,MAAM,KAAK,SAAS,IAAI,EAAE,KAAK,GAAG,KAAK,QAAQ,CAAC;AAC/D,YAAM,MAAwB,CAAC;AAC/B,iBAAW,SAAS,OAAO,QAAQ;AACjC,mBAAW,WAAW,MAAM,UAAU;AACpC,cAAI,KAAK,iBAAiB,IAAI,MAAM,MAAM,OAAO,OAAO,CAAC;AAAA,QAC3D;AAGA,YAAI,MAAM,WAAW,UAAU,MAAM,WAAW,WAAW,MAAM,WAAW,WAAW;AACrF,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA,SAAS,MAAM;AAAA,cACf,OAAO,UAAU,MAAM,KAAK,KAAK,MAAM,MAAM,KAAK,MAAM,UAAU,iBAAiB;AAAA,cACnF,UACE,MAAM,WAAW,UAAU,SAAS,MAAM,WAAW,YAAY,WAAW;AAAA,cAC9E,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU;AAAA,gBACR,cAAc,MAAM;AAAA,gBACpB,aAAa,MAAM;AAAA,gBACnB,OAAO,MAAM;AAAA,gBACb,aAAa,MAAM;AAAA,cACrB;AAAA,YACF,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,qBAAqB;AAAA,QAC7B,QAAQ,OAAO,OAAO;AAAA,QACtB,SAAS,OAAO;AAAA,QAChB,UAAU,OAAO;AAAA,MACnB,CAAC;AACD,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAEA,SAAS,iBACP,YACA,MACA,OACA,GACgB;AAChB,SAAO,YAAY;AAAA,IACjB;AAAA,IACA;AAAA,IACA,SAAS,EAAE,SAAS;AAAA,IACpB,OAAO,EAAE;AAAA,IACT,UAAU,aAAa,EAAE,QAAQ;AAAA,IACjC,YAAY;AAAA,IACZ,eAAe,EAAE,WACb,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC,IAClE,CAAC;AAAA,IACL,UAAU,EAAE;AAAA,EACd,CAAC;AACH;AAYO,SAAS,uBAAuB,OAA6B,CAAC,GAAsB;AACzF,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,QAAM,SAAS,KAAK,UAAU,IAAI,UAAU;AAC5C,QAAM,YAAY,KAAK,aAAa;AACpC,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,gBAAgB;AAAA,IAC9B,SAAS,cAAc,WAAW;AAAA,IAClC,MAAM,QAAQ,OAAO;AACnB,YAAM,QAAQ,OAAO,WAAW,KAAK;AACrC,YAAM,MAAwB,CAAC;AAC/B,YAAM,OAA6D;AAAA,QACjE,CAAC,WAAW,YAAY,mCAAmC;AAAA,QAC3D,CAAC,gBAAgB,QAAQ,sBAAsB;AAAA,QAC/C,CAAC,oBAAoB,QAAQ,6CAA6C;AAAA,QAC1E,CAAC,kBAAkB,UAAU,yBAAyB;AAAA,QACtD,CAAC,gBAAgB,UAAU,6BAA6B;AAAA,QACxD,CAAC,eAAe,QAAQ,6BAA6B;AAAA,QACrD,CAAC,aAAa,YAAY,wBAAwB;AAAA,MACpD;AACA,iBAAW,CAAC,KAAK,KAAK,GAAG,KAAK,MAAM;AAClC,cAAM,QAAQ,MAAM,GAAG;AACvB,YAAI,OAAO,UAAU,YAAY,QAAQ,WAAW;AAClD,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA,SAAS;AAAA,cACT,OAAO;AAAA,cACP,WAAW,GAAG,GAAG,IAAI,MAAM,QAAQ,CAAC,CAAC,oBAAoB,SAAS;AAAA,cAClE,UAAU;AAAA,cACV,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU,EAAE,WAAW,KAAK,OAAO,WAAW,QAAQ,MAAM,IAAI,MAAM;AAAA,YACxE,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,UAAI,MAAM,eAAe,IAAI,WAAW;AACtC,YAAI;AAAA,UACF,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,SAAS;AAAA,YACT,OAAO;AAAA,YACP,WAAW,gBAAgB,MAAM,aAAa,QAAQ,CAAC,CAAC;AAAA,YACxD,UAAU;AAAA,YACV,YAAY;AAAA,YACZ,eAAe,CAAC;AAAA,YAChB,UAAU,EAAE,eAAe,MAAM,cAAc,OAAO,MAAM,MAAM;AAAA,UACpE,CAAC;AAAA,QACH;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAgBO,SAAS,mBAAmB,MAA6C;AAC9E,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,QAAM,YAAY,KAAK,aAAa;AACpC,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,KAAK,QAAQ,EAAE,MAAM,MAAM;AAAA,IACjC,SAAS,SAAS,WAAW;AAAA,IAC7B,MAAM,QAAQ,OAAO;AACnB,YAAM,SAAS,MAAM,KAAK,MAAM,KAAK,QAAQ,KAAK;AAClD,aAAO,OACJ,OAAO,CAAC,MAAM,YAAY,EAAE,KAAK,IAAI,SAAS,EAC9C,IAAI,CAAC,MAAM,eAAe,IAAI,MAAM,CAAC,CAAC;AAAA,IAC3C;AAAA,EACF;AACF;AAEA,SAAS,YAAY,GAAmB;AAEtC,SAAO,KAAK,IAAI,IAAI,KAAK;AAC3B;AAEA,SAAS,eAAe,YAAoB,MAAc,GAA+B;AACvF,QAAM,UAAU,YAAY,EAAE,KAAK;AACnC,QAAM,WACJ,UAAU,IAAI,aAAa,UAAU,IAAI,SAAS,UAAU,IAAI,WAAW;AAC7E,SAAO,YAAY;AAAA,IACjB;AAAA,IACA;AAAA,IACA,SAAS,EAAE;AAAA,IACX,OAAO,GAAG,EAAE,SAAS,IAAI,EAAE,SAAS,WAAW,QAAQ,QAAQ,CAAC,CAAC;AAAA,IACjE,WAAW,EAAE;AAAA,IACb;AAAA,IACA,YAAY;AAAA,IACZ,eAAe,EAAE,WACb,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC,IAClE,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA,IAKL,oBAAoB;AAAA,IACpB,UAAU,EAAE,YAAY,EAAE,WAAW,WAAW,EAAE,WAAW,UAAU,QAAQ;AAAA,EACjF,CAAC;AACH;AAUO,SAAS,kCACd,OAAwC,CAAC,GACL;AACpC,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,OAAO,QAAQ,KAAK,SAAS,QAAQ,CAAC,KAAK,QAAQ,KAAK,IAAI,OAAU;AAAA,IACpF,SAAS,GAAG,8BAA8B,YAAY,WAAW;AAAA,IACjE,MAAM,QAAQ,OAAO;AACnB,YAAM,SAAS,MAAM,wBAAwB,OAAO,KAAK,OAAO;AAChE,UAAI,CAAC,OAAO,WAAW;AACrB,eAAO;AAAA,UACL,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,OAAO;AAAA,YACP,WAAW,OAAO;AAAA,YAClB,UAAU;AAAA,YACV,YAAY;AAAA,YACZ,eAAe,CAAC;AAAA,YAChB,UAAU,EAAE,QAAQ,OAAO,MAAM;AAAA,UACnC,CAAC;AAAA,QACH;AAAA,MACF;AACA,YAAM,MAAwB,CAAC;AAC/B,iBAAW,KAAK,OAAO,UAAU;AAG/B,YAAI,EAAE,WAAW,EAAE,SAAS,EAAG;AAC/B,YAAI;AAAA,UACF,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,SAAS,EAAE;AAAA,YACX,OAAO,EAAE,UACL,YAAY,EAAE,OAAO,cAAc,EAAE,KAAK,SAC1C,YAAY,EAAE,OAAO;AAAA,YACzB,WAAW,EAAE;AAAA,YACb,UAAU,aAAa,EAAE,QAAQ;AAAA,YACjC,YAAY;AAAA,YACZ,eAAe,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC;AAAA,YACjF,UAAU;AAAA,cACR,SAAS,EAAE;AAAA,cACX,SAAS,EAAE;AAAA,cACX,UAAU,EAAE;AAAA,cACZ,UAAU,OAAO,WAAW;AAAA,YAC9B;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}

package/dist/{analyze-runs-DJYpep3L.d.ts → analyze-runs-Dmz6LA9e.d.ts}

@@ -1,7 +1,7 @@
-import { A as AnalystRegistry } from './default-registry-BswHCXnU.js';
-import { a as DatasetScenario } from './dataset-DS7ytHZU.js';
-import { R as RunRecord } from './run-record-I-Z3JNvO.js';
-import { I as InsightReport } from './insight-report-B4xrdwEK.js';
+import { A as AnalystRegistry } from './default-registry-Bcf1uKVI.js';
+import { D as DatasetScenario } from './dataset-NENEzRgk.js';
+import { R as RunRecord } from './run-record-DksGsfgv.js';
+import { I as InsightReport } from './insight-report-D4cXFsLt.js';
 
 /**
  * # `analyzeRuns()` — turn a set of agent runs into an actionable decision packet.

package/dist/{baseline-Bbid3WoO.d.ts → baseline-DsNteOgR.d.ts}

@@ -1,4 +1,5 @@
-import { T as TraceStore } from './store-BcFXE6LG.js';
+import { T as TraceStore } from './store-BsVi7ncX.js';
+import { S as Span, e as TraceEvent } from './schema-SGWcK9wa.js';
 
 /**
  * Tool-use metrics — derived purely from trace data.
@@ -33,6 +34,35 @@ interface ToolUseOptions {
 }
 declare function computeToolUseMetrics(store: TraceStore, runId: string, options?: ToolUseOptions): Promise<ToolUseMetrics>;
 
+/**
+ * Trajectory — ordered, structured view over a run's spans.
+ *
+ * A pure function `buildTrajectory(store, runId) → Trajectory` returns
+ * a topologically ordered list of `TrajectoryStep` with parent-child
+ * grouping collapsed into a single line-of-agent-work. Separate
+ * analyzers (stuck-loop detection, waste ratio) live in
+ * `pipelines/` and consume the trajectory.
+ */
+
+interface TrajectoryStep {
+    index: number;
+    span: Span;
+    /** Depth in the span tree from the root. 0 = top-level. */
+    depth: number;
+    /** Events attached to this span. */
+    events: TraceEvent[];
+}
+interface Trajectory {
+    runId: string;
+    steps: TrajectoryStep[];
+    llmTurns: number;
+    toolCalls: number;
+    judgeVerdicts: number;
+    retrievals: number;
+    totalDurationMs: number;
+}
+declare function buildTrajectory(store: TraceStore, runId: string): Promise<Trajectory>;
+
 /**
  * Baseline regression detection.
  *
@@ -105,4 +135,4 @@ declare function welchsTTest(a: number[], b: number[]): {
     p: number;
 };
 
-export { type BaselineOptions as B, type MetricSamples as M, type ToolStats as T, type BaselineReport as a, type MetricVerdict as b, computeToolUseMetrics as c, type ToolUseMetrics as d, type ToolUseOptions as e, compareToBaseline as f, iqr as i, welchsTTest as w };
+export { type BaselineOptions as B, type MetricSamples as M, type TrajectoryStep as T, type BaselineReport as a, type Trajectory as b, computeToolUseMetrics as c, type MetricVerdict as d, type ToolStats as e, type ToolUseMetrics as f, type ToolUseOptions as g, buildTrajectory as h, compareToBaseline as i, iqr as j, welchsTTest as w };

package/dist/belief-state/index.d.ts

@@ -1,10 +1,10 @@
-import { c as CalibrationReport } from '../calibration-BPmzuVPk.js';
+import { c as CalibrationReport } from '../calibration-Dz8TQV4y.js';
 import { O as OffPolicyEstimate, a as OffPolicyOptions, b as OffPolicyTrajectory } from '../off-policy-DiwuKKg7.js';
-import { d as CodeAgentSessionSource, a as CodeAgentSessionIntakeOptions, c as CodeAgentSessionMetrics, C as CodeAgentSessionDiagnostic } from '../code-agent-session-rnJKlqmT.js';
-import { R as RunRecord, b as RunSplitTag } from '../run-record-I-Z3JNvO.js';
-import { T as TraceStore } from '../store-BcFXE6LG.js';
-import { R as RuntimeTrajectoryRecord, P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection } from '../runtime-trajectory-iW9IhV3e.js';
-import '../schema-m0gsnbt3.js';
+import { d as CodeAgentSessionSource, a as CodeAgentSessionIntakeOptions, c as CodeAgentSessionMetrics, C as CodeAgentSessionDiagnostic } from '../code-agent-session-yitf9I-F.js';
+import { R as RunRecord, b as RunSplitTag } from '../run-record-DksGsfgv.js';
+import { T as TraceStore } from '../store-BsVi7ncX.js';
+import { R as RuntimeTrajectoryRecord, P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection } from '../runtime-trajectory-h5i0SZUj.js';
+import '../schema-SGWcK9wa.js';
 import '../outcome-store-rnXLEqSn.js';
 import '@tangle-network/agent-interface';
 import '../errors-oeQrLqXC.js';

package/dist/benchmarks/index.d.ts

@@ -1,7 +1,7 @@
-export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, k as BenchmarkDistribution, c as BenchmarkEvaluation, d as BenchmarkFamily, l as BenchmarkMetricCalibrationOptions, m as BenchmarkMetricCalibrationResult, n as BenchmarkReport, e as BenchmarkResponder, o as BenchmarkRunOptions, p as BenchmarkRunResult, f as BenchmarkScenario, q as BenchmarkSliceSummary, g as BenchmarkSource, h as BenchmarkTaskKind, r as BuildStandardRetrievalItemsOptions, R as RetrievalIdAdapterOptions, S as StandardRetrievalArtifact, s as StandardRetrievalDocument, t as StandardRetrievalEvaluationOptions, u as StandardRetrievalPayload, v as StandardRetrievalQrel, w as StandardRetrievalQuery, x as StandardRetrievalResult, y as buildStandardRetrievalItems, z as calibrateBenchmarkMetric, A as createRetrievalIdBenchmarkAdapter, i as deterministicSplit, C as evaluateStandardRetrieval, D as normalizeRetrievedDocumentIds, E as parseBeirCorpusJsonl, F as parseBeirQueriesJsonl, G as parseJsonlRows, H as parseQrels, I as parseTsvRows, J as renderBenchmarkReportMarkdown, K as retrievalMetricsAtCutoff, L as routing, M as runBenchmarkAdapter, N as summarizeBenchmarkCampaign } from '../index-pPtfoIJO.js';
-import '../types-BdIv5dvA.js';
-import '../run-record-I-Z3JNvO.js';
+export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, k as BenchmarkDistribution, c as BenchmarkEvaluation, d as BenchmarkFamily, l as BenchmarkMetricCalibrationOptions, m as BenchmarkMetricCalibrationResult, n as BenchmarkReport, e as BenchmarkResponder, o as BenchmarkRunOptions, p as BenchmarkRunResult, f as BenchmarkScenario, q as BenchmarkSliceSummary, g as BenchmarkSource, h as BenchmarkTaskKind, r as BuildStandardRetrievalItemsOptions, R as RetrievalIdAdapterOptions, S as StandardRetrievalArtifact, s as StandardRetrievalDocument, t as StandardRetrievalEvaluationOptions, u as StandardRetrievalPayload, v as StandardRetrievalQrel, w as StandardRetrievalQuery, x as StandardRetrievalResult, y as buildStandardRetrievalItems, z as calibrateBenchmarkMetric, A as createRetrievalIdBenchmarkAdapter, i as deterministicSplit, C as evaluateStandardRetrieval, D as normalizeRetrievedDocumentIds, E as parseBeirCorpusJsonl, F as parseBeirQueriesJsonl, G as parseJsonlRows, H as parseQrels, I as parseTsvRows, J as renderBenchmarkReportMarkdown, K as retrievalMetricsAtCutoff, L as routing, M as runBenchmarkAdapter, N as summarizeBenchmarkCampaign } from '../index-Dc3VLGhp.js';
+import '../types-v--ctu-b.js';
+import '../run-record-DksGsfgv.js';
 import '@tangle-network/agent-interface';
 import '../errors-oeQrLqXC.js';
-import '../schema-m0gsnbt3.js';
+import '../schema-SGWcK9wa.js';
 import '../storage-Dw_f7WMt.js';

package/dist/benchmarks/index.js

@@ -17,23 +17,22 @@ import {
   runBenchmarkAdapter,
   summarizeBenchmarkCampaign
 } from "../chunk-T6W5ADLG.js";
-import "../chunk-6SKVFBTR.js";
-import "../chunk-OVPVM4JC.js";
-import "../chunk-3PFZBGMR.js";
+import "../chunk-RNB2NICW.js";
+import "../chunk-GTERJI6Q.js";
 import "../chunk-VI2UW6B6.js";
-import "../chunk-LOJ2QVCE.js";
-import "../chunk-2OGPXHOB.js";
+import "../chunk-3PFZBGMR.js";
+import "../chunk-2IY4ILP4.js";
+import "../chunk-7NX6ZSBG.js";
 import "../chunk-N22ZO7FV.js";
-import "../chunk-FUCQVFMU.js";
-import "../chunk-45EEMHTC.js";
 import "../chunk-XMSYF4A7.js";
 import "../chunk-RPDDVKI7.js";
 import "../chunk-GGE4NNQT.js";
 import "../chunk-LNQEP766.js";
-import "../chunk-PC4UYEBM.js";
 import "../chunk-VK6HBGAE.js";
 import "../chunk-XJYR7XFV.js";
 import "../chunk-VSMTAMNK.js";
+import "../chunk-FUCQVFMU.js";
+import "../chunk-PC4UYEBM.js";
 import "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";
 export {

package/dist/builder-eval/index.d.ts

@@ -1,7 +1,7 @@
-import { S as SandboxDriver, H as HarnessConfig, a as SandboxHarnessResult, T as TestGradedScenario, b as TestGradedRunResult } from '../test-graded-scenario-DeODGLra.js';
-import { T as TraceEmitter } from '../emitter-C2rqGH_l.js';
-import { R as Run } from '../schema-m0gsnbt3.js';
-import { T as TraceStore } from '../store-BcFXE6LG.js';
+import { S as SandboxDriver, H as HarnessConfig, a as SandboxHarnessResult, T as TestGradedScenario, b as TestGradedRunResult } from '../test-graded-scenario-mzYBKspu.js';
+import { T as TraceEmitter } from '../emitter-BRchAAAx.js';
+import { R as Run } from '../schema-SGWcK9wa.js';
+import { T as TraceStore } from '../store-BsVi7ncX.js';
 
 /**
  * BuilderSession — ties a builder-of-builders workflow together.

package/dist/builder-eval/index.js

@@ -8,8 +8,7 @@ import {
 } from "../chunk-XMSYF4A7.js";
 import {
   judgeSpans
-} from "../chunk-JZXGWLK5.js";
-import "../chunk-5BKGXME7.js";
+} from "../chunk-MHNQWM4I.js";
 import {
   TraceEmitter
 } from "../chunk-TVVP3ZZQ.js";

package/dist/builder-eval/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/builder-eval/builder-session.ts","../../src/builder-eval/correlation.ts","../../src/builder-eval/project-registry.ts","../../src/builder-eval/three-layer-eval.ts"],"sourcesContent":["/**\n * BuilderSession — ties a builder-of-builders workflow together.\n *\n * Models agent-builder's shape: Project → Chat → Edit → Ship → App →\n * AppAgent. Each layer is a Run (linked via parentRunId). The\n * framework-enforced invariants:\n *\n *   - One Project → many Chats; chatId scopes runs within a project.\n *   - One Chat = one builder Run with `layer='builder'`.\n *   - One Ship = one child Run with `layer='app-build'` + SandboxHarness.\n *   - One AppScenario = one grandchild Run with `layer='app-runtime'`.\n *\n * Consumers obtain a BuilderSession, call `startChat`, drive the\n * builder agent (emitting spans), and call `ship` / `runAppScenario`\n * as the workflow progresses. The session reconstructs itself from\n * trace data via `resume(store, projectId)`.\n */\n\nimport type { HarnessConfig, SandboxDriver, SandboxHarnessResult } from '../sandbox-harness'\nimport { SandboxHarness } from '../sandbox-harness'\nimport type { TestGradedRunResult, TestGradedScenario } from '../test-graded-scenario'\nimport { runTestGradedScenario } from '../test-graded-scenario'\nimport { TraceEmitter } from '../trace/emitter'\nimport type { Run } from '../trace/schema'\nimport type { TraceStore } from '../trace/store'\n\nexport interface BuilderSessionInit {\n  projectId: string\n  chatId?: string\n  /** Free-form: user's task description, project name, etc. Stored on the builder Run. */\n  tags?: Record<string, string>\n}\n\nexport interface ShipOptions {\n  harness: HarnessConfig\n  driver?: SandboxDriver\n  /** scenarioId of this app-build run. Defaults to `${projectId}/build`. */\n  scenarioId?: string\n}\n\nexport interface RunAppScenarioOptions {\n  scenario: TestGradedScenario\n  /** Harness driver override; defaults to the one the session was created with. */\n  driver?: SandboxDriver\n}\n\nexport class BuilderSession {\n  private store: TraceStore\n  private builderEmitter: TraceEmitter\n  readonly projectId: string\n  readonly chatId: string\n  private builderRunId?: string\n  private lastBuildRunId?: string\n  private defaultDriver?: SandboxDriver\n\n  constructor(store: TraceStore, init: BuilderSessionInit, driver?: SandboxDriver) {\n    this.store = store\n    this.projectId = init.projectId\n    this.chatId = init.chatId ?? cryptoId()\n    this.defaultDriver = driver\n    this.builderEmitter = new TraceEmitter(store)\n  }\n\n  /** Start the builder (L0) run for this chat. Returns the runId. */\n  async startChat(scenarioId = `${this.projectId}/chat`): Promise<string> {\n    await this.builderEmitter.startRun({\n      scenarioId,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      layer: 'builder',\n    })\n    this.builderRunId = this.builderEmitter.runId\n    return this.builderRunId\n  }\n\n  /** The emitter for builder-level spans (edits, LLM calls, tool invocations). */\n  get emitter(): TraceEmitter {\n    if (!this.builderRunId) throw new Error('BuilderSession.emitter: call startChat() first')\n    return this.builderEmitter\n  }\n\n  /**\n   * Ship the project's generated app: run the sandbox harness as a child\n   * Run (`layer='app-build'`). Returns the build result + runId.\n   */\n  async ship(options: ShipOptions): Promise<{ runId: string; result: SandboxHarnessResult }> {\n    if (!this.builderRunId) throw new Error('BuilderSession.ship: call startChat() first')\n    const buildEmitter = new TraceEmitter(this.store)\n    await buildEmitter.startRun({\n      scenarioId: options.scenarioId ?? `${this.projectId}/build`,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      parentRunId: this.builderRunId,\n      layer: 'app-build',\n    })\n    const harness = new SandboxHarness(options.driver ?? this.defaultDriver)\n    const result = await harness.run(options.harness, buildEmitter)\n    await buildEmitter.endRun({\n      pass: result.passed,\n      score: result.score,\n      failureClass: result.passed ? 'success' : 'sandbox_failure',\n    })\n    this.lastBuildRunId = buildEmitter.runId\n    return { runId: buildEmitter.runId, result }\n  }\n\n  /**\n   * Run a domain scenario against the just-built app as a grandchild Run\n   * (`layer='app-runtime'`). The `ship` call must precede this so the\n   * parent is set correctly; if no build exists yet the session attaches\n   * directly to the builder run (useful for prototypes).\n   */\n  async runAppScenario(options: RunAppScenarioOptions): Promise<TestGradedRunResult> {\n    const parentRunId = this.lastBuildRunId ?? this.builderRunId\n    if (!parentRunId)\n      throw new Error('BuilderSession.runAppScenario: call startChat() + ship() first')\n    const { scenario, driver } = options\n    const result = await runTestGradedScenario(scenario, this.store, {\n      driver: driver ?? this.defaultDriver,\n      provenance: { codeSha: undefined, promptSha: undefined, modelFingerprint: undefined },\n    })\n    // Attach to the parent chain by updating the stored Run in place.\n    await this.store.updateRun(result.runId, {\n      parentRunId,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      layer: 'app-runtime',\n    })\n    return result\n  }\n\n  /** Record an end-of-chat meta score (judge verdict on whether the builder\n   *  served the user's intent). Accepts a numeric score + optional rationale. */\n  async recordMetaScore(score: number, rationale?: string): Promise<void> {\n    if (!this.builderRunId)\n      throw new Error('BuilderSession.recordMetaScore: call startChat() first')\n    await this.builderEmitter.recordJudge({\n      judgeId: 'builder-meta',\n      targetSpanId: this.builderRunId, // attach to the builder run itself\n      dimension: 'user_intent_satisfaction',\n      score,\n      rationale,\n      name: 'builder-meta',\n    })\n  }\n\n  /** Close the builder Run with a final outcome. */\n  async endChat(outcome: { pass: boolean; score?: number; notes?: string }): Promise<void> {\n    await this.builderEmitter.endRun({\n      pass: outcome.pass,\n      score: outcome.score,\n      notes: outcome.notes,\n    })\n  }\n\n  /**\n   * Inline app-runtime run — for cases where the \"scenario\" isn't a\n   * SWE-bench-style test suite but a live agent interaction (LLM chat,\n   * domain flow). Returns an emitter bound to a fresh Run in the\n   * `app-runtime` layer; caller emits spans inside and calls\n   * `.endRun()` with the final verdict.\n   */\n  async startAppRuntime(scenarioId: string): Promise<TraceEmitter> {\n    const parentRunId = this.lastBuildRunId ?? this.builderRunId\n    if (!parentRunId)\n      throw new Error(\n        'BuilderSession.startAppRuntime: call startChat() + (optionally) ship() first',\n      )\n    const emitter = new TraceEmitter(this.store)\n    await emitter.startRun({\n      scenarioId,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      parentRunId,\n      layer: 'app-runtime',\n    })\n    return emitter\n  }\n\n  /**\n   * Lightweight \"ship marker\" — record an app-build Run with a caller-\n   * provided verdict. Use when there isn't a sandbox harness to run but\n   * you still want to mark the build state at publish time.\n   */\n  async recordShipMarker(args: {\n    pass: boolean\n    score: number\n    scenarioId?: string\n    notes?: string\n  }): Promise<string> {\n    if (!this.builderRunId)\n      throw new Error('BuilderSession.recordShipMarker: call startChat() first')\n    const emitter = new TraceEmitter(this.store)\n    await emitter.startRun({\n      scenarioId: args.scenarioId ?? `${this.projectId}/ship`,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      parentRunId: this.builderRunId,\n      layer: 'app-build',\n    })\n    await emitter.endRun({\n      pass: args.pass,\n      score: args.score,\n      failureClass: args.pass ? 'success' : 'sandbox_failure',\n      notes: args.notes,\n    })\n    this.lastBuildRunId = emitter.runId\n    return emitter.runId\n  }\n\n  get lastBuildRunIdValue(): string | undefined {\n    return this.lastBuildRunId\n  }\n  get builderRunIdValue(): string | undefined {\n    return this.builderRunId\n  }\n}\n\n/**\n * Reconstruct the most recent BuilderSession state for a given project —\n * returns { builderRunId, lastBuildRunId, chatRuns }. For chat-first UIs\n * this is how a resumed session finds its place in the edit history.\n */\nexport async function resumeBuilderSession(\n  store: TraceStore,\n  projectId: string,\n): Promise<{\n  projectId: string\n  chatRuns: Run[]\n  lastBuilderRun?: Run\n  lastBuildRun?: Run\n  lastAppRuntimeRuns: Run[]\n}> {\n  const runs = await store.listRuns({ projectId })\n  const chatRuns = runs\n    .filter((r) => r.layer === 'builder')\n    .sort((a, b) => b.startedAt - a.startedAt)\n  const buildRuns = runs\n    .filter((r) => r.layer === 'app-build')\n    .sort((a, b) => b.startedAt - a.startedAt)\n  const appRuntimeRuns = runs\n    .filter((r) => r.layer === 'app-runtime')\n    .sort((a, b) => b.startedAt - a.startedAt)\n  return {\n    projectId,\n    chatRuns,\n    lastBuilderRun: chatRuns[0],\n    lastBuildRun: buildRuns[0],\n    lastAppRuntimeRuns: appRuntimeRuns,\n  }\n}\n\nfunction cryptoId(): 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 * Meta-eval correlation — the highest-leverage signal in the framework.\n *\n * Given a corpus of three-layer project reports, compute how well each\n * pair of layers correlates. The question we care about most:\n *\n *   Does `metaScore` (what the builder thinks it did) predict\n *   `runtimeScore` (what the user actually gets)?\n *\n * If r < ~0.4, the builder's self-scoring is broken — it's optimizing\n * for something other than real-world success. If r > 0.7, meta_score\n * is a usable proxy and can drive CI gates cheaply.\n *\n * Non-parametric rank correlation (Spearman) is also reported because\n * meta scores are often ordinal-ish.\n */\n\nimport { pearsonR, spearmanR } from '../statistics'\nimport type { ThreeLayerProjectReport } from './three-layer-eval'\n\nexport interface LayerCorrelation {\n  n: number\n  pearson: number\n  spearman: number\n}\n\nexport interface CorrelationReport {\n  /** Pairs present in the corpus (layers with ≥ 2 matched data points). */\n  metaVsBuild?: LayerCorrelation\n  metaVsRuntime?: LayerCorrelation\n  buildVsRuntime?: LayerCorrelation\n  /** Number of complete projects (all 3 scores present). */\n  completeProjects: number\n}\n\nexport function correlateLayers(reports: ThreeLayerProjectReport[]): CorrelationReport {\n  const completeProjects = reports.filter((r) => r.complete).length\n  return {\n    metaVsBuild: pairwise(\n      reports,\n      (r) => r.metaScore,\n      (r) => r.buildScore,\n    ),\n    metaVsRuntime: pairwise(\n      reports,\n      (r) => r.metaScore,\n      (r) => r.runtimeScore,\n    ),\n    buildVsRuntime: pairwise(\n      reports,\n      (r) => r.buildScore,\n      (r) => r.runtimeScore,\n    ),\n    completeProjects,\n  }\n}\n\nfunction pairwise(\n  reports: ThreeLayerProjectReport[],\n  a: (r: ThreeLayerProjectReport) => number | null,\n  b: (r: ThreeLayerProjectReport) => number | null,\n): LayerCorrelation | undefined {\n  const xs: number[] = []\n  const ys: number[] = []\n  for (const r of reports) {\n    const x = a(r)\n    const y = b(r)\n    if (x !== null && y !== null && Number.isFinite(x) && Number.isFinite(y)) {\n      xs.push(x)\n      ys.push(y)\n    }\n  }\n  if (xs.length < 2) return undefined\n  return {\n    n: xs.length,\n    pearson: pearsonR(xs, ys),\n    spearman: spearmanR(xs, ys),\n  }\n}\n","/**\n * ProjectRegistry — project-level aggregation over the trace corpus.\n *\n * Thin reader over TraceStore that answers the questions a chat-first,\n * resumable UI needs:\n *   - listProjects() → project IDs with latest activity\n *   - projectTimeline(id) → chats + builds + runtime runs, chronological\n *   - projectChats(id) → chat-level summaries (turn count, outcome)\n *\n * All queries are pure reads; no state duplication.\n */\n\nimport type { Run } from '../trace/schema'\nimport type { TraceStore } from '../trace/store'\n\nexport interface ProjectSummary {\n  projectId: string\n  chatCount: number\n  buildCount: number\n  appRuntimeCount: number\n  lastActivityAt: number\n  latestChatId?: string\n  latestOutcome?: { pass: boolean; score?: number }\n}\n\nexport interface ChatSummary {\n  chatId: string\n  projectId: string\n  builderRunId: string\n  startedAt: number\n  endedAt?: number\n  status: Run['status']\n  outcome?: Run['outcome']\n  /** Counts of spans emitted during the chat. */\n  llmTurns?: number\n  toolCalls?: number\n  buildRunId?: string\n  appRuntimeRunIds: string[]\n}\n\nexport interface ProjectTimelineEntry {\n  run: Run\n  layerBucket: 'chat' | 'build' | 'runtime' | 'other'\n}\n\nexport class ProjectRegistry {\n  constructor(private store: TraceStore) {}\n\n  async listProjects(): Promise<ProjectSummary[]> {\n    const runs = await this.store.listRuns()\n    const byProject = new Map<string, Run[]>()\n    for (const r of runs) {\n      if (!r.projectId) continue\n      const arr = byProject.get(r.projectId) ?? []\n      arr.push(r)\n      byProject.set(r.projectId, arr)\n    }\n    const summaries: ProjectSummary[] = []\n    for (const [projectId, projectRuns] of byProject) {\n      const sorted = projectRuns.slice().sort((a, b) => b.startedAt - a.startedAt)\n      const chats = projectRuns.filter((r) => r.layer === 'builder')\n      const builds = projectRuns.filter((r) => r.layer === 'app-build')\n      const runtimes = projectRuns.filter((r) => r.layer === 'app-runtime')\n      const latest = sorted[0]\n      if (!latest) continue\n      summaries.push({\n        projectId,\n        chatCount: chats.length,\n        buildCount: builds.length,\n        appRuntimeCount: runtimes.length,\n        lastActivityAt: latest.startedAt,\n        latestChatId: chats[0]?.chatId,\n        latestOutcome: latest.outcome\n          ? { pass: latest.outcome.pass ?? false, score: latest.outcome.score }\n          : undefined,\n      })\n    }\n    return summaries.sort((a, b) => b.lastActivityAt - a.lastActivityAt)\n  }\n\n  async projectTimeline(projectId: string): Promise<ProjectTimelineEntry[]> {\n    const runs = await this.store.listRuns({ projectId })\n    const ordered = runs.slice().sort((a, b) => a.startedAt - b.startedAt)\n    return ordered.map((run) => ({\n      run,\n      layerBucket:\n        run.layer === 'builder'\n          ? 'chat'\n          : run.layer === 'app-build'\n            ? 'build'\n            : run.layer === 'app-runtime'\n              ? 'runtime'\n              : 'other',\n    }))\n  }\n\n  async projectChats(projectId: string): Promise<ChatSummary[]> {\n    const builderRuns = (await this.store.listRuns({ projectId, layer: 'builder' })).sort(\n      (a, b) => b.startedAt - a.startedAt,\n    )\n    const childrenFor = async (runId: string) => this.store.listRuns({ parentRunId: runId })\n    const out: ChatSummary[] = []\n    for (const run of builderRuns) {\n      const spans = await this.store.spans({ runId: run.runId })\n      const children = await childrenFor(run.runId)\n      const build = children.find((c) => c.layer === 'app-build')\n      const runtime: string[] = []\n      // Runtime runs can be grandchildren (attached to the build) or siblings\n      // when shipped skipped.\n      if (build) {\n        const grands = await childrenFor(build.runId)\n        for (const g of grands) if (g.layer === 'app-runtime') runtime.push(g.runId)\n      }\n      for (const c of children) if (c.layer === 'app-runtime') runtime.push(c.runId)\n      out.push({\n        chatId: run.chatId ?? run.runId,\n        projectId,\n        builderRunId: run.runId,\n        startedAt: run.startedAt,\n        endedAt: run.endedAt,\n        status: run.status,\n        outcome: run.outcome,\n        llmTurns: spans.filter((s) => s.kind === 'llm').length,\n        toolCalls: spans.filter((s) => s.kind === 'tool').length,\n        buildRunId: build?.runId,\n        appRuntimeRunIds: runtime,\n      })\n    }\n    return out\n  }\n}\n","/**\n * Three-layer evaluation — the canonical scoring breakdown for\n * builder-of-builders workflows.\n *\n *   meta_score:    did the builder understand + satisfy user intent?\n *                  (judge verdict attached to the builder run)\n *   build_score:   did the generated scaffold build + pass its own tests?\n *                  (outcome.score on the app-build child run)\n *   runtime_score: did the generated agent pass its domain scenarios?\n *                  (mean outcome.score over app-runtime grandchild runs)\n *\n * Returns a structured report per project. The cross-layer correlation\n * is the highest-leverage signal the framework computes — if\n * meta_score doesn't predict runtime_score, the builder's self-scoring\n * is broken.\n *\n * Scaffold-only mode: when a project has no `app-runtime` runs (e.g. a\n * scaffold-builder eval that grades compose + build without driving a\n * runtime scenario), `kind` is `'scaffold-only'` and `complete` measures\n * meta + build only. Consumers can tell the two apart without having to\n * interpret null-runtime as either \"not yet computed\" or \"N/A for this\n * project shape\".\n */\n\nimport { judgeSpans } from '../trace/query'\nimport type { Run } from '../trace/schema'\nimport type { TraceStore } from '../trace/store'\n\nexport type ProjectKind = 'full' | 'scaffold-only'\n\nexport interface ThreeLayerProjectReport {\n  projectId: string\n  /**\n   * `'full'` when the project has at least one `app-runtime` run;\n   * `'scaffold-only'` when it only has meta + build layers. Lets\n   * downstream consumers treat a null runtime score as expected\n   * (scaffold-only) vs. missing (full, pipeline broke).\n   */\n  kind: ProjectKind\n  builderRunId?: string\n  /** Judge-verdict score on the builder run (0..1 after normalization). */\n  metaScore: number | null\n  buildRunId?: string\n  /** 0..1 from the sandbox harness (testsPassed / testsTotal). */\n  buildScore: number | null\n  appRuntimeRunIds: string[]\n  /** Mean of outcome.score over app-runtime runs, 0..1. Always null in scaffold-only mode. */\n  runtimeScore: number | null\n  runtimePassRate: number | null\n  /**\n   * Layer-aware completeness:\n   *   - `kind='full'`: all three layers scored\n   *   - `kind='scaffold-only'`: meta + build scored (runtime not applicable)\n   */\n  complete: boolean\n}\n\nexport async function scoreProject(\n  store: TraceStore,\n  projectId: string,\n): Promise<ThreeLayerProjectReport> {\n  const allRuns = await store.listRuns({ projectId })\n  const builder = latestByLayer(allRuns, 'builder')\n  const build = latestByLayer(allRuns, 'app-build')\n  const runtime = allRuns.filter((r) => r.layer === 'app-runtime')\n\n  const metaScore = builder ? await extractMetaScore(store, builder.runId) : null\n  const buildScore = build?.outcome?.score ?? null\n  const runtimeScores = runtime\n    .map((r) => r.outcome?.score)\n    .filter((s): s is number => typeof s === 'number')\n  const runtimeScore =\n    runtimeScores.length > 0\n      ? runtimeScores.reduce((a, b) => a + b, 0) / runtimeScores.length\n      : null\n  const runtimePassed = runtime.filter((r) => r.outcome?.pass === true).length\n  const runtimePassRate = runtime.length > 0 ? runtimePassed / runtime.length : null\n\n  const kind: ProjectKind = runtime.length === 0 ? 'scaffold-only' : 'full'\n  const complete =\n    kind === 'scaffold-only'\n      ? metaScore !== null && buildScore !== null\n      : metaScore !== null && buildScore !== null && runtimeScore !== null\n\n  return {\n    projectId,\n    kind,\n    builderRunId: builder?.runId,\n    metaScore,\n    buildRunId: build?.runId,\n    buildScore,\n    appRuntimeRunIds: runtime.map((r) => r.runId),\n    runtimeScore,\n    runtimePassRate,\n    complete,\n  }\n}\n\n/** Aggregate scoring across every project in a corpus. */\nexport async function scoreAllProjects(store: TraceStore): Promise<ThreeLayerProjectReport[]> {\n  const runs = await store.listRuns()\n  const projectIds = [...new Set(runs.map((r) => r.projectId).filter((p): p is string => !!p))]\n  return Promise.all(projectIds.map((p) => scoreProject(store, p)))\n}\n\nfunction latestByLayer(runs: Run[], layer: Run['layer']): Run | undefined {\n  const filtered = runs.filter((r) => r.layer === layer).sort((a, b) => b.startedAt - a.startedAt)\n  return filtered[0]\n}\n\nasync function extractMetaScore(store: TraceStore, builderRunId: string): Promise<number | null> {\n  const js = await judgeSpans(store, builderRunId)\n  const meta = js.find(\n    (s) => s.judgeId === 'builder-meta' && s.dimension === 'user_intent_satisfaction',\n  )\n  if (!meta) return null\n  // Normalize score to 0..1. Accept 0-1 natively; 0-10 scale is also common.\n  if (meta.score >= 0 && meta.score <= 1) return meta.score\n  if (meta.score >= 0 && meta.score <= 10) return meta.score / 10\n  return null\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;AA8CO,IAAM,iBAAN,MAAqB;AAAA,EAClB;AAAA,EACA;AAAA,EACC;AAAA,EACA;AAAA,EACD;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,OAAmB,MAA0B,QAAwB;AAC/E,SAAK,QAAQ;AACb,SAAK,YAAY,KAAK;AACtB,SAAK,SAAS,KAAK,UAAU,SAAS;AACtC,SAAK,gBAAgB;AACrB,SAAK,iBAAiB,IAAI,aAAa,KAAK;AAAA,EAC9C;AAAA;AAAA,EAGA,MAAM,UAAU,aAAa,GAAG,KAAK,SAAS,SAA0B;AACtE,UAAM,KAAK,eAAe,SAAS;AAAA,MACjC;AAAA,MACA,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,OAAO;AAAA,IACT,CAAC;AACD,SAAK,eAAe,KAAK,eAAe;AACxC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,UAAwB;AAC1B,QAAI,CAAC,KAAK,aAAc,OAAM,IAAI,MAAM,gDAAgD;AACxF,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,KAAK,SAAgF;AACzF,QAAI,CAAC,KAAK,aAAc,OAAM,IAAI,MAAM,6CAA6C;AACrF,UAAM,eAAe,IAAI,aAAa,KAAK,KAAK;AAChD,UAAM,aAAa,SAAS;AAAA,MAC1B,YAAY,QAAQ,cAAc,GAAG,KAAK,SAAS;AAAA,MACnD,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,aAAa,KAAK;AAAA,MAClB,OAAO;AAAA,IACT,CAAC;AACD,UAAM,UAAU,IAAI,eAAe,QAAQ,UAAU,KAAK,aAAa;AACvE,UAAM,SAAS,MAAM,QAAQ,IAAI,QAAQ,SAAS,YAAY;AAC9D,UAAM,aAAa,OAAO;AAAA,MACxB,MAAM,OAAO;AAAA,MACb,OAAO,OAAO;AAAA,MACd,cAAc,OAAO,SAAS,YAAY;AAAA,IAC5C,CAAC;AACD,SAAK,iBAAiB,aAAa;AACnC,WAAO,EAAE,OAAO,aAAa,OAAO,OAAO;AAAA,EAC7C;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,MAAM,eAAe,SAA8D;AACjF,UAAM,cAAc,KAAK,kBAAkB,KAAK;AAChD,QAAI,CAAC;AACH,YAAM,IAAI,MAAM,gEAAgE;AAClF,UAAM,EAAE,UAAU,OAAO,IAAI;AAC7B,UAAM,SAAS,MAAM,sBAAsB,UAAU,KAAK,OAAO;AAAA,MAC/D,QAAQ,UAAU,KAAK;AAAA,MACvB,YAAY,EAAE,SAAS,QAAW,WAAW,QAAW,kBAAkB,OAAU;AAAA,IACtF,CAAC;AAED,UAAM,KAAK,MAAM,UAAU,OAAO,OAAO;AAAA,MACvC;AAAA,MACA,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,OAAO;AAAA,IACT,CAAC;AACD,WAAO;AAAA,EACT;AAAA;AAAA;AAAA,EAIA,MAAM,gBAAgB,OAAe,WAAmC;AACtE,QAAI,CAAC,KAAK;AACR,YAAM,IAAI,MAAM,wDAAwD;AAC1E,UAAM,KAAK,eAAe,YAAY;AAAA,MACpC,SAAS;AAAA,MACT,cAAc,KAAK;AAAA;AAAA,MACnB,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,MAAM;AAAA,IACR,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,MAAM,QAAQ,SAA2E;AACvF,UAAM,KAAK,eAAe,OAAO;AAAA,MAC/B,MAAM,QAAQ;AAAA,MACd,OAAO,QAAQ;AAAA,MACf,OAAO,QAAQ;AAAA,IACjB,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAM,gBAAgB,YAA2C;AAC/D,UAAM,cAAc,KAAK,kBAAkB,KAAK;AAChD,QAAI,CAAC;AACH,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AACF,UAAM,UAAU,IAAI,aAAa,KAAK,KAAK;AAC3C,UAAM,QAAQ,SAAS;AAAA,MACrB;AAAA,MACA,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb;AAAA,MACA,OAAO;AAAA,IACT,CAAC;AACD,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,iBAAiB,MAKH;AAClB,QAAI,CAAC,KAAK;AACR,YAAM,IAAI,MAAM,yDAAyD;AAC3E,UAAM,UAAU,IAAI,aAAa,KAAK,KAAK;AAC3C,UAAM,QAAQ,SAAS;AAAA,MACrB,YAAY,KAAK,cAAc,GAAG,KAAK,SAAS;AAAA,MAChD,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,aAAa,KAAK;AAAA,MAClB,OAAO;AAAA,IACT,CAAC;AACD,UAAM,QAAQ,OAAO;AAAA,MACnB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,MACZ,cAAc,KAAK,OAAO,YAAY;AAAA,MACtC,OAAO,KAAK;AAAA,IACd,CAAC;AACD,SAAK,iBAAiB,QAAQ;AAC9B,WAAO,QAAQ;AAAA,EACjB;AAAA,EAEA,IAAI,sBAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,oBAAwC;AAC1C,WAAO,KAAK;AAAA,EACd;AACF;AAOA,eAAsB,qBACpB,OACA,WAOC;AACD,QAAM,OAAO,MAAM,MAAM,SAAS,EAAE,UAAU,CAAC;AAC/C,QAAM,WAAW,KACd,OAAO,CAAC,MAAM,EAAE,UAAU,SAAS,EACnC,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3C,QAAM,YAAY,KACf,OAAO,CAAC,MAAM,EAAE,UAAU,WAAW,EACrC,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3C,QAAM,iBAAiB,KACpB,OAAO,CAAC,MAAM,EAAE,UAAU,aAAa,EACvC,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3C,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA,gBAAgB,SAAS,CAAC;AAAA,IAC1B,cAAc,UAAU,CAAC;AAAA,IACzB,oBAAoB;AAAA,EACtB;AACF;AAEA,SAAS,WAAmB;AAC1B,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;;;AC5NO,SAAS,gBAAgB,SAAuD;AACrF,QAAM,mBAAmB,QAAQ,OAAO,CAAC,MAAM,EAAE,QAAQ,EAAE;AAC3D,SAAO;AAAA,IACL,aAAa;AAAA,MACX;AAAA,MACA,CAAC,MAAM,EAAE;AAAA,MACT,CAAC,MAAM,EAAE;AAAA,IACX;AAAA,IACA,eAAe;AAAA,MACb;AAAA,MACA,CAAC,MAAM,EAAE;AAAA,MACT,CAAC,MAAM,EAAE;AAAA,IACX;AAAA,IACA,gBAAgB;AAAA,MACd;AAAA,MACA,CAAC,MAAM,EAAE;AAAA,MACT,CAAC,MAAM,EAAE;AAAA,IACX;AAAA,IACA;AAAA,EACF;AACF;AAEA,SAAS,SACP,SACA,GACA,GAC8B;AAC9B,QAAM,KAAe,CAAC;AACtB,QAAM,KAAe,CAAC;AACtB,aAAW,KAAK,SAAS;AACvB,UAAM,IAAI,EAAE,CAAC;AACb,UAAM,IAAI,EAAE,CAAC;AACb,QAAI,MAAM,QAAQ,MAAM,QAAQ,OAAO,SAAS,CAAC,KAAK,OAAO,SAAS,CAAC,GAAG;AACxE,SAAG,KAAK,CAAC;AACT,SAAG,KAAK,CAAC;AAAA,IACX;AAAA,EACF;AACA,MAAI,GAAG,SAAS,EAAG,QAAO;AAC1B,SAAO;AAAA,IACL,GAAG,GAAG;AAAA,IACN,SAAS,SAAS,IAAI,EAAE;AAAA,IACxB,UAAU,UAAU,IAAI,EAAE;AAAA,EAC5B;AACF;;;ACjCO,IAAM,kBAAN,MAAsB;AAAA,EAC3B,YAAoB,OAAmB;AAAnB;AAAA,EAAoB;AAAA,EAApB;AAAA,EAEpB,MAAM,eAA0C;AAC9C,UAAM,OAAO,MAAM,KAAK,MAAM,SAAS;AACvC,UAAM,YAAY,oBAAI,IAAmB;AACzC,eAAW,KAAK,MAAM;AACpB,UAAI,CAAC,EAAE,UAAW;AAClB,YAAM,MAAM,UAAU,IAAI,EAAE,SAAS,KAAK,CAAC;AAC3C,UAAI,KAAK,CAAC;AACV,gBAAU,IAAI,EAAE,WAAW,GAAG;AAAA,IAChC;AACA,UAAM,YAA8B,CAAC;AACrC,eAAW,CAAC,WAAW,WAAW,KAAK,WAAW;AAChD,YAAM,SAAS,YAAY,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3E,YAAM,QAAQ,YAAY,OAAO,CAAC,MAAM,EAAE,UAAU,SAAS;AAC7D,YAAM,SAAS,YAAY,OAAO,CAAC,MAAM,EAAE,UAAU,WAAW;AAChE,YAAM,WAAW,YAAY,OAAO,CAAC,MAAM,EAAE,UAAU,aAAa;AACpE,YAAM,SAAS,OAAO,CAAC;AACvB,UAAI,CAAC,OAAQ;AACb,gBAAU,KAAK;AAAA,QACb;AAAA,QACA,WAAW,MAAM;AAAA,QACjB,YAAY,OAAO;AAAA,QACnB,iBAAiB,SAAS;AAAA,QAC1B,gBAAgB,OAAO;AAAA,QACvB,cAAc,MAAM,CAAC,GAAG;AAAA,QACxB,eAAe,OAAO,UAClB,EAAE,MAAM,OAAO,QAAQ,QAAQ,OAAO,OAAO,OAAO,QAAQ,MAAM,IAClE;AAAA,MACN,CAAC;AAAA,IACH;AACA,WAAO,UAAU,KAAK,CAAC,GAAG,MAAM,EAAE,iBAAiB,EAAE,cAAc;AAAA,EACrE;AAAA,EAEA,MAAM,gBAAgB,WAAoD;AACxE,UAAM,OAAO,MAAM,KAAK,MAAM,SAAS,EAAE,UAAU,CAAC;AACpD,UAAM,UAAU,KAAK,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AACrE,WAAO,QAAQ,IAAI,CAAC,SAAS;AAAA,MAC3B;AAAA,MACA,aACE,IAAI,UAAU,YACV,SACA,IAAI,UAAU,cACZ,UACA,IAAI,UAAU,gBACZ,YACA;AAAA,IACZ,EAAE;AAAA,EACJ;AAAA,EAEA,MAAM,aAAa,WAA2C;AAC5D,UAAM,eAAe,MAAM,KAAK,MAAM,SAAS,EAAE,WAAW,OAAO,UAAU,CAAC,GAAG;AAAA,MAC/E,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE;AAAA,IAC5B;AACA,UAAM,cAAc,OAAO,UAAkB,KAAK,MAAM,SAAS,EAAE,aAAa,MAAM,CAAC;AACvF,UAAM,MAAqB,CAAC;AAC5B,eAAW,OAAO,aAAa;AAC7B,YAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,EAAE,OAAO,IAAI,MAAM,CAAC;AACzD,YAAM,WAAW,MAAM,YAAY,IAAI,KAAK;AAC5C,YAAM,QAAQ,SAAS,KAAK,CAAC,MAAM,EAAE,UAAU,WAAW;AAC1D,YAAM,UAAoB,CAAC;AAG3B,UAAI,OAAO;AACT,cAAM,SAAS,MAAM,YAAY,MAAM,KAAK;AAC5C,mBAAW,KAAK,OAAQ,KAAI,EAAE,UAAU,cAAe,SAAQ,KAAK,EAAE,KAAK;AAAA,MAC7E;AACA,iBAAW,KAAK,SAAU,KAAI,EAAE,UAAU,cAAe,SAAQ,KAAK,EAAE,KAAK;AAC7E,UAAI,KAAK;AAAA,QACP,QAAQ,IAAI,UAAU,IAAI;AAAA,QAC1B;AAAA,QACA,cAAc,IAAI;AAAA,QAClB,WAAW,IAAI;AAAA,QACf,SAAS,IAAI;AAAA,QACb,QAAQ,IAAI;AAAA,QACZ,SAAS,IAAI;AAAA,QACb,UAAU,MAAM,OAAO,CAAC,MAAM,EAAE,SAAS,KAAK,EAAE;AAAA,QAChD,WAAW,MAAM,OAAO,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE;AAAA,QAClD,YAAY,OAAO;AAAA,QACnB,kBAAkB;AAAA,MACpB,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AACF;;;ACzEA,eAAsB,aACpB,OACA,WACkC;AAClC,QAAM,UAAU,MAAM,MAAM,SAAS,EAAE,UAAU,CAAC;AAClD,QAAM,UAAU,cAAc,SAAS,SAAS;AAChD,QAAM,QAAQ,cAAc,SAAS,WAAW;AAChD,QAAM,UAAU,QAAQ,OAAO,CAAC,MAAM,EAAE,UAAU,aAAa;AAE/D,QAAM,YAAY,UAAU,MAAM,iBAAiB,OAAO,QAAQ,KAAK,IAAI;AAC3E,QAAM,aAAa,OAAO,SAAS,SAAS;AAC5C,QAAM,gBAAgB,QACnB,IAAI,CAAC,MAAM,EAAE,SAAS,KAAK,EAC3B,OAAO,CAAC,MAAmB,OAAO,MAAM,QAAQ;AACnD,QAAM,eACJ,cAAc,SAAS,IACnB,cAAc,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI,cAAc,SACzD;AACN,QAAM,gBAAgB,QAAQ,OAAO,CAAC,MAAM,EAAE,SAAS,SAAS,IAAI,EAAE;AACtE,QAAM,kBAAkB,QAAQ,SAAS,IAAI,gBAAgB,QAAQ,SAAS;AAE9E,QAAM,OAAoB,QAAQ,WAAW,IAAI,kBAAkB;AACnE,QAAM,WACJ,SAAS,kBACL,cAAc,QAAQ,eAAe,OACrC,cAAc,QAAQ,eAAe,QAAQ,iBAAiB;AAEpE,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA,cAAc,SAAS;AAAA,IACvB;AAAA,IACA,YAAY,OAAO;AAAA,IACnB;AAAA,IACA,kBAAkB,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK;AAAA,IAC5C;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAGA,eAAsB,iBAAiB,OAAuD;AAC5F,QAAM,OAAO,MAAM,MAAM,SAAS;AAClC,QAAM,aAAa,CAAC,GAAG,IAAI,IAAI,KAAK,IAAI,CAAC,MAAM,EAAE,SAAS,EAAE,OAAO,CAAC,MAAmB,CAAC,CAAC,CAAC,CAAC,CAAC;AAC5F,SAAO,QAAQ,IAAI,WAAW,IAAI,CAAC,MAAM,aAAa,OAAO,CAAC,CAAC,CAAC;AAClE;AAEA,SAAS,cAAc,MAAa,OAAsC;AACxE,QAAM,WAAW,KAAK,OAAO,CAAC,MAAM,EAAE,UAAU,KAAK,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC/F,SAAO,SAAS,CAAC;AACnB;AAEA,eAAe,iBAAiB,OAAmB,cAA8C;AAC/F,QAAM,KAAK,MAAM,WAAW,OAAO,YAAY;AAC/C,QAAM,OAAO,GAAG;AAAA,IACd,CAAC,MAAM,EAAE,YAAY,kBAAkB,EAAE,cAAc;AAAA,EACzD;AACA,MAAI,CAAC,KAAM,QAAO;AAElB,MAAI,KAAK,SAAS,KAAK,KAAK,SAAS,EAAG,QAAO,KAAK;AACpD,MAAI,KAAK,SAAS,KAAK,KAAK,SAAS,GAAI,QAAO,KAAK,QAAQ;AAC7D,SAAO;AACT;","names":[]}
+{"version":3,"sources":["../../src/builder-eval/builder-session.ts","../../src/builder-eval/correlation.ts","../../src/builder-eval/project-registry.ts","../../src/builder-eval/three-layer-eval.ts"],"sourcesContent":["/**\n * BuilderSession — ties a builder-of-builders workflow together.\n *\n * Models agent-builder's shape: Project → Chat → Edit → Ship → App →\n * AppAgent. Each layer is a Run (linked via parentRunId). The\n * framework-enforced invariants:\n *\n *   - One Project → many Chats; chatId scopes runs within a project.\n *   - One Chat = one builder Run with `layer='builder'`.\n *   - One Ship = one child Run with `layer='app-build'` + SandboxHarness.\n *   - One AppScenario = one grandchild Run with `layer='app-runtime'`.\n *\n * Consumers obtain a BuilderSession, call `startChat`, drive the\n * builder agent (emitting spans), and call `ship` / `runAppScenario`\n * as the workflow progresses. The session reconstructs itself from\n * trace data via `resume(store, projectId)`.\n */\n\nimport type { HarnessConfig, SandboxDriver, SandboxHarnessResult } from '../sandbox-harness'\nimport { SandboxHarness } from '../sandbox-harness'\nimport type { TestGradedRunResult, TestGradedScenario } from '../test-graded-scenario'\nimport { runTestGradedScenario } from '../test-graded-scenario'\nimport { TraceEmitter } from '../trace/emitter'\nimport type { Run } from '../trace/schema'\nimport type { TraceStore } from '../trace/store'\n\nexport interface BuilderSessionInit {\n  projectId: string\n  chatId?: string\n  /** Free-form: user's task description, project name, etc. Stored on the builder Run. */\n  tags?: Record<string, string>\n}\n\nexport interface ShipOptions {\n  harness: HarnessConfig\n  driver?: SandboxDriver\n  /** scenarioId of this app-build run. Defaults to `${projectId}/build`. */\n  scenarioId?: string\n}\n\nexport interface RunAppScenarioOptions {\n  scenario: TestGradedScenario\n  /** Harness driver override; defaults to the one the session was created with. */\n  driver?: SandboxDriver\n}\n\nexport class BuilderSession {\n  private store: TraceStore\n  private builderEmitter: TraceEmitter\n  readonly projectId: string\n  readonly chatId: string\n  private builderRunId?: string\n  private lastBuildRunId?: string\n  private defaultDriver?: SandboxDriver\n\n  constructor(store: TraceStore, init: BuilderSessionInit, driver?: SandboxDriver) {\n    this.store = store\n    this.projectId = init.projectId\n    this.chatId = init.chatId ?? cryptoId()\n    this.defaultDriver = driver\n    this.builderEmitter = new TraceEmitter(store)\n  }\n\n  /** Start the builder (L0) run for this chat. Returns the runId. */\n  async startChat(scenarioId = `${this.projectId}/chat`): Promise<string> {\n    await this.builderEmitter.startRun({\n      scenarioId,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      layer: 'builder',\n    })\n    this.builderRunId = this.builderEmitter.runId\n    return this.builderRunId\n  }\n\n  /** The emitter for builder-level spans (edits, LLM calls, tool invocations). */\n  get emitter(): TraceEmitter {\n    if (!this.builderRunId) throw new Error('BuilderSession.emitter: call startChat() first')\n    return this.builderEmitter\n  }\n\n  /**\n   * Ship the project's generated app: run the sandbox harness as a child\n   * Run (`layer='app-build'`). Returns the build result + runId.\n   */\n  async ship(options: ShipOptions): Promise<{ runId: string; result: SandboxHarnessResult }> {\n    if (!this.builderRunId) throw new Error('BuilderSession.ship: call startChat() first')\n    const buildEmitter = new TraceEmitter(this.store)\n    await buildEmitter.startRun({\n      scenarioId: options.scenarioId ?? `${this.projectId}/build`,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      parentRunId: this.builderRunId,\n      layer: 'app-build',\n    })\n    const harness = new SandboxHarness(options.driver ?? this.defaultDriver)\n    const result = await harness.run(options.harness, buildEmitter)\n    await buildEmitter.endRun({\n      pass: result.passed,\n      score: result.score,\n      failureClass: result.passed ? 'success' : 'sandbox_failure',\n    })\n    this.lastBuildRunId = buildEmitter.runId\n    return { runId: buildEmitter.runId, result }\n  }\n\n  /**\n   * Run a domain scenario against the just-built app as a grandchild Run\n   * (`layer='app-runtime'`). The `ship` call must precede this so the\n   * parent is set correctly; if no build exists yet the session attaches\n   * directly to the builder run (useful for prototypes).\n   */\n  async runAppScenario(options: RunAppScenarioOptions): Promise<TestGradedRunResult> {\n    const parentRunId = this.lastBuildRunId ?? this.builderRunId\n    if (!parentRunId)\n      throw new Error('BuilderSession.runAppScenario: call startChat() + ship() first')\n    const { scenario, driver } = options\n    const result = await runTestGradedScenario(scenario, this.store, {\n      driver: driver ?? this.defaultDriver,\n      provenance: { codeSha: undefined, promptSha: undefined, modelFingerprint: undefined },\n    })\n    // Attach to the parent chain by updating the stored Run in place.\n    await this.store.updateRun(result.runId, {\n      parentRunId,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      layer: 'app-runtime',\n    })\n    return result\n  }\n\n  /** Record an end-of-chat meta score (judge verdict on whether the builder\n   *  served the user's intent). Accepts a numeric score + optional rationale. */\n  async recordMetaScore(score: number, rationale?: string): Promise<void> {\n    if (!this.builderRunId)\n      throw new Error('BuilderSession.recordMetaScore: call startChat() first')\n    await this.builderEmitter.recordJudge({\n      judgeId: 'builder-meta',\n      targetSpanId: this.builderRunId, // attach to the builder run itself\n      dimension: 'user_intent_satisfaction',\n      score,\n      rationale,\n      name: 'builder-meta',\n    })\n  }\n\n  /** Close the builder Run with a final outcome. */\n  async endChat(outcome: { pass: boolean; score?: number; notes?: string }): Promise<void> {\n    await this.builderEmitter.endRun({\n      pass: outcome.pass,\n      score: outcome.score,\n      notes: outcome.notes,\n    })\n  }\n\n  /**\n   * Inline app-runtime run — for cases where the \"scenario\" isn't a\n   * SWE-bench-style test suite but a live agent interaction (LLM chat,\n   * domain flow). Returns an emitter bound to a fresh Run in the\n   * `app-runtime` layer; caller emits spans inside and calls\n   * `.endRun()` with the final verdict.\n   */\n  async startAppRuntime(scenarioId: string): Promise<TraceEmitter> {\n    const parentRunId = this.lastBuildRunId ?? this.builderRunId\n    if (!parentRunId)\n      throw new Error(\n        'BuilderSession.startAppRuntime: call startChat() + (optionally) ship() first',\n      )\n    const emitter = new TraceEmitter(this.store)\n    await emitter.startRun({\n      scenarioId,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      parentRunId,\n      layer: 'app-runtime',\n    })\n    return emitter\n  }\n\n  /**\n   * Lightweight \"ship marker\" — record an app-build Run with a caller-\n   * provided verdict. Use when there isn't a sandbox harness to run but\n   * you still want to mark the build state at publish time.\n   */\n  async recordShipMarker(args: {\n    pass: boolean\n    score: number\n    scenarioId?: string\n    notes?: string\n  }): Promise<string> {\n    if (!this.builderRunId)\n      throw new Error('BuilderSession.recordShipMarker: call startChat() first')\n    const emitter = new TraceEmitter(this.store)\n    await emitter.startRun({\n      scenarioId: args.scenarioId ?? `${this.projectId}/ship`,\n      projectId: this.projectId,\n      chatId: this.chatId,\n      parentRunId: this.builderRunId,\n      layer: 'app-build',\n    })\n    await emitter.endRun({\n      pass: args.pass,\n      score: args.score,\n      failureClass: args.pass ? 'success' : 'sandbox_failure',\n      notes: args.notes,\n    })\n    this.lastBuildRunId = emitter.runId\n    return emitter.runId\n  }\n\n  get lastBuildRunIdValue(): string | undefined {\n    return this.lastBuildRunId\n  }\n  get builderRunIdValue(): string | undefined {\n    return this.builderRunId\n  }\n}\n\n/**\n * Reconstruct the most recent BuilderSession state for a given project —\n * returns { builderRunId, lastBuildRunId, chatRuns }. For chat-first UIs\n * this is how a resumed session finds its place in the edit history.\n */\nexport async function resumeBuilderSession(\n  store: TraceStore,\n  projectId: string,\n): Promise<{\n  projectId: string\n  chatRuns: Run[]\n  lastBuilderRun?: Run\n  lastBuildRun?: Run\n  lastAppRuntimeRuns: Run[]\n}> {\n  const runs = await store.listRuns({ projectId })\n  const chatRuns = runs\n    .filter((r) => r.layer === 'builder')\n    .sort((a, b) => b.startedAt - a.startedAt)\n  const buildRuns = runs\n    .filter((r) => r.layer === 'app-build')\n    .sort((a, b) => b.startedAt - a.startedAt)\n  const appRuntimeRuns = runs\n    .filter((r) => r.layer === 'app-runtime')\n    .sort((a, b) => b.startedAt - a.startedAt)\n  return {\n    projectId,\n    chatRuns,\n    lastBuilderRun: chatRuns[0],\n    lastBuildRun: buildRuns[0],\n    lastAppRuntimeRuns: appRuntimeRuns,\n  }\n}\n\nfunction cryptoId(): 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 * Meta-eval correlation — the highest-leverage signal in the framework.\n *\n * Given a corpus of three-layer project reports, compute how well each\n * pair of layers correlates. The question we care about most:\n *\n *   Does `metaScore` (what the builder thinks it did) predict\n *   `runtimeScore` (what the user actually gets)?\n *\n * If r < ~0.4, the builder's self-scoring is broken — it's optimizing\n * for something other than real-world success. If r > 0.7, meta_score\n * is a usable proxy and can drive CI gates cheaply.\n *\n * Non-parametric rank correlation (Spearman) is also reported because\n * meta scores are often ordinal-ish.\n */\n\nimport { pearsonR, spearmanR } from '../statistics'\nimport type { ThreeLayerProjectReport } from './three-layer-eval'\n\nexport interface LayerCorrelation {\n  n: number\n  pearson: number\n  spearman: number\n}\n\nexport interface CorrelationReport {\n  /** Pairs present in the corpus (layers with ≥ 2 matched data points). */\n  metaVsBuild?: LayerCorrelation\n  metaVsRuntime?: LayerCorrelation\n  buildVsRuntime?: LayerCorrelation\n  /** Number of complete projects (all 3 scores present). */\n  completeProjects: number\n}\n\nexport function correlateLayers(reports: ThreeLayerProjectReport[]): CorrelationReport {\n  const completeProjects = reports.filter((r) => r.complete).length\n  return {\n    metaVsBuild: pairwise(\n      reports,\n      (r) => r.metaScore,\n      (r) => r.buildScore,\n    ),\n    metaVsRuntime: pairwise(\n      reports,\n      (r) => r.metaScore,\n      (r) => r.runtimeScore,\n    ),\n    buildVsRuntime: pairwise(\n      reports,\n      (r) => r.buildScore,\n      (r) => r.runtimeScore,\n    ),\n    completeProjects,\n  }\n}\n\nfunction pairwise(\n  reports: ThreeLayerProjectReport[],\n  a: (r: ThreeLayerProjectReport) => number | null,\n  b: (r: ThreeLayerProjectReport) => number | null,\n): LayerCorrelation | undefined {\n  const xs: number[] = []\n  const ys: number[] = []\n  for (const r of reports) {\n    const x = a(r)\n    const y = b(r)\n    if (x !== null && y !== null && Number.isFinite(x) && Number.isFinite(y)) {\n      xs.push(x)\n      ys.push(y)\n    }\n  }\n  if (xs.length < 2) return undefined\n  return {\n    n: xs.length,\n    pearson: pearsonR(xs, ys),\n    spearman: spearmanR(xs, ys),\n  }\n}\n","/**\n * ProjectRegistry — project-level aggregation over the trace corpus.\n *\n * Thin reader over TraceStore that answers the questions a chat-first,\n * resumable UI needs:\n *   - listProjects() → project IDs with latest activity\n *   - projectTimeline(id) → chats + builds + runtime runs, chronological\n *   - projectChats(id) → chat-level summaries (turn count, outcome)\n *\n * All queries are pure reads; no state duplication.\n */\n\nimport type { Run } from '../trace/schema'\nimport type { TraceStore } from '../trace/store'\n\nexport interface ProjectSummary {\n  projectId: string\n  chatCount: number\n  buildCount: number\n  appRuntimeCount: number\n  lastActivityAt: number\n  latestChatId?: string\n  latestOutcome?: { pass: boolean; score?: number }\n}\n\nexport interface ChatSummary {\n  chatId: string\n  projectId: string\n  builderRunId: string\n  startedAt: number\n  endedAt?: number\n  status: Run['status']\n  outcome?: Run['outcome']\n  /** Counts of spans emitted during the chat. */\n  llmTurns?: number\n  toolCalls?: number\n  buildRunId?: string\n  appRuntimeRunIds: string[]\n}\n\nexport interface ProjectTimelineEntry {\n  run: Run\n  layerBucket: 'chat' | 'build' | 'runtime' | 'other'\n}\n\nexport class ProjectRegistry {\n  constructor(private store: TraceStore) {}\n\n  async listProjects(): Promise<ProjectSummary[]> {\n    const runs = await this.store.listRuns()\n    const byProject = new Map<string, Run[]>()\n    for (const r of runs) {\n      if (!r.projectId) continue\n      const arr = byProject.get(r.projectId) ?? []\n      arr.push(r)\n      byProject.set(r.projectId, arr)\n    }\n    const summaries: ProjectSummary[] = []\n    for (const [projectId, projectRuns] of byProject) {\n      const sorted = projectRuns.slice().sort((a, b) => b.startedAt - a.startedAt)\n      const chats = projectRuns.filter((r) => r.layer === 'builder')\n      const builds = projectRuns.filter((r) => r.layer === 'app-build')\n      const runtimes = projectRuns.filter((r) => r.layer === 'app-runtime')\n      const latest = sorted[0]\n      if (!latest) continue\n      summaries.push({\n        projectId,\n        chatCount: chats.length,\n        buildCount: builds.length,\n        appRuntimeCount: runtimes.length,\n        lastActivityAt: latest.startedAt,\n        latestChatId: chats[0]?.chatId,\n        latestOutcome: latest.outcome\n          ? { pass: latest.outcome.pass ?? false, score: latest.outcome.score }\n          : undefined,\n      })\n    }\n    return summaries.sort((a, b) => b.lastActivityAt - a.lastActivityAt)\n  }\n\n  async projectTimeline(projectId: string): Promise<ProjectTimelineEntry[]> {\n    const runs = await this.store.listRuns({ projectId })\n    const ordered = runs.slice().sort((a, b) => a.startedAt - b.startedAt)\n    return ordered.map((run) => ({\n      run,\n      layerBucket:\n        run.layer === 'builder'\n          ? 'chat'\n          : run.layer === 'app-build'\n            ? 'build'\n            : run.layer === 'app-runtime'\n              ? 'runtime'\n              : 'other',\n    }))\n  }\n\n  async projectChats(projectId: string): Promise<ChatSummary[]> {\n    const builderRuns = (await this.store.listRuns({ projectId, layer: 'builder' })).sort(\n      (a, b) => b.startedAt - a.startedAt,\n    )\n    const childrenFor = async (runId: string) => this.store.listRuns({ parentRunId: runId })\n    const out: ChatSummary[] = []\n    for (const run of builderRuns) {\n      const spans = await this.store.spans({ runId: run.runId })\n      const children = await childrenFor(run.runId)\n      const build = children.find((c) => c.layer === 'app-build')\n      const runtime: string[] = []\n      // Runtime runs can be grandchildren (attached to the build) or siblings\n      // when shipped skipped.\n      if (build) {\n        const grands = await childrenFor(build.runId)\n        for (const g of grands) if (g.layer === 'app-runtime') runtime.push(g.runId)\n      }\n      for (const c of children) if (c.layer === 'app-runtime') runtime.push(c.runId)\n      out.push({\n        chatId: run.chatId ?? run.runId,\n        projectId,\n        builderRunId: run.runId,\n        startedAt: run.startedAt,\n        endedAt: run.endedAt,\n        status: run.status,\n        outcome: run.outcome,\n        llmTurns: spans.filter((s) => s.kind === 'llm').length,\n        toolCalls: spans.filter((s) => s.kind === 'tool').length,\n        buildRunId: build?.runId,\n        appRuntimeRunIds: runtime,\n      })\n    }\n    return out\n  }\n}\n","/**\n * Three-layer evaluation — the canonical scoring breakdown for\n * builder-of-builders workflows.\n *\n *   meta_score:    did the builder understand + satisfy user intent?\n *                  (judge verdict attached to the builder run)\n *   build_score:   did the generated scaffold build + pass its own tests?\n *                  (outcome.score on the app-build child run)\n *   runtime_score: did the generated agent pass its domain scenarios?\n *                  (mean outcome.score over app-runtime grandchild runs)\n *\n * Returns a structured report per project. The cross-layer correlation\n * is the highest-leverage signal the framework computes — if\n * meta_score doesn't predict runtime_score, the builder's self-scoring\n * is broken.\n *\n * Scaffold-only mode: when a project has no `app-runtime` runs (e.g. a\n * scaffold-builder eval that grades compose + build without driving a\n * runtime scenario), `kind` is `'scaffold-only'` and `complete` measures\n * meta + build only. Consumers can tell the two apart without having to\n * interpret null-runtime as either \"not yet computed\" or \"N/A for this\n * project shape\".\n */\n\nimport { judgeSpans } from '../trace/query'\nimport type { Run } from '../trace/schema'\nimport type { TraceStore } from '../trace/store'\n\nexport type ProjectKind = 'full' | 'scaffold-only'\n\nexport interface ThreeLayerProjectReport {\n  projectId: string\n  /**\n   * `'full'` when the project has at least one `app-runtime` run;\n   * `'scaffold-only'` when it only has meta + build layers. Lets\n   * downstream consumers treat a null runtime score as expected\n   * (scaffold-only) vs. missing (full, pipeline broke).\n   */\n  kind: ProjectKind\n  builderRunId?: string\n  /** Judge-verdict score on the builder run (0..1 after normalization). */\n  metaScore: number | null\n  buildRunId?: string\n  /** 0..1 from the sandbox harness (testsPassed / testsTotal). */\n  buildScore: number | null\n  appRuntimeRunIds: string[]\n  /** Mean of outcome.score over app-runtime runs, 0..1. Always null in scaffold-only mode. */\n  runtimeScore: number | null\n  runtimePassRate: number | null\n  /**\n   * Layer-aware completeness:\n   *   - `kind='full'`: all three layers scored\n   *   - `kind='scaffold-only'`: meta + build scored (runtime not applicable)\n   */\n  complete: boolean\n}\n\nexport async function scoreProject(\n  store: TraceStore,\n  projectId: string,\n): Promise<ThreeLayerProjectReport> {\n  const allRuns = await store.listRuns({ projectId })\n  const builder = latestByLayer(allRuns, 'builder')\n  const build = latestByLayer(allRuns, 'app-build')\n  const runtime = allRuns.filter((r) => r.layer === 'app-runtime')\n\n  const metaScore = builder ? await extractMetaScore(store, builder.runId) : null\n  const buildScore = build?.outcome?.score ?? null\n  const runtimeScores = runtime\n    .map((r) => r.outcome?.score)\n    .filter((s): s is number => typeof s === 'number')\n  const runtimeScore =\n    runtimeScores.length > 0\n      ? runtimeScores.reduce((a, b) => a + b, 0) / runtimeScores.length\n      : null\n  const runtimePassed = runtime.filter((r) => r.outcome?.pass === true).length\n  const runtimePassRate = runtime.length > 0 ? runtimePassed / runtime.length : null\n\n  const kind: ProjectKind = runtime.length === 0 ? 'scaffold-only' : 'full'\n  const complete =\n    kind === 'scaffold-only'\n      ? metaScore !== null && buildScore !== null\n      : metaScore !== null && buildScore !== null && runtimeScore !== null\n\n  return {\n    projectId,\n    kind,\n    builderRunId: builder?.runId,\n    metaScore,\n    buildRunId: build?.runId,\n    buildScore,\n    appRuntimeRunIds: runtime.map((r) => r.runId),\n    runtimeScore,\n    runtimePassRate,\n    complete,\n  }\n}\n\n/** Aggregate scoring across every project in a corpus. */\nexport async function scoreAllProjects(store: TraceStore): Promise<ThreeLayerProjectReport[]> {\n  const runs = await store.listRuns()\n  const projectIds = [...new Set(runs.map((r) => r.projectId).filter((p): p is string => !!p))]\n  return Promise.all(projectIds.map((p) => scoreProject(store, p)))\n}\n\nfunction latestByLayer(runs: Run[], layer: Run['layer']): Run | undefined {\n  const filtered = runs.filter((r) => r.layer === layer).sort((a, b) => b.startedAt - a.startedAt)\n  return filtered[0]\n}\n\nasync function extractMetaScore(store: TraceStore, builderRunId: string): Promise<number | null> {\n  const js = await judgeSpans(store, builderRunId)\n  const meta = js.find(\n    (s) => s.judgeId === 'builder-meta' && s.dimension === 'user_intent_satisfaction',\n  )\n  if (!meta) return null\n  // Normalize score to 0..1. Accept 0-1 natively; 0-10 scale is also common.\n  if (meta.score >= 0 && meta.score <= 1) return meta.score\n  if (meta.score >= 0 && meta.score <= 10) return meta.score / 10\n  return null\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AA8CO,IAAM,iBAAN,MAAqB;AAAA,EAClB;AAAA,EACA;AAAA,EACC;AAAA,EACA;AAAA,EACD;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,OAAmB,MAA0B,QAAwB;AAC/E,SAAK,QAAQ;AACb,SAAK,YAAY,KAAK;AACtB,SAAK,SAAS,KAAK,UAAU,SAAS;AACtC,SAAK,gBAAgB;AACrB,SAAK,iBAAiB,IAAI,aAAa,KAAK;AAAA,EAC9C;AAAA;AAAA,EAGA,MAAM,UAAU,aAAa,GAAG,KAAK,SAAS,SAA0B;AACtE,UAAM,KAAK,eAAe,SAAS;AAAA,MACjC;AAAA,MACA,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,OAAO;AAAA,IACT,CAAC;AACD,SAAK,eAAe,KAAK,eAAe;AACxC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,UAAwB;AAC1B,QAAI,CAAC,KAAK,aAAc,OAAM,IAAI,MAAM,gDAAgD;AACxF,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,KAAK,SAAgF;AACzF,QAAI,CAAC,KAAK,aAAc,OAAM,IAAI,MAAM,6CAA6C;AACrF,UAAM,eAAe,IAAI,aAAa,KAAK,KAAK;AAChD,UAAM,aAAa,SAAS;AAAA,MAC1B,YAAY,QAAQ,cAAc,GAAG,KAAK,SAAS;AAAA,MACnD,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,aAAa,KAAK;AAAA,MAClB,OAAO;AAAA,IACT,CAAC;AACD,UAAM,UAAU,IAAI,eAAe,QAAQ,UAAU,KAAK,aAAa;AACvE,UAAM,SAAS,MAAM,QAAQ,IAAI,QAAQ,SAAS,YAAY;AAC9D,UAAM,aAAa,OAAO;AAAA,MACxB,MAAM,OAAO;AAAA,MACb,OAAO,OAAO;AAAA,MACd,cAAc,OAAO,SAAS,YAAY;AAAA,IAC5C,CAAC;AACD,SAAK,iBAAiB,aAAa;AACnC,WAAO,EAAE,OAAO,aAAa,OAAO,OAAO;AAAA,EAC7C;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,MAAM,eAAe,SAA8D;AACjF,UAAM,cAAc,KAAK,kBAAkB,KAAK;AAChD,QAAI,CAAC;AACH,YAAM,IAAI,MAAM,gEAAgE;AAClF,UAAM,EAAE,UAAU,OAAO,IAAI;AAC7B,UAAM,SAAS,MAAM,sBAAsB,UAAU,KAAK,OAAO;AAAA,MAC/D,QAAQ,UAAU,KAAK;AAAA,MACvB,YAAY,EAAE,SAAS,QAAW,WAAW,QAAW,kBAAkB,OAAU;AAAA,IACtF,CAAC;AAED,UAAM,KAAK,MAAM,UAAU,OAAO,OAAO;AAAA,MACvC;AAAA,MACA,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,OAAO;AAAA,IACT,CAAC;AACD,WAAO;AAAA,EACT;AAAA;AAAA;AAAA,EAIA,MAAM,gBAAgB,OAAe,WAAmC;AACtE,QAAI,CAAC,KAAK;AACR,YAAM,IAAI,MAAM,wDAAwD;AAC1E,UAAM,KAAK,eAAe,YAAY;AAAA,MACpC,SAAS;AAAA,MACT,cAAc,KAAK;AAAA;AAAA,MACnB,WAAW;AAAA,MACX;AAAA,MACA;AAAA,MACA,MAAM;AAAA,IACR,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,MAAM,QAAQ,SAA2E;AACvF,UAAM,KAAK,eAAe,OAAO;AAAA,MAC/B,MAAM,QAAQ;AAAA,MACd,OAAO,QAAQ;AAAA,MACf,OAAO,QAAQ;AAAA,IACjB,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAM,gBAAgB,YAA2C;AAC/D,UAAM,cAAc,KAAK,kBAAkB,KAAK;AAChD,QAAI,CAAC;AACH,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AACF,UAAM,UAAU,IAAI,aAAa,KAAK,KAAK;AAC3C,UAAM,QAAQ,SAAS;AAAA,MACrB;AAAA,MACA,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb;AAAA,MACA,OAAO;AAAA,IACT,CAAC;AACD,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,iBAAiB,MAKH;AAClB,QAAI,CAAC,KAAK;AACR,YAAM,IAAI,MAAM,yDAAyD;AAC3E,UAAM,UAAU,IAAI,aAAa,KAAK,KAAK;AAC3C,UAAM,QAAQ,SAAS;AAAA,MACrB,YAAY,KAAK,cAAc,GAAG,KAAK,SAAS;AAAA,MAChD,WAAW,KAAK;AAAA,MAChB,QAAQ,KAAK;AAAA,MACb,aAAa,KAAK;AAAA,MAClB,OAAO;AAAA,IACT,CAAC;AACD,UAAM,QAAQ,OAAO;AAAA,MACnB,MAAM,KAAK;AAAA,MACX,OAAO,KAAK;AAAA,MACZ,cAAc,KAAK,OAAO,YAAY;AAAA,MACtC,OAAO,KAAK;AAAA,IACd,CAAC;AACD,SAAK,iBAAiB,QAAQ;AAC9B,WAAO,QAAQ;AAAA,EACjB;AAAA,EAEA,IAAI,sBAA0C;AAC5C,WAAO,KAAK;AAAA,EACd;AAAA,EACA,IAAI,oBAAwC;AAC1C,WAAO,KAAK;AAAA,EACd;AACF;AAOA,eAAsB,qBACpB,OACA,WAOC;AACD,QAAM,OAAO,MAAM,MAAM,SAAS,EAAE,UAAU,CAAC;AAC/C,QAAM,WAAW,KACd,OAAO,CAAC,MAAM,EAAE,UAAU,SAAS,EACnC,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3C,QAAM,YAAY,KACf,OAAO,CAAC,MAAM,EAAE,UAAU,WAAW,EACrC,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3C,QAAM,iBAAiB,KACpB,OAAO,CAAC,MAAM,EAAE,UAAU,aAAa,EACvC,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3C,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA,gBAAgB,SAAS,CAAC;AAAA,IAC1B,cAAc,UAAU,CAAC;AAAA,IACzB,oBAAoB;AAAA,EACtB;AACF;AAEA,SAAS,WAAmB;AAC1B,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;;;AC5NO,SAAS,gBAAgB,SAAuD;AACrF,QAAM,mBAAmB,QAAQ,OAAO,CAAC,MAAM,EAAE,QAAQ,EAAE;AAC3D,SAAO;AAAA,IACL,aAAa;AAAA,MACX;AAAA,MACA,CAAC,MAAM,EAAE;AAAA,MACT,CAAC,MAAM,EAAE;AAAA,IACX;AAAA,IACA,eAAe;AAAA,MACb;AAAA,MACA,CAAC,MAAM,EAAE;AAAA,MACT,CAAC,MAAM,EAAE;AAAA,IACX;AAAA,IACA,gBAAgB;AAAA,MACd;AAAA,MACA,CAAC,MAAM,EAAE;AAAA,MACT,CAAC,MAAM,EAAE;AAAA,IACX;AAAA,IACA;AAAA,EACF;AACF;AAEA,SAAS,SACP,SACA,GACA,GAC8B;AAC9B,QAAM,KAAe,CAAC;AACtB,QAAM,KAAe,CAAC;AACtB,aAAW,KAAK,SAAS;AACvB,UAAM,IAAI,EAAE,CAAC;AACb,UAAM,IAAI,EAAE,CAAC;AACb,QAAI,MAAM,QAAQ,MAAM,QAAQ,OAAO,SAAS,CAAC,KAAK,OAAO,SAAS,CAAC,GAAG;AACxE,SAAG,KAAK,CAAC;AACT,SAAG,KAAK,CAAC;AAAA,IACX;AAAA,EACF;AACA,MAAI,GAAG,SAAS,EAAG,QAAO;AAC1B,SAAO;AAAA,IACL,GAAG,GAAG;AAAA,IACN,SAAS,SAAS,IAAI,EAAE;AAAA,IACxB,UAAU,UAAU,IAAI,EAAE;AAAA,EAC5B;AACF;;;ACjCO,IAAM,kBAAN,MAAsB;AAAA,EAC3B,YAAoB,OAAmB;AAAnB;AAAA,EAAoB;AAAA,EAApB;AAAA,EAEpB,MAAM,eAA0C;AAC9C,UAAM,OAAO,MAAM,KAAK,MAAM,SAAS;AACvC,UAAM,YAAY,oBAAI,IAAmB;AACzC,eAAW,KAAK,MAAM;AACpB,UAAI,CAAC,EAAE,UAAW;AAClB,YAAM,MAAM,UAAU,IAAI,EAAE,SAAS,KAAK,CAAC;AAC3C,UAAI,KAAK,CAAC;AACV,gBAAU,IAAI,EAAE,WAAW,GAAG;AAAA,IAChC;AACA,UAAM,YAA8B,CAAC;AACrC,eAAW,CAAC,WAAW,WAAW,KAAK,WAAW;AAChD,YAAM,SAAS,YAAY,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC3E,YAAM,QAAQ,YAAY,OAAO,CAAC,MAAM,EAAE,UAAU,SAAS;AAC7D,YAAM,SAAS,YAAY,OAAO,CAAC,MAAM,EAAE,UAAU,WAAW;AAChE,YAAM,WAAW,YAAY,OAAO,CAAC,MAAM,EAAE,UAAU,aAAa;AACpE,YAAM,SAAS,OAAO,CAAC;AACvB,UAAI,CAAC,OAAQ;AACb,gBAAU,KAAK;AAAA,QACb;AAAA,QACA,WAAW,MAAM;AAAA,QACjB,YAAY,OAAO;AAAA,QACnB,iBAAiB,SAAS;AAAA,QAC1B,gBAAgB,OAAO;AAAA,QACvB,cAAc,MAAM,CAAC,GAAG;AAAA,QACxB,eAAe,OAAO,UAClB,EAAE,MAAM,OAAO,QAAQ,QAAQ,OAAO,OAAO,OAAO,QAAQ,MAAM,IAClE;AAAA,MACN,CAAC;AAAA,IACH;AACA,WAAO,UAAU,KAAK,CAAC,GAAG,MAAM,EAAE,iBAAiB,EAAE,cAAc;AAAA,EACrE;AAAA,EAEA,MAAM,gBAAgB,WAAoD;AACxE,UAAM,OAAO,MAAM,KAAK,MAAM,SAAS,EAAE,UAAU,CAAC;AACpD,UAAM,UAAU,KAAK,MAAM,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AACrE,WAAO,QAAQ,IAAI,CAAC,SAAS;AAAA,MAC3B;AAAA,MACA,aACE,IAAI,UAAU,YACV,SACA,IAAI,UAAU,cACZ,UACA,IAAI,UAAU,gBACZ,YACA;AAAA,IACZ,EAAE;AAAA,EACJ;AAAA,EAEA,MAAM,aAAa,WAA2C;AAC5D,UAAM,eAAe,MAAM,KAAK,MAAM,SAAS,EAAE,WAAW,OAAO,UAAU,CAAC,GAAG;AAAA,MAC/E,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE;AAAA,IAC5B;AACA,UAAM,cAAc,OAAO,UAAkB,KAAK,MAAM,SAAS,EAAE,aAAa,MAAM,CAAC;AACvF,UAAM,MAAqB,CAAC;AAC5B,eAAW,OAAO,aAAa;AAC7B,YAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,EAAE,OAAO,IAAI,MAAM,CAAC;AACzD,YAAM,WAAW,MAAM,YAAY,IAAI,KAAK;AAC5C,YAAM,QAAQ,SAAS,KAAK,CAAC,MAAM,EAAE,UAAU,WAAW;AAC1D,YAAM,UAAoB,CAAC;AAG3B,UAAI,OAAO;AACT,cAAM,SAAS,MAAM,YAAY,MAAM,KAAK;AAC5C,mBAAW,KAAK,OAAQ,KAAI,EAAE,UAAU,cAAe,SAAQ,KAAK,EAAE,KAAK;AAAA,MAC7E;AACA,iBAAW,KAAK,SAAU,KAAI,EAAE,UAAU,cAAe,SAAQ,KAAK,EAAE,KAAK;AAC7E,UAAI,KAAK;AAAA,QACP,QAAQ,IAAI,UAAU,IAAI;AAAA,QAC1B;AAAA,QACA,cAAc,IAAI;AAAA,QAClB,WAAW,IAAI;AAAA,QACf,SAAS,IAAI;AAAA,QACb,QAAQ,IAAI;AAAA,QACZ,SAAS,IAAI;AAAA,QACb,UAAU,MAAM,OAAO,CAAC,MAAM,EAAE,SAAS,KAAK,EAAE;AAAA,QAChD,WAAW,MAAM,OAAO,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE;AAAA,QAClD,YAAY,OAAO;AAAA,QACnB,kBAAkB;AAAA,MACpB,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AACF;;;ACzEA,eAAsB,aACpB,OACA,WACkC;AAClC,QAAM,UAAU,MAAM,MAAM,SAAS,EAAE,UAAU,CAAC;AAClD,QAAM,UAAU,cAAc,SAAS,SAAS;AAChD,QAAM,QAAQ,cAAc,SAAS,WAAW;AAChD,QAAM,UAAU,QAAQ,OAAO,CAAC,MAAM,EAAE,UAAU,aAAa;AAE/D,QAAM,YAAY,UAAU,MAAM,iBAAiB,OAAO,QAAQ,KAAK,IAAI;AAC3E,QAAM,aAAa,OAAO,SAAS,SAAS;AAC5C,QAAM,gBAAgB,QACnB,IAAI,CAAC,MAAM,EAAE,SAAS,KAAK,EAC3B,OAAO,CAAC,MAAmB,OAAO,MAAM,QAAQ;AACnD,QAAM,eACJ,cAAc,SAAS,IACnB,cAAc,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI,cAAc,SACzD;AACN,QAAM,gBAAgB,QAAQ,OAAO,CAAC,MAAM,EAAE,SAAS,SAAS,IAAI,EAAE;AACtE,QAAM,kBAAkB,QAAQ,SAAS,IAAI,gBAAgB,QAAQ,SAAS;AAE9E,QAAM,OAAoB,QAAQ,WAAW,IAAI,kBAAkB;AACnE,QAAM,WACJ,SAAS,kBACL,cAAc,QAAQ,eAAe,OACrC,cAAc,QAAQ,eAAe,QAAQ,iBAAiB;AAEpE,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA,cAAc,SAAS;AAAA,IACvB;AAAA,IACA,YAAY,OAAO;AAAA,IACnB;AAAA,IACA,kBAAkB,QAAQ,IAAI,CAAC,MAAM,EAAE,KAAK;AAAA,IAC5C;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAGA,eAAsB,iBAAiB,OAAuD;AAC5F,QAAM,OAAO,MAAM,MAAM,SAAS;AAClC,QAAM,aAAa,CAAC,GAAG,IAAI,IAAI,KAAK,IAAI,CAAC,MAAM,EAAE,SAAS,EAAE,OAAO,CAAC,MAAmB,CAAC,CAAC,CAAC,CAAC,CAAC;AAC5F,SAAO,QAAQ,IAAI,WAAW,IAAI,CAAC,MAAM,aAAa,OAAO,CAAC,CAAC,CAAC;AAClE;AAEA,SAAS,cAAc,MAAa,OAAsC;AACxE,QAAM,WAAW,KAAK,OAAO,CAAC,MAAM,EAAE,UAAU,KAAK,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAC/F,SAAO,SAAS,CAAC;AACnB;AAEA,eAAe,iBAAiB,OAAmB,cAA8C;AAC/F,QAAM,KAAK,MAAM,WAAW,OAAO,YAAY;AAC/C,QAAM,OAAO,GAAG;AAAA,IACd,CAAC,MAAM,EAAE,YAAY,kBAAkB,EAAE,cAAc;AAAA,EACzD;AACA,MAAI,CAAC,KAAM,QAAO;AAElB,MAAI,KAAK,SAAS,KAAK,KAAK,SAAS,EAAG,QAAO,KAAK;AACpD,MAAI,KAAK,SAAS,KAAK,KAAK,SAAS,GAAI,QAAO,KAAK,QAAQ;AAC7D,SAAO;AACT;","names":[]}

package/dist/{calibration-BPmzuVPk.d.ts → calibration-Dz8TQV4y.d.ts}

@@ -1,5 +1,5 @@
-import { T as TraceStore } from './store-BcFXE6LG.js';
-import { R as Run } from './schema-m0gsnbt3.js';
+import { T as TraceStore } from './store-BsVi7ncX.js';
+import { R as Run } from './schema-SGWcK9wa.js';
 import { O as OutcomeFilter, b as OutcomeStore } from './outcome-store-rnXLEqSn.js';
 
 /**