@tangle-network/agent-eval 0.116.0 → 0.117.1

package/CHANGELOG.md

@@ -4,6 +4,44 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ---
 
+## [0.117.1] — 2026-07-13 — retry-safe code-candidate cleanup
+
+### Fixed
+
+- `gitWorktreeAdapter().discard()` now reconciles worktree and branch removal independently.
+  Repeated cleanup is safe, partial cleanup can be retried, and a Git command that reports an error after completing its mutation no longer strands candidate branches or worktrees.
+
+## [0.117.0] — 2026-07-13 — durable cost and bounded behavioral evidence
+
+### Added
+
+- `createReferenceEquivalenceJudge()` and `runReferenceEquivalenceJudge()` score whether an answer preserves the meaning of one or more references, with the same cost and transport accounting as other judges.
+
+### Changed
+
+- `CostLedger.runPaidCall()` is now the single paid-call path across campaigns, proposers, judges, analysts, and distillation.
+  It durably reserves maximum spend before dispatch, records provider receipts, blocks unresolved crash state, and enforces the run ceiling before another paid call starts.
+- `ToolSpan.argsCaptured` distinguishes a call with unavailable arguments from a captured no-argument call.
+  Repeated-call analysis, failure clustering, tool-use metrics, and per-step redundancy grading no longer compare uncaptured arguments.
+  Every OTLP export path uses one mapping that preserves this distinction.
+
+### Breaking
+
+- `CostLedger.record()` is removed because recording spend after a provider call cannot enforce a cost limit or survive a crash.
+  Use `CostLedger.runPaidCall()` for billable work, `CostLedger` receipt import for already-settled calls, or `costForUsage()` for pure estimates.
+- `computeTraceMetrics()` now rejects mixed-trace input, and `BehavioralMetrics` adds required `traceId` and `tokenSequences` fields.
+  The convenience token trajectories now expose the longest proven-serial sequence instead of flattening parallel branches.
+- `ToolUseMetrics` and `ToolStats` add required `callsWithCapturedArgs` fields.
+  `duplicateRate` now uses captured-argument calls as its denominator.
+
+### Fixed
+
+- Repeated-call findings now require a contiguous, time-bounded, serial episode within one agent branch instead of grouping identical or concurrent calls across an entire run.
+- Behavioral token findings now analyze each trace and serial agent timeline independently, use numeric time ordering across accepted timestamp formats, and only attribute output decay to context that actually grew.
+- Behavioral issue IDs remain stable across trace runs while evidence retains exact trace identities and sampled prevalence.
+- Partial timing isolates only the uncertain interval, and same-named root spans retain independent structural identity.
+- Multi-trace behavioral findings use pattern-level claims while each trace's exact values remain in its evidence reference.
+
 ## [0.116.0] — 2026-07-12 — evidence-linked AgentProfile optimization
 
 ### Added

package/dist/analyst/index.d.ts

@@ -1,21 +1,24 @@
 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-CKjePUMh.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 FINDING_SUBJECT_SYNTAX, g as FindingSubject, h as FindingSubjectKind, i as FindingSubjectStringSchema, j as FindingsDiff, k as FindingsStore, I as IMPROVEMENT_KIND_SPEC, K as KIND_EXPECTED_SUBJECTS, l as KNOWLEDGE_GAP_KIND_SPEC, m as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, n as SKILL_USAGE_ANALYST, o as SkillUsageAnalyst, p as SkillUsageRecord, q as SkillUsageReport, r as SkillUsageScanConfig, s as buildSkillUsageReport, t as createAnalystAi, u as defaultIsMaterial, v as diffFindings, w as emitSkillUsageFindings, x as findingSubjectGrammarPromptFor, y as parseFindingSubject, z as renderFindingSubject } from '../semantic-concept-judge-CKjePUMh.js';
-import { b as JudgeFn, a as JudgeInput } from '../types-C7DGg5ex.js';
-import { A as Analyst, i as AnalystSeverity, d as AnalystFinding, L as LlmClientOptions } from '../policy-edit-Clb2v6Oa.js';
-export { b as AnalystContext, h as AnalystCost, j as AnalystInputKind, k as AnalystRequirements, g as AnalystRunEvent, f as AnalystRunInputs, e 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, D as DirectProviderTransportOpts, E as EvidenceRef, F as FindingToPolicyEditOptions, M as MockTransportOpts, r as POLICY_EDIT_AXES, s as POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, t as POLICY_EDIT_TARGET_SURFACES, u as PolicyEdit, v as PolicyEditAdmission, w as PolicyEditAdmissionOptions, x as PolicyEditAxis, P as PolicyEditCandidateRecord, y as PolicyEditChange, z as PolicyEditExpectedGain, B as PolicyEditGainDirection, G as PolicyEditGainUnit, H as PolicyEditInit, I as PolicyEditRisk, J as PolicyEditSchemaVersion, K as PolicyEditSource, N as PolicyEditTarget, O as PolicyEditTargetSurface, Q as PolicyEditValidationError, R as RouterTransportOpts, S as SandboxSdkTransportOpts, T as admitPolicyEdit, U as applyPolicyEditToSurface, V as computeFindingId, W as computePolicyEditId, X as createChatClient, Y as isPolicyEdit, Z as makeFinding, _ as makePolicyEdit, $ as makePolicyEditCandidateRecord, a0 as policyEditFromFinding, a1 as policyEditsFromFindings, a2 as scorePolicyEditReadiness, a3 as validatePolicyEdit, a4 as validatePolicyEditCandidateRecord } from '../policy-edit-Clb2v6Oa.js';
+import { R as RunCritic, S as SemanticConceptJudgeOptions, a as RunTrace, b as SemanticConceptJudgeInput, B as BehavioralMetrics } from '../semantic-concept-judge-CXnPEJbf.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 FINDING_SUBJECT_SYNTAX, g as FindingSubject, h as FindingSubjectKind, i as FindingSubjectStringSchema, j as FindingsDiff, k as FindingsStore, I as IMPROVEMENT_KIND_SPEC, K as KIND_EXPECTED_SUBJECTS, l as KNOWLEDGE_GAP_KIND_SPEC, m as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, n as SKILL_USAGE_ANALYST, o as SkillUsageAnalyst, p as SkillUsageRecord, q as SkillUsageReport, r as SkillUsageScanConfig, s as buildSkillUsageReport, t as createAnalystAi, u as defaultIsMaterial, v as diffFindings, w as emitSkillUsageFindings, x as findingSubjectGrammarPromptFor, y as parseFindingSubject, z as renderFindingSubject } from '../semantic-concept-judge-CXnPEJbf.js';
+import { b as JudgeFn, a as JudgeInput } from '../types-BkfcQnxV.js';
+import { A as Analyst, h as AnalystSeverity, c as AnalystFinding } from '../policy-edit-wG9uFEFm.js';
+export { a as AnalystContext, g as AnalystCost, i as AnalystInputKind, j as AnalystRequirements, f as AnalystRunEvent, e as AnalystRunInputs, d as AnalystRunResult, b 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, F as FindingToPolicyEditOptions, M as MockTransportOpts, q as POLICY_EDIT_AXES, r as POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, s as POLICY_EDIT_TARGET_SURFACES, t as PolicyEdit, u as PolicyEditAdmission, v as PolicyEditAdmissionOptions, w as PolicyEditAxis, P as PolicyEditCandidateRecord, x as PolicyEditChange, y as PolicyEditExpectedGain, z as PolicyEditGainDirection, B as PolicyEditGainUnit, G as PolicyEditInit, H as PolicyEditRisk, I as PolicyEditSchemaVersion, J as PolicyEditSource, K as PolicyEditTarget, L as PolicyEditTargetSurface, N as PolicyEditValidationError, R as RouterTransportOpts, S as SandboxSdkTransportOpts, O as admitPolicyEdit, Q as applyPolicyEditToSurface, T as computeFindingId, U as computePolicyEditId, V as createChatClient, W as isPolicyEdit, X as makeFinding, Y as makePolicyEdit, Z as makePolicyEditCandidateRecord, _ as policyEditFromFinding, $ as policyEditsFromFindings, a0 as scorePolicyEditReadiness, a1 as validatePolicyEdit, a2 as validatePolicyEditCandidateRecord } from '../policy-edit-wG9uFEFm.js';
 import { TCloud } from '@tangle-network/tcloud';
-import { T as TraceAnalysisStore } from '../store-9cAScOcb.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-DDfv22MQ.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-DWOvXjR_.js';
+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-DaK8b3fv.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-ClZmO25A.js';
+import { C as CostLedger } from '../cost-ledger-DWy3XdJc.js';
+import { a as LlmClientOptions } from '../llm-client-qoDd18Qz.js';
 import { AxFunction } from '@ax-llm/ax';
 import '../verdict-C9MlYujm.js';
 import 'zod';
-import '../schema-SGWcK9wa.js';
-import '../store-BsVi7ncX.js';
-import '../run-record-CZmcpWPo.js';
+import '../schema-B3Q3l9Z_.js';
+import '../store-DGqD0Pyo.js';
+import '../run-record-BDH49H2E.js';
 import '@tangle-network/agent-interface';
 import '../errors-oeQrLqXC.js';
+import '../raw-provider-sink-C46HDghv.js';
 
 /**
  * Adapter factories — lift each existing agent-eval primitive into the
@@ -186,6 +189,10 @@ interface StructureFindingsOptions {
     model: string;
     baseUrl: string;
     apiKey?: string;
+    /** Optional ledger for direct use. */
+    costLedger?: CostLedger;
+    costPhase?: string;
+    maxTokens?: number;
     /** Max reask attempts after a zero/invalid extraction. Default 1. */
     maxReasks?: number;
     /** Test seam: inject a fetch (no network in unit tests). */

package/dist/analyst/index.js

@@ -6,18 +6,19 @@ import {
   SkillUsageAnalyst,
   buildSkillUsageReport,
   createAnalystAi,
-  createChatClient,
   defaultIsMaterial,
   diffFindings,
   emitSkillUsageFindings,
   runSemanticConceptJudge
-} from "../chunk-NBSS5NDZ.js";
+} from "../chunk-CCZIVI3F.js";
 import {
   behavioralAnalyst,
   buildDefaultAnalystRegistry,
+  createChatClient,
   deriveEfficiencyFindings
-} from "../chunk-7GKEAIAD.js";
-import "../chunk-MPHTT5HE.js";
+} from "../chunk-VF3XSYTI.js";
+import "../chunk-HHWE3POT.js";
+import "../chunk-3YYRZDON.js";
 import {
   ANALYST_SEVERITIES,
   AnalystRegistry,
@@ -64,11 +65,13 @@ import {
   structureFindings,
   validatePolicyEdit,
   validatePolicyEditCandidateRecord
-} from "../chunk-CIUOICJT.js";
-import "../chunk-LNQEP766.js";
+} from "../chunk-MGEHEHSN.js";
+import "../chunk-S2F4J57L.js";
 import "../chunk-XJYR7XFV.js";
 import "../chunk-VSMTAMNK.js";
-import "../chunk-GY4SYVPJ.js";
+import "../chunk-NJC7U437.js";
+import "../chunk-VCTY3W6J.js";
+import "../chunk-VI2UW6B6.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/{analyst-CFBc14Wc.d.ts → analyst-C8HHvfJp.d.ts}

@@ -1,5 +1,5 @@
 import { AxAIService } from '@ax-llm/ax';
-import { T as TraceAnalysisStore } from './store-9cAScOcb.js';
+import { T as TraceAnalysisStore } from './store-C1YxJDEK.js';
 
 interface AnalyzeTracesInput {
     /** The user-facing question. Domain framing belongs here, not in the

package/dist/{analyze-runs-0rz_m29H.d.ts → analyze-runs--2x39HZ7.d.ts}

@@ -1,7 +1,7 @@
-import { A as AnalystRegistry } from './default-registry-DDfv22MQ.js';
+import { A as AnalystRegistry } from './default-registry-DaK8b3fv.js';
 import { D as DatasetScenario } from './dataset-NENEzRgk.js';
-import { R as RunRecord } from './run-record-CZmcpWPo.js';
-import { I as InsightReport } from './insight-report-oMVxDTxl.js';
+import { R as RunRecord } from './run-record-BDH49H2E.js';
+import { I as InsightReport } from './insight-report-DY4nDW9Q.js';
 
 /**
  * # `analyzeRuns()` — turn a set of agent runs into an actionable decision packet.

package/dist/{baseline-DsNteOgR.d.ts → baseline-DKq3gJpP.d.ts}

@@ -1,5 +1,5 @@
-import { T as TraceStore } from './store-BsVi7ncX.js';
-import { S as Span, e as TraceEvent } from './schema-SGWcK9wa.js';
+import { T as TraceStore } from './store-DGqD0Pyo.js';
+import { S as Span, e as TraceEvent } from './schema-B3Q3l9Z_.js';
 
 /**
  * Tool-use metrics — derived purely from trace data.
@@ -13,9 +13,11 @@ import { S as Span, e as TraceEvent } from './schema-SGWcK9wa.js';
 interface ToolUseMetrics {
     runId: string;
     totalCalls: number;
+    /** Calls whose arguments were captured and can be compared for duplication. */
+    callsWithCapturedArgs: number;
     byTool: Record<string, ToolStats>;
     errorRate: number;
-    /** Ratio of calls with identical (toolName, argHash) already seen earlier in the same run. */
+    /** Ratio of captured-argument calls already seen with the same tool name and arguments. */
     duplicateRate: number;
     /** Ratio of error calls followed by ≥1 retry on same tool. */
     retryRate: number;
@@ -24,6 +26,7 @@ interface ToolUseMetrics {
 }
 interface ToolStats {
     calls: number;
+    callsWithCapturedArgs: number;
     errors: number;
     avgLatencyMs: number;
     duplicates: number;

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

@@ -1,10 +1,10 @@
-import { c as CalibrationReport } from '../calibration-Dz8TQV4y.js';
+import { c as CalibrationReport } from '../calibration-C8MTS7cw.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-CdxteG0y.js';
-import { R as RunRecord, a as RunSplitTag } from '../run-record-CZmcpWPo.js';
-import { T as TraceStore } from '../store-BsVi7ncX.js';
-import { R as RuntimeTrajectoryRecord, P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection } from '../runtime-trajectory-CC0jx9ql.js';
-import '../schema-SGWcK9wa.js';
+import { d as CodeAgentSessionSource, a as CodeAgentSessionIntakeOptions, c as CodeAgentSessionMetrics, C as CodeAgentSessionDiagnostic } from '../code-agent-session-CjZsVd19.js';
+import { R as RunRecord, a as RunSplitTag } from '../run-record-BDH49H2E.js';
+import { T as TraceStore } from '../store-DGqD0Pyo.js';
+import { R as RuntimeTrajectoryRecord, P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection } from '../runtime-trajectory-DGBIUt4B.js';
+import '../schema-B3Q3l9Z_.js';
 import '../outcome-store-rnXLEqSn.js';
 import '@tangle-network/agent-interface';
 import '../errors-oeQrLqXC.js';

package/dist/belief-state/index.js

@@ -11,13 +11,13 @@ import {
 import {
   projectRuntimeTrajectoryEvidence
 } from "../chunk-T4SQEITX.js";
-import "../chunk-VI2UW6B6.js";
 import {
   offPolicyEstimateAll
 } from "../chunk-DTJ6QUQB.js";
 import {
   confidenceInterval
 } from "../chunk-PJQFMIOX.js";
+import "../chunk-VI2UW6B6.js";
 import {
   ValidationError
 } from "../chunk-ONWEPEDO.js";

package/dist/benchmarks/index.d.ts

@@ -1,11 +1,14 @@
-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-DbCXJfZ1.js';
-import '../types-Ca_63YSD.js';
-import '../policy-edit-Clb2v6Oa.js';
-import '../run-record-CZmcpWPo.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-PdX4VnPA.js';
+import '../types-BSw1rOUB.js';
+import '../policy-edit-wG9uFEFm.js';
+import '../run-record-BDH49H2E.js';
 import '@tangle-network/agent-interface';
 import '../errors-oeQrLqXC.js';
-import '../schema-SGWcK9wa.js';
-import '../store-9cAScOcb.js';
-import '../types-C7DGg5ex.js';
+import '../schema-B3Q3l9Z_.js';
+import '../store-C1YxJDEK.js';
+import '../types-BkfcQnxV.js';
+import '../cost-ledger-DWy3XdJc.js';
 import '@tangle-network/tcloud';
-import '../storage-Dw_f7WMt.js';
+import '../llm-client-qoDd18Qz.js';
+import '../raw-provider-sink-C46HDghv.js';
+import '../storage-DrX3v_5B.js';

package/dist/benchmarks/index.js

@@ -16,22 +16,23 @@ import {
   routing_exports,
   runBenchmarkAdapter,
   summarizeBenchmarkCampaign
-} from "../chunk-3LXTCTWL.js";
-import "../chunk-GSW3OBHK.js";
-import "../chunk-3274WNK7.js";
-import "../chunk-VI2UW6B6.js";
-import "../chunk-FAOEFFRT.js";
-import "../chunk-MPHTT5HE.js";
-import "../chunk-CIUOICJT.js";
+} from "../chunk-JSDVRFAP.js";
+import "../chunk-JSJZ4PJ6.js";
+import "../chunk-HQPHZGL6.js";
+import "../chunk-IDZTTFRR.js";
+import "../chunk-3YYRZDON.js";
+import "../chunk-MGEHEHSN.js";
 import "../chunk-ARU2PZFM.js";
 import "../chunk-PJQFMIOX.js";
-import "../chunk-RPDDVKI7.js";
+import "../chunk-4JLWXDYA.js";
 import "../chunk-GGE4NNQT.js";
-import "../chunk-LNQEP766.js";
+import "../chunk-S2F4J57L.js";
 import "../chunk-5UF54T55.js";
 import "../chunk-XJYR7XFV.js";
 import "../chunk-VSMTAMNK.js";
-import "../chunk-GY4SYVPJ.js";
+import "../chunk-NJC7U437.js";
+import "../chunk-VCTY3W6J.js";
+import "../chunk-VI2UW6B6.js";
 import "../chunk-PC4UYEBM.js";
 import "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";

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-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';
+import { S as SandboxDriver, H as HarnessConfig, a as SandboxHarnessResult, T as TestGradedScenario, b as TestGradedRunResult } from '../test-graded-scenario-B0ybnPY7.js';
+import { T as TraceEmitter } from '../emitter-CjD7vUwv.js';
+import { R as Run } from '../schema-B3Q3l9Z_.js';
+import { T as TraceStore } from '../store-DGqD0Pyo.js';
 
 /**
  * BuilderSession — ties a builder-of-builders workflow together.

package/dist/builder-eval/index.js

@@ -8,7 +8,7 @@ import {
 } from "../chunk-PJQFMIOX.js";
 import {
   judgeSpans
-} from "../chunk-MHNQWM4I.js";
+} from "../chunk-LQUTGLOZ.js";
 import {
   TraceEmitter
 } from "../chunk-TVVP3ZZQ.js";

package/dist/{calibration-Dz8TQV4y.d.ts → calibration-C8MTS7cw.d.ts}

@@ -1,5 +1,5 @@
-import { T as TraceStore } from './store-BsVi7ncX.js';
-import { R as Run } from './schema-SGWcK9wa.js';
+import { T as TraceStore } from './store-DGqD0Pyo.js';
+import { R as Run } from './schema-B3Q3l9Z_.js';
 import { O as OutcomeFilter, b as OutcomeStore } from './outcome-store-rnXLEqSn.js';
 
 /**

package/dist/campaign/index.d.ts

@@ -1,34 +1,37 @@
-import { L as LlmClientOptions, w as PolicyEditAdmissionOptions, v as PolicyEditAdmission, u as PolicyEdit, d as AnalystFinding, F as FindingToPolicyEditOptions } from '../policy-edit-Clb2v6Oa.js';
-export { s as POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, P as PolicyEditCandidateRecord, a4 as validatePolicyEditCandidateRecord } from '../policy-edit-Clb2v6Oa.js';
-import { P as PairedArmsComparison, S as SignedManifest, B as BackendIntegrityReport, C as CompletionRequirement, R as RuntimeEventLike, a as CompletionVerdict, b as ProducedState, c as CorrectnessChecker } from '../pre-registration--vU0mMtD.js';
-export { L as LlmJudgeDimension, d as LlmJudgeOptions, l as llmJudge } from '../pre-registration--vU0mMtD.js';
-import { A as AnalyzeTracesOptions, a as AnalyzeTracesInput, b as AnalyzeTracesResult } from '../analyst-CFBc14Wc.js';
-import { S as Scenario, M as MutableSurface, D as DispatchContext, b as JudgeConfig, g as Gate, e as GenerationRecord, J as JudgeScore, L as LabeledScenarioStore, s as LabeledScenarioWrite, t as LabeledScenarioSampleArgs, u as LabeledScenarioRecord, v as LabelTrust, f as SurfaceProposer, w as ProposedCandidate, x as ProposeContext, m as CodeSurface, y as LabeledScenarioSource, C as CampaignResult } from '../types-Ca_63YSD.js';
-export { i as CampaignAggregates, j as CampaignArtifactWriter, k as CampaignCellResult, l as CampaignCostMeter, z as CampaignTokenUsage, d as CampaignTraceWriter, c as DispatchFn, n as GateContext, h as GateDecision, G as GateResult, o as GenerationCandidate, A as JudgeAggregate, a as JudgeDimension, p as Mutator, O as OptimizationProposer, q as OptimizerConfig, P as ParetoParent, R as RedactionStatus, B as ScenarioAggregate, E as ScoredSurfaceOutcome, r as SessionScript, T as TraceSpan, F as isProposedCandidate, H as labelTrustRank } from '../types-Ca_63YSD.js';
-import { C as CampaignRunPlan, P as PlanCampaignRunOptions, b as RunCampaignOptions, c as RunImprovementLoopOptions } from '../gepa-CQelRtuC.js';
-export { f as CampaignRunPlanCell, h as GepaProposerConstraints, G as GepaProposerOptions, O as OpenAutoPrOptions, i as OpenAutoPrResult, a as RunImprovementLoopResult, R as RunOptimizationOptions, j as RunOptimizationResult, k as countSentenceEdits, l as defaultRenderDiff, m as extractH2Sections, g as gepaProposer, o as openAutoPr, p as planCampaignRun, r as runCampaign, d as runImprovementLoop, n as runOptimization } from '../gepa-CQelRtuC.js';
-import { a as PairedBootstrapResult, E as EProcessState } from '../statistics-oUbOJe-S.js';
-import { C as CampaignStorage } from '../storage-Dw_f7WMt.js';
-export { f as fsCampaignStorage, i as inMemoryCampaignStorage } from '../storage-Dw_f7WMt.js';
-export { A as AxisEvidence, a as AxisVerdict, B as BuildEvidenceVectorOptions, l as BuildLoopProvenanceArgs, D as DefaultProductionGateOptions, m as EmitLoopProvenanceArgs, n as EmitLoopProvenanceResult, E as EvidenceVector, b as EvolutionaryProposerOptions, H as HeldOutGateOptions, o as LoopProvenanceBackend, q as LoopProvenanceCandidate, L as LoopProvenanceRecord, O as ObjectiveSource, c as ParetoSignificanceGateOptions, P as PowerPreflight, s as PowerPreflightOptions, d as PromotionObjective, e as PromotionPolicy, R as RunEvalOptions, f as buildEvidenceVector, t as buildLoopProvenanceRecord, g as composeGate, h as defaultProductionGate, u as emitLoopProvenance, i as evolutionaryProposer, j as heldOutGate, v as loopProvenanceSpans, p as paretoPolicy, k as paretoSignificanceGate, w as powerPreflight, x as provenanceRecordPath, y as provenanceSpansPath, r as runEval } from '../provenance-BbVagC68.js';
+import { v as PolicyEditAdmissionOptions, u as PolicyEditAdmission, t as PolicyEdit, c as AnalystFinding, F as FindingToPolicyEditOptions } from '../policy-edit-wG9uFEFm.js';
+export { r as POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, P as PolicyEditCandidateRecord, a2 as validatePolicyEditCandidateRecord } from '../policy-edit-wG9uFEFm.js';
+import { P as PairedArmsComparison, S as SignedManifest, B as BackendIntegrityReport, C as CompletionRequirement, R as RuntimeEventLike, a as CompletionVerdict, b as ProducedState, c as CorrectnessChecker } from '../pre-registration-BWQhJ3vz.js';
+export { L as LlmJudgeDimension, d as LlmJudgeOptions, l as llmJudge } from '../pre-registration-BWQhJ3vz.js';
+import { C as CampaignRunPlan, P as PlanCampaignRunOptions, h as RunCampaignOptions, i as RunImprovementLoopOptions } from '../gepa-eESocoDi.js';
+export { o as CampaignRunPlanCell, p as GepaProposerConstraints, G as GepaProposerOptions, O as OpenAutoPrOptions, q as OpenAutoPrResult, e as ReferenceEquivalenceJudgeOptions, g as ReferenceEquivalenceScenario, a as RunImprovementLoopResult, R as RunOptimizationOptions, s as RunOptimizationResult, t as countSentenceEdits, j as createReferenceEquivalenceJudge, u as defaultRenderDiff, v as extractH2Sections, k as gepaProposer, w as openAutoPr, x as planCampaignRun, r as runCampaign, l as runImprovementLoop, y as runOptimization } from '../gepa-eESocoDi.js';
+import { A as AnalyzeTracesOptions, a as AnalyzeTracesInput, b as AnalyzeTracesResult } from '../analyst-C8HHvfJp.js';
+import { S as Scenario, M as MutableSurface, D as DispatchContext, b as JudgeConfig, G as Gate, o as GenerationRecord, J as JudgeScore, L as LabeledScenarioStore, s as LabeledScenarioWrite, t as LabeledScenarioSampleArgs, u as LabeledScenarioRecord, v as LabelTrust, c as SurfaceProposer, w as ProposedCandidate, x as ProposeContext, j as CodeSurface, y as LabeledScenarioSource, C as CampaignResult } from '../types-BSw1rOUB.js';
+export { e as CampaignAggregates, f as CampaignArtifactWriter, g as CampaignCellResult, h as CampaignCostMeter, z as CampaignTokenUsage, i as CampaignTraceWriter, k as DispatchFn, l as GateContext, d as GateDecision, m as GateResult, n as GenerationCandidate, A as JudgeAggregate, a as JudgeDimension, p as Mutator, O as OptimizationProposer, q as OptimizerConfig, P as ParetoParent, R as RedactionStatus, B as ScenarioAggregate, E as ScoredSurfaceOutcome, r as SessionScript, T as TraceSpan, F as isProposedCandidate, H as labelTrustRank } from '../types-BSw1rOUB.js';
+import { a as PairedBootstrapResult, E as EProcessState } from '../statistics-KUnG73jH.js';
+import { C as CampaignStorage } from '../storage-DrX3v_5B.js';
+export { c as createRunCostLedger, f as fsCampaignStorage, i as inMemoryCampaignStorage } from '../storage-DrX3v_5B.js';
+export { A as AxisEvidence, a as AxisVerdict, B as BuildEvidenceVectorOptions, l as BuildLoopProvenanceArgs, D as DefaultProductionGateOptions, m as EmitLoopProvenanceArgs, n as EmitLoopProvenanceResult, E as EvidenceVector, b as EvolutionaryProposerOptions, H as HeldOutGateOptions, o as LoopProvenanceBackend, q as LoopProvenanceCandidate, L as LoopProvenanceRecord, O as ObjectiveSource, c as ParetoSignificanceGateOptions, P as PowerPreflight, s as PowerPreflightOptions, d as PromotionObjective, e as PromotionPolicy, R as RunEvalOptions, f as buildEvidenceVector, t as buildLoopProvenanceRecord, g as composeGate, h as defaultProductionGate, u as emitLoopProvenance, i as evolutionaryProposer, j as heldOutGate, v as loopProvenanceSpans, p as paretoPolicy, k as paretoSignificanceGate, w as powerPreflight, x as provenanceRecordPath, y as provenanceSpansPath, r as runEval } from '../provenance-DpjwyseI.js';
+import { a as LlmClientOptions } from '../llm-client-qoDd18Qz.js';
 import { AgentProfile } from '@tangle-network/agent-interface';
 import { A as AgentEvalError, V as ValidationError } from '../errors-oeQrLqXC.js';
-import { a as RunSplitTag, R as RunRecord } from '../run-record-CZmcpWPo.js';
-import { T as TraceAnalystKindSpec } from '../kind-factory-DWOvXjR_.js';
-import '../store-9cAScOcb.js';
-import '../types-C7DGg5ex.js';
+import { a as RunSplitTag, R as RunRecord } from '../run-record-BDH49H2E.js';
+import { C as CostLedger, b as CostLedgerSummary, M as MaximumCharge, c as CostReceiptInput } from '../cost-ledger-DWy3XdJc.js';
+import { T as TraceAnalystKindSpec } from '../kind-factory-ClZmO25A.js';
+import '../store-C1YxJDEK.js';
+import '../types-BkfcQnxV.js';
 import '@tangle-network/tcloud';
+import 'zod';
+import '../raw-provider-sink-C46HDghv.js';
 import '../verdict-C9MlYujm.js';
-import '@ax-llm/ax';
 import '../dataset-NENEzRgk.js';
-import '../store-BsVi7ncX.js';
-import '../schema-SGWcK9wa.js';
+import '../store-DGqD0Pyo.js';
+import '../schema-B3Q3l9Z_.js';
+import '@ax-llm/ax';
 import '../judge-calibration-7C-IDmKr.js';
 import '../hosted/index.js';
-import '../insight-report-oMVxDTxl.js';
-import '../summary-report-DTNgQycC.js';
-import '../failure-cluster-C48PiReX.js';
-import 'zod';
+import '../insight-report-DY4nDW9Q.js';
+import '../summary-report-C5bKFfm-.js';
+import '../failure-cluster-DOAcSJ87.js';
 
 /**
  * Lineage DAG — a git-graph of improvement candidates.
@@ -1468,12 +1471,12 @@ declare function fapoEscalationEntry<TScenario extends Scenario, TArtifact>(conf
  * It runs `runCampaign` once per profile (reusing its seeds, reps, bootstrap
  * CIs, resumability, and the `LabeledScenarioStore` capture flywheel), maps
  * every cell to a validated `RunRecord` carrying the real `tokenUsage` the
- * dispatch reported via `ctx.cost.observeTokens`, and runs `assertRealBackend`
+ * dispatch committed via `ctx.cost.runPaidCall`, and runs `assertRealBackend`
  * BY CONSTRUCTION before returning — so a stub-backend run fails loudly instead
  * of reporting a clean 0/N leaderboard.
  *
  * Dispatch contract: a dispatch that calls an LLM MUST report usage via
- * `ctx.cost.observeTokens({ input, output })` (and cost via `ctx.cost.observe`).
+ * `ctx.cost.runPaidCall({ execute, receipt })`.
  * A dispatch that reports zero tokens is indistinguishable from a stub and the
  * integrity guard treats it as one.
  */
@@ -1485,8 +1488,8 @@ declare class ProfileMatrixError extends AgentEvalError {
     constructor(message: string);
 }
 /** Dispatch for one cell: render `profile` against `scenario`, returning the
- *  artifact the judges score. Report LLM usage via `ctx.cost.observeTokens`
- *  and `ctx.cost.observe` — the integrity guard depends on it. */
+ *  artifact the judges score. Run LLM work through `ctx.cost.runPaidCall` —
+ *  the integrity check depends on its receipt. */
 type ProfileDispatchFn<TScenario extends Scenario, TArtifact> = (profile: AgentProfile, scenario: TScenario, ctx: DispatchContext) => Promise<TArtifact>;
 interface RunProfileMatrixOptions<TScenario extends Scenario, TArtifact> {
     /** Axis 3 — the agent-under-test configurations. Each is one column. */
@@ -1637,7 +1640,7 @@ interface PlaybackContext extends DispatchContext {
  * `SandboxPlaybackDriver` (real API / sandbox workspace) and
  * `PlaywrightPlaybackDriver` (real UI) — because they depend on runtime /
  * browser infra the substrate must not import. The driver MUST report LLM
- * usage via `ctx.cost.observeTokens` so the backend-integrity guard sees real
+ * usage through `ctx.cost.runPaidCall` so the backend-integrity check sees real
  * tokens (a run that never reports tokens reads as a stub).
  */
 interface PlaybackDriver<TStory extends UserStory = UserStory> {
@@ -1950,10 +1953,14 @@ interface ProposePatchesArgs {
     /** How many candidate patches to propose. */
     count: number;
     signal: AbortSignal;
+    costLedger?: CostLedger;
+    costPhase?: string;
 }
 interface SkillOptProposerOptions {
     llm: LlmClientOptions;
     model: string;
+    /** Optional ledger for direct proposer use. Campaign context takes precedence. */
+    costLedger?: CostLedger;
     /** What the skill document governs — orients the prompt. */
     target: string;
     /** Default ops-per-patch cap when used as a bare `SurfaceProposer`. The
@@ -2077,6 +2084,8 @@ interface RunSkillOptResult {
     /** Total cost across every scoring campaign (train evidence + holdout
      *  acceptance) the hill-climb ran. */
     totalCostUsd: number;
+    /** Run-wide spend, including scoring, proposals, and judges. */
+    cost: CostLedgerSummary;
 }
 /**
  * SkillOpt sequential hill-climb: each epoch reflects on train-scenario weaknesses, proposes bounded patches, accepts the first patch that strictly improves the held-out composite, and anneals the edit budget on consecutive rejections.
@@ -2191,6 +2200,11 @@ interface HaloProposerOptions {
     model?: string;
     /** Model used to APPLY halo's findings to the prompt surface. Default = `model`. */
     applyModel?: string;
+    /** Optional ledger for direct proposer use. Campaign context takes precedence. */
+    costLedger?: CostLedger;
+    analysisMaximumCharge?: MaximumCharge;
+    analysisReceipt?: (report: string) => CostReceiptInput;
+    applyMaxTokens?: number;
     /** The real halo binary. Default 'halo' (from `pip install halo-engine`). */
     haloBin?: string;
     /** Resolve the OTLP traces (JSONL string) halo should analyze for THIS
@@ -2323,6 +2337,8 @@ interface PolicyEditHistoryGenerationContext {
 interface LlmPolicyEditProposerOptions {
     llm: LlmClientOptions;
     model: string;
+    /** Optional ledger for direct proposer use. Campaign context takes precedence. */
+    costLedger?: CostLedger;
     /** Plain-language description of the JSON surface being improved. */
     target: string;
     /** PolicyEdit target surface every authored edit must retain. */
@@ -2396,6 +2412,8 @@ declare function projectPolicyEditHistory(history: readonly GenerationRecord[],
  */
 
 interface MemoryCurationProposerOptions {
+    /** Optional ledger for direct proposer use. Campaign context takes precedence. */
+    costLedger?: CostLedger;
     /** Top-K lessons retained in the surface memory block. Default 12. */
     maxEntries?: number;
     /** Heading rendered above the lessons inside the block. Default below. */
@@ -2408,6 +2426,7 @@ interface MemoryCurationProposerOptions {
         baseUrl: string;
         apiKey?: string;
         model: string;
+        maxTokens?: number;
         fetchImpl?: LlmClientOptions['fetch'];
     };
 }
@@ -2496,6 +2515,11 @@ interface TraceAnalystProposerOptions {
     /** Model used to APPLY findings to the prompt surface. Default = `model`.
      *  Keep this EQUAL to haloProposer's `applyModel` for an apples-to-apples run. */
     applyModel?: string;
+    /** Optional ledger for direct proposer use. Campaign context takes precedence. */
+    costLedger?: CostLedger;
+    analysisMaximumCharge?: MaximumCharge;
+    analysisReceipt?: (report: string) => CostReceiptInput;
+    applyMaxTokens?: number;
     /** Ax provider name. Default 'openai' — works for any OpenAI-compatible base
      *  via `apiURL`. Use 'deepseek' to hit DeepSeek's native provider. */
     provider?: string;
@@ -3045,7 +3069,7 @@ interface WorktreeAdapter {
     /** Commit pending changes, freeze the exact Git objects + binary patch, and
      *  verify the worktree still matches that identity. */
     finalize(worktree: Worktree, summary: string): Promise<CodeSurface>;
-    /** Remove the worktree (and its branch) — called for losing candidates. */
+    /** Idempotently remove the worktree and branch. Safe to retry after partial cleanup. */
     discard(worktree: Worktree): Promise<void>;
 }
 /** Typed failure from a `WorktreeAdapter` operation (create/finalize/discard) — wraps the underlying git error as `cause`. */