@tangle-network/agent-eval 0.115.3 → 0.117.0

package/CHANGELOG.md

@@ -4,6 +4,61 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ---
 
+## [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
+
+- `llmPolicyEditProposer()` converts attributed trace findings and bounded search history into typed JSON edits over caller-approved AgentProfile paths.
+- Author context selection retains promoted candidates plus outcome extremes, selects task rows by difficulty and change from parent, enforces an exact serialized size limit, and pseudonymizes known task identifiers before model dispatch.
+- Policy-edit history and provenance retain the exact edit, measured parent, observed score change, coverage, eligibility, surface bytes, and final winner chain needed for credit assignment.
+
+### Changed
+
+- `runOptimization()` now keeps one best complete surface across every generation.
+  Baselines and candidates must cover the exact designed task-by-repetition count, and partial, failed, or non-finite results cannot be promoted.
+- Model-authored confidence and gain forecasts no longer suppress evidence-linked candidates by default.
+  Forecasts must describe increasing raw search scores, respect the declared range and current headroom, and enter residual history only when their units match the measured outcome.
+- GEPA reflection now uses evidence from the measured incumbent that is actually being edited instead of the latest losing candidate.
+
+### Breaking
+
+- `runOptimization({ promoteTopK })` accepts only `1`; multiple concurrent incumbents were never represented by the optimizer state and now fail before dispatch.
+- `ScoredSurfaceOutcome` requires `split: 'search'` and the actual `generation` that measured the surface.
+- `llmPolicyEditProposer()` requires explicit raw-score objectives and `PolicyEditFindingInput` rows whose source is either an exact measured surface-generation pair or an explicitly global finding.
+- `LoopProvenanceRecord.schema` is now `tangle.loop-provenance.v3`.
+  Consumers of v2 records must migrate to the v3 baseline score, parent chain, coverage, eligibility, and exact surface fields; no compatibility shim is provided.
+
 ## [0.115.3] — 2026-07-12 — fail-closed structured output parsing
 
 ### Fixed

package/dist/analyst/index.d.ts

@@ -1,20 +1,21 @@
 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-DxJmRkyJ.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-DxJmRkyJ.js';
-import { b as JudgeFn, a as JudgeInput } from '../types-C7DGg5ex.js';
-import { a as Analyst, g as AnalystSeverity, A as AnalystFinding } from '../kind-factory-DcNg13sZ.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-DcNg13sZ.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-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-DltpYR5u.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-RLn8GWof.js';
-import { L as LlmClientOptions } from '../llm-client-DyqEH4jH.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-B7RTi_ix.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';
@@ -188,6 +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

@@ -6,34 +6,19 @@ import {
   SkillUsageAnalyst,
   buildSkillUsageReport,
   createAnalystAi,
-  createChatClient,
   defaultIsMaterial,
   diffFindings,
   emitSkillUsageFindings,
   runSemanticConceptJudge
-} from "../chunk-WSBUZMBU.js";
+} from "../chunk-CCZIVI3F.js";
 import {
   behavioralAnalyst,
   buildDefaultAnalystRegistry,
+  createChatClient,
   deriveEfficiencyFindings
-} from "../chunk-I6LVHOV3.js";
-import {
-  POLICY_EDIT_AXES,
-  POLICY_EDIT_TARGET_SURFACES,
-  PolicyEditValidationError,
-  admitPolicyEdit,
-  applyPolicyEditToSurface,
-  assertNoJudgeVerdict,
-  computePolicyEditId,
-  isJudgeVerdict,
-  isPolicyEdit,
-  isTraceObservable,
-  makePolicyEdit,
-  policyEditFromFinding,
-  policyEditsFromFindings,
-  scorePolicyEditReadiness,
-  validatePolicyEdit
-} from "../chunk-QMXXSNC4.js";
+} from "../chunk-VF3XSYTI.js";
+import "../chunk-HHWE3POT.js";
+import "../chunk-3YYRZDON.js";
 import {
   ANALYST_SEVERITIES,
   AnalystRegistry,
@@ -47,26 +32,46 @@ import {
   KIND_EXPECTED_SUBJECTS,
   KNOWLEDGE_GAP_KIND_SPEC,
   KNOWLEDGE_POISONING_KIND_SPEC,
+  POLICY_EDIT_AXES,
+  POLICY_EDIT_CANDIDATE_RECORD_SCHEMA,
+  POLICY_EDIT_TARGET_SURFACES,
+  PolicyEditValidationError,
   RAW_FINDING_SCHEMA_PROMPT,
   RawAnalystFindingSchema,
+  admitPolicyEdit,
+  applyPolicyEditToSurface,
+  assertNoJudgeVerdict,
   buildTraceToolsForGroup,
   coerceJson,
   coerceToFindingRows,
   computeFindingId,
+  computePolicyEditId,
   createTraceAnalystKind,
   findingSubjectGrammarPromptFor,
+  isJudgeVerdict,
+  isPolicyEdit,
+  isTraceObservable,
   makeFinding,
+  makePolicyEdit,
+  makePolicyEditCandidateRecord,
   parseFindingSubject,
   parseRawFinding,
+  policyEditFromFinding,
+  policyEditsFromFindings,
   renderFindingSubject,
   renderPriorFindings,
+  scorePolicyEditReadiness,
   stripCodeFences,
-  structureFindings
-} from "../chunk-5S5NJ63F.js";
-import "../chunk-LNQEP766.js";
+  structureFindings,
+  validatePolicyEdit,
+  validatePolicyEditCandidateRecord
+} 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";
@@ -307,6 +312,7 @@ export {
   KNOWLEDGE_GAP_KIND_SPEC,
   KNOWLEDGE_POISONING_KIND_SPEC,
   POLICY_EDIT_AXES,
+  POLICY_EDIT_CANDIDATE_RECORD_SCHEMA,
   POLICY_EDIT_TARGET_SURFACES,
   PolicyEditValidationError,
   RAW_FINDING_SCHEMA_PROMPT,
@@ -342,6 +348,7 @@ export {
   liftSeverity,
   makeFinding,
   makePolicyEdit,
+  makePolicyEditCandidateRecord,
   parseFindingSubject,
   parseRawFinding,
   policyEditFromFinding,
@@ -351,6 +358,7 @@ export {
   scorePolicyEditReadiness,
   stripCodeFences,
   structureFindings,
-  validatePolicyEdit
+  validatePolicyEdit,
+  validatePolicyEditCandidateRecord
 };
 //# sourceMappingURL=index.js.map

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-BYHg6Irm.d.ts → analyze-runs--2x39HZ7.d.ts}

@@ -1,7 +1,7 @@
-import { A as AnalystRegistry } from './default-registry-DltpYR5u.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-B7RTi_ix.js';
-import { I as InsightReport } from './insight-report-IwwvqZZv.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-D-g04tcy.js';
-import { R as RunRecord, b as RunSplitTag } from '../run-record-B7RTi_ix.js';
-import { T as TraceStore } from '../store-BsVi7ncX.js';
-import { R as RuntimeTrajectoryRecord, P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection } from '../runtime-trajectory-Dws7Kpgi.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

@@ -1,7 +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-BTEpx9He.js';
-import '../types-C5gJrOVT.js';
-import '../run-record-B7RTi_ix.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 '../storage-Dw_f7WMt.js';
+import '../schema-B3Q3l9Z_.js';
+import '../store-C1YxJDEK.js';
+import '../types-BkfcQnxV.js';
+import '../cost-ledger-DWy3XdJc.js';
+import '@tangle-network/tcloud';
+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-KG4TD7EQ.js";
-import "../chunk-ADYLPOSX.js";
-import "../chunk-VI2UW6B6.js";
-import "../chunk-FAOEFFRT.js";
-import "../chunk-QMXXSNC4.js";
-import "../chunk-5S5NJ63F.js";
+} from "../chunk-JSDVRFAP.js";
+import "../chunk-ZUXV7UWZ.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';
 
 /**