@tangle-network/agent-eval 0.103.0 → 0.103.2

package/dist/adapters/http.d.ts

@@ -1,7 +1,7 @@
-import { S as Scenario, D as DispatchFn, b as DispatchContext } from '../types-DFI_Z-ZL.js';
-import '../run-record-MRdJ-Kq2.js';
+import { S as Scenario, D as DispatchFn, b as DispatchContext } from '../types-Mh6I44ub.js';
+import '../run-record-I-Z3JNvO.js';
 import '@tangle-network/agent-interface';
-import '../errors-CzMUYo7b.js';
+import '../errors-oeQrLqXC.js';
 import '../schema-m0gsnbt3.js';
 
 /**

package/dist/adapters/langchain.d.ts

@@ -1,7 +1,7 @@
-import { S as Scenario, J as JudgeScore, D as DispatchFn, a as JudgeConfig } from '../types-DFI_Z-ZL.js';
-import '../run-record-MRdJ-Kq2.js';
+import { S as Scenario, J as JudgeScore, D as DispatchFn, a as JudgeConfig } from '../types-Mh6I44ub.js';
+import '../run-record-I-Z3JNvO.js';
 import '@tangle-network/agent-interface';
-import '../errors-CzMUYo7b.js';
+import '../errors-oeQrLqXC.js';
 import '../schema-m0gsnbt3.js';
 
 /**

package/dist/adapters/otel.d.ts

@@ -1,14 +1,14 @@
 import { TraceSpanEvent, HostedClient } from '../hosted/index.js';
-import '../types-DFI_Z-ZL.js';
-import '../run-record-MRdJ-Kq2.js';
+import '../types-Mh6I44ub.js';
+import '../run-record-I-Z3JNvO.js';
 import '@tangle-network/agent-interface';
-import '../errors-CzMUYo7b.js';
+import '../errors-oeQrLqXC.js';
 import '../schema-m0gsnbt3.js';
-import '../insight-report-k0sRTzKg.js';
-import '../summary-report-C0nnxOD8.js';
+import '../insight-report-B4xrdwEK.js';
+import '../summary-report-QMZVe3P-.js';
 import '../failure-cluster-DH9Flgcf.js';
 import '../store-BcFXE6LG.js';
-import '../judge-calibration-0p2QcWNE.js';
+import '../judge-calibration-7C-IDmKr.js';
 
 /**
  * # `@tangle-network/agent-eval/adapters/otel` — OTel→hosted bridge.

package/dist/analyst/index.d.ts

@@ -1,24 +1,24 @@
-import { M as MultiLayerVerifier, V as VerifyOptions, S as Severity } from '../multi-layer-verifier-DUZXrPDA.js';
+import { M as MultiLayerVerifier, V as VerifyOptions, S as Severity } from '../multi-layer-verifier-CI4jdX-q.js';
 import { c as RunCritic, a as RunTrace } from '../run-critic-CmMf05uV.js';
-import { S as SemanticConceptJudgeOptions, a as SemanticConceptJudgeInput, B as BehavioralMetrics } from '../semantic-concept-judge-D-IlH5v1.js';
-export { C as CreateAnalystAiConfig, D as DEFAULT_TRACE_ANALYST_KINDS, b as DiffPolicy, F as FAILURE_MODE_KIND_SPEC, c as FINDING_SUBJECT_GRAMMAR_PROMPT, d as FINDING_SUBJECT_KINDS, e as FindingSubject, f as FindingSubjectKind, g as FindingSubjectStringSchema, h as FindingsDiff, i as FindingsStore, I as IMPROVEMENT_KIND_SPEC, K as KIND_EXPECTED_SUBJECTS, j as KNOWLEDGE_GAP_KIND_SPEC, k as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, l as SKILL_USAGE_ANALYST, m as SkillUsageAnalyst, n as SkillUsageRecord, o as SkillUsageReport, p as SkillUsageScanConfig, q as buildSkillUsageReport, r as createAnalystAi, s as defaultIsMaterial, t as diffFindings, u as emitSkillUsageFindings, v as parseFindingSubject, w as renderFindingSubject } from '../semantic-concept-judge-D-IlH5v1.js';
+import { S as SemanticConceptJudgeOptions, a as SemanticConceptJudgeInput, B as BehavioralMetrics } from '../semantic-concept-judge-BmNZPB_j.js';
+export { C as CreateAnalystAiConfig, D as DEFAULT_TRACE_ANALYST_KINDS, b as DiffPolicy, F as FAILURE_MODE_KIND_SPEC, c as FINDING_SUBJECT_GRAMMAR_PROMPT, d as FINDING_SUBJECT_KINDS, e as FindingSubject, f as FindingSubjectKind, g as FindingSubjectStringSchema, h as FindingsDiff, i as FindingsStore, I as IMPROVEMENT_KIND_SPEC, K as KIND_EXPECTED_SUBJECTS, j as KNOWLEDGE_GAP_KIND_SPEC, k as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, l as SKILL_USAGE_ANALYST, m as SkillUsageAnalyst, n as SkillUsageRecord, o as SkillUsageReport, p as SkillUsageScanConfig, q as buildSkillUsageReport, r as createAnalystAi, s as defaultIsMaterial, t as diffFindings, u as emitSkillUsageFindings, v as parseFindingSubject, w as renderFindingSubject } from '../semantic-concept-judge-BmNZPB_j.js';
 import { b as JudgeFn, a as JudgeInput } from '../types-C7DGg5ex.js';
-import { a as Analyst, h as AnalystSeverity, A as AnalystFinding } from '../types-Dz9cKF0g.js';
-export { b as AnalystContext, g as AnalystCost, i as AnalystInputKind, j as AnalystRequirements, f as AnalystRunEvent, e as AnalystRunInputs, d as AnalystRunResult, c as AnalystRunSummary, k as ChatCallOpts, C as ChatClient, l as ChatRequest, m as ChatResponse, n as ChatTransport, o as CliBridgeTransportOpts, p as CreateChatClientOpts, D as DirectProviderTransportOpts, E as EvidenceRef, M as MockTransportOpts, R as RouterTransportOpts, S as SandboxSdkTransportOpts, q as computeFindingId, r as createChatClient, s as makeFinding } from '../types-Dz9cKF0g.js';
+import { a as Analyst, h as AnalystSeverity, A as AnalystFinding } from '../types-D1ytG0Yg.js';
+export { b as AnalystContext, g as AnalystCost, i as AnalystInputKind, j as AnalystRequirements, f as AnalystRunEvent, e as AnalystRunInputs, d as AnalystRunResult, c as AnalystRunSummary, k as ChatCallOpts, C as ChatClient, l as ChatRequest, m as ChatResponse, n as ChatTransport, o as CliBridgeTransportOpts, p as CreateChatClientOpts, D as DirectProviderTransportOpts, E as EvidenceRef, M as MockTransportOpts, R as RouterTransportOpts, S as SandboxSdkTransportOpts, q as computeFindingId, r as createChatClient, s as makeFinding } from '../types-D1ytG0Yg.js';
 import { 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-ZhqsTr4K.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-D0nk7AKV.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-BDQzzsBU.js';
-import { L as LlmClientOptions } from '../llm-client-Bj7g0rqu.js';
+export { a as AnalystHooks, A as AnalystRegistry, b as AnalystRegistryOptions, B as BudgetPolicy, D as DefaultAnalystRegistryOptions, R as RegistryRunOpts, c as buildDefaultAnalystRegistry } from '../default-registry-BswHCXnU.js';
+export { A as ANALYST_SEVERITIES, C as CreateTraceAnalystKindOpts, R as RAW_FINDING_SCHEMA_PROMPT, a as RawAnalystFinding, b as RawAnalystFindingSchema, c as TraceAnalystGolden, T as TraceAnalystKindSpec, d as createTraceAnalystKind, p as parseRawFinding, r as renderPriorFindings } from '../kind-factory-DvIGo_cP.js';
+export { F as FindingToPolicyEditOptions, P as POLICY_EDIT_AXES, a as POLICY_EDIT_TARGET_SURFACES, b as PolicyEdit, c as PolicyEditAdmission, d as PolicyEditAdmissionOptions, e as PolicyEditAxis, f as PolicyEditChange, g as PolicyEditExpectedGain, h as PolicyEditGainDirection, i as PolicyEditGainUnit, j as PolicyEditInit, k as PolicyEditRisk, l as PolicyEditSchemaVersion, m as PolicyEditSource, n as PolicyEditTarget, o as PolicyEditTargetSurface, p as PolicyEditValidationError, q as admitPolicyEdit, r as applyPolicyEditToSurface, s as computePolicyEditId, t as isPolicyEdit, u as makePolicyEdit, v as policyEditFromFinding, w as policyEditsFromFindings, x as scorePolicyEditReadiness, y as validatePolicyEdit } from '../policy-edit-DQUXYMDm.js';
+import { L as LlmClientOptions } from '../llm-client-DyqEH4jH.js';
 import { AxFunction } from '@ax-llm/ax';
 import '../verdict-C9MlYujm.js';
 import '../schema-m0gsnbt3.js';
 import '../store-BcFXE6LG.js';
 import 'zod';
-import '../run-record-MRdJ-Kq2.js';
+import '../run-record-I-Z3JNvO.js';
 import '@tangle-network/agent-interface';
-import '../errors-CzMUYo7b.js';
+import '../errors-oeQrLqXC.js';
 import '../raw-provider-sink-C46HDghv.js';
 
 /**

package/dist/analyst/index.js

@@ -11,13 +11,13 @@ import {
   diffFindings,
   emitSkillUsageFindings,
   runSemanticConceptJudge
-} from "../chunk-JU6ZX3CX.js";
+} from "../chunk-5PK3626Q.js";
 import "../chunk-DJWX3GVS.js";
 import {
   behavioralAnalyst,
   buildDefaultAnalystRegistry,
   deriveEfficiencyFindings
-} from "../chunk-L5TVEZFT.js";
+} from "../chunk-QRVS7MX4.js";
 import {
   POLICY_EDIT_AXES,
   POLICY_EDIT_TARGET_SURFACES,
@@ -34,7 +34,7 @@ import {
   policyEditsFromFindings,
   scorePolicyEditReadiness,
   validatePolicyEdit
-} from "../chunk-IN3SHQML.js";
+} from "../chunk-LOJ2QVCE.js";
 import {
   ANALYST_SEVERITIES,
   AnalystRegistry,
@@ -59,17 +59,17 @@ import {
   renderPriorFindings,
   stripCodeFences,
   structureFindings
-} from "../chunk-FRI6RG3P.js";
-import "../chunk-CWNP4DV4.js";
+} from "../chunk-2OGPXHOB.js";
+import "../chunk-FUCQVFMU.js";
 import {
   computeFindingId,
   makeFinding
 } from "../chunk-45EEMHTC.js";
-import "../chunk-YBIGNSCZ.js";
+import "../chunk-LNQEP766.js";
 import "../chunk-PC4UYEBM.js";
-import "../chunk-ABOIVNXL.js";
+import "../chunk-XJYR7XFV.js";
 import "../chunk-VSMTAMNK.js";
-import "../chunk-3BFEG2F6.js";
+import "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";
 
 // src/analyst/adapters.ts

package/dist/{analyze-runs-Cd-A_K4l.d.ts → analyze-runs-DJYpep3L.d.ts}

@@ -1,7 +1,7 @@
-import { A as AnalystRegistry } from './default-registry-ZhqsTr4K.js';
-import { a as DatasetScenario } from './dataset-BbGkaN2I.js';
-import { R as RunRecord } from './run-record-MRdJ-Kq2.js';
-import { I as InsightReport } from './insight-report-k0sRTzKg.js';
+import { A as AnalystRegistry } from './default-registry-BswHCXnU.js';
+import { a as DatasetScenario } from './dataset-DS7ytHZU.js';
+import { R as RunRecord } from './run-record-I-Z3JNvO.js';
+import { I as InsightReport } from './insight-report-B4xrdwEK.js';
 
 /**
  * # `analyzeRuns()` — turn a set of agent runs into an actionable decision packet.

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

@@ -1,13 +1,13 @@
 import { c as CalibrationReport } from '../calibration-BPmzuVPk.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-Ce-9u7YM.js';
-import { R as RunRecord, b as RunSplitTag } from '../run-record-MRdJ-Kq2.js';
+import { d as CodeAgentSessionSource, a as CodeAgentSessionIntakeOptions, c as CodeAgentSessionMetrics, C as CodeAgentSessionDiagnostic } from '../code-agent-session-rnJKlqmT.js';
+import { R as RunRecord, b as RunSplitTag } from '../run-record-I-Z3JNvO.js';
 import { T as TraceStore } from '../store-BcFXE6LG.js';
-import { R as RuntimeTrajectoryRecord, P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection } from '../runtime-trajectory-8w0_jmtR.js';
+import { R as RuntimeTrajectoryRecord, P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection } from '../runtime-trajectory-iW9IhV3e.js';
 import '../schema-m0gsnbt3.js';
 import '../outcome-store-rnXLEqSn.js';
 import '@tangle-network/agent-interface';
-import '../errors-CzMUYo7b.js';
+import '../errors-oeQrLqXC.js';
 
 declare const BELIEF_DECISION_KINDS: readonly ["continue", "verify", "ask", "retry", "stop", "memory-write", "memory-read", "tool-select", "skill-select", "workflow-select", "surface-promote"];
 type BeliefDecisionKind = (typeof BELIEF_DECISION_KINDS)[number];

package/dist/belief-state/index.js

@@ -14,13 +14,13 @@ import {
 import "../chunk-VI2UW6B6.js";
 import {
   offPolicyEstimateAll
-} from "../chunk-4DIJWVUT.js";
+} from "../chunk-DTJ6QUQB.js";
 import {
   confidenceInterval
-} from "../chunk-HRGUJTER.js";
+} from "../chunk-XMSYF4A7.js";
 import {
   ValidationError
-} from "../chunk-3BFEG2F6.js";
+} from "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";
 
 // src/belief-state/calibration.ts

package/dist/benchmarks/index.d.ts

@@ -1,5 +1,5 @@
-export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, c as BenchmarkEvaluation, d as deterministicSplit, e as routing } from '../index-B-bFgiAF.js';
-import '../run-record-MRdJ-Kq2.js';
+export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, c as BenchmarkEvaluation, d as deterministicSplit, e as routing } from '../index-C2CC_dry.js';
+import '../run-record-I-Z3JNvO.js';
 import '@tangle-network/agent-interface';
-import '../errors-CzMUYo7b.js';
+import '../errors-oeQrLqXC.js';
 import '../schema-m0gsnbt3.js';

package/dist/builder-eval/index.js

@@ -1,19 +1,19 @@
 import {
   SandboxHarness,
   runTestGradedScenario
-} from "../chunk-ZOPLCJSY.js";
+} from "../chunk-HZHNRYHK.js";
 import {
   pearsonR,
   spearmanR
-} from "../chunk-HRGUJTER.js";
+} from "../chunk-XMSYF4A7.js";
 import {
   judgeSpans
-} from "../chunk-UFSG7ACU.js";
+} from "../chunk-JZXGWLK5.js";
 import "../chunk-5BKGXME7.js";
 import {
   TraceEmitter
 } from "../chunk-TVVP3ZZQ.js";
-import "../chunk-3BFEG2F6.js";
+import "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";
 
 // src/builder-eval/builder-session.ts

package/dist/campaign/index.d.ts

@@ -1,34 +1,34 @@
-import { S as SignedManifest, B as BackendIntegrityReport, C as CompletionRequirement, R as RuntimeEventLike, a as CompletionVerdict, P as ProducedState, b as CorrectnessChecker } from '../pre-registration-mWG2w8d-.js';
-export { L as LlmJudgeDimension, c as LlmJudgeOptions, l as llmJudge } from '../pre-registration-mWG2w8d-.js';
+import { S as SignedManifest, B as BackendIntegrityReport, C as CompletionRequirement, R as RuntimeEventLike, a as CompletionVerdict, P as ProducedState, b as CorrectnessChecker } from '../pre-registration-D8Zr8-rr.js';
+export { L as LlmJudgeDimension, c as LlmJudgeOptions, l as llmJudge } from '../pre-registration-D8Zr8-rr.js';
 import { A as AnalyzeTracesOptions, a as AnalyzeTracesInput, b as AnalyzeTracesResult } from '../analyst-C8HHvfJp.js';
-import { S as Scenario, M as MutableSurface, b as DispatchContext, a as JudgeConfig, e as GenerationRecord, g as Gate, 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, y as LabeledScenarioSource, C as CampaignResult, o as CodeSurface } from '../types-DFI_Z-ZL.js';
-export { k as CampaignAggregates, l as CampaignArtifactWriter, m as CampaignCellResult, n as CampaignCostMeter, z as CampaignTokenUsage, d as CampaignTraceWriter, D as DispatchFn, h as GateContext, j as GateDecision, G as GateResult, p as GenerationCandidate, A as JudgeAggregate, c as JudgeDimension, i as Mutator, O as OptimizationProposer, q as OptimizerConfig, P as ParetoParent, R as RedactionStatus, B as ScenarioAggregate, r as SessionScript, T as TraceSpan, E as isProposedCandidate, F as labelTrustRank } from '../types-DFI_Z-ZL.js';
-import { e as CampaignRunPlan, P as PlanCampaignRunOptions, C as CampaignStorage, R as RunCampaignOptions, c as RunImprovementLoopOptions } from '../gepa-DeyPTlvx.js';
-export { h as CampaignRunPlanCell, j as GepaProposerConstraints, G as GepaProposerOptions, O as OpenAutoPrOptions, k as OpenAutoPrResult, b as RunImprovementLoopResult, a as RunOptimizationOptions, l as RunOptimizationResult, m as countSentenceEdits, n as defaultRenderDiff, o as extractH2Sections, f as fsCampaignStorage, g as gepaProposer, i as inMemoryCampaignStorage, p as openAutoPr, q as planCampaignRun, r as runCampaign, d as runImprovementLoop, s as runOptimization, t as surfaceHash } from '../gepa-DeyPTlvx.js';
-export { A as AxisEvidence, a as AxisVerdict, B as BuildEvidenceVectorOptions, k as BuildLoopProvenanceArgs, D as DefaultProductionGateOptions, l as EmitLoopProvenanceArgs, m as EmitLoopProvenanceResult, E as EvidenceVector, b as EvolutionaryProposerOptions, H as HeldOutGateOptions, n as LoopProvenanceBackend, o as LoopProvenanceCandidate, L as LoopProvenanceRecord, O as ObjectiveSource, P as ParetoSignificanceGateOptions, c as PromotionObjective, d as PromotionPolicy, R as RunEvalOptions, e as buildEvidenceVector, q as buildLoopProvenanceRecord, f as composeGate, g as defaultProductionGate, s as emitLoopProvenance, h as evolutionaryProposer, i as heldOutGate, t as loopProvenanceSpans, p as paretoPolicy, j as paretoSignificanceGate, u as provenanceRecordPath, v as provenanceSpansPath, r as runEval, w as surfaceContentHash } from '../provenance-BhJm32vN.js';
-import { E as EProcessState, a as PairedBootstrapResult } from '../statistics-xP-cWc5k.js';
-import { L as LlmClientOptions } from '../llm-client-Bj7g0rqu.js';
+import { S as Scenario, M as MutableSurface, b as DispatchContext, a as JudgeConfig, e as GenerationRecord, g as Gate, 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, y as LabeledScenarioSource, C as CampaignResult, o as CodeSurface } from '../types-Mh6I44ub.js';
+export { k as CampaignAggregates, l as CampaignArtifactWriter, m as CampaignCellResult, n as CampaignCostMeter, z as CampaignTokenUsage, d as CampaignTraceWriter, D as DispatchFn, h as GateContext, j as GateDecision, G as GateResult, p as GenerationCandidate, A as JudgeAggregate, c as JudgeDimension, i as Mutator, O as OptimizationProposer, q as OptimizerConfig, P as ParetoParent, R as RedactionStatus, B as ScenarioAggregate, r as SessionScript, T as TraceSpan, E as isProposedCandidate, F as labelTrustRank } from '../types-Mh6I44ub.js';
+import { e as CampaignRunPlan, P as PlanCampaignRunOptions, C as CampaignStorage, R as RunCampaignOptions, c as RunImprovementLoopOptions } from '../gepa-Dn0gtj9n.js';
+export { h as CampaignRunPlanCell, j as GepaProposerConstraints, G as GepaProposerOptions, O as OpenAutoPrOptions, k as OpenAutoPrResult, b as RunImprovementLoopResult, a as RunOptimizationOptions, l as RunOptimizationResult, m as countSentenceEdits, n as defaultRenderDiff, o as extractH2Sections, f as fsCampaignStorage, g as gepaProposer, i as inMemoryCampaignStorage, p as openAutoPr, q as planCampaignRun, r as runCampaign, d as runImprovementLoop, s as runOptimization, t as surfaceHash } from '../gepa-Dn0gtj9n.js';
+export { A as AxisEvidence, a as AxisVerdict, B as BuildEvidenceVectorOptions, k as BuildLoopProvenanceArgs, D as DefaultProductionGateOptions, l as EmitLoopProvenanceArgs, m as EmitLoopProvenanceResult, E as EvidenceVector, b as EvolutionaryProposerOptions, H as HeldOutGateOptions, n as LoopProvenanceBackend, o as LoopProvenanceCandidate, L as LoopProvenanceRecord, O as ObjectiveSource, P as ParetoSignificanceGateOptions, c as PromotionObjective, d as PromotionPolicy, R as RunEvalOptions, e as buildEvidenceVector, q as buildLoopProvenanceRecord, f as composeGate, g as defaultProductionGate, s as emitLoopProvenance, h as evolutionaryProposer, i as heldOutGate, t as loopProvenanceSpans, p as paretoPolicy, j as paretoSignificanceGate, u as provenanceRecordPath, v as provenanceSpansPath, r as runEval, w as surfaceContentHash } from '../provenance-phkRBl4h.js';
+import { E as EProcessState, a as PairedBootstrapResult } from '../statistics-D88peojY.js';
+import { L as LlmClientOptions } from '../llm-client-DyqEH4jH.js';
 import { AgentProfile } from '@tangle-network/agent-interface';
-import { A as AgentEvalError } from '../errors-CzMUYo7b.js';
-import { b as RunSplitTag, R as RunRecord } from '../run-record-MRdJ-Kq2.js';
-import { b as PolicyEdit, F as FindingToPolicyEditOptions, d as PolicyEditAdmissionOptions, c as PolicyEditAdmission } from '../policy-edit-BDQzzsBU.js';
-import { T as TraceAnalystKindSpec } from '../kind-factory-D0nk7AKV.js';
-import { A as AnalystFinding } from '../types-Dz9cKF0g.js';
+import { A as AgentEvalError } from '../errors-oeQrLqXC.js';
+import { b as RunSplitTag, R as RunRecord } from '../run-record-I-Z3JNvO.js';
+import { b as PolicyEdit, F as FindingToPolicyEditOptions, d as PolicyEditAdmissionOptions, c as PolicyEditAdmission } from '../policy-edit-DQUXYMDm.js';
+import { T as TraceAnalystKindSpec } from '../kind-factory-DvIGo_cP.js';
+import { A as AnalystFinding } from '../types-D1ytG0Yg.js';
 import '@tangle-network/tcloud';
 import '../raw-provider-sink-C46HDghv.js';
 import '../verdict-C9MlYujm.js';
 import '@ax-llm/ax';
 import '../store-C1YxJDEK.js';
-import '../red-team-BWdoyleI.js';
-import '../dataset-BbGkaN2I.js';
+import '../red-team-KmmiqBlY.js';
+import '../dataset-DS7ytHZU.js';
 import '../store-BcFXE6LG.js';
 import '../schema-m0gsnbt3.js';
 import '../pareto-E-pembql.js';
 import '../hosted/index.js';
-import '../insight-report-k0sRTzKg.js';
-import '../summary-report-C0nnxOD8.js';
+import '../insight-report-B4xrdwEK.js';
+import '../summary-report-QMZVe3P-.js';
 import '../failure-cluster-DH9Flgcf.js';
-import '../judge-calibration-0p2QcWNE.js';
+import '../judge-calibration-7C-IDmKr.js';
 import '../types-C7DGg5ex.js';
 import 'zod';
 
@@ -184,9 +184,19 @@ interface PlanEvalFixtureRunOptions<TArtifact = unknown> extends Pick<PlanCampai
 type EvalFixtureRunPlan = CampaignRunPlan & {
     fixtures: Array<Pick<EvalFixtureScenario, 'fixtureName' | 'fixturePath' | 'fingerprint'>>;
 };
+/** Walk `evalsDir` and return the relative name of every fixture directory (one containing an exact-case `PROMPT.md`). */
 declare function discoverEvalFixtures(evalsDir: string): string[];
+/**
+ * Load ONE fixture by name: reads `PROMPT.md` (plus `EVAL.ts`/`EVAL.tsx` and `package.json` under
+ * `vitest` validation) and content-fingerprints the full file set for cache identity.
+ */
 declare function loadEvalFixture(evalsDir: string, name: string, options?: EvalFixtureLoadOptions): EvalFixture;
+/** Load fixtures (all discovered, or just `names`) as campaign `Scenario`s tagged `eval-fixture`. */
 declare function loadEvalFixtureScenarios(evalsDir: string, options?: LoadEvalFixtureScenariosOptions): EvalFixtureScenario[];
+/**
+ * Dry-run planner for a fixture campaign: loads the scenarios, delegates to `planCampaignRun`,
+ * and returns the plan plus each fixture's name/path/fingerprint.
+ */
 declare function planEvalFixtureRun<TArtifact = unknown>(options: PlanEvalFixtureRunOptions<TArtifact>): EvalFixtureRunPlan;
 
 /**
@@ -467,10 +477,15 @@ interface FsLabeledScenarioStoreOptions {
     /** Test seam — override `Date.now()` for deterministic tests. */
     now?: () => number;
 }
+/** Typed rejection from a labeled-scenario store (bad provenance, rate limit, invalid sample args) — carries a stable string `code`. */
 declare class LabeledScenarioStoreError extends Error {
     readonly code: string;
     constructor(code: string, message: string);
 }
+/**
+ * Filesystem `LabeledScenarioStore`: appends one JSONL file per source with provenance and
+ * rate-limit guards. For tests, local dev, and small workloads — high-throughput lands in Turso.
+ */
 declare class FsLabeledScenarioStore implements LabeledScenarioStore {
     private readonly options;
     private readonly now;
@@ -583,6 +598,9 @@ interface FapoProposerOptions<TFindings = unknown> {
 }
 /** Build a FAPO policy proposer from level-specific candidate generators. */
 declare function fapoProposer<TFindings = unknown>(opts: FapoProposerOptions<TFindings>): SurfaceProposer<TFindings>;
+/**
+ * Scan a findings array and extract FAPO attribution signals — per-level counts and failure clusters used to decide which optimization level to escalate to next.
+ */
 declare function extractFapoAttributionSignals(findings: readonly unknown[]): FapoAttributionSignals;
 type JsonPrimitive = string | number | boolean | null;
 type JsonValue = JsonPrimitive | JsonValue[] | {
@@ -687,6 +705,9 @@ interface CompareProposersOptions<TScenario extends Scenario, TArtifact> extends
     /** CI confidence. Default 0.95. */
     confidence?: number;
 }
+/**
+ * Run a head-to-head lift benchmark across surface proposers on a shared holdout, returning per-proposer lift CIs and pairwise "who wins" verdicts.
+ */
 declare function compareProposers<TScenario extends Scenario, TArtifact>(opts: CompareProposersOptions<TScenario, TArtifact>): Promise<ProposerComparison>;
 /** Shared corpus + transport for the three built-in optimizer entries. */
 interface OptimizerEntryConfig<TScenario extends Scenario, TArtifact> {
@@ -750,6 +771,9 @@ interface FapoEntryConfig<TScenario extends Scenario, TArtifact> extends Optimiz
     /** FAPO policy knobs: scope, reviewer, plateau thresholds. */
     fapo?: Omit<FapoProposerOptions, 'proposers' | 'promptProposer' | 'parameterProposer' | 'structuralProposer'>;
 }
+/**
+ * Build a `ProposerEntry` that runs the full FAPO escalation policy (prompt → parameter → structural) as a single comparable optimizer entry.
+ */
 declare function fapoEscalationEntry<TScenario extends Scenario, TArtifact>(config: FapoEntryConfig<TScenario, TArtifact>, name?: string): ProposerEntry;
 
 /**
@@ -896,6 +920,9 @@ interface RunProfileMatrixResult<TArtifact, TScenario extends Scenario> {
     /** The raw per-profile campaign results, keyed by profile id. */
     campaigns: Record<string, CampaignResult<TArtifact, TScenario>>;
 }
+/**
+ * Profile × scenario matrix runner: fan N agent profiles across M scenarios, project each cell to a validated `RunRecord` with real token usage, and enforce the backend-integrity guard before returning.
+ */
 declare function runProfileMatrix<TScenario extends Scenario, TArtifact>(opts: RunProfileMatrixOptions<TScenario, TArtifact>): Promise<RunProfileMatrixResult<TArtifact, TScenario>>;
 
 /**
@@ -1151,6 +1178,9 @@ interface SkillOptProposer extends SurfaceProposer {
      *  acceptance/budget/buffer). Returns structured patches, NOT surfaces. */
     proposePatches(args: ProposePatchesArgs): Promise<SkillPatch[]>;
 }
+/**
+ * SkillOpt proposer: proposes bounded, anchored patch operations (add/delete/replace) on a skill document, conforming to both the patch-native `SkillOptProposer` and the generic `SurfaceProposer` interfaces.
+ */
 declare function skillOptProposer(opts: SkillOptProposerOptions): SkillOptProposer;
 /** Parse + validate the patch response. Throws `SkillPatchParseError` when the
  *  response is not valid JSON at all (a router/model failure the caller must
@@ -1160,6 +1190,9 @@ declare function skillOptProposer(opts: SkillOptProposerOptions): SkillOptPropos
 declare class SkillPatchParseError extends Error {
     constructor(message: string);
 }
+/**
+ * Parse a SkillOpt LLM response into validated `SkillPatch` objects, throwing `SkillPatchParseError` on malformed JSON and silently dropping ops that violate the edit budget.
+ */
 declare function parseSkillPatchResponse(raw: string, maxPatches: number, editBudget: number): SkillPatch[];
 
 /**
@@ -1254,6 +1287,9 @@ interface RunSkillOptResult {
      *  acceptance) the hill-climb ran. */
     totalCostUsd: number;
 }
+/**
+ * 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.
+ */
 declare function runSkillOpt<TScenario extends Scenario, TArtifact>(opts: RunSkillOptOptions<TScenario, TArtifact>): Promise<RunSkillOptResult>;
 
 /**
@@ -1287,6 +1323,9 @@ interface AceProposerOptions {
     /** Heading rendered above the bullets inside the block. */
     sectionHeading?: string;
 }
+/**
+ * Append-only context engineering proposer: grows a skill playbook by appending generation-tagged lessons without merging or overwriting prior entries.
+ */
 declare function aceProposer(opts?: AceProposerOptions): SurfaceProposer;
 
 /**
@@ -1399,6 +1438,10 @@ interface PolicyEditProposerOptions {
     /** Optional callback for audit UIs/tests that need rejected edit reasons. */
     onAdmission?: (admission: PolicyEditAdmission) => void;
 }
+/**
+ * `SurfaceProposer` that admission-checks typed analyst `PolicyEdit`s and applies each
+ * admitted edit to the current surface as one measured candidate.
+ */
 declare function policyEditProposer(opts?: PolicyEditProposerOptions): SurfaceProposer;
 
 /**
@@ -1517,6 +1560,7 @@ interface WorktreeAdapter {
     /** Remove the worktree (and its branch) — called for losing candidates. */
     discard(worktree: Worktree): Promise<void>;
 }
+/** Typed failure from a `WorktreeAdapter` operation (create/finalize/discard) — wraps the underlying git error as `cause`. */
 declare class WorktreeAdapterError extends Error {
     readonly cause?: unknown | undefined;
     constructor(message: string, cause?: unknown | undefined);
@@ -1531,6 +1575,9 @@ interface GitWorktreeAdapterOptions {
     /** Test seam — defaults to a real `git` runner. */
     git?: (args: string[], cwd: string) => string;
 }
+/**
+ * Git-backed `WorktreeAdapter`: creates isolated worktrees on fresh branches, commits agent changes, and discards losers.
+ */
 declare function gitWorktreeAdapter(opts: GitWorktreeAdapterOptions): WorktreeAdapter;
 /** Resolve a `CodeSurface`'s worktreeRef to a directory the measurement can
  *  run the worker in. A path ref is returned as-is; anything else is treated

package/dist/campaign/index.js

@@ -10,7 +10,7 @@ import {
   paretoPolicy,
   paretoSignificanceGate,
   runEval
-} from "../chunk-HV5PBTJF.js";
+} from "../chunk-Z7FK73R5.js";
 import {
   HARNESS_NATIVE_MODEL,
   agentProfileHash,
@@ -20,7 +20,7 @@ import {
   harnessAxisOf,
   llmJudge,
   verifyCompletion
-} from "../chunk-IXOV77YF.js";
+} from "../chunk-BBQILVBH.js";
 import {
   buildLoopProvenanceRecord,
   campaignBreakdown,
@@ -42,7 +42,7 @@ import {
   runOptimization,
   surfaceContentHash,
   surfaceHash
-} from "../chunk-NTVWIH24.js";
+} from "../chunk-ZKIL3VOD.js";
 import {
   assertRealBackend,
   contentHash,
@@ -53,7 +53,7 @@ import {
   runCampaign,
   summarizeBackendIntegrity,
   tangleTracesRoot
-} from "../chunk-XKA6ZGEY.js";
+} from "../chunk-3PFZBGMR.js";
 import {
   estimateCost,
   isModelPriced
@@ -63,43 +63,43 @@ import {
   applyPolicyEditToSurface,
   isPolicyEdit,
   policyEditsFromFindings
-} from "../chunk-IN3SHQML.js";
+} from "../chunk-LOJ2QVCE.js";
 import {
   AnalystRegistry,
   DEFAULT_TRACE_ANALYST_KINDS,
   createTraceAnalystKind
-} from "../chunk-FRI6RG3P.js";
-import "../chunk-AIGWQEME.js";
+} from "../chunk-2OGPXHOB.js";
+import "../chunk-N22ZO7FV.js";
 import {
   callLlm
-} from "../chunk-CWNP4DV4.js";
+} from "../chunk-FUCQVFMU.js";
 import "../chunk-45EEMHTC.js";
 import {
   eProcess,
   mulberry32,
   pairedBootstrap
-} from "../chunk-HRGUJTER.js";
+} from "../chunk-XMSYF4A7.js";
 import {
   analyzeTraces
-} from "../chunk-2MLIEQSN.js";
+} from "../chunk-RPDDVKI7.js";
 import "../chunk-GGE4NNQT.js";
 import {
   OtlpFileTraceStore
-} from "../chunk-YBIGNSCZ.js";
+} from "../chunk-LNQEP766.js";
 import "../chunk-PC4UYEBM.js";
 import {
   modelHasSnapshot,
   validateRunRecord
-} from "../chunk-2NSLDY4B.js";
+} from "../chunk-VK6HBGAE.js";
 import {
   buildAgentProfileCell
-} from "../chunk-ABOIVNXL.js";
+} from "../chunk-XJYR7XFV.js";
 import {
   canonicalize
 } from "../chunk-VSMTAMNK.js";
 import {
   AgentEvalError
-} from "../chunk-3BFEG2F6.js";
+} from "../chunk-ONWEPEDO.js";
 import "../chunk-PZ5AY32C.js";
 
 // src/campaign/analyst-surface.ts