@tangle-network/agent-eval 0.103.0 → 0.103.1

package/dist/{control-C8RmK9H4.d.ts → control-KofK3gfG.d.ts}

@@ -3,7 +3,7 @@ import { C as ControlEvalResult, a as ControlRunResult, h as ControlRuntimeConfi
 import { T as TraceEmitter } from './emitter-C2rqGH_l.js';
 import { F as FailureClass } from './schema-m0gsnbt3.js';
 import { T as TraceStore } from './store-BcFXE6LG.js';
-import { b as RunSplitTag, c as RunTokenUsage, R as RunRecord } from './run-record-MRdJ-Kq2.js';
+import { b as RunSplitTag, c as RunTokenUsage, R as RunRecord } from './run-record-CPfd1ARZ.js';
 
 interface ActionExecutionPolicy {
     allowedTypes?: string[];

package/dist/control.d.ts

@@ -1,4 +1,4 @@
-export { A as ActionExecutionPolicy, a as ActionPolicyDecision, C as ControlRunToRunRecordOptions, e as ProposeReviewConfig, f as ProposeReviewControlAction, g as ProposeReviewControlConfig, h as ProposeReviewControlResult, i as ProposeReviewControlState, j as ProposeReviewReport, p as RunEvidenceMetadata, s as controlRunToRunRecord, u as evaluateActionPolicy, x as runProposeReview, y as runProposeReviewAsControlLoop, z as scoreFromEvals } from './control-C8RmK9H4.js';
+export { A as ActionExecutionPolicy, a as ActionPolicyDecision, C as ControlRunToRunRecordOptions, e as ProposeReviewConfig, f as ProposeReviewControlAction, g as ProposeReviewControlConfig, h as ProposeReviewControlResult, i as ProposeReviewControlState, j as ProposeReviewReport, p as RunEvidenceMetadata, s as controlRunToRunRecord, u as evaluateActionPolicy, x as runProposeReview, y as runProposeReviewAsControlLoop, z as scoreFromEvals } from './control-KofK3gfG.js';
 export { c as ControlActionFailureMode, d as ControlActionOutcome, e as ControlBudget, f as ControlContext, g as ControlDecision, C as ControlEvalResult, a as ControlRunResult, h as ControlRuntimeConfig, i as ControlRuntimeError, j as ControlSeverity, b as ControlStep, k as ControlStopPolicies, S as StopDecision, l as allCriticalPassed, o as objectiveEval, r as runAgentControlLoop, s as stopOnNoProgress, m as stopOnRepeatedAction, n as subjectiveEval } from './control-runtime-Acf9CGhw.js';
 import './feedback-trajectory-BxY0cKfs.js';
 import './dataset-BbGkaN2I.js';
@@ -6,5 +6,5 @@ import './errors-CzMUYo7b.js';
 import './emitter-C2rqGH_l.js';
 import './schema-m0gsnbt3.js';
 import './store-BcFXE6LG.js';
-import './run-record-MRdJ-Kq2.js';
+import './run-record-CPfd1ARZ.js';
 import '@tangle-network/agent-interface';

package/dist/control.js

@@ -4,7 +4,7 @@ import {
   runProposeReview,
   runProposeReviewAsControlLoop,
   scoreFromEvals
-} from "./chunk-B2TMQM62.js";
+} from "./chunk-BHCFJGL4.js";
 import {
   allCriticalPassed,
   objectiveEval,
@@ -13,9 +13,9 @@ import {
   stopOnRepeatedAction,
   subjectiveEval
 } from "./chunk-YEHAEDUD.js";
-import "./chunk-2NSLDY4B.js";
+import "./chunk-J5MUFXEY.js";
 import "./chunk-TVVP3ZZQ.js";
-import "./chunk-ABOIVNXL.js";
+import "./chunk-AGYDMORK.js";
 import "./chunk-VSMTAMNK.js";
 import "./chunk-3BFEG2F6.js";
 import "./chunk-PZ5AY32C.js";

package/dist/{corpus-CiSzzLa5.d.ts → corpus-CysLCiK4.d.ts}

@@ -1,4 +1,4 @@
-import { R as RunRecord, b as RunSplitTag } from './run-record-MRdJ-Kq2.js';
+import { R as RunRecord, b as RunSplitTag } from './run-record-CPfd1ARZ.js';
 import { S as Span } from './schema-m0gsnbt3.js';
 import { T as TraceStore } from './store-BcFXE6LG.js';
 

package/dist/{default-registry-ZhqsTr4K.d.ts → default-registry-Ez4cxuZ4.d.ts}

@@ -1,6 +1,6 @@
 import { AxAIService } from '@ax-llm/ax';
-import { T as TraceAnalystKindSpec } from './kind-factory-D0nk7AKV.js';
-import { a as Analyst, b as AnalystContext, c as AnalystRunSummary, A as AnalystFinding, d as AnalystRunResult, C as ChatClient, e as AnalystRunInputs, f as AnalystRunEvent } from './types-Dz9cKF0g.js';
+import { T as TraceAnalystKindSpec } from './kind-factory-BLHxwX71.js';
+import { a as Analyst, b as AnalystContext, c as AnalystRunSummary, A as AnalystFinding, d as AnalystRunResult, C as ChatClient, e as AnalystRunInputs, f as AnalystRunEvent } from './types-DA9yj-Jd.js';
 
 /**
  * AnalystRegistry — orchestrate N analysts against one run.

package/dist/diagnose.d.ts

@@ -3,9 +3,9 @@ export { c as CounterfactualContext, d as CounterfactualResult } from './counter
 import { S as Span } from './schema-m0gsnbt3.js';
 import { T as TraceStore } from './store-BcFXE6LG.js';
 import { a as TrajectoryStep, T as Trajectory } from './trajectory-2TkpSEVh.js';
-import { h as AnalystSeverity, A as AnalystFinding } from './types-Dz9cKF0g.js';
-import { C as CorpusRecord } from './corpus-CiSzzLa5.js';
-import { R as RunRecord } from './run-record-MRdJ-Kq2.js';
+import { h as AnalystSeverity, A as AnalystFinding } from './types-DA9yj-Jd.js';
+import { C as CorpusRecord } from './corpus-CysLCiK4.js';
+import { R as RunRecord } from './run-record-CPfd1ARZ.js';
 import './emitter-C2rqGH_l.js';
 import './store-C1YxJDEK.js';
 import './types-C7DGg5ex.js';

package/dist/diagnose.js

@@ -10,12 +10,12 @@ import {
 } from "./chunk-45EEMHTC.js";
 import {
   confidenceInterval
-} from "./chunk-HRGUJTER.js";
+} from "./chunk-HYGRFL7C.js";
 import {
   validateRunRecord
-} from "./chunk-2NSLDY4B.js";
+} from "./chunk-J5MUFXEY.js";
 import "./chunk-TVVP3ZZQ.js";
-import "./chunk-ABOIVNXL.js";
+import "./chunk-AGYDMORK.js";
 import "./chunk-VSMTAMNK.js";
 import {
   ValidationError

package/dist/{gepa-DeyPTlvx.d.ts → gepa-COlCAkHN.d.ts}

@@ -1,4 +1,4 @@
-import { S as Scenario, C as CampaignResult, G as GateResult, D as DispatchFn, a as JudgeConfig, L as LabeledScenarioStore, d as CampaignTraceWriter, e as GenerationRecord, M as MutableSurface, P as ParetoParent, f as SurfaceProposer, g as Gate } from './types-DFI_Z-ZL.js';
+import { S as Scenario, C as CampaignResult, G as GateResult, D as DispatchFn, a as JudgeConfig, L as LabeledScenarioStore, d as CampaignTraceWriter, e as GenerationRecord, M as MutableSurface, P as ParetoParent, f as SurfaceProposer, g as Gate } from './types-Bihq6-a3.js';
 import { L as LlmClientOptions } from './llm-client-Bj7g0rqu.js';
 
 /**
@@ -45,6 +45,9 @@ interface OpenAutoPrResult {
     dryRun: boolean;
     reason: string;
 }
+/**
+ * Open a GitHub PR for a gate-approved surface promotion, attaching the manifest hash, gate verdict, and diff as the PR body.
+ */
 declare function openAutoPr<TArtifact, TScenario extends Scenario>(options: OpenAutoPrOptions<TArtifact, TScenario>): OpenAutoPrResult;
 
 /**
@@ -183,6 +186,9 @@ interface RunCampaignOptions<TScenario extends Scenario, TArtifact> {
         generation?: number;
     }) => string | undefined;
 }
+/**
+ * Core campaign orchestrator: fan scenarios through dispatch, score with judges, aggregate bootstrap CIs, and persist reproducible `CampaignResult` records.
+ */
 declare function runCampaign<TScenario extends Scenario, TArtifact>(opts: RunCampaignOptions<TScenario, TArtifact>): Promise<CampaignResult<TArtifact, TScenario>>;
 interface CampaignRunPlanCell {
     cellId: string;
@@ -303,7 +309,13 @@ interface RunOptimizationResult<TArtifact, TScenario extends Scenario> {
      *  NOT the winner is uniquely best on some scenario the winner loses on. */
     paretoFrontier: ParetoParent[];
 }
+/**
+ * Improvement loop body: N generations of propose → campaign → rank, maintaining a Pareto frontier and promoting the top-scoring candidates to the next generation.
+ */
 declare function runOptimization<TScenario extends Scenario, TArtifact>(opts: RunOptimizationOptions<TScenario, TArtifact>): Promise<RunOptimizationResult<TArtifact, TScenario>>;
+/**
+ * Short (16-char) sha256 fingerprint of a `MutableSurface`: hashes text content for prompt surfaces, or the worktree + base ref pair for code surfaces.
+ */
 declare function surfaceHash(surface: MutableSurface): string;
 
 /**
@@ -364,7 +376,13 @@ interface RunImprovementLoopResult<TArtifact, TScenario extends Scenario> extend
     promotedDiff: string;
     prResult?: ReturnType<typeof openAutoPr>;
 }
+/**
+ * Gated-promotion shell over `runOptimization`: scores the winner against the baseline on a holdout set, runs the release gate, and optionally opens a PR.
+ */
 declare function runImprovementLoop<TScenario extends Scenario, TArtifact>(opts: RunImprovementLoopOptions<TScenario, TArtifact>): Promise<RunImprovementLoopResult<TArtifact, TScenario>>;
+/**
+ * Default surface diff renderer: produces a unified baseline/winner text diff for prompt surfaces or a worktree-ref summary for code surfaces.
+ */
 declare function defaultRenderDiff(winnerSurface: MutableSurface, baselineSurface: MutableSurface): string;
 
 /**
@@ -443,6 +461,9 @@ interface GepaProposerOptions {
      *  composite first), to bound prompt size. Default 4. */
     combineMaxParents?: number;
 }
+/**
+ * GEPA reflective proposer: each generation reflects on the weakest scenarios and dimensions to produce targeted prompt rewrites, optionally combining Pareto-frontier parents.
+ */
 declare function gepaProposer(opts: GepaProposerOptions): SurfaceProposer;
 /** Extract H2 headings (`## Foo`) from a markdown surface. Exported for
  *  consumers building custom mutators that share the same invariant. */

package/dist/governance/index.d.ts

@@ -1,5 +1,5 @@
 import { c as DatasetManifest } from '../dataset-BbGkaN2I.js';
-import { a as CalibrationResult } from '../judge-calibration-0p2QcWNE.js';
+import { a as CalibrationResult } from '../judge-calibration-7C-IDmKr.js';
 import { b as OutcomeStore } from '../outcome-store-rnXLEqSn.js';
 import { d as RedTeamReport } from '../red-team-BWdoyleI.js';
 import { T as TraceStore } from '../store-BcFXE6LG.js';

package/dist/hosted/index.d.ts

@@ -1,13 +1,13 @@
-import { M as MutableSurface, j as GateDecision } from '../types-DFI_Z-ZL.js';
-import { I as InsightReport } from '../insight-report-k0sRTzKg.js';
-import '../run-record-MRdJ-Kq2.js';
+import { M as MutableSurface, j as GateDecision } from '../types-Bihq6-a3.js';
+import { I as InsightReport } from '../insight-report-DumCfEur.js';
+import '../run-record-CPfd1ARZ.js';
 import '@tangle-network/agent-interface';
 import '../errors-CzMUYo7b.js';
 import '../schema-m0gsnbt3.js';
-import '../summary-report-C0nnxOD8.js';
+import '../summary-report-Fc_YFJat.js';
 import '../failure-cluster-DH9Flgcf.js';
 import '../store-BcFXE6LG.js';
-import '../judge-calibration-0p2QcWNE.js';
+import '../judge-calibration-7C-IDmKr.js';
 
 /**
  * # Hosted-tier wire format — the schema that EVERY orchestrator (ours,

package/dist/{index-B-bFgiAF.d.ts → index-unCSYRJJ.d.ts}

@@ -1,4 +1,4 @@
-import { b as RunSplitTag } from './run-record-MRdJ-Kq2.js';
+import { b as RunSplitTag } from './run-record-CPfd1ARZ.js';
 
 /**
  * Shared types for the reference benchmark wrappers under

package/dist/index.d.ts

@@ -1,13 +1,13 @@
-export { A as ActionExecutionPolicy, a as ActionPolicyDecision, C as ControlRunToRunRecordOptions, L as LlmJsonCall, b as LlmReviewerConfig, P as ProposeFn, c as ProposeInput, d as ProposeOutput, e as ProposeReviewConfig, f as ProposeReviewControlAction, g as ProposeReviewControlConfig, h as ProposeReviewControlResult, i as ProposeReviewControlState, j as ProposeReviewReport, k as ProposeReviewShot, R as Review, l as ReviewFn, m as ReviewInput, n as ReviewMemoryEntry, o as ReviewMemoryStore, p as RunEvidenceMetadata, V as Verification, q as VerifyFn, r as controlFailureClassFromVerification, s as controlRunToRunRecord, t as createLlmReviewer, u as evaluateActionPolicy, v as inMemoryReviewStore, w as jsonlReviewStore, x as runProposeReview, y as runProposeReviewAsControlLoop, z as scoreFromEvals } from './control-C8RmK9H4.js';
-import { R as RunRecord, b as RunSplitTag } from './run-record-MRdJ-Kq2.js';
-export { f as AGENT_PROFILE_KINDS, g as AgentInterfaceProfileLike, A as AgentProfileCell, e as AgentProfileCellInput, h as AgentProfileCellSchemaVersion, i as AgentProfileCellValidationError, j as AgentProfileDimensionValue, k as AgentProfileHarness, a as AgentProfileJson, l as AgentProfileKind, m as AgentProfileSource, n as AgentProfileSourceInput, J as JudgeScoresRecord, d as RunJudgeMetadata, o as RunOutcome, p as RunRecordValidationError, c as RunTokenUsage, q as agentProfileCellHashMaterial, r as agentProfileCellKey, s as assertRunAgentProfileCell, t as buildAgentInterfaceProfileCell, u as buildAgentProfileCell, v as groupRunsByAgentProfileCell, w as isRunRecord, x as modelHasSnapshot, y as parseRunRecordSafe, z as requireAgentProfileCell, B as roundTripRunRecord, C as toAgentProfileJson, D as validateAgentProfileCell, E as validateRunRecord, F as verifyAgentProfileCell } from './run-record-MRdJ-Kq2.js';
-import { B as BehavioralMetrics } from './semantic-concept-judge-D-IlH5v1.js';
-export { x as ConceptComplexity, y as ConceptFinding, z as ConceptSpec, A as ConceptWeightStrategy, C as CreateAnalystAiConfig, E as DEFAULT_COMPLEXITY_WEIGHTS, D as DEFAULT_TRACE_ANALYST_KINDS, b as DiffPolicy, F as FAILURE_MODE_KIND_SPEC, e as FindingSubject, f as FindingSubjectKind, h as FindingsDiff, i as FindingsStore, I as IMPROVEMENT_KIND_SPEC, j as KNOWLEDGE_GAP_KIND_SPEC, k as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, G as SEMANTIC_CONCEPT_JUDGE_VERSION, l as SKILL_USAGE_ANALYST, a as SemanticConceptJudgeInput, S as SemanticConceptJudgeOptions, H as SemanticConceptJudgeResult, m as SkillUsageAnalyst, J as SuboptimalCode, L as SuboptimalSignal, M as computeTraceMetrics, r as createAnalystAi, N as createSemanticConceptJudge, s as defaultIsMaterial, t as diffFindings, O as runSemanticConceptJudge } from './semantic-concept-judge-D-IlH5v1.js';
-import { l as ChatRequest, p as CreateChatClientOpts } from './types-Dz9cKF0g.js';
-export { a as Analyst, b as AnalystContext, g as AnalystCost, A as AnalystFinding, i as AnalystInputKind, j as AnalystRequirements, f as AnalystRunEvent, e as AnalystRunInputs, d as AnalystRunResult, c as AnalystRunSummary, h as AnalystSeverity, k as ChatCallOpts, C as ChatClient, m as ChatResponse, n as ChatTransport, o as CliBridgeTransportOpts, 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';
-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 { C as CreateTraceAnalystKindOpts, a as RawAnalystFinding, c as TraceAnalystGolden, T as TraceAnalystKindSpec, d as createTraceAnalystKind, 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';
+export { A as ActionExecutionPolicy, a as ActionPolicyDecision, C as ControlRunToRunRecordOptions, L as LlmJsonCall, b as LlmReviewerConfig, P as ProposeFn, c as ProposeInput, d as ProposeOutput, e as ProposeReviewConfig, f as ProposeReviewControlAction, g as ProposeReviewControlConfig, h as ProposeReviewControlResult, i as ProposeReviewControlState, j as ProposeReviewReport, k as ProposeReviewShot, R as Review, l as ReviewFn, m as ReviewInput, n as ReviewMemoryEntry, o as ReviewMemoryStore, p as RunEvidenceMetadata, V as Verification, q as VerifyFn, r as controlFailureClassFromVerification, s as controlRunToRunRecord, t as createLlmReviewer, u as evaluateActionPolicy, v as inMemoryReviewStore, w as jsonlReviewStore, x as runProposeReview, y as runProposeReviewAsControlLoop, z as scoreFromEvals } from './control-KofK3gfG.js';
+import { R as RunRecord, b as RunSplitTag } from './run-record-CPfd1ARZ.js';
+export { f as AGENT_PROFILE_KINDS, g as AgentInterfaceProfileLike, A as AgentProfileCell, e as AgentProfileCellInput, h as AgentProfileCellSchemaVersion, i as AgentProfileCellValidationError, j as AgentProfileDimensionValue, k as AgentProfileHarness, a as AgentProfileJson, l as AgentProfileKind, m as AgentProfileSource, n as AgentProfileSourceInput, J as JudgeScoresRecord, d as RunJudgeMetadata, o as RunOutcome, p as RunRecordValidationError, c as RunTokenUsage, q as agentProfileCellHashMaterial, r as agentProfileCellKey, s as assertRunAgentProfileCell, t as buildAgentInterfaceProfileCell, u as buildAgentProfileCell, v as groupRunsByAgentProfileCell, w as isRunRecord, x as modelHasSnapshot, y as parseRunRecordSafe, z as requireAgentProfileCell, B as roundTripRunRecord, C as toAgentProfileJson, D as validateAgentProfileCell, E as validateRunRecord, F as verifyAgentProfileCell } from './run-record-CPfd1ARZ.js';
+import { B as BehavioralMetrics } from './semantic-concept-judge-C6mDEBIo.js';
+export { x as ConceptComplexity, y as ConceptFinding, z as ConceptSpec, A as ConceptWeightStrategy, C as CreateAnalystAiConfig, E as DEFAULT_COMPLEXITY_WEIGHTS, D as DEFAULT_TRACE_ANALYST_KINDS, b as DiffPolicy, F as FAILURE_MODE_KIND_SPEC, e as FindingSubject, f as FindingSubjectKind, h as FindingsDiff, i as FindingsStore, I as IMPROVEMENT_KIND_SPEC, j as KNOWLEDGE_GAP_KIND_SPEC, k as KNOWLEDGE_POISONING_KIND_SPEC, P as PersistedFinding, G as SEMANTIC_CONCEPT_JUDGE_VERSION, l as SKILL_USAGE_ANALYST, a as SemanticConceptJudgeInput, S as SemanticConceptJudgeOptions, H as SemanticConceptJudgeResult, m as SkillUsageAnalyst, J as SuboptimalCode, L as SuboptimalSignal, M as computeTraceMetrics, r as createAnalystAi, N as createSemanticConceptJudge, s as defaultIsMaterial, t as diffFindings, O as runSemanticConceptJudge } from './semantic-concept-judge-C6mDEBIo.js';
+import { l as ChatRequest, p as CreateChatClientOpts } from './types-DA9yj-Jd.js';
+export { a as Analyst, b as AnalystContext, g as AnalystCost, A as AnalystFinding, i as AnalystInputKind, j as AnalystRequirements, f as AnalystRunEvent, e as AnalystRunInputs, d as AnalystRunResult, c as AnalystRunSummary, h as AnalystSeverity, k as ChatCallOpts, C as ChatClient, m as ChatResponse, n as ChatTransport, o as CliBridgeTransportOpts, 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-DA9yj-Jd.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-Ez4cxuZ4.js';
+export { C as CreateTraceAnalystKindOpts, a as RawAnalystFinding, c as TraceAnalystGolden, T as TraceAnalystKindSpec, d as createTraceAnalystKind, r as renderPriorFindings } from './kind-factory-BLHxwX71.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-nUhuFtLF.js';
 import { TCloud } from '@tangle-network/tcloud';
 import { B as BenchmarkRunnerConfig, S as Scenario, c as BenchmarkReport, P as ProductClientConfig, C as CheckResult, T as TestResult, d as PersonaConfig, D as DriverResult, e as DriverState, b as JudgeFn, f as CollectedArtifacts, g as ScenarioResult, h as TurnMetrics, i as ScenarioFile, j as CompletionCriterion } from './types-C7DGg5ex.js';
 export { A as ArtifactCheck, k as ArtifactResult, E as EvalResult, F as FeedbackPattern, l as JudgeConfig, a as JudgeInput, m as JudgeRubric, J as JudgeScore, n as PersonaRigor, R as RouteMap, o as RubricDimension, p as Turn, q as TurnResult } from './types-C7DGg5ex.js';
@@ -18,20 +18,20 @@ import { A as AgentEvalError, J as JudgeError, a as ConfigError } from './errors
 export { b as AgentEvalErrorCode, C as CaptureIntegrityError, N as NotFoundError, R as ReplayError, V as ValidationError, c as VerificationError } from './errors-CzMUYo7b.js';
 import { b as FeedbackLabel, F as FeedbackTrajectoryStore, a as FeedbackTrajectory } from './feedback-trajectory-BxY0cKfs.js';
 export { c as FeedbackArtifactType, d as FeedbackAttempt, e as FeedbackLabelKind, f as FeedbackLabelSource, g as FeedbackOptimizerRow, h as FeedbackOutcome, i as FeedbackReplayAdapter, j as FeedbackReplayResult, k as FeedbackSeverity, l as FeedbackSplitPolicy, m as FeedbackTask, n as FeedbackTrajectoryFilter, o as FileSystemFeedbackTrajectoryStore, I as InMemoryFeedbackTrajectoryStore, P as PreferenceMemoryEntry, p as ProposedSideEffect, q as assignFeedbackSplit, r as controlRunToFeedbackTrajectory, s as createFeedbackTrajectory, t as feedbackTrajectoriesToDatasetScenarios, u as feedbackTrajectoriesToOptimizerRows, v as feedbackTrajectoryToDatasetScenario, w as feedbackTrajectoryToOptimizerRow, x as parseFeedbackTrajectoriesJsonl, y as renderPreferenceMemoryMarkdown, z as replayFeedbackTrajectories, A as replayFeedbackTrajectory, B as serializeFeedbackTrajectoriesJsonl, C as summarizePreferenceMemory, D as withAssignedFeedbackSplit } from './feedback-trajectory-BxY0cKfs.js';
-import { b as CorrectnessChecker } from './pre-registration-mWG2w8d-.js';
-export { A as ArtifactCheckArtifact, d as ArtifactEventLike, e as ArtifactValidator, f as BackendIntegrityError, B as BackendIntegrityReport, C as CompletionRequirement, a as CompletionVerdict, H as HypothesisManifest, g as HypothesisResult, h as LlmCorrectnessCheckerOpts, L as LlmJudgeDimension, c as LlmJudgeOptions, i as ProducedProposal, P as ProducedState, j as ProposalEventLike, k as RequirementCheck, R as RuntimeEventLike, m as SatisfiedBy, S as SignedManifest, n as SignedManifestAlgo, T as TaskGold, o as ToolCallEventLike, V as ValidationContext, p as ValidationIssue, q as ValidationResult, r as assertRealBackend, s as byteLengthRange, t as canonicalize, u as completionVerdict, v as composeValidators, w as containsAll, x as createLlmCorrectnessChecker, y as createTokenRecallChecker, z as evaluateHypothesis, D as extractProducedState, E as hashJson, F as jsonHasKeys, l as llmJudge, G as parseCorrectnessResponse, I as regexMatch, J as signManifest, K as summarizeBackendIntegrity, M as verifyCompletion, N as verifyManifest } from './pre-registration-mWG2w8d-.js';
+import { b as CorrectnessChecker } from './pre-registration-CUOSGAZK.js';
+export { A as ArtifactCheckArtifact, d as ArtifactEventLike, e as ArtifactValidator, f as BackendIntegrityError, B as BackendIntegrityReport, C as CompletionRequirement, a as CompletionVerdict, H as HypothesisManifest, g as HypothesisResult, h as LlmCorrectnessCheckerOpts, L as LlmJudgeDimension, c as LlmJudgeOptions, i as ProducedProposal, P as ProducedState, j as ProposalEventLike, k as RequirementCheck, R as RuntimeEventLike, m as SatisfiedBy, S as SignedManifest, n as SignedManifestAlgo, T as TaskGold, o as ToolCallEventLike, V as ValidationContext, p as ValidationIssue, q as ValidationResult, r as assertRealBackend, s as byteLengthRange, t as canonicalize, u as completionVerdict, v as composeValidators, w as containsAll, x as createLlmCorrectnessChecker, y as createTokenRecallChecker, z as evaluateHypothesis, D as extractProducedState, E as hashJson, F as jsonHasKeys, l as llmJudge, G as parseCorrectnessResponse, I as regexMatch, J as signManifest, K as summarizeBackendIntegrity, M as verifyCompletion, N as verifyManifest } from './pre-registration-CUOSGAZK.js';
 export { DataAcquisitionPlan, KnowledgeAcquisitionMode, KnowledgeBundle, KnowledgeFallbackPolicy, KnowledgeFreshness, KnowledgeImportance, KnowledgeReadinessReport, KnowledgeRecommendedAction, KnowledgeRequirement, KnowledgeRequirementCategory, KnowledgeResponsibleSurface, KnowledgeSensitivity, ScoreKnowledgeReadinessOptions, UserQuestion, acquisitionPlansForKnowledgeGaps, blockingKnowledgeEval, knowledgeReadinessTracePayload, scoreKnowledgeReadiness, userQuestionsForKnowledgeGaps } from './knowledge/index.js';
-import { h as ReleaseConfidenceThresholds, f as ReleaseConfidenceScorecard } from './release-report-BQ1Ziyu-.js';
-export { A as ActionableSideInfo, o as AsiSeverity, B as BootstrapOptions, a as BootstrapResult, J as JudgeReplayGateArgs, R as ReleaseConfidenceAxis, b as ReleaseConfidenceAxisName, c as ReleaseConfidenceInput, d as ReleaseConfidenceIssue, e as ReleaseConfidenceMetrics, g as ReleaseConfidenceStatus, i as ReleaseTraceEvidence, j as RenderReleaseReportOptions, V as Verdict, k as assertReleaseConfidence, l as bootstrapCi, m as evaluateReleaseConfidence, n as judgeReplayGate, r as renderReleaseReport } from './release-report-BQ1Ziyu-.js';
-export { c as CliffsMagnitude, d as CorpusAgreementOptions, e as CorpusAgreementPerDimension, C as CorpusAgreementReport, f as CorpusScoreRecord, g as EProcess, h as EProcessOptions, E as EProcessState, i as EProcessStep, M as McNemarResult, P as PairedBootstrapOptions, a as PairedBootstrapResult, j as ProportionInterval, R as RiskDifferenceResult, W as WeightedCompositeInput, k as WeightedCompositeResult, b as benjaminiHochberg, l as bonferroni, m as cliffsDelta, n as cohensD, o as confidenceInterval, q as corpusInterRaterAgreement, r as corpusInterRaterAgreementFromJudgeScores, s as eProcess, t as interRaterReliability, u as interpretCliffs, v as mannWhitneyU, x as mcnemar, y as mcnemarPower, z as mcnemarRequiredN, A as mulberry32, B as normalizeScores, p as pairedBootstrap, D as pairedMde, F as pairedRiskDifference, G as pairedTTest, H as partialCredit, I as passAtK, J as pearsonR, K as ranks, L as requiredSampleSize, N as spearmanR, O as weightedComposite, Q as weightedMean, w as wilcoxonSignedRank, S as wilson } from './statistics-xP-cWc5k.js';
+import { h as ReleaseConfidenceThresholds, f as ReleaseConfidenceScorecard } from './release-report-DkaCZ9k4.js';
+export { A as ActionableSideInfo, o as AsiSeverity, B as BootstrapOptions, a as BootstrapResult, J as JudgeReplayGateArgs, R as ReleaseConfidenceAxis, b as ReleaseConfidenceAxisName, c as ReleaseConfidenceInput, d as ReleaseConfidenceIssue, e as ReleaseConfidenceMetrics, g as ReleaseConfidenceStatus, i as ReleaseTraceEvidence, j as RenderReleaseReportOptions, V as Verdict, k as assertReleaseConfidence, l as bootstrapCi, m as evaluateReleaseConfidence, n as judgeReplayGate, r as renderReleaseReport } from './release-report-DkaCZ9k4.js';
+export { c as CliffsMagnitude, d as CorpusAgreementOptions, e as CorpusAgreementPerDimension, C as CorpusAgreementReport, f as CorpusScoreRecord, g as EProcess, h as EProcessOptions, E as EProcessState, i as EProcessStep, M as McNemarResult, P as PairedBootstrapOptions, a as PairedBootstrapResult, j as ProportionInterval, R as RiskDifferenceResult, W as WeightedCompositeInput, k as WeightedCompositeResult, b as benjaminiHochberg, l as bonferroni, m as cliffsDelta, n as cohensD, o as confidenceInterval, q as corpusInterRaterAgreement, r as corpusInterRaterAgreementFromJudgeScores, s as eProcess, t as interRaterReliability, u as interpretCliffs, v as mannWhitneyU, x as mcnemar, y as mcnemarPower, z as mcnemarRequiredN, A as mulberry32, B as normalizeScores, p as pairedBootstrap, D as pairedMde, F as pairedRiskDifference, G as pairedTTest, H as partialCredit, I as passAtK, J as pearsonR, K as ranks, L as requiredSampleSize, N as spearmanR, O as weightedComposite, Q as weightedMean, w as wilcoxonSignedRank, S as wilson } from './statistics-D88peojY.js';
 import { OtelExporter, OtelExportConfig } from './traces.js';
 export { CaptureFetchContext, CaptureFetchOptions, ExportableSpan, ExtractedUsage, FlattenOtlpOptions, LLM_CACHED_TOKENS, LLM_CACHED_TOKEN_ATTR_KEYS, LLM_COST_ATTR_KEYS, LLM_COST_USD, LLM_INPUT_TOKENS, LLM_INPUT_TOKEN_ATTR_KEYS, LLM_MODEL_ATTR_KEYS, LLM_MODEL_NAME, LLM_OUTPUT_TOKENS, LLM_OUTPUT_TOKEN_ATTR_KEYS, OPENINFERENCE_SPAN_KIND, OTEL_AGENT_EVAL_SCOPE, OtlpExport, OtlpFileTraceStore, OtlpFileTraceStoreOptions, OtlpFlatLine, OtlpResourceSpans, OtlpSpan, OtlpToRunRecordsOptions, OtlpTraceRunRecord, ProjectedOtlpSpan, ReplayCache, ReplayCacheEntry, ReplayCacheMissError, ReplayCacheStats, ReplayFetchOptions, SPAN_KIND_ATTR_KEYS, SpanNotFoundError, TOOL_NAME, TOOL_NAME_ATTR_KEYS, TRACE_ANALYST_ACTOR_DESCRIPTION, TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, TRACE_ANALYST_SUBAGENT_DESCRIPTION, TraceAggregate, TraceAnalystHookOptions, TraceFileMissingError, TraceInsightContext, TraceInsightFinding, TraceInsightPanelRole, TraceInsightPromptInput, TraceInsightQualityGate, TraceInsightQuestion, TraceInsightReadiness, TraceInsightSuite, TraceInsightTask, TraceNotFoundError, TraceStoreSource, TraceStoreToOtlpOptions, TracesToOtlpResult, asNumber, asString, buildTraceAnalystTools, buildTraceInsightContext, buildTraceInsightPrompt, captureFetchToRawSink, convertTraceStoresToOtlp, createOtelExporter, createOtelTracingStore, createReplayFetch, defaultTraceInsightPanel, describeTraceInsightScope, domainEvidencePattern, exportRunAsOtlp, extractOtlpAttributes, extractUsage, extractUsageFromResponse, extractUsageFromSse, firstNumberAttr, firstStringAttr, flattenOtlpExportToNdjson, inferDomainKeywords, inferOtlpKind, iterateRawCalls, otelRunCompleteHook, otlpToRunRecords, otlpToTraceRunRecords, planTraceInsightQuestions, projectOtlpFlatLine, readOtlpStatus, scoreTraceInsightReadiness, stringField, tokenizeDomainWords, traceAnalystFunctionGroup, traceAnalystOnRunComplete, traceSpanKindToOpenInferenceKind } from './traces.js';
 import { a as AnalyzeTracesInput, A as AnalyzeTracesOptions, b as AnalyzeTracesResult } from './analyst-C8HHvfJp.js';
 export { c as AnalyzeTracesTurnSnapshot, d as analyzeTraces } from './analyst-C8HHvfJp.js';
 import { a as TraceAnalystSpan } from './store-C1YxJDEK.js';
 export { D as DEFAULT_TRACE_ANALYST_BUDGETS, b as DatasetOverview, E as ErrorCluster, Q as QueryTracesPage, S as SearchSpanResult, c as SearchTraceResult, d as SpanMatchRecord, e as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, T as TraceAnalysisStore, f as TraceAnalystByteBudgets, g as TraceAnalystFilters, h as TraceAnalystSpanKind, i as TraceAnalystSpanStatus, j as TraceAnalystTraceSummary, V as ViewSpansResult, k as ViewTraceOversized, l as ViewTraceResult } from './store-C1YxJDEK.js';
-import { a as JudgeConfig, J as JudgeScore, S as Scenario$1, g as Gate } from './types-DFI_Z-ZL.js';
-import { A as AnalyzeRunsOptions } from './analyze-runs-Cd-A_K4l.js';
+import { a as JudgeConfig, J as JudgeScore, S as Scenario$1, g as Gate } from './types-Bihq6-a3.js';
+import { A as AnalyzeRunsOptions } from './analyze-runs-rF2eGxZ_.js';
 import { S as SteeringBundle } from './harness-optimizer-mOl9XX_O.js';
 export { D as DEFAULT_HARNESS_OBJECTIVES, H as HarnessAdapter, a as HarnessExperimentConfig, b as HarnessExperimentResult, c as HarnessIntervention, d as HarnessRunRequest, e as HarnessRunResult, f as HarnessScenario, g as HarnessSelection, h as HarnessVariant, i as HarnessVariantReport, M as MeasurementPolicy, j as SteeringDelta, k as SteeringRolePrompt, W as WorkflowTopology, m as mergeSteeringBundle, r as renderSteeringText, l as runHarnessExperiment, s as selectHarnessVariant, n as summarizeHarnessResults } from './harness-optimizer-mOl9XX_O.js';
 import { S as SandboxDriver, H as HarnessConfig, a as SandboxHarnessResult } from './test-graded-scenario-DeODGLra.js';
@@ -41,13 +41,13 @@ export { D as DEFAULT_RUN_SCORE_WEIGHTS, c as RunCritic, d as RunCriticOptions,
 import { T as TraceEmitter } from './emitter-C2rqGH_l.js';
 export { R as RunCompleteHook, a as RunCompleteHookContext, S as SpanHandle, b as TraceEmitterOptions, l as llmSpanFromProvider } from './emitter-C2rqGH_l.js';
 export { b as RunIntegrityError, R as RunIntegrityExpectations, c as RunIntegrityIssue, d as RunIntegrityIssueCode, a as RunIntegrityReport, e as assertRunCaptured, t as throwIfRunIncomplete } from './integrity-D2t12mMw.js';
-export { a as aggregateLlm, b as argHash, g as groupBy, j as judgeSpans, l as llmSpans, r as runFailureClass, c as runsForScenario, t as toolSpans } from './query-B7GGjRox.js';
+export { a as aggregateLlm, b as argHash, g as groupBy, j as judgeSpans, l as llmSpans, r as runFailureClass, c as runsForScenario, t as toolSpans } from './query-0aTmbmQe.js';
 export { F as FileSystemRawProviderSink, a as FileSystemRawProviderSinkOptions, I as InMemoryRawProviderSink, b as InMemoryRawProviderSinkOptions, N as NoopRawProviderSink, P as ProviderRedactor, c as RawProviderDirection, d as RawProviderEvent, R as RawProviderSink, e as RawProviderSinkFilter, f as defaultProviderRedactor, p as providerFromBaseUrl } from './raw-provider-sink-C46HDghv.js';
 export { D as DEFAULT_REDACTION_RULES, b as REDACTION_VERSION, a as RedactionReport, R as RedactionRule, r as redactString, c as redactValue } from './redact-B40YG2M_.js';
 import { T as TraceStore, R as RunFilter } from './store-BcFXE6LG.js';
 export { E as EventFilter, F as FileSystemTraceStore, a as FileSystemTraceStoreOptions, I as InMemoryTraceStore, S as SpanFilter } from './store-BcFXE6LG.js';
 export { D as DEFAULT_FAILURE_RULES, b as FailureClassification, c as FailureContext, d as FailureRule, e as classifyFailure } from './failure-cluster-DH9Flgcf.js';
-export { P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection, b as RuntimeTrajectoryEvidenceSummary, c as RuntimeTrajectoryHookEvent, R as RuntimeTrajectoryRecord, d as RuntimeTrajectoryRunRecord, p as parseRuntimeTrajectoryHookEvent, e as projectRuntimeTrajectoryEvidence } from './runtime-trajectory-8w0_jmtR.js';
+export { P as ProjectRuntimeTrajectoryEvidenceOptions, a as RuntimeTrajectoryEvidenceProjection, b as RuntimeTrajectoryEvidenceSummary, c as RuntimeTrajectoryHookEvent, R as RuntimeTrajectoryRecord, d as RuntimeTrajectoryRunRecord, p as parseRuntimeTrajectoryHookEvent, e as projectRuntimeTrajectoryEvidence } from './runtime-trajectory-DZ8ei-Jo.js';
 import { a as BaselineReport } from './baseline-Bbid3WoO.js';
 export { B as BaselineOptions, M as MetricSamples, b as MetricVerdict, T as ToolStats, d as ToolUseMetrics, e as ToolUseOptions, f as compareToBaseline, c as computeToolUseMetrics, i as iqr, w as welchsTTest } from './baseline-Bbid3WoO.js';
 import { a as TrajectoryStep, T as Trajectory } from './trajectory-2TkpSEVh.js';
@@ -61,26 +61,26 @@ export { S as SeriesConvergenceOptions, a as SeriesConvergenceResult, b as analy
 import { D as DefaultVerdict } from './verdict-C9MlYujm.js';
 import { a as DatasetScenario, b as Dataset } from './dataset-BbGkaN2I.js';
 export { d as DatasetDifficulty, c as DatasetManifest, e as DatasetProvenance, D as DatasetSplit, H as HoldoutLockedError, S as SliceOptions, h as hashScenarios } from './dataset-BbGkaN2I.js';
-export { a as CalibrationResult, c as CandidateScore, C as ContinuousAgreement, d as ContinuousAgreementOptions, b as ContinuousCalibrationResult, G as GoldenItem, P as PositionalBiasResult, S as SelfPreferenceResult, V as VerbosityBiasResult, e as calibrateJudge, f as calibrateJudgeContinuous, g as continuousAgreement, p as positionalBias, s as selfPreference, v as verbosityBias } from './judge-calibration-0p2QcWNE.js';
+export { a as CalibrationResult, c as CandidateScore, C as ContinuousAgreement, d as ContinuousAgreementOptions, b as ContinuousCalibrationResult, G as GoldenItem, P as PositionalBiasResult, S as SelfPreferenceResult, V as VerbosityBiasResult, e as calibrateJudge, f as calibrateJudgeContinuous, g as continuousAgreement, p as positionalBias, s as selfPreference, v as verbosityBias } from './judge-calibration-7C-IDmKr.js';
 export { D as DEFAULT_RED_TEAM_CORPUS, R as RedTeamCase, a as RedTeamCategory, b as RedTeamFinding, c as RedTeamPayload, d as RedTeamReport, r as redTeamDataset, e as redTeamReport, s as scoreRedTeamOutput, t as toolNamesForRun } from './red-team-BWdoyleI.js';
 export { c as CounterfactualContext, C as CounterfactualMutation, d as CounterfactualResult, b as CounterfactualRunner, a as attributeCounterfactuals, r as runCounterfactual } from './counterfactual-DlOz8PBx.js';
 import { a as PrmGrader } from './rubric-Cc6UHvUb.js';
 export { EuRiskClass, GovernanceContext, GovernanceFinding, GovernanceReport, UseCaseSignals, classifyEuAiRisk, euAiActReport, nistAiRmfReport, renderMarkdown, soc2Report, summarize } from './governance/index.js';
-import { b as Layer, S as Severity, L as LayerResult, c as VerifyContext } from './multi-layer-verifier-DUZXrPDA.js';
-export { F as Finding, d as LayerStatus, M as MultiLayerVerifier, a as VerificationReport, V as VerifyOptions, g as gradeSemanticStatus } from './multi-layer-verifier-DUZXrPDA.js';
+import { b as Layer, S as Severity, L as LayerResult, c as VerifyContext } from './multi-layer-verifier-CI4jdX-q.js';
+export { F as Finding, d as LayerStatus, M as MultiLayerVerifier, a as VerificationReport, V as VerifyOptions, g as gradeSemanticStatus } from './multi-layer-verifier-CI4jdX-q.js';
 import { L as LlmClientOptions } from './llm-client-Bj7g0rqu.js';
 export { d as LlmCallError, b as LlmCallRequest, c as LlmCallResult, e as LlmClient, f as LlmMessage, g as LlmRouteAssertionError, a as LlmRouteRequirements, h as LlmUsage, i as assertLlmRoute, j as backoffMs, k as callLlm, l as callLlmJson, m as isTransientLlmError, p as probeLlm, s as stripFencedJson } from './llm-client-Bj7g0rqu.js';
-export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, c as BenchmarkEvaluation, d as benchmarkDeterministicSplit, i as benchmarks } from './index-B-bFgiAF.js';
-export { C as CallbackResearcher, d as CallbackResearcherOptions, e as CampaignFactoryParams, f as CampaignIntegrityPolicy, g as CampaignRunContext, h as CampaignRunOutcome, i as CampaignRunner, j as CampaignScenario, k as CampaignVariant, c as EvalCampaignOptions, b as EvalCampaignResult, E as ExperimentPlan, a as ExperimentResult, l as FailedRun, F as FailureMode, N as NoopResearcher, R as Researcher, S as SteeringChange, r as runEvalCampaign } from './researcher-B_ODTAJs.js';
-export { G as GainDistributionBin, a as GainDistributionFigureSpec, b as GainDistributionOptions, m as GateDecision, n as GateEvidence, H as HeldOutGate, o as HeldOutGateConfig, q as HeldOutGateRejectionCode, P as ParetoFigureSpec, c as ParetoPoint, R as RESEARCH_REPORT_HARD_PAIR_FLOOR, d as ResearchReport, e as ResearchReportCandidate, f as ResearchReportDecision, g as ResearchReportMethodology, h as ResearchReportOptions, i as ResearchReportRecommendation, S as SummaryTable, j as SummaryTableOptions, k as SummaryTableRow, l as gainHistogram, p as paretoChart, r as researchReport, s as summaryTable } from './summary-report-C0nnxOD8.js';
+export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, c as BenchmarkEvaluation, d as benchmarkDeterministicSplit, i as benchmarks } from './index-unCSYRJJ.js';
+export { C as CallbackResearcher, d as CallbackResearcherOptions, e as CampaignFactoryParams, f as CampaignIntegrityPolicy, g as CampaignRunContext, h as CampaignRunOutcome, i as CampaignRunner, j as CampaignScenario, k as CampaignVariant, c as EvalCampaignOptions, b as EvalCampaignResult, E as ExperimentPlan, a as ExperimentResult, l as FailedRun, F as FailureMode, N as NoopResearcher, R as Researcher, S as SteeringChange, r as runEvalCampaign } from './researcher-SDAezfML.js';
+export { G as GainDistributionBin, a as GainDistributionFigureSpec, b as GainDistributionOptions, m as GateDecision, n as GateEvidence, H as HeldOutGate, o as HeldOutGateConfig, q as HeldOutGateRejectionCode, P as ParetoFigureSpec, c as ParetoPoint, R as RESEARCH_REPORT_HARD_PAIR_FLOOR, d as ResearchReport, e as ResearchReportCandidate, f as ResearchReportDecision, g as ResearchReportMethodology, h as ResearchReportOptions, i as ResearchReportRecommendation, S as SummaryTable, j as SummaryTableOptions, k as SummaryTableRow, l as gainHistogram, p as paretoChart, r as researchReport, s as summaryTable } from './summary-report-Fc_YFJat.js';
 export { L as LockedJsonlAppender } from './testing-C21CHsq2.js';
 export { I as InterimReleaseConfidence, a as InterimReleaseConfidenceInput, P as PairedEvalueOptions, b as PairedEvalueSequence, c as PairedEvalueStep, S as SequentialDecision, e as evaluateInterimReleaseConfidence, p as pairedEvalueSequence } from './sequential-5iSVfzl2.js';
-import { j as GepaProposerConstraints, b as RunImprovementLoopResult } from './gepa-DeyPTlvx.js';
+import { j as GepaProposerConstraints, b as RunImprovementLoopResult } from './gepa-COlCAkHN.js';
 export { IntegrityResult, IntegrityViolation, JourneySpec, PerfBaseline, PerfGateResult, PerfRegression, PerfScenario, PerfStat, ScenarioAxes, assertRecordIntegrity, checkRecordIntegrity, expandMatrix, gatePerf, scenarioKey, summarizeRecords } from './perf/index.js';
 export { AgentProfileRuntimeReceipt, ProductBenchmarkArm, ProductBenchmarkArtifactPaths, ProductBenchmarkBudgets, ProductBenchmarkExportOptions, ProductBenchmarkExportResult, ProductBenchmarkManifest, ProductBenchmarkProfileRef, ProductBenchmarkRecord, ProductBenchmarkRepoRef, ProductBenchmarkRunInput, ProductBenchmarkScenario, ProductBenchmarkSingleRunExportOptions, ProductBenchmarkSplit, ProductBenchmarkSubstrateVersions, ProductBenchmarkValidationReport, RuntimeResolution, assertProductBenchmarkRun, buildProductBenchmarkManifest, exportProductBenchmark, exportProductBenchmarkRuns, findProductBenchmarkArtifacts, productBenchmarkIntegrityFailures, productBenchmarkMutableSurfaces, productBenchmarkRepoIdentity, productBenchmarkSplits, readProductBenchmarkManifest, readProductBenchmarkRecords, runRecordToProductBenchmarkRecord, validateProductBenchmarkManifest, validateProductBenchmarkRecord, validateProductBenchmarkRun } from './product-benchmark/index.js';
 import '@ax-llm/ax';
 import 'zod';
-import './insight-report-k0sRTzKg.js';
+import './insight-report-DumCfEur.js';
 import './outcome-store-rnXLEqSn.js';
 
 /**
@@ -1554,9 +1554,13 @@ declare class JudgeRunner {
     run(spec: SandboxJudgeSpec): Promise<SandboxJudgeResult>;
 }
 declare function runJudgeFleet(specs: SandboxJudgeSpec[], options?: JudgeFleetOptions): Promise<SandboxJudgeResult[]>;
+/** Build a `SandboxJudgeSpec` that scores whether the harness compiles without errors. */
 declare function compilerJudge(id: string, config: HarnessConfig): SandboxJudgeSpec;
+/** Build a `SandboxJudgeSpec` that scores the harness by its test-suite pass rate. */
 declare function testJudge(id: string, config: HarnessConfig): SandboxJudgeSpec;
+/** Build a `SandboxJudgeSpec` that scores the harness by linter rule violations. */
 declare function linterJudge(id: string, config: HarnessConfig): SandboxJudgeSpec;
+/** Build a `SandboxJudgeSpec` that scores the harness output for security issues via a security scanner. */
 declare function securityJudge(id: string, config: HarnessConfig): SandboxJudgeSpec;
 
 interface PlaybookEntry {

package/dist/index.js

@@ -9,7 +9,7 @@ import {
   checkBehavioralCanary,
   checkCanaries,
   runBehavioralCanaries
-} from "./chunk-U3IDYATS.js";
+} from "./chunk-TCPP4Z5Q.js";
 import {
   classifyEuAiRisk,
   euAiActReport,
@@ -72,7 +72,7 @@ import {
   llmJudge,
   parseCorrectnessResponse,
   verifyCompletion
-} from "./chunk-IXOV77YF.js";
+} from "./chunk-YFYIOSNV.js";
 import {
   DEFAULT_MUTATION_PRIMITIVES,
   DEFAULT_RED_TEAM_CORPUS,
@@ -96,7 +96,7 @@ import {
   scoreRedTeamOutput,
   surfaceContentHash,
   toolNamesForRun
-} from "./chunk-NTVWIH24.js";
+} from "./chunk-2B4HPFXN.js";
 import {
   BackendIntegrityError,
   assertRealBackend,
@@ -106,7 +106,7 @@ import {
   fileVerdictCache,
   inMemoryVerdictCache,
   summarizeBackendIntegrity
-} from "./chunk-XKA6ZGEY.js";
+} from "./chunk-VNM52AGA.js";
 import {
   MODEL_PRICING,
   MetricsCollector,
@@ -128,7 +128,7 @@ import {
   computeToolUseMetrics,
   iqr,
   welchsTTest
-} from "./chunk-IH7LYRHL.js";
+} from "./chunk-BXZVBX3D.js";
 import {
   analyzeSeries
 } from "./chunk-BOD4O7OF.js";
@@ -159,7 +159,7 @@ import {
   defaultIsMaterial,
   diffFindings,
   runSemanticConceptJudge
-} from "./chunk-JU6ZX3CX.js";
+} from "./chunk-SMCACT4Z.js";
 import {
   LockedJsonlAppender,
   Mutex
@@ -184,7 +184,7 @@ import {
   policyEditsFromFindings,
   scorePolicyEditReadiness,
   validatePolicyEdit
-} from "./chunk-IN3SHQML.js";
+} from "./chunk-LMSQ6EFA.js";
 import {
   AnalystRegistry,
   DEFAULT_TRACE_ANALYST_KINDS,
@@ -205,7 +205,7 @@ import {
   runProposeReview,
   runProposeReviewAsControlLoop,
   scoreFromEvals
-} from "./chunk-B2TMQM62.js";
+} from "./chunk-BHCFJGL4.js";
 import {
   allCriticalPassed,
   errorStreakDetector,
@@ -224,10 +224,10 @@ import {
   evaluateReleaseConfidence,
   judgeReplayGate,
   renderReleaseReport
-} from "./chunk-VDGPPGE3.js";
+} from "./chunk-UMEAR2FI.js";
 import {
   runEvalCampaign
-} from "./chunk-6FIAJHCU.js";
+} from "./chunk-6YVAN6R3.js";
 import {
   LlmCallError,
   LlmClient,
@@ -250,7 +250,7 @@ import {
   paretoChart,
   researchReport,
   summaryTable
-} from "./chunk-6GT4NI4V.js";
+} from "./chunk-CLS3374R.js";
 import {
   attributeCounterfactuals,
   runCounterfactual
@@ -299,7 +299,7 @@ import {
   weightedMean,
   wilcoxonSignedRank,
   wilson
-} from "./chunk-HRGUJTER.js";
+} from "./chunk-HYGRFL7C.js";
 import {
   FileSystemTraceStore,
   InMemoryTraceStore,
@@ -330,7 +330,7 @@ import {
   scoreTraceInsightReadiness,
   tokenizeDomainWords,
   traceAnalystOnRunComplete
-} from "./chunk-7RBJANJD.js";
+} from "./chunk-3722PKPB.js";
 import {
   TRACE_ANALYST_ACTOR_DESCRIPTION,
   TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION,
@@ -352,7 +352,7 @@ import {
   runFailureClass,
   runsForScenario,
   toolSpans
-} from "./chunk-UFSG7ACU.js";
+} from "./chunk-JZXGWLK5.js";
 import {
   FAILURE_CLASSES,
   TRACE_SCHEMA_VERSION,
@@ -415,7 +415,7 @@ import {
   parseRunRecordSafe,
   roundTripRunRecord,
   validateRunRecord
-} from "./chunk-2NSLDY4B.js";
+} from "./chunk-J5MUFXEY.js";
 import {
   TraceEmitter,
   llmSpanFromProvider
@@ -433,7 +433,7 @@ import {
   toAgentProfileJson,
   validateAgentProfileCell,
   verifyAgentProfileCell
-} from "./chunk-ABOIVNXL.js";
+} from "./chunk-AGYDMORK.js";
 import {
   canonicalize,
   evaluateHypothesis,