@tangle-network/agent-eval 0.148.0 → 0.149.0

package/dist/benchmarks/index.d.ts

@@ -2,7 +2,7 @@ import { s as RunSplitTag } from "../run-record-DVV82Gwh.js";
 import { a as CampaignResult } from "../types-C1Bmeb8X.js";
 import { a as BenchmarkFamily, c as BenchmarkSource, i as BenchmarkEvaluation, l as BenchmarkTaskKind, n as BenchmarkAdapter, o as BenchmarkResponder, r as BenchmarkDatasetItem, s as BenchmarkScenario, t as BENCHMARK_SPLIT_SEED, u as deterministicSplit } from "../types-ColZlZtc.js";
 import { Cn as CampaignStorage } from "../provenance-DA-Pmyfv.js";
-import "../index-Dzn8Q3C2.js";
+import "../index-BCDyf_kT.js";
 //#region src/benchmarks/calibration.d.ts
 interface BenchmarkMetricCalibrationOptions<TPayload = unknown, TArtifact = string> {
   adapter: BenchmarkAdapter<BenchmarkDatasetItem<TPayload>, TPayload, TArtifact>;

package/dist/campaign/index.d.ts

@@ -3,8 +3,8 @@ import { _ as ProposalFinding, v as ProposalFindingOrigin, x as makeProposalFind
 import { a as ExternalOptimizerEndpointFormat, c as ExternalOptimizerModelBudget, d as ExternalOptimizerModelCallResult, f as ExternalOptimizerModelExecutionObservation, g as ExternalTextCandidate, h as ExternalOptimizerRunnerCommand, i as ExternalOptimizerChatRequest, l as ExternalOptimizerModelCall, n as DEFAULT_EXTERNAL_OPTIMIZER_PROCESS_LIMITS, o as ExternalOptimizerEvaluationObservation, p as ExternalOptimizerProcessLimits, r as ExternalOptimizerCallbackLimits, s as ExternalOptimizerEvaluationRefusalReason, t as DEFAULT_EXTERNAL_OPTIMIZER_CALLBACK_LIMITS, u as ExternalOptimizerModelCallRequest, v as resolveExternalOptimizerCallbackLimits, y as resolveExternalOptimizerProcessLimits } from "../external-optimizer-contracts-CSjDLLmr.js";
 import { A as LabeledScenarioWrite, B as ScoredSurfaceOutcome, C as JudgeDimension, D as LabeledScenarioSampleArgs, E as LabeledScenarioRecord, F as ProposeContext, G as labelTrustRank, H as SurfaceProposer, I as ProposedCandidate, L as RedactionStatus, M as OptimizerConfig, N as ParetoParent, O as LabeledScenarioSource, P as ProposalTrackContext, R as Scenario, S as JudgeConfig, T as LabelTrust, U as TraceSpan, V as SessionScript, W as isProposedCandidate, _ as GateDecision, a as CampaignResult, b as GenerationRecord, c as CampaignTraceWriter, d as DispatchContext, f as DispatchFn, g as GateContribution, h as GateContext, i as CampaignCostMeter, j as MutableSurface, k as LabeledScenarioStore, l as CodeSurface, m as GateCheckStatus, n as CampaignArtifactWriter, o as CampaignScenarioIdentity, p as Gate, r as CampaignCellResult, s as CampaignTokenUsage, t as CampaignAggregates, u as ComponentSurface, v as GateResult, w as JudgeScore, x as JudgeAggregate, y as GenerationCandidate, z as ScenarioAggregate } from "../types-C1Bmeb8X.js";
 import { $ as OptimizationTokenUsage, $t as SearchTokenAccounting, An as LlmJudgeDimension, At as SearchCompletedEvent, Bt as SearchModelIdentity, C as RunOptimizationOptions, Cn as CampaignStorage, Ct as SearchAttemptAccounting, D as runEval, Dn as OpenAutoPrOptions, Dt as SearchCandidateSlot, E as RunEvalOptions, En as inMemoryCampaignStorage, Et as SearchCandidateRegisteredEvent, Ft as SearchLedgerEntry, G as OptimizationMethodComparison, Gt as SearchPlannedOperation, H as CompareOptimizationMethodsOptions, Ht as SearchOperationRecordedEvent, It as SearchLedgerEvent, J as OptimizationMethodProvenance, Jt as SearchSurfaceEffect, K as OptimizationMethodInput, Kt as SearchPlannedTask, Lt as SearchLedgerHash, Mn as llmJudge, Mt as SearchFailureReason, Nt as SearchLedger, On as OpenAutoPrResult, Ot as SearchCandidateSlotClosedEvent, Pt as SearchLedgerAppendResult, Q as OptimizationPackageSource, Qt as SearchTaskOutcome, Rt as SearchLedgerReplay, S as PremeasuredOptimizationBaseline, Sn as planCampaignRun, St as SearchArtifactRef, T as runOptimization, Tn as fsCampaignStorage, Tt as SearchCandidateLineage, U as ComparisonCost, Ut as SearchPlan, Vt as SearchOperationKind, W as OptimizationMethod, Wt as SearchPlannedEvent, X as OptimizationMethodRunOptions, Xt as SearchSurfaceKind, Y as OptimizationMethodResult, Yt as SearchSurfaceEvidence, Z as OptimizationMethodScore, Zt as SearchTaskAttemptedEvent, _ as provenanceSpansPath, _n as RunCampaignOptions, _t as verifySearchHistoryReceipt, a as LoopProvenanceBackend, an as GepaCandidateSelectionScore, at as SEARCH_HISTORY_RECEIPT_DIGEST_ALGORITHM, b as RunImprovementLoopResult, bn as CampaignRunPlanCell, bt as SEARCH_LEDGER_SCHEMA, c as LoopProvenanceOptimizationMethod, cn as ExternalOptimizerObservationArtifact, ct as SearchHistoryCoverage, d as campaignMeasurementDigest, dn as readExternalOptimizerObservationArtifact, dt as SearchHistoryReceipt, en as openSearchLedger, et as combineComparisonCosts, f as canonicalDigest, fn as CacheIssueReason, ft as SearchHistoryRequiredError, g as provenanceRecordPath, gn as CampaignCellFailureReceipt, gt as searchHistoryCoverageRow, h as loopProvenanceSpans, hn as cellCachePath, ht as createSearchHistoryReceipt, i as LoopProvenanceArgsFromResult, in as GepaCandidatePopulationSummary, it as CreateSearchHistoryReceiptInput, jn as LlmJudgeOptions, jt as SearchCostAccounting, kn as openAutoPr, kt as SearchCandidateSurface, l as LoopProvenanceRecord, ln as ExternalOptimizerObservationSummary, lt as SearchHistoryCoverageRow, m as loopProvenanceArgsFromResult, mn as readCachedCell, mt as assertSearchHistoryMatchesReplay, n as EmitLoopProvenanceArgs, nn as GepaCandidatePopulationArtifact, nt as costFromLedgerSummary, o as LoopProvenanceCandidate, on as readGepaCandidatePopulationArtifact, ot as SEARCH_HISTORY_RECEIPT_SCHEMA_VERSION, p as emitLoopProvenance, pn as CacheRead, pt as assertCompleteSearchHistory, q as OptimizationMethodPairwise, qt as SearchSourceRef, r as EmitLoopProvenanceResult, rn as GepaCandidatePopulationCandidate, rt as optimizationTokenUsageFromSummary, s as LoopProvenanceEvidence, sn as ExternalOptimizerExecutionSummary, st as SearchHistoryAuditSummary, t as BuildLoopProvenanceArgs, tn as validateSearchLedgerEvent, tt as compareOptimizationMethods, u as buildLoopProvenanceRecord, un as ExternalOptimizerSubmittedCandidate, ut as SearchHistoryPolicy, v as verifyLoopProvenanceRecord, vn as runCampaign, vt as FileSearchLedger, w as RunOptimizationResult, wn as createRunCostLedger, wt as SearchCandidateDecidedEvent, x as runImprovementLoop, xn as PlanCampaignRunOptions, xt as SearchAccountingAudit, y as RunImprovementLoopOptions, yn as CampaignRunPlan, yt as OpenSearchLedgerOptions, zt as SearchLedgerTrustedHeadMode } from "../provenance-DA-Pmyfv.js";
-import { A as campaignScenarioIdentity, C as ExternalTextOptimizerContext, D as decodeExternalTextCandidate, E as ExternalTextEvaluationResponse, I as ReferenceEquivalenceJudgeOptions, M as campaignSplitDigestFromIdentities, O as assertCampaignDesign, R as ReferenceEquivalenceScenario, S as ExternalTextOptimizationMethodConfig, T as ExternalOptimizationExample, _ as DefaultProductionGateOptions, a as GepaAdaptiveEngineRun, b as composeGate, c as GepaOptimizationMethodConfig, d as gepaOptimizationMethod, f as OpenAICompatibleOptimizerModel, g as DefaultProductionGateCheck, h as heldOutGate, i as skillOptOptimizationMethod, j as campaignSplitDigest, k as assertCampaignSplitIdentity, l as GepaOptimizationRecipe, m as HeldOutGateOptions, n as SkillOptRunnerCommand, o as GepaEngineOptions, p as OptimizerModelBudget, r as SkillOptTrainerConfig, s as GepaEngineRun, t as SkillOptOptimizationMethodConfig, u as GepaRunnerCommand, v as DefaultProductionRewardHackingOptions, w as ExternalTextOptimizerResult, x as externalTextOptimizationMethod, y as defaultProductionGate, z as createReferenceEquivalenceJudge } from "../skillopt-optimization-method-CrY0OG17.js";
-import { $ as ScoreboardRow, $t as CrossSurfaceCandidateOutcome, A as compareRankKeys, At as EvalFixtureLoadOptions, B as ProfileMatrixCoverage, Bt as SearchLedgerConflictError, C as surfaceHash, Cn as BuildTraceAnalystSurfaceDispatchOptions, Ct as UngroundedLiteralReport, D as CampaignBreakdown, Dn as traceAnalystQualityJudge, Dt as neutralizationGate, E as acquireSingleRunLock, En as buildTraceAnalystSurfaceDispatch, Et as NeutralizationGateOptions, F as resolveRunDir, Ft as PlanEvalFixtureRunOptions, G as SegmentedProfileMatrixError, Gt as CrossSurfaceAdditionDecision, H as ProfileMatrixRow, Ht as SearchLedgerIntegrityError, I as tangleTracesRoot, It as discoverEvalFixtures, J as runProfileMatrixSegment, Jt as CrossSurfaceBestSingleSelection, K as createProfileMatrixPlan, Kt as CrossSurfaceAdditionRejectionReason, L as CreateProfileMatrixPlanOptions, Lt as loadEvalFixture, M as ScenarioSignal, Mt as EvalFixtureScenario, N as scoreDiscrimination, Nt as EvalFixtureValidationMode, O as campaignBreakdown, Ot as EvalFixture, P as selectDiscriminative, Pt as LoadEvalFixtureScenariosOptions, Q as ScoreboardRenderOptions, Qt as CrossSurfaceCandidateEvidence, R as FinalizeProfileMatrixOptions, Rt as loadEvalFixtureScenarios, S as surfaceContentHash, Sn as CrossSurfaceTaskRow, St as ScoredRollout, T as SingleRunLockOptions, Tn as TraceAnalystScenario, Tt as rolloutArgumentDiff, U as ProfileMatrixSegmentResult, Ut as analyzeCrossSurfaceInteractions, V as ProfileMatrixPlan, Vt as SearchLedgerError, W as RunProfileMatrixSegmentOptions, Wt as AnalyzeCrossSurfaceInteractionsInput, X as PlaybackDriver, Xt as CrossSurfaceCandidate, Y as PlaybackContext, Yt as CrossSurfaceBootstrapPolicy, Z as PlaybackStep, Zt as CrossSurfaceCandidateComparison, _ as assertCodeSurfaceIdentity, _n as CrossSurfacePairwiseEntry, _t as FsLabeledScenarioStoreOptions, a as WorktreeAdapterError, an as CrossSurfaceEligibility, at as scoreUserStory, b as componentSurfaceIdentityMaterial, bn as CrossSurfaceSelectionPolicy, bt as RolloutArgumentDiffOptions, c as verifyCodeSurface, cn as CrossSurfaceInteractionAwareSelection, ct as ProfileDispatchFn, d as PhoenixEvaluationResultLike, dn as CrossSurfaceInteractionReport, dt as RunProfileMatrixOptions, en as CrossSurfaceCandidateSummary, et as ScoreboardSummary, f as PhoenixEvaluatorLike, fn as CrossSurfaceInteractionTask, ft as RunProfileMatrixResult, g as isTransientTransportFailure, gn as CrossSurfacePairIncompatibilityReason, gt as FsLabeledScenarioStore, h as TransientFailureOptions, hn as CrossSurfacePairEvidence, ht as neutralizeText, i as WorktreeAdapter, in as CrossSurfaceDistribution, it as renderScoreboardMarkdown, j as DiscriminationScore, jt as EvalFixtureRunPlan, k as campaignMeanComposite, kt as EvalFixtureFile, l as AutoevalsScoreLike, ln as CrossSurfaceInteractionEffect, lt as ProfileMatrixError, m as phoenixEvaluatorJudge, mn as CrossSurfacePairCompatibility, mt as runProfileMatrix, n as GitWorktreeAdapterOptions, nn as CrossSurfaceComponentEvidence, nt as UserStoryVerdict, o as gitWorktreeAdapter, on as CrossSurfaceEvidenceBreakdown, ot as scoreboardSummary, p as autoevalsScorerJudge, pn as CrossSurfaceNaiveStackSelection, pt as ScenarioRollup, q as finalizeProfileMatrix, qt as CrossSurfaceAttemptCompleteness, r as Worktree, rn as CrossSurfaceCompositionStep, rt as makePlaybackDispatch, s as resolveWorktreePath, sn as CrossSurfaceIneligibilityReason, st as userStoryScoreboard, t as CodeSurfaceVerification, tn as CrossSurfaceComponent, tt as UserStory, u as AutoevalsScorerLike, un as CrossSurfaceInteractionPath, ut as ProfileSummary, v as assertComponentSurface, vn as CrossSurfaceRankedSingle, vt as LabeledScenarioStoreError, w as SingleRunLock, wn as TraceAnalystArtifact, wt as classifyUngroundedLiterals, x as renderSurfaceDiff, xn as CrossSurfaceSelections, xt as RolloutCall, y as codeSurfaceIdentityMaterial, yn as CrossSurfaceRelativeCost, yt as RolloutArgumentDiff, z as FinalizedProfileMatrixResult, zt as planEvalFixtureRun } from "../index-Dzn8Q3C2.js";
+import { A as campaignScenarioIdentity, C as ExternalTextOptimizerContext, D as decodeExternalTextCandidate, E as ExternalTextEvaluationResponse, I as ReferenceEquivalenceJudgeOptions, M as campaignSplitDigestFromIdentities, O as assertCampaignDesign, R as ReferenceEquivalenceScenario, S as ExternalTextOptimizationMethodConfig, T as ExternalOptimizationExample, _ as DefaultProductionGateOptions, a as GepaAdaptiveEngineRun, b as composeGate, c as GepaOptimizationMethodConfig, d as gepaOptimizationMethod, f as OpenAICompatibleOptimizerModel, g as DefaultProductionGateCheck, h as heldOutGate, i as skillOptOptimizationMethod, j as campaignSplitDigest, k as assertCampaignSplitIdentity, l as GepaOptimizationRecipe, m as HeldOutGateOptions, n as SkillOptRunnerCommand, o as GepaEngineOptions, p as OptimizerModelBudget, r as SkillOptTrainerConfig, s as GepaEngineRun, t as SkillOptOptimizationMethodConfig, u as GepaRunnerCommand, v as DefaultProductionRewardHackingOptions, w as ExternalTextOptimizerResult, x as externalTextOptimizationMethod, y as defaultProductionGate, z as createReferenceEquivalenceJudge } from "../skillopt-optimization-method-ChCTN4FD.js";
+import { $ as ScoreboardRow, $t as CrossSurfaceCandidateOutcome, A as compareRankKeys, At as EvalFixtureLoadOptions, B as ProfileMatrixCoverage, Bt as SearchLedgerConflictError, C as surfaceHash, Cn as BuildTraceAnalystSurfaceDispatchOptions, Ct as UngroundedLiteralReport, D as CampaignBreakdown, Dn as traceAnalystQualityJudge, Dt as neutralizationGate, E as acquireSingleRunLock, En as buildTraceAnalystSurfaceDispatch, Et as NeutralizationGateOptions, F as resolveRunDir, Ft as PlanEvalFixtureRunOptions, G as SegmentedProfileMatrixError, Gt as CrossSurfaceAdditionDecision, H as ProfileMatrixRow, Ht as SearchLedgerIntegrityError, I as tangleTracesRoot, It as discoverEvalFixtures, J as runProfileMatrixSegment, Jt as CrossSurfaceBestSingleSelection, K as createProfileMatrixPlan, Kt as CrossSurfaceAdditionRejectionReason, L as CreateProfileMatrixPlanOptions, Lt as loadEvalFixture, M as ScenarioSignal, Mt as EvalFixtureScenario, N as scoreDiscrimination, Nt as EvalFixtureValidationMode, O as campaignBreakdown, Ot as EvalFixture, P as selectDiscriminative, Pt as LoadEvalFixtureScenariosOptions, Q as ScoreboardRenderOptions, Qt as CrossSurfaceCandidateEvidence, R as FinalizeProfileMatrixOptions, Rt as loadEvalFixtureScenarios, S as surfaceContentHash, Sn as CrossSurfaceTaskRow, St as ScoredRollout, T as SingleRunLockOptions, Tn as TraceAnalystScenario, Tt as rolloutArgumentDiff, U as ProfileMatrixSegmentResult, Ut as analyzeCrossSurfaceInteractions, V as ProfileMatrixPlan, Vt as SearchLedgerError, W as RunProfileMatrixSegmentOptions, Wt as AnalyzeCrossSurfaceInteractionsInput, X as PlaybackDriver, Xt as CrossSurfaceCandidate, Y as PlaybackContext, Yt as CrossSurfaceBootstrapPolicy, Z as PlaybackStep, Zt as CrossSurfaceCandidateComparison, _ as assertCodeSurfaceIdentity, _n as CrossSurfacePairwiseEntry, _t as FsLabeledScenarioStoreOptions, a as WorktreeAdapterError, an as CrossSurfaceEligibility, at as scoreUserStory, b as componentSurfaceIdentityMaterial, bn as CrossSurfaceSelectionPolicy, bt as RolloutArgumentDiffOptions, c as verifyCodeSurface, cn as CrossSurfaceInteractionAwareSelection, ct as ProfileDispatchFn, d as PhoenixEvaluationResultLike, dn as CrossSurfaceInteractionReport, dt as RunProfileMatrixOptions, en as CrossSurfaceCandidateSummary, et as ScoreboardSummary, f as PhoenixEvaluatorLike, fn as CrossSurfaceInteractionTask, ft as RunProfileMatrixResult, g as isTransientTransportFailure, gn as CrossSurfacePairIncompatibilityReason, gt as FsLabeledScenarioStore, h as TransientFailureOptions, hn as CrossSurfacePairEvidence, ht as neutralizeText, i as WorktreeAdapter, in as CrossSurfaceDistribution, it as renderScoreboardMarkdown, j as DiscriminationScore, jt as EvalFixtureRunPlan, k as campaignMeanComposite, kt as EvalFixtureFile, l as AutoevalsScoreLike, ln as CrossSurfaceInteractionEffect, lt as ProfileMatrixError, m as phoenixEvaluatorJudge, mn as CrossSurfacePairCompatibility, mt as runProfileMatrix, n as GitWorktreeAdapterOptions, nn as CrossSurfaceComponentEvidence, nt as UserStoryVerdict, o as gitWorktreeAdapter, on as CrossSurfaceEvidenceBreakdown, ot as scoreboardSummary, p as autoevalsScorerJudge, pn as CrossSurfaceNaiveStackSelection, pt as ScenarioRollup, q as finalizeProfileMatrix, qt as CrossSurfaceAttemptCompleteness, r as Worktree, rn as CrossSurfaceCompositionStep, rt as makePlaybackDispatch, s as resolveWorktreePath, sn as CrossSurfaceIneligibilityReason, st as userStoryScoreboard, t as CodeSurfaceVerification, tn as CrossSurfaceComponent, tt as UserStory, u as AutoevalsScorerLike, un as CrossSurfaceInteractionPath, ut as ProfileSummary, v as assertComponentSurface, vn as CrossSurfaceRankedSingle, vt as LabeledScenarioStoreError, w as SingleRunLock, wn as TraceAnalystArtifact, wt as classifyUngroundedLiterals, x as renderSurfaceDiff, xn as CrossSurfaceSelections, xt as RolloutCall, y as codeSurfaceIdentityMaterial, yn as CrossSurfaceRelativeCost, yt as RolloutArgumentDiff, z as FinalizedProfileMatrixResult, zt as planEvalFixtureRun } from "../index-BCDyf_kT.js";
 import { c as powerPreflight, o as PowerPreflight, s as PowerPreflightOptions } from "../pareto-BqNW3LJR.js";
 import { a as ObjectiveSource, c as PromotionPolicy, d as paretoSignificanceGate, i as EvidenceVector, l as buildEvidenceVector, n as AxisVerdict, o as ParetoSignificanceGateOptions, r as BuildEvidenceVectorOptions, s as PromotionObjective, t as AxisEvidence, u as paretoPolicy } from "../promotion-policy-CewdjfCJ.js";
 import { a as detectScale, c as pairHoldout, d as SequentialDecision, f as SequentialObservation, g as sequentialPairedGate, h as sequentialDecide, i as PairedHoldout, l as SequentialDecideFn, m as SequentialPairedGateOptions, n as HeldoutSignificance, o as dimensionRegressions, p as SequentialPairedGate, r as HeldoutSignificanceOptions, s as heldoutSignificance, t as DimensionRegression, u as SequentialDecideOptions } from "../statistical-heldout-uxSpFEjm.js";

package/dist/cli.js

@@ -1,6 +1,6 @@
 #!/usr/bin/env node
 import { o as runRolloutReleaseCli } from "./hf-dataset-XggBupCr.js";
-import { n as runAnalystBenchmarkCommand } from "./benchmark-command-Cz4fDmsk.js";
+import { n as runAnalystBenchmarkCommand } from "./benchmark-command-gufocQ-N.js";
 import { a as runRpcBatch, o as runRpcOnce, p as handleVersion, r as startServerAsync, s as buildOpenApi } from "./server-D9wQclzG.js";
 import { writeFileSync } from "node:fs";
 //#region src/cli-config.ts

package/dist/contract/index.d.ts

@@ -7,11 +7,11 @@ import { C as JudgeDimension, H as SurfaceProposer, M as OptimizerConfig, R as S
 import { a as FailureClusterInsight, c as JudgeInsight, d as Recommendation, f as ReleaseSummary, i as FailureClassTally, l as LiftInsight, m as TokenUsageInsight, n as ExecutionErrorOutcomeCell, o as InsightReport, p as ScalarDistribution, r as ExecutionInsight, s as InterRaterInsight, t as CostProvenanceSummary, u as OutcomeCorrelationInsight } from "../insight-report-BeT8KCgI.js";
 import { C as analyzeRuns, S as SummarizeExecutionOptions, b as AnalyzeRunsOptions, w as summarizeExecution, x as ExecutionReport } from "../engine-CIy18RTX.js";
 import { $ as OptimizationTokenUsage, An as LlmJudgeDimension, Cn as CampaignStorage, D as runEval, E as RunEvalOptions, En as inMemoryCampaignStorage, G as OptimizationMethodComparison, H as CompareOptimizationMethodsOptions, J as OptimizationMethodProvenance, K as OptimizationMethodInput, Mn as llmJudge, Q as OptimizationPackageSource, Tn as fsCampaignStorage, U as ComparisonCost, W as OptimizationMethod, Y as OptimizationMethodResult, _n as RunCampaignOptions, b as RunImprovementLoopResult, gn as CampaignCellFailureReceipt, jn as LlmJudgeOptions, tt as compareOptimizationMethods, vn as runCampaign, x as runImprovementLoop, y as RunImprovementLoopOptions } from "../provenance-DA-Pmyfv.js";
-import { B as runReferenceEquivalenceJudge, C as ExternalTextOptimizerContext, E as ExternalTextEvaluationResponse, F as ReferenceEquivalenceJudgeInput, I as ReferenceEquivalenceJudgeOptions, L as ReferenceEquivalenceJudgeResult, N as REFERENCE_EQUIVALENCE_INPUT_LIMITS, P as REFERENCE_EQUIVALENCE_JUDGE_VERSION, R as ReferenceEquivalenceScenario, S as ExternalTextOptimizationMethodConfig, T as ExternalOptimizationExample, _ as DefaultProductionGateOptions, a as GepaAdaptiveEngineRun, b as composeGate, c as GepaOptimizationMethodConfig, d as gepaOptimizationMethod, f as OpenAICompatibleOptimizerModel, g as DefaultProductionGateCheck, h as heldOutGate, i as skillOptOptimizationMethod, j as campaignSplitDigest, l as GepaOptimizationRecipe, m as HeldOutGateOptions, n as SkillOptRunnerCommand, o as GepaEngineOptions, p as OptimizerModelBudget, r as SkillOptTrainerConfig, s as GepaEngineRun, t as SkillOptOptimizationMethodConfig, u as GepaRunnerCommand, v as DefaultProductionRewardHackingOptions, w as ExternalTextOptimizerResult, x as externalTextOptimizationMethod, y as defaultProductionGate, z as createReferenceEquivalenceJudge } from "../skillopt-optimization-method-CrY0OG17.js";
+import { B as runReferenceEquivalenceJudge, C as ExternalTextOptimizerContext, E as ExternalTextEvaluationResponse, F as ReferenceEquivalenceJudgeInput, I as ReferenceEquivalenceJudgeOptions, L as ReferenceEquivalenceJudgeResult, N as REFERENCE_EQUIVALENCE_INPUT_LIMITS, P as REFERENCE_EQUIVALENCE_JUDGE_VERSION, R as ReferenceEquivalenceScenario, S as ExternalTextOptimizationMethodConfig, T as ExternalOptimizationExample, _ as DefaultProductionGateOptions, a as GepaAdaptiveEngineRun, b as composeGate, c as GepaOptimizationMethodConfig, d as gepaOptimizationMethod, f as OpenAICompatibleOptimizerModel, g as DefaultProductionGateCheck, h as heldOutGate, i as skillOptOptimizationMethod, j as campaignSplitDigest, l as GepaOptimizationRecipe, m as HeldOutGateOptions, n as SkillOptRunnerCommand, o as GepaEngineOptions, p as OptimizerModelBudget, r as SkillOptTrainerConfig, s as GepaEngineRun, t as SkillOptOptimizationMethodConfig, u as GepaRunnerCommand, v as DefaultProductionRewardHackingOptions, w as ExternalTextOptimizerResult, x as externalTextOptimizationMethod, y as defaultProductionGate, z as createReferenceEquivalenceJudge } from "../skillopt-optimization-method-ChCTN4FD.js";
 import { a as ObjectiveSource, c as PromotionPolicy, d as paretoSignificanceGate, i as EvidenceVector, l as buildEvidenceVector, n as AxisVerdict, o as ParetoSignificanceGateOptions, r as BuildEvidenceVectorOptions, s as PromotionObjective, t as AxisEvidence, u as paretoPolicy } from "../promotion-policy-CewdjfCJ.js";
 import { c as EvalRunGenerationSnapshot, g as TraceSpanEvent, n as HostedTenant, o as EvalRunCellScore, s as EvalRunEvent } from "../client-BV7icIfy.js";
 import { i as InMemoryOutcomeStore, n as FileSystemOutcomeStore, o as OutcomeStore, r as FileSystemOutcomeStoreOptions, t as DeploymentOutcome } from "../outcome-store-BYHIuO0e.js";
-import { a as DefinedAgentEval, c as SelfImproveOptions, d as SelfImproveRunError, f as selfImprove, i as DefineAgentEvalOptions, l as SelfImproveProgressEvent, n as AgentEvalEvaluateOptions, o as defineAgentEval, r as AgentEvalImproveOptions, s as SelfImproveBudget, t as AgentEvalAgent, u as SelfImproveResult } from "../define-agent-eval-hXLUBKtb.js";
+import { a as DefinedAgentEval, c as SelfImproveOptions, d as SelfImproveRunError, f as selfImprove, i as DefineAgentEvalOptions, l as SelfImproveProgressEvent, n as AgentEvalEvaluateOptions, o as defineAgentEval, r as AgentEvalImproveOptions, s as SelfImproveBudget, t as AgentEvalAgent, u as SelfImproveResult } from "../define-agent-eval-BWuk2O_N.js";
 import { AgentCandidateBenchmarkCellRef, AgentCandidateBenchmarkSuiteInputs, AgentCandidateBenchmarkTask, AgentCandidateBenchmarkTaskMaterial, AgentCandidateBundle, AgentCandidateEvaluationPolicy, AgentCandidateExperiment, AgentCandidateExperimentMaterial, AgentCandidateExperimentMeasurement, AgentImprovementCost, AgentImprovementMeasuredComparison, AgentProfileImprovementExperiment, AgentProfileImprovementExperimentMaterial, AgentProfileImprovementMeasuredComparison, AgentProfileImprovementMeasurement, AgentProfileImprovementRunCell, AgentProfileImprovementRunReceipt, AgentProfileImprovementSuiteInputs, AgentProfileImprovementTask, AgentProfileImprovementTaskMaterial, CandidateExecutionEvidence, Sha256Digest } from "@tangle-network/agent-interface";
 //#region src/contract/measured-comparison.d.ts
 interface SealCandidateBenchmarkSuiteOptions {

package/dist/{define-agent-eval-hXLUBKtb.d.ts → define-agent-eval-BWuk2O_N.d.ts}

@@ -99,8 +99,7 @@ interface SelfImproveOptions<TScenario extends Scenario, TArtifact> {
   /** Scenarios to evaluate against. Train/holdout split is computed from
    *  these unless `budget.holdoutScenarios` is set explicitly. */
   scenarios: TScenario[];
-  /** Judge that scores artifacts. Bring your own; use `langchainJudge`
-   *  from `/adapters/langchain` for a Runnable-shaped one. */
+  /** Judge that scores artifacts. Bring your own, or wrap `llmJudge`. */
   judge: JudgeConfig<TArtifact, TScenario>;
   /** Starting surface — system prompt, JSON config, anything `MutableSurface`
    *  accepts. The proposer mutates this each generation. */
@@ -385,4 +384,4 @@ interface DefinedAgentEval<TScenario extends Scenario, TArtifact> {
 declare function defineAgentEval<TScenario extends Scenario, TArtifact>(defaults: DefineAgentEvalOptions<TScenario, TArtifact>): DefinedAgentEval<TScenario, TArtifact>;
 //#endregion
 export { DefinedAgentEval as a, SelfImproveOptions as c, SelfImproveRunError as d, selfImprove as f, DefineAgentEvalOptions as i, SelfImproveProgressEvent as l, AgentEvalEvaluateOptions as n, defineAgentEval as o, AgentEvalImproveOptions as r, SelfImproveBudget as s, AgentEvalAgent as t, SelfImproveResult as u };
-//# sourceMappingURL=define-agent-eval-hXLUBKtb.d.ts.map
+//# sourceMappingURL=define-agent-eval-BWuk2O_N.d.ts.map

package/dist/define-agent-eval-BWuk2O_N.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"define-agent-eval-BWuk2O_N.d.ts","names":[],"sources":["../src/contract/self-improve.ts","../src/contract/define-agent-eval.ts"],"mappings":";;;;;;;;UA+DiB;;;EAGf;;;;EAIA;;EAEA;;EAEA;;;EAGA;;;EAGA;;;;EAIA;;EAEA,mBAAmB;;;;;;;;;EASnB;;EAEA;;;;;EAKA;;KAGU;EACN;EAA0B;;EAC1B;EAA4B;EAAuB;;EACnD;EAA4B;EAAe;;EAC3C;EAA8B;EAAe;EAAuB;;EAGpE;EAAsB;EAAkB;;EACxC;EAAyB;EAAW;EAAY;EAAa;;UAElD,mBAAmB,kBAAkB,UAAU;;;;;;;;;;;;;;EAc9D,QAAQ,SAAS,gBAAgB,UAAU,WAAW,KAAK,oBAAoB,QAAQ;;;;;;;EAQvF;;;EAIA,WAAW;;EAGX,OAAO,YAAY,WAAW;;;EAI9B,iBAAiB;;EAGjB,SAAS;;;;;;;;;;;;EAaT,sBAAsB,gCAAgC,WAAW;;;;;EAMjE,WAAW,gBAAgB;;;;;;EAO3B,SAAS,mBAAmB,WAAW;;;EAIvC,qBAAqB;;;EAIrB,OAAO,KAAK,WAAW;;;;;;;;EASvB,cAAc,eAAe,gBAAgB,iBAAiB,mBAAmB;;;;EAKjF,UAAU;;;;EAKV;;;EAIA,gBAAgB,QAAQ;;;;;EAMxB,iBAAiB;IACf,UAAU;IACV;IACA;;;;EAKF;;;EAIA,cAAc,OAAO;;;;EAKrB;EACA;EACA;;;;;;;;;;;;;;EAeA,eAAe;;;EAIf,eAAe;;;;EAKf,eAAe;;EAGf;;;;;;;;EASA;;;;;;;;;EAUA,oBAAoB,uBAAuB,WAAW;;;EAItD,WAAW;;;;;;;EAQX,mBAAmB,uBAAuB,WAAW;;UAGtC,kBAAkB,kBAAkB,UAAU;;;;EAI7D;IACE;IACA,aAAa;;;;;EAKf;IACE;IACA,aAAa;IACb,SAAS;;;IAGT;;;;IAIA;;;;;;EAMF;;;EAGA;;;;EAIA,YAAY;;EAEZ;;;EAGA;;EAEA;;EAEA;;EAEA,MAAM;;;EAGN,UAAU;;EAEV;IACE;IACA,MAAM;IACN;IACA,aAAa;;;;;;;;EAQf,SAAS;;;;;EAKT,QAAQ;;;;;;EAMR,KAAK,yBAAyB,WAAW;;;cAI9B,4BAA4B;WAC9B,MAAM;WACN,UAAU;EAEnB,YAAY,gBAAgB,QAAQ;;;;;;;;;;;;;;;;;;;;;;;;;;;;iBA+LhB,YAAY,kBAAkB,UAAU,WAC5D,MAAM,mBAAmB,WAAW,aACnC,QAAQ,kBAAkB,WAAW;;;KCtiB5B,eAAe,kBAAkB,UAAU,cACrD,SAAS,gBACT,UAAU,WACV,KAAK,oBACF,QAAQ;KAED,uBAAuB,kBAAkB,UAAU,aAAa,mBAC1E,WACA;UAGe,yBAAyB,kBAAkB,UAAU,mBAC5D,KACN,eAAe,WAAW;;EAI5B,YAAY;;EAEZ,UAAU;;EAEV,QAAQ,eAAe,WAAW;;EAElC,QAAQ,YAAY,WAAW;;EAE/B,SAAS,YAAY,WAAW;;EAEhC;;KAGU,wBAAwB,kBAAkB,UAAU,aAAa,KAC3E,QAAQ,mBAAmB,WAAW;EAGtC,SAAS,QAAQ;EACjB,eAAe,QAAQ;;UAGR,iBAAiB,kBAAkB,UAAU;;WAEnD,oBAAoB;;WAEpB,iBAAiB;;;;;EAK1B,SACE,OAAO,yBAAyB,WAAW,aAC1C,QAAQ,eAAe,WAAW;;;;;;;EAOrC,QACE,OAAO,wBAAwB,WAAW,aACzC,QAAQ,kBAAkB,WAAW;;;;;;;;;iBAU1B,gBAAgB,kBAAkB,UAAU,WAC1D,UAAU,uBAAuB,WAAW,aAC3C,iBAAiB,WAAW"}