@tangle-network/agent-eval 0.221.7 → 0.222.0

package/dist/index.d.ts

@@ -1,59 +1,55 @@
-import { _ as PendingCostCall, a as CostLedgerFilter, c as CostLedgerPersistence, d as CostReceipt, f as CostReceiptInput, g as PaidCallResult, h as MaximumCharge, i as CostChannel, l as CostLedgerSummary, m as CustomTokenPricing, n as ChannelRollup, o as CostLedgerHandle, p as CostUsage, s as CostLedgerOptions, t as CostLedger, u as CostProvenance, v as PendingCostCallView, y as RunPaidCallInput } from "./cost-ledger-C1W3ICmP.js";
-import { a as JudgeError, c as ValidationError, i as ConfigError, n as AgentEvalErrorCode, s as NotFoundError, t as AgentEvalError } from "./errors-DEE6u6ot.js";
-import { _ as toAgentProfileJson, g as groupRunsByAgentProfileCell, h as buildAgentProfileCell, i as AgentProfileCellInput, l as AgentProfileJson, m as agentProfileCellKey, p as AgentProfileSourceInput, r as AgentProfileCell, s as AgentProfileDimensionValue, t as AGENT_PROFILE_KINDS, y as verifyAgentProfileCell } from "./agent-profile-cell-BcpFbNd5.js";
-import { E as isToolSpan, S as TraceEvent, T as isLlmSpan, _ as Span, a as FAILURE_CLASSES, c as JudgeSpan, f as Run, h as RunStatus, l as LlmSpan, n as BudgetLedgerEntry, o as FailureClass, p as RunCaptureReport, r as BudgetSpec, s as GenericSpan, t as Artifact, w as isJudgeSpan, x as ToolSpan } from "./schema-D6EVKzTB.js";
-import { _ as runTaskScore, a as RunRecord, c as RunSplitTag, d as RunTokenUsage, f as RunTraceRef, g as parseRunRecordSafe, h as modelHasSnapshot, i as RunOutcome, l as RunTaskFailure, m as isRunRecord, n as RunCostProvenance, o as RunRecordValidationError, p as UNKNOWN_MODEL, r as RunJudgeMetadata, s as RunSearchCoordinates, t as JudgeScoresRecord, u as RunTerminalOutcome, v as validateRunRecord } from "./run-record-C3awiXPf.js";
-import { $ as QueryTracesPage, H as createBoundedTraceAnalysisStore, J as TraceAnalysisStore, Q as ErrorCluster, S as makeProposalFinding, X as DEFAULT_TRACE_ANALYST_BUDGETS, Z as DatasetOverview, at as TraceAnalystSpan, b as computeFindingId, c as AnalystRunEvent, ct as TraceAnalystTraceSummary, d as AnalystRunSummary, dt as ViewTraceResult, et as SearchSpanResult, f as AnalystSeverity, i as AnalystFinding, it as TraceAnalystFilters, l as AnalystRunInputs, lt as ViewSpansResult, m as EvidenceRef, n as AnalystContext, nt as SpanMatchRecord, ot as TraceAnalystSpanKind, p as AnalystUsageReceipt, rt as TraceAnalystByteBudgets, st as TraceAnalystSpanStatus, t as Analyst, tt as SearchTraceResult, u as AnalystRunResult, ut as ViewTraceOversized, v as ProposalFinding, x as makeFinding } from "./types-6Lfdlshp.js";
-import { $ as AssertCrossFamilyServedOptions, A as createChatClient, B as LlmToolChoice, C as ChatResponse, D as MockTransportOpts, E as CustomTransportOpts, G as InMemoryRawProviderSink, H as LlmUsage, I as LlmMessage, M as LlmCallRequest, N as LlmCallResult, O as OpenAiCompatibleTransportOpts, P as LlmChargeBounds, R as LlmTokenLogprob, S as ChatRequest, T as CreateChatClientOpts, U as FileSystemRawProviderSink, V as LlmToolDefinition, X as RawProviderEvent, Z as RawProviderSink, _ as ScenarioResult, a as CompletionCriterion, at as assertCrossFamilyServed, b as ChatCallOpts, c as JudgeFn, ct as checkServedModel, d as PersonaConfig, dt as CrossFamilyError, et as AssertServedModelOptions, f as ProductClientConfig, ft as JudgeFamily, g as ScenarioFile, h as RubricDimension, i as CollectedArtifacts, it as ServedModelVerdict, j as LlmCallMetadata, k as SandboxSdkTransportOpts, l as JudgeInput, lt as describeServedModelCheck, m as RouteMap, mt as judgeFamily, n as ArtifactResult, nt as ServedModelCheck, o as DimensionJudgeScore, ot as assertServedModel, p as ProductScenario, pt as assertCrossFamily, q as NoopRawProviderSink, r as CheckResult, rt as ServedModelPolicy, s as DriverState, st as assertServedModels, t as ArtifactCheck, tt as ModelSubstitutionError, u as JudgeRubric, ut as AssertCrossFamilyOptions, v as Turn, w as ChatTransport, x as ChatClient, y as TurnResult, z as LlmToolCall } from "./types-B_IFcKp5.js";
-import { a as ContinuousCalibrationResult, d as calibrateJudgeContinuous, f as continuousAgreement, h as verbosityBias, i as ContinuousAgreementOptions, l as VerbosityBiasResult, n as CandidateScore, o as GoldenItem, r as ContinuousAgreement, t as CalibrationResult, u as calibrateJudge } from "./judge-calibration-DFtEMlde.js";
-import { $ as measureJudgeAgreement, B as JudgeConfusion, F as CalibrationExample, G as JudgeGateRefusedError, H as JudgeGateDecision, I as CalibrationMerge, J as JudgeRunMeta, K as JudgeGateStatus, L as DEFAULT_JUDGE_GATE_POLICY, Q as judgeGateDecision, R as JudgeAgreement, St as analyzeSeries, U as JudgeGateInput, V as JudgeExampleAgreement, W as JudgeGatePolicy, X as addCalibrationVerdicts, Y as OwnerVerdict, Z as assertJudgeMayGate, an as CorpusAgreementPerDimension, bt as SeriesConvergenceOptions, cn as corpusInterRaterAgreement, et as registerCalibrationSet, in as CorpusAgreementOptions, ln as corpusInterRaterAgreementFromJudgeScores, on as CorpusAgreementReport, q as JudgeRun, sn as CorpusScoreRecord, tt as snapshotFromJudgeAgreement, un as interRaterReliability, xt as SeriesConvergenceResult, z as JudgeCalibrationSet } from "./index-CfJULcW5.js";
-import { a as partialCredit, c as spearmanR, d as weightedMean, i as confidenceInterval, l as summarizeNumberSeries, n as WeightedCompositeInput, o as pearsonR, r as WeightedCompositeResult, s as ranks, t as SeriesDistribution, u as weightedComposite } from "./descriptive-B2iPaT9J.js";
-import { n as bonferroni, r as holm, t as benjaminiHochberg } from "./multiplicity-DIWHvysC.js";
-import { A as PairRunRecordsResult, At as EProcess, B as ArmPopulationSummary, Bt as mcnemar, D as MatchedRunRecordPair, E as MatchedPair, F as comparePairedArms, Ft as ExactRiskDifferenceResult, Gt as wilson, Ht as pairedRiskDifferenceExact, I as pairArms, It as McNemarResult, L as pairRunRecords, Lt as ProportionInterval, M as PairedArmsComparison, Mt as EProcessState, N as PairedCorrectness, Nt as EProcessStep, O as PairArmsOptions, P as PairedMetricDelta, Pt as eProcess, R as ArmAssignment, Rt as RiskDifferenceResult, S as manifestContentDigest, T as ComparePairedArmsOptions, Ut as pairedRiskDifferenceScore, Vt as pairedRiskDifference, Wt as passAtK, _t as AttestationProvenance, bt as attest, gt as ATTESTATION_ALGORITHM, it as paretoFrontier, j as PairedArmRow, jt as EProcessOptions, k as PairArmsResult, nt as ParetoResult, rt as dominates, tt as Objective, vt as AttestationVerification, x as hashJson, xt as verifyAttestation, yt as AttestedReport, z as ArmPopulationReading, zt as ScoreRiskDifferenceResult } from "./statistical-heldout-BjzSOmLH.js";
-import { a as PairedPromotionDecision, c as BOOTSTRAP_GATE_MIN_N, d as PairedSignTestResult, f as PairedTTestResult, g as pairedTTest, h as pairedSignTest, i as PairedMcNemarEvidence, l as PairedBootstrapOptions, m as pairedBootstrap, n as PairedDecisionShape, o as PairedPromotionDecisionOptions, p as SignTestAlternative, r as PairedDecisionStatistic, s as decidePairedPromotion, t as PairedDecisionMethod, u as PairedBootstrapResult } from "./paired-promotion-decision-CQMJ5xGN.js";
-import { a as requiredPairedSampleSize, i as pairedMde, n as mcnemarPower, o as requiredSampleSize, r as mcnemarRequiredN, t as mulberry32 } from "./random-DGEyr3zG.js";
-import { C as mannWhitneyU, S as WilcoxonSignedRankResult, _ as evaluateReleaseConfidence, a as bootstrapCi, b as RankTestMethodRequest, d as ReleaseConfidenceIssue, f as ReleaseConfidenceMetrics, g as ReleaseTraceEvidence, i as Verdict, n as BootstrapResult, p as ReleaseConfidenceScorecard, s as ActionableSideInfo, t as BootstrapOptions, u as ReleaseConfidenceInput, v as MannWhitneyResult, w as wilcoxonSignedRank, x as RankTestOptions, y as RankTestMethod } from "./promotion-gate-CkVZHagC.js";
-import { $ as summarizeBackendIntegrity, Bi as runCampaign, Ct as RequirementCheck, Dt as createLlmCorrectnessChecker, Et as completionVerdict, Li as CampaignCellFailureReceipt, Ot as verifyCompletion, Q as assertRealBackend, Ri as CampaignCellRetryPolicy, Sr as runCanaries, St as ProducedState, Tt as TaskGold, X as BackendIntegrityError, Z as BackendIntegrityReport, _r as CanaryAlert, _t as CompletionRequirement, bi as LlmJudgeOptions, br as CanaryOptions, bt as LlmCorrectnessCheckerOpts, dr as RedTeamCategory, fr as RedTeamFinding, ft as ArtifactEventLike, gr as scoreRedTeamOutput, gt as extractProducedState, hr as redTeamReport, ht as ToolCallEventLike, lr as DEFAULT_RED_TEAM_CORPUS, mr as redTeamDataset, mt as RuntimeEventLike, pr as RedTeamReport, pt as ProposalEventLike, ur as RedTeamCase, vr as CanaryEvaluation, vt as CompletionVerdict, wi as transientDispatchFailure, wt as SatisfiedBy, xi as llmJudge, xr as CanaryReport, xt as ProducedProposal, yr as CanaryKind, yt as CorrectnessChecker, zi as RunCampaignOptions } from "./index-CdEVr8-O.js";
-import { R as Scenario, S as JudgeConfig, _ as GateDecision, a as CampaignResult, w as JudgeScore } from "./types-M6eF9m-r.js";
-import { a as RunFilter, i as InMemoryTraceStore, n as FileSystemTraceStore, o as SpanFilter, s as TraceStore, t as EventFilter } from "./store-B3nWRy3B.js";
-import { a as MintedRolloutLine } from "./schema-C-ziyJD7.js";
-import { n as ScorePreference } from "./reward-BeUgVVC_.js";
-import { a as mintRolloutRows, i as RolloutSearchLineage, n as MintRolloutResult, t as MintRolloutOptions } from "./mint-CkzpVPdz.js";
-import { a as ProfileAxisSpec, c as expandProfileAxes, i as HarnessType, l as harnessAxisOf, n as CODING_HARNESSES, o as agentProfileHash, r as HARNESS_NATIVE_MODEL, s as agentProfileId, t as AgentProfile } from "./agent-profile-CNmwfaqq.js";
-import { i as DatasetSplit, n as DatasetManifest, r as DatasetScenario, t as Dataset } from "./dataset-DQqhOCPt.js";
-import { C as runEquivalenceCheck, S as defineEquivalenceCheck, _ as StrategyChecker, a as CheckerIdentity, b as VerificationStrategyProfile, c as EquivalenceCheckDefinition, d as EquivalenceCheckerInput, f as EquivalenceCheckerResult, g as EquivalenceRecord, h as EquivalenceProtocolError, i as equivalenceVerdict, l as EquivalenceCheckSpec, m as EquivalenceObligationStatus, n as VerdictCertification, o as CheckerOutcome, p as EquivalenceObligation, r as certificationEvidenceDigest, s as EquivalenceArm, t as DefaultVerdict, u as EquivalenceChecker, v as VERIFICATION_STRATEGIES, x as VerificationStrategySource, y as VERIFICATION_STRATEGY_SOURCES } from "./verdict-C24NELtN.js";
-import { a as MultiLayerVerifier, c as VerifyOptions, i as LayerStatus, l as gradeSemanticStatus, n as Layer, o as Severity, r as LayerResult, s as VerificationReport, t as Finding } from "./multi-layer-verifier-DEgMUk6v.js";
-import { C as HeldOutGateRejectionCode, S as HeldOutGateDecision, _ as gainHistogram, a as ResearchReportOptions, b as HeldOutGate, d as GainDistributionOptions, g as SummaryTableRow, h as SummaryTableOptions, l as GainDistributionBin, m as SummaryTable, p as ParetoPoint, t as ResearchReport, u as GainDistributionFigureSpec, v as summaryTable, w as SplitCoverage, x as HeldOutGateConfig, y as GateEvidence } from "./summary-report-C8V49Jt3.js";
-import { a as FailureClassification, c as FailureRule, d as failureBlame, i as FailureBlame, l as classifyFailure, o as FailureContext, r as failureClusterView, s as FailureReasonRule, u as classifyFailureReason } from "./failure-cluster-DN0t63Z0.js";
-import { d as InsightReport, n as analyzeRuns, t as AnalyzeRunsOptions } from "./analyze-runs-DfX6dHP0.js";
-import { _ as SelfImproveMethodResult, a as DefinedAgentEval, c as SelfImproveMethodOptions, d as SelfImproveProposerOptions, f as SelfImproveProposerResult, g as SelfImproveMethodProvenance, h as selfImprove, i as DefineAgentEvalOptions, l as SelfImproveOptions, n as AgentEvalEvaluateOptions, o as defineAgentEval, p as SelfImproveResult, r as AgentEvalImproveOptions, s as SelfImproveBudget, t as AgentEvalAgent, u as SelfImproveProgressEvent } from "./define-agent-eval-JKt3gMQr.js";
-import { B as FindingsDiff, H as PersistedFinding, U as diffFindings, V as FindingsStore, z as DiffPolicy } from "./systemone-CHnqwmYg.js";
-import { i as ExactAnalystRunEvent, l as ExactCapableAnalyst, o as ExactAnalystRunResult } from "./exact-types-Do63KoKn.js";
-import { a as createTraceAnalyst, i as TraceAnalystDefinition, n as buildDefaultAnalystRegistry, r as CreateTraceAnalystOptions, t as DefaultAnalystRegistryOptions } from "./default-registry-CLG5YqGS.js";
-import { r as TraceAnalysisEngineResult, s as RawAnalystFinding, t as TraceAnalysisEngine } from "./engine-dtRVeuQs.js";
-import { a as BudgetPolicy, c as RegistryRunOpts, i as AnalystRegistryOptions, s as ExactRegistryRunOpts, t as AnalystRegistry } from "./registry-BcdJmyoU.js";
-import { C as SEMANTIC_CONCEPT_JUDGE_VERSION, D as runSemanticConceptJudge, E as SemanticConceptJudgeResult, S as ConceptSpec, T as SemanticConceptJudgeOptions, _ as DspyRlmTraceEngineOptions, a as FAILURE_MODE_KIND_SPEC, b as createChatTraceEngine, d as FindingSubject, f as FindingSubjectKind, t as DEFAULT_TRACE_ANALYST_KINDS, v as createDspyRlmTraceEngine, w as SemanticConceptJudgeInput, x as ConceptFinding, y as ChatTraceEngineOptions } from "./index-CTIRz-AB.js";
-import { i as BenchmarkEvaluation } from "./types-DZ0mYVw9.js";
-import { a as SpanHandle, n as CaptureWrite, o as TraceEmitter, s as TraceEmitterOptions, t as CaptureErrorHandler } from "./emitter-DABYYTcH.js";
-import { n as SandboxDriver } from "./sandbox-harness-BlSOu4LX.js";
-import { _ as shareAllowed, a as RedactionProfile, d as assessShareSafety, g as redactText, h as redactForShare, l as ShareSafetyStatus, m as redact, n as REDACTION_VERSION, o as RedactionReport, r as RedactOptions, u as ShareSafetyVerdict } from "./redact-DXdlFb_U.js";
-import { a as PairedEvalueStep, c as SequentialDecision, i as PairedEvalueSequence, l as evaluateInterimReleaseConfidence, n as InterimReleaseConfidenceInput, r as PairedEvalueOptions, t as InterimReleaseConfidence, u as pairedEvalueSequence } from "./sequential-KwGspfeR.js";
-import { S as welchsTTest, _ as computeToolUseMetrics, b as WelchTestStatus, d as Trajectory, f as TrajectoryStep, g as ToolUseOptions, h as ToolUseMetrics, i as toolWasteView, m as ToolStats, p as buildTrajectory, s as judgeAgreementView, u as budgetBreachView, x as iqr, y as WelchTestResult } from "./tool-waste-CkjzWmeY.js";
-import { A as AnalystFindingDigest, B as ControlRuntimeConfig, C as createFeedbackTrajectory, D as renderPreferenceMemoryMarkdown, E as feedbackTrajectoryToOptimizerRow, F as ControlBudget, H as ControlSeverity, I as ControlContext, L as ControlDecision, M as AnalystRunDigest, N as analystRunDigest, O as summarizePreferenceMemory, P as ControlActionOutcome, R as ControlEvalResult, S as controlRunToFeedbackTrajectory, T as feedbackTrajectoriesToOptimizerRows, U as ControlStep, V as ControlRuntimeError, W as StopDecision, _ as PreferenceMemoryEntry, a as FeedbackLabel, b as analystRunToReviewRequests, c as FeedbackOptimizerRow, d as FeedbackTask, f as FeedbackTrajectory, g as InMemoryFeedbackTrajectoryStore, h as FileSystemFeedbackTrajectoryStore, i as FeedbackAttempt, j as AnalystReviewDecision, k as withAssignedFeedbackSplit, l as FeedbackOutcome, m as FeedbackTrajectoryStore, n as AnalystReviewRequest, o as FeedbackLabelKind, p as FeedbackTrajectoryFilter, r as FeedbackArtifactType, s as FeedbackLabelSource, t as AnalystFeedbackTrajectoryOptions, u as FeedbackSplitPolicy, v as ProposedSideEffect, w as feedbackTrajectoriesToDatasetScenarios, x as assignFeedbackSplit, y as analystRunToFeedbackTrajectory, z as ControlRunResult } from "./feedback-trajectory-CYVJXolG.js";
-import { A as TraceInsightReadiness, F as describeTraceInsightScope, G as AnalyzeTracesResult, H as tokenizeDomainWords, I as domainEvidencePattern, K as analyzeTraces, L as inferDomainKeywords, N as buildTraceInsightContext, P as buildTraceInsightPrompt, V as scoreTraceInsightReadiness, a as captureFetchToRawSink, b as OtlpSpan, c as ExtractedUsage, d as extractUsageFromSse, f as OtlpFileTraceStore, g as spanRecordsToTraceAnalysisStore, h as otlpTextToTraceAnalysisStore, j as TraceInsightSuite, m as SpanRecordsToTraceAnalysisStoreOptions, n as toolSpansToTraceAnalysisStore, o as OtlpFlatLine, u as extractUsage, v as OtlpExport, x as exportRunAsOtlp } from "./store-tool-spans-DxsFwtdr.js";
-import { a as RunIntegrityReport, o as assertRunCaptured, t as RunIntegrityError } from "./integrity-D9lBnv_K.js";
-import { a as SteeringChange, c as CampaignRunContext, d as CampaignVariant, f as EvalCampaignOptions, h as runEvalCampaign, i as Researcher, l as CampaignRunOutcome, m as FailedRun, n as ExperimentResult, o as CampaignFactoryParams, p as EvalCampaignResult, r as FailureMode, s as CampaignIntegrityPolicy, t as ExperimentPlan, u as CampaignScenario } from "./researcher-DKWtoMWk.js";
-import "./index-D7fmC-1K.js";
-import { y as OUTPUT_VALUE } from "./attribute-vocabulary-CcXqHODT.js";
-import { i as judgeSpans, n as argHash, s as runsForScenario } from "./query-Cmv_U8vf.js";
+import { _ as PendingCostCall, a as CostLedgerFilter, c as CostLedgerPersistence, d as CostReceipt, f as CostReceiptInput, g as PaidCallResult, h as MaximumCharge, i as CostChannel, l as CostLedgerSummary, m as CustomTokenPricing, n as ChannelRollup, o as CostLedgerHandle, p as CostUsage, s as CostLedgerOptions, t as CostLedger, u as CostProvenance, v as PendingCostCallView, y as RunPaidCallInput } from "./cost-ledger-Czro9wmo.js";
+import { a as JudgeError, c as ValidationError, i as ConfigError, n as AgentEvalErrorCode, s as NotFoundError, t as AgentEvalError } from "./errors-E9I2hldP.js";
+import { _ as toAgentProfileJson, g as groupRunsByAgentProfileCell, h as buildAgentProfileCell, i as AgentProfileCellInput, l as AgentProfileJson, m as agentProfileCellKey, p as AgentProfileSourceInput, r as AgentProfileCell, s as AgentProfileDimensionValue, t as AGENT_PROFILE_KINDS, y as verifyAgentProfileCell } from "./agent-profile-cell-3kYxUsK2.js";
+import { E as isToolSpan, S as TraceEvent, T as isLlmSpan, _ as Span, a as FAILURE_CLASSES, c as JudgeSpan, f as Run, h as RunStatus, l as LlmSpan, n as BudgetLedgerEntry, o as FailureClass, p as RunCaptureReport, r as BudgetSpec, s as GenericSpan, t as Artifact, w as isJudgeSpan, x as ToolSpan } from "./schema-CkIvnmLU.js";
+import { _ as runTaskScore, a as RunRecord, c as RunSplitTag, d as RunTokenUsage, f as RunTraceRef, g as parseRunRecordSafe, h as modelHasSnapshot, i as RunOutcome, l as RunTaskFailure, m as isRunRecord, n as RunCostProvenance, o as RunRecordValidationError, p as UNKNOWN_MODEL, r as RunJudgeMetadata, s as RunSearchCoordinates, t as JudgeScoresRecord, u as RunTerminalOutcome, v as validateRunRecord } from "./run-record-D2BFMPGl.js";
+import { $ as QueryTracesPage, H as createBoundedTraceAnalysisStore, J as TraceAnalysisStore, Q as ErrorCluster, S as makeProposalFinding, X as DEFAULT_TRACE_ANALYST_BUDGETS, Z as DatasetOverview, at as TraceAnalystSpan, b as computeFindingId, c as AnalystRunEvent, ct as TraceAnalystTraceSummary, d as AnalystRunSummary, dt as ViewTraceResult, et as SearchSpanResult, f as AnalystSeverity, i as AnalystFinding, it as TraceAnalystFilters, l as AnalystRunInputs, lt as ViewSpansResult, m as EvidenceRef, n as AnalystContext, nt as SpanMatchRecord, ot as TraceAnalystSpanKind, p as AnalystUsageReceipt, rt as TraceAnalystByteBudgets, st as TraceAnalystSpanStatus, t as Analyst, tt as SearchTraceResult, u as AnalystRunResult, ut as ViewTraceOversized, v as ProposalFinding, x as makeFinding } from "./types-DzCO-S1m.js";
+import { $ as AssertCrossFamilyServedOptions, A as createChatClient, B as LlmToolChoice, C as ChatResponse, D as MockTransportOpts, E as CustomTransportOpts, G as InMemoryRawProviderSink, H as LlmUsage, I as LlmMessage, M as LlmCallRequest, N as LlmCallResult, O as OpenAiCompatibleTransportOpts, P as LlmChargeBounds, R as LlmTokenLogprob, S as ChatRequest, T as CreateChatClientOpts, U as FileSystemRawProviderSink, V as LlmToolDefinition, X as RawProviderEvent, Z as RawProviderSink, _ as ScenarioResult, a as CompletionCriterion, at as assertCrossFamilyServed, b as ChatCallOpts, c as JudgeFn, ct as checkServedModel, d as PersonaConfig, dt as CrossFamilyError, et as AssertServedModelOptions, f as ProductClientConfig, ft as JudgeFamily, g as ScenarioFile, h as RubricDimension, i as CollectedArtifacts, it as ServedModelVerdict, j as LlmCallMetadata, k as SandboxSdkTransportOpts, l as JudgeInput, lt as describeServedModelCheck, m as RouteMap, mt as judgeFamily, n as ArtifactResult, nt as ServedModelCheck, o as DimensionJudgeScore, ot as assertServedModel, p as ProductScenario, pt as assertCrossFamily, q as NoopRawProviderSink, r as CheckResult, rt as ServedModelPolicy, s as DriverState, st as assertServedModels, t as ArtifactCheck, tt as ModelSubstitutionError, u as JudgeRubric, ut as AssertCrossFamilyOptions, v as Turn, w as ChatTransport, x as ChatClient, y as TurnResult, z as LlmToolCall } from "./types-0KglnQ9o.js";
+import { a as ContinuousCalibrationResult, d as calibrateJudgeContinuous, f as continuousAgreement, h as verbosityBias, i as ContinuousAgreementOptions, l as VerbosityBiasResult, n as CandidateScore, o as GoldenItem, r as ContinuousAgreement, t as CalibrationResult, u as calibrateJudge } from "./judge-calibration-DnrbQaXI.js";
+import { $ as measureJudgeAgreement, B as JudgeConfusion, F as CalibrationExample, G as JudgeGateRefusedError, H as JudgeGateDecision, I as CalibrationMerge, J as JudgeRunMeta, K as JudgeGateStatus, L as DEFAULT_JUDGE_GATE_POLICY, Q as judgeGateDecision, R as JudgeAgreement, St as analyzeSeries, U as JudgeGateInput, V as JudgeExampleAgreement, W as JudgeGatePolicy, X as addCalibrationVerdicts, Y as OwnerVerdict, Z as assertJudgeMayGate, an as CorpusAgreementPerDimension, bt as SeriesConvergenceOptions, cn as corpusInterRaterAgreement, et as registerCalibrationSet, in as CorpusAgreementOptions, ln as corpusInterRaterAgreementFromJudgeScores, on as CorpusAgreementReport, q as JudgeRun, sn as CorpusScoreRecord, tt as snapshotFromJudgeAgreement, un as interRaterReliability, xt as SeriesConvergenceResult, z as JudgeCalibrationSet } from "./index-DAk2quYu.js";
+import { B as WeightedCompositeResult, G as spearmanR, H as partialCredit, J as weightedMean, K as summarizeNumberSeries, L as SeriesDistribution, R as SeriesDistributionOptions, U as pearsonR, V as confidenceInterval, W as ranks, a as MintedRolloutLine, q as weightedComposite, z as WeightedCompositeInput } from "./schema-DkFJr3Y8.js";
+import { n as bonferroni, r as holm, t as benjaminiHochberg } from "./multiplicity-D_SEsc5j.js";
+import { A as PairRunRecordsResult, At as EProcess, B as ArmPopulationSummary, Bt as mcnemar, D as MatchedRunRecordPair, E as MatchedPair, F as comparePairedArms, Ft as ExactRiskDifferenceResult, Gt as wilson, Ht as pairedRiskDifferenceExact, I as pairArms, It as McNemarResult, L as pairRunRecords, Lt as ProportionInterval, M as PairedArmsComparison, Mt as EProcessState, N as PairedCorrectness, Nt as EProcessStep, O as PairArmsOptions, P as PairedMetricDelta, Pt as eProcess, R as ArmAssignment, Rt as RiskDifferenceResult, S as manifestContentDigest, T as ComparePairedArmsOptions, Ut as pairedRiskDifferenceScore, Vt as pairedRiskDifference, Wt as passAtK, _t as AttestationProvenance, bt as attest, gt as ATTESTATION_ALGORITHM, it as paretoFrontier, j as PairedArmRow, jt as EProcessOptions, k as PairArmsResult, nt as ParetoResult, rt as dominates, tt as Objective, vt as AttestationVerification, x as hashJson, xt as verifyAttestation, yt as AttestedReport, z as ArmPopulationReading, zt as ScoreRiskDifferenceResult } from "./statistical-heldout-Cei036nE.js";
+import { a as PairedPromotionDecision, c as BOOTSTRAP_GATE_MIN_N, d as PairedSignTestResult, f as PairedTTestResult, g as pairedTTest, h as pairedSignTest, i as PairedMcNemarEvidence, l as PairedBootstrapOptions, m as pairedBootstrap, n as PairedDecisionShape, o as PairedPromotionDecisionOptions, p as SignTestAlternative, r as PairedDecisionStatistic, s as decidePairedPromotion, t as PairedDecisionMethod, u as PairedBootstrapResult } from "./paired-promotion-decision-BDnqepZ1.js";
+import { a as requiredPairedSampleSize, i as pairedMde, n as mcnemarPower, o as requiredSampleSize, r as mcnemarRequiredN, t as mulberry32 } from "./random-BN150Q_X.js";
+import { C as mannWhitneyU, S as WilcoxonSignedRankResult, _ as evaluateReleaseConfidence, a as bootstrapCi, b as RankTestMethodRequest, d as ReleaseConfidenceIssue, f as ReleaseConfidenceMetrics, g as ReleaseTraceEvidence, i as Verdict, n as BootstrapResult, p as ReleaseConfidenceScorecard, s as ActionableSideInfo, t as BootstrapOptions, u as ReleaseConfidenceInput, v as MannWhitneyResult, w as wilcoxonSignedRank, x as RankTestOptions, y as RankTestMethod } from "./promotion-gate-CixshGI8.js";
+import { $ as summarizeBackendIntegrity, $n as RedTeamCategory, Di as RunCampaignOptions, Ei as CampaignCellRetryPolicy, Oi as runCampaign, Q as assertRealBackend, Qn as RedTeamCase, Ti as CampaignCellFailureReceipt, X as BackendIntegrityError, Z as BackendIntegrityReport, Zn as DEFAULT_RED_TEAM_CORPUS, ar as CanaryAlert, ci as LlmJudgeOptions, cr as CanaryOptions, er as RedTeamFinding, fi as transientDispatchFailure, ft as ArtifactEventLike, gt as extractProducedState, ht as ToolCallEventLike, ir as scoreRedTeamOutput, li as llmJudge, lr as CanaryReport, mt as RuntimeEventLike, nr as redTeamDataset, or as CanaryEvaluation, pt as ProposalEventLike, rr as redTeamReport, sr as CanaryKind, tr as RedTeamReport, ur as runCanaries } from "./index-B7v2QHlD.js";
+import { $ as MintRolloutResult, Q as MintRolloutOptions, R as Scenario, S as JudgeConfig, _ as GateDecision, a as CampaignResult, nt as mintRolloutRows, tt as RolloutSearchLineage, w as JudgeScore } from "./types-C54Uk_a1.js";
+import { a as RunFilter, i as InMemoryTraceStore, n as FileSystemTraceStore, o as SpanFilter, s as TraceStore, t as EventFilter } from "./store-CMs5wGXP.js";
+import { n as ScorePreference } from "./reward-ST4PXNXF.js";
+import { a as ProfileAxisSpec, c as expandProfileAxes, i as HarnessType, l as harnessAxisOf, n as CODING_HARNESSES, o as agentProfileHash, r as HARNESS_NATIVE_MODEL, s as agentProfileId, t as AgentProfile } from "./agent-profile-CoD1Dotf.js";
+import { i as DatasetSplit, n as DatasetManifest, r as DatasetScenario, t as Dataset } from "./dataset-DThC1S6G.js";
+import { C as runEquivalenceCheck, S as defineEquivalenceCheck, _ as StrategyChecker, a as CheckerIdentity, b as VerificationStrategyProfile, c as EquivalenceCheckDefinition, d as EquivalenceCheckerInput, f as EquivalenceCheckerResult, g as EquivalenceRecord, h as EquivalenceProtocolError, i as equivalenceVerdict, l as EquivalenceCheckSpec, m as EquivalenceObligationStatus, n as VerdictCertification, o as CheckerOutcome, p as EquivalenceObligation, r as certificationEvidenceDigest, s as EquivalenceArm, t as DefaultVerdict, u as EquivalenceChecker, v as VERIFICATION_STRATEGIES, x as VerificationStrategySource, y as VERIFICATION_STRATEGY_SOURCES } from "./verdict-D_qiBQ6_.js";
+import { a as MultiLayerVerifier, c as VerifyOptions, i as LayerStatus, l as gradeSemanticStatus, n as Layer, o as Severity, r as LayerResult, s as VerificationReport, t as Finding } from "./multi-layer-verifier-lM4aVN78.js";
+import { C as HeldOutGateRejectionCode, S as HeldOutGateDecision, _ as gainHistogram, a as ResearchReportOptions, b as HeldOutGate, d as GainDistributionOptions, g as SummaryTableRow, h as SummaryTableOptions, l as GainDistributionBin, m as SummaryTable, p as ParetoPoint, t as ResearchReport, u as GainDistributionFigureSpec, v as summaryTable, w as SplitCoverage, x as HeldOutGateConfig, y as GateEvidence } from "./summary-report-0La5vejY.js";
+import { a as FailureClassification, c as FailureRule, d as failureBlame, i as FailureBlame, l as classifyFailure, o as FailureContext, r as failureClusterView, s as FailureReasonRule, u as classifyFailureReason } from "./failure-cluster-DOoa7CBg.js";
+import { d as InsightReport, n as analyzeRuns, t as AnalyzeRunsOptions } from "./analyze-runs-C-pP1sK9.js";
+import { a as ProducedProposal, c as SatisfiedBy, d as createLlmCorrectnessChecker, f as verifyCompletion, i as LlmCorrectnessCheckerOpts, l as TaskGold, n as CompletionVerdict, o as ProducedState, r as CorrectnessChecker, s as RequirementCheck, t as CompletionRequirement, u as completionVerdict } from "./completion-verifier-7KEfroqx.js";
+import { _ as SelfImproveMethodResult, a as DefinedAgentEval, c as SelfImproveMethodOptions, d as SelfImproveProposerOptions, f as SelfImproveProposerResult, g as SelfImproveMethodProvenance, h as selfImprove, i as DefineAgentEvalOptions, l as SelfImproveOptions, n as AgentEvalEvaluateOptions, o as defineAgentEval, p as SelfImproveResult, r as AgentEvalImproveOptions, s as SelfImproveBudget, t as AgentEvalAgent, u as SelfImproveProgressEvent } from "./define-agent-eval-CAffiFbN.js";
+import { B as FindingsDiff, H as PersistedFinding, U as diffFindings, V as FindingsStore, z as DiffPolicy } from "./systemone-Cw2txK5m.js";
+import { i as ExactAnalystRunEvent, l as ExactCapableAnalyst, o as ExactAnalystRunResult } from "./exact-types-D98BKpxo.js";
+import { a as createTraceAnalyst, i as TraceAnalystDefinition, n as buildDefaultAnalystRegistry, r as CreateTraceAnalystOptions, t as DefaultAnalystRegistryOptions } from "./default-registry-B1DstPYg.js";
+import { r as TraceAnalysisEngineResult, s as RawAnalystFinding, t as TraceAnalysisEngine } from "./engine-CaHaCc3W.js";
+import { a as BudgetPolicy, c as RegistryRunOpts, i as AnalystRegistryOptions, s as ExactRegistryRunOpts, t as AnalystRegistry } from "./registry-dhdDoAV8.js";
+import { C as SEMANTIC_CONCEPT_JUDGE_VERSION, D as runSemanticConceptJudge, E as SemanticConceptJudgeResult, S as ConceptSpec, T as SemanticConceptJudgeOptions, _ as DspyRlmTraceEngineOptions, a as FAILURE_MODE_KIND_SPEC, b as createChatTraceEngine, d as FindingSubject, f as FindingSubjectKind, t as DEFAULT_TRACE_ANALYST_KINDS, v as createDspyRlmTraceEngine, w as SemanticConceptJudgeInput, x as ConceptFinding, y as ChatTraceEngineOptions } from "./index-CUjeTDPv.js";
+import { i as BenchmarkEvaluation } from "./types-0oMg7zee.js";
+import { a as SpanHandle, n as CaptureWrite, o as TraceEmitter, s as TraceEmitterOptions, t as CaptureErrorHandler } from "./emitter-mAXC5PSN.js";
+import { n as SandboxDriver } from "./sandbox-harness-B3Kx1fOA.js";
+import { _ as shareAllowed, a as RedactionProfile, d as assessShareSafety, g as redactText, h as redactForShare, l as ShareSafetyStatus, m as redact, n as REDACTION_VERSION, o as RedactionReport, r as RedactOptions, u as ShareSafetyVerdict } from "./redact-C_wzC4c_.js";
+import { a as PairedEvalueStep, c as SequentialDecision, i as PairedEvalueSequence, l as evaluateInterimReleaseConfidence, n as InterimReleaseConfidenceInput, r as PairedEvalueOptions, t as InterimReleaseConfidence, u as pairedEvalueSequence } from "./sequential-Bb_pdTqG.js";
+import { S as welchsTTest, _ as computeToolUseMetrics, b as WelchTestStatus, d as Trajectory, f as TrajectoryStep, g as ToolUseOptions, h as ToolUseMetrics, i as toolWasteView, m as ToolStats, p as buildTrajectory, s as judgeAgreementView, u as budgetBreachView, x as iqr, y as WelchTestResult } from "./tool-waste-BFG1vwTV.js";
+import { A as AnalystFindingDigest, B as ControlRuntimeConfig, C as createFeedbackTrajectory, D as renderPreferenceMemoryMarkdown, E as feedbackTrajectoryToOptimizerRow, F as ControlBudget, H as ControlSeverity, I as ControlContext, L as ControlDecision, M as AnalystRunDigest, N as analystRunDigest, O as summarizePreferenceMemory, P as ControlActionOutcome, R as ControlEvalResult, S as controlRunToFeedbackTrajectory, T as feedbackTrajectoriesToOptimizerRows, U as ControlStep, V as ControlRuntimeError, W as StopDecision, _ as PreferenceMemoryEntry, a as FeedbackLabel, b as analystRunToReviewRequests, c as FeedbackOptimizerRow, d as FeedbackTask, f as FeedbackTrajectory, g as InMemoryFeedbackTrajectoryStore, h as FileSystemFeedbackTrajectoryStore, i as FeedbackAttempt, j as AnalystReviewDecision, k as withAssignedFeedbackSplit, l as FeedbackOutcome, m as FeedbackTrajectoryStore, n as AnalystReviewRequest, o as FeedbackLabelKind, p as FeedbackTrajectoryFilter, r as FeedbackArtifactType, s as FeedbackLabelSource, t as AnalystFeedbackTrajectoryOptions, u as FeedbackSplitPolicy, v as ProposedSideEffect, w as feedbackTrajectoriesToDatasetScenarios, x as assignFeedbackSplit, y as analystRunToFeedbackTrajectory, z as ControlRunResult } from "./feedback-trajectory-SBGncrCP.js";
+import { A as TraceInsightReadiness, F as describeTraceInsightScope, G as AnalyzeTracesResult, H as tokenizeDomainWords, I as domainEvidencePattern, K as analyzeTraces, L as inferDomainKeywords, N as buildTraceInsightContext, P as buildTraceInsightPrompt, V as scoreTraceInsightReadiness, a as captureFetchToRawSink, b as OtlpSpan, c as ExtractedUsage, d as extractUsageFromSse, f as OtlpFileTraceStore, g as spanRecordsToTraceAnalysisStore, h as otlpTextToTraceAnalysisStore, j as TraceInsightSuite, m as SpanRecordsToTraceAnalysisStoreOptions, n as toolSpansToTraceAnalysisStore, o as OtlpFlatLine, u as extractUsage, v as OtlpExport, x as exportRunAsOtlp } from "./store-tool-spans-B3i3fl0A.js";
+import { a as RunIntegrityReport, o as assertRunCaptured, t as RunIntegrityError } from "./integrity-BAdVzT8u.js";
+import { a as SteeringChange, c as CampaignRunContext, d as CampaignVariant, f as EvalCampaignOptions, h as runEvalCampaign, i as Researcher, l as CampaignRunOutcome, m as FailedRun, n as ExperimentResult, o as CampaignFactoryParams, p as EvalCampaignResult, r as FailureMode, s as CampaignIntegrityPolicy, t as ExperimentPlan, u as CampaignScenario } from "./researcher-CLnNJ7Or.js";
+import "./index-DUn0GNiZ.js";
+import { y as OUTPUT_VALUE } from "./attribute-vocabulary-5oo8ykIx.js";
+import { i as judgeSpans, n as argHash, s as runsForScenario } from "./query-UD0B08Sb.js";
 import { SpanKind } from "@tangle-network/agent-trace-contract";
-//#region src/cost-ledger/errors.d.ts
 export declare class CostCeilingReachedError extends ValidationError {
   constructor(ceilingUsd: number, committedUsd: number, requestedUsd: number, phase: string, actor: string);
 }
 export declare class CostAccountingIncompleteError extends ValidationError {}
-//#endregion
-//#region src/cost-ledger/pricing.d.ts
 interface CostResult {
   costUsd: number;
   costUnknown: boolean;
@@ -61,8 +57,6 @@ interface CostResult {
 export declare function costForUsage(model: string, usage: CostUsage): CostResult;
 /** Price token counts with caller-supplied per-million rates. */
 export declare function costForTokenPricing(pricing: CustomTokenPricing, usage: Pick<CostUsage, 'inputTokens' | 'outputTokens' | 'cachedTokens' | 'cacheWriteTokens'>): number;
-//#endregion
-//#region src/llm-client/accounting.d.ts
 export declare function maximumChargeForLlmRequest(request: Pick<LlmCallRequest, 'model' | 'messages' | 'jsonSchema' | 'tools' | 'toolChoice' | 'maxTokens' | 'thinking'>, options?: LlmChargeBounds): MaximumCharge | undefined;
 /** Convert a provider result into the canonical paid-call receipt input.
  * The receipt is JSON-clean: absent optional fields are omitted, never carried
@@ -72,8 +66,6 @@ export declare function maximumChargeForLlmRequest(request: Pick<LlmCallRequest,
 export declare function costReceiptFromLlm(result: LlmCallResult, customTokenPricing?: CustomTokenPricing): CostReceiptInput;
 /** Structured-response failures retain their completed provider receipt. */
 export declare function costReceiptFromLlmError(error: Error, customTokenPricing?: CustomTokenPricing): CostReceiptInput | undefined;
-//#endregion
-//#region src/llm-client/errors.d.ts
 export declare class LlmCallError extends AgentEvalError {
   readonly status: number;
   readonly body: string;
@@ -89,16 +81,12 @@ export declare class LlmResponseError extends AgentEvalError {
     cause?: unknown;
   });
 }
-//#endregion
-//#region src/llm-client/json.d.ts
 /**
  * Strip a ```json / ``` code fence if the model emitted one.
  * Idempotent for naked JSON. Some models (claude-code via router, certain
  * deepseek models) wrap output even under json_object.
  */
 export declare function stripFencedJson(raw: string): string;
-//#endregion
-//#region src/llm-client/resilience.d.ts
 /**
  * True when an error is a transient transport/network fault worth retrying,
  * as opposed to a deterministic failure (4xx schema reject, JSON parse) that
@@ -111,8 +99,6 @@ export declare function stripFencedJson(raw: string): string;
  * consistently across transports.
  */
 export declare function isTransientLlmError(err: unknown): boolean;
-//#endregion
-//#region src/statistics/effect-sizes.d.ts
 /**
  * Standardized effect sizes: Cohen's d family and Cliff's delta, with the
  * shared degenerate-input contract (null where the effect is undefined).
@@ -157,8 +143,6 @@ export declare function cliffsDelta(before: number[], after: number[]): number;
  * <0.474 medium, else large.
  */
 export declare function interpretCliffs(delta: number): CliffsMagnitude;
-//#endregion
-//#region src/statistics/inference.d.ts
 /** Two-sided normal tail, evaluated without subtracting a near-one CDF. */
 export declare function normalTwoSidedPValue(z: number): number;
 /** Two-sided Student-t tail, preserving small tail probabilities. */
@@ -181,8 +165,6 @@ export declare function linearRegressionSlope(points: readonly {
   x: number;
   y: number;
 }[]): LinearRegressionResult;
-//#endregion
-//#region src/statistics/samples.d.ts
 interface SampleSummaryOptions {
   confidence?: number;
   seed?: number;
@@ -215,12 +197,10 @@ interface SampleSummary {
 export declare function sampleMedian(values: readonly number[]): number | null;
 /** Descriptive facts, not stability policy or a significance test. */
 export declare function summarizeSample(values: readonly number[], options?: SampleSummaryOptions): SampleSummary | null;
-//#endregion
-//#region src/analyst/policy-edit.d.ts
 type PolicyEditSchemaVersion = 'policy-edit/v1';
-declare const POLICY_EDIT_AXES: readonly ['carrier', 'representation', 'budget', 'sampling', 'output_contract', 'tool_contract', 'routing', 'memory', 'agent_profile', 'deployment_target'];
+declare const POLICY_EDIT_AXES: readonly ["carrier", "representation", "budget", "sampling", "output_contract", "tool_contract", "routing", "memory", "agent_profile", "deployment_target"];
 type PolicyEditAxis = (typeof POLICY_EDIT_AXES)[number];
-declare const POLICY_EDIT_TARGET_SURFACES: readonly ['prompt', 'tool-contract', 'runtime-config', 'memory', 'agent-profile', 'code', 'deployment'];
+declare const POLICY_EDIT_TARGET_SURFACES: readonly ["prompt", "tool-contract", "runtime-config", "memory", "agent-profile", "code", "deployment"];
 type PolicyEditTargetSurface = (typeof POLICY_EDIT_TARGET_SURFACES)[number];
 type PolicyEditRisk = 'low' | 'medium' | 'high' | 'unknown';
 type PolicyEditGainDirection = 'increase' | 'decrease';
@@ -279,7 +259,7 @@ interface PolicyEdit {
   validationPlan?: string;
   metadata?: Record<string, unknown>;
 }
-declare const POLICY_EDIT_CANDIDATE_RECORD_SCHEMA: 'tangle.policy-edit-candidate.v1';
+declare const POLICY_EDIT_CANDIDATE_RECORD_SCHEMA: "tangle.policy-edit-candidate.v1";
 /** JSON-safe attribution carried with a measured candidate and its scores. */
 interface PolicyEditCandidateRecord {
   schema: typeof POLICY_EDIT_CANDIDATE_RECORD_SCHEMA;
@@ -308,8 +288,6 @@ interface PolicyEditAdmission {
   reasons: string[];
 }
 export declare function makePolicyEditCandidateRecord(edit: PolicyEdit): PolicyEditCandidateRecord;
-//#endregion
-//#region src/capability-headroom.d.ts
 /**
  * Capability-headroom gate — "can this task set even SEE the capability
  * you're about to A/B?"
@@ -401,8 +379,6 @@ interface AssertCapabilityHeadroomOptions {
  * calibration run and the treatment run.
  */
 export declare function assertCapabilityHeadroom(result: CapabilityHeadroomResult, opts?: AssertCapabilityHeadroomOptions): void;
-//#endregion
-//#region src/cost-tracker.d.ts
 /**
  * Cost tracker — token + USD accounting per scenario and per run.
  *
@@ -478,8 +454,6 @@ interface CostSummary {
   /** Total USD / completed scenarios — null when nothing completed. */
   costPerCompletedTaskUsd: number | null;
 }
-//#endregion
-//#region src/detectors/index.d.ts
 type DetectorSeverity = 'info' | 'warn' | 'critical';
 /** A normalized step a detector folds over. Fields are optional; each detector reads only what it
  *  needs (repeated-action → `actionFingerprint`; no-progress → `stateFingerprint` + `score`;
@@ -535,8 +509,6 @@ export declare function errorStreakDetector(opts?: ErrorStreakOptions): Streamin
 /** Fold one event through many detectors at once; returns every signal that fired this step. The
  *  natural shape for an online pipe watching with a whole panel of detectors. */
 export declare function observeAll(detectors: ReadonlyArray<StreamingDetector>, event: DetectorEvent): DetectorSignal[];
-//#endregion
-//#region src/error-count-extractor.d.ts
 /**
  * Toolchain error-count extractor.
  *
@@ -583,8 +555,6 @@ interface ExtractResult {
  * NOT treat that as "zero TS errors" because the toolchain is wrong.
  */
 export declare function extractErrorCount(text: string, opts?: ExtractOptions): ExtractResult;
-//#endregion
-//#region src/knowledge/types.d.ts
 type KnowledgeRequirementCategory = 'user_specific' | 'company_specific' | 'domain_specific' | 'codebase_specific' | 'market_specific' | 'regulatory' | 'tool_api' | 'credential_or_secret' | 'runtime_environment' | 'preference' | 'historical_context';
 type KnowledgeAcquisitionMode = 'ask_user' | 'search_web' | 'query_connector' | 'inspect_repo' | 'run_command' | 'infer_low_confidence' | 'not_available';
 type KnowledgeImportance = 'blocking' | 'high' | 'medium' | 'low';
@@ -657,8 +627,6 @@ interface DataAcquisitionPlan {
   questions?: UserQuestion[];
   metadata?: Record<string, unknown>;
 }
-//#endregion
-//#region src/leaderboard.d.ts
 interface LeaderboardRow {
   /** Group key — the agent-profile cellId by default, else a caller dimension. */
   key: string;
@@ -691,8 +659,6 @@ interface LeaderboardOptions {
  * Pure projection: no I/O, deterministic, safe to call on any RunRecord[].
  */
 export declare function leaderboard(records: RunRecord[], opts: LeaderboardOptions): LeaderboardRow[];
-//#endregion
-//#region src/paired-delta-test.d.ts
 interface PairedDeltaTestOptions extends PairedBootstrapOptions {
   /** Smallest candidate-minus-baseline delta that counts as improvement. Default 0. */
   threshold?: number;
@@ -753,8 +719,6 @@ export declare function minimumPairsForPairedDeltaTest(confidence?: number): num
  * interval, rather than thresholding this function's bootstrap directly.
  */
 export declare function pairedDeltaTest(before: number[], after: number[], options?: PairedDeltaTestOptions): PairedDeltaTestResult;
-//#endregion
-//#region src/partition-held-out.d.ts
 /**
  * Deterministic held-out partitioning over stable ids.
  *
@@ -837,8 +801,6 @@ interface TrainSelectionTestPartition<T> {
   selectionFraction: number;
   testFraction: number;
 }
-//#endregion
-//#region src/product-benchmark/export.d.ts
 interface ProductBenchmarkExportOptions {
   /** Source eval run directories, each containing a `records.jsonl` of RunRecord rows. */
   readonly runDirs: readonly string[];
@@ -900,9 +862,7 @@ export declare function exportProductBenchmark(options: ProductBenchmarkSingleRu
  * run through the contract validators before anything is written.
  */
 export declare function exportProductBenchmarkRuns(options: ProductBenchmarkExportOptions): ProductBenchmarkExportResult;
-//#endregion
-//#region src/product-benchmark/index.d.ts
-declare const productBenchmarkSplits: readonly ['practice', 'dev', 'holdout', 'safety', 'sentinel'];
+declare const productBenchmarkSplits: readonly ["practice", "dev", "holdout", "safety", "sentinel"];
 type ProductBenchmarkSplit = (typeof productBenchmarkSplits)[number];
 interface ProductBenchmarkRepoRef {
   readonly url: string;
@@ -979,8 +939,6 @@ export declare function readProductBenchmarkManifest(path: string): ProductBench
  * integrity, and artifact failure listed. Returns the report when clean.
  */
 export declare function assertProductBenchmarkRun(runDir: string): ProductBenchmarkValidationReport;
-//#endregion
-//#region src/runtime-trajectory.d.ts
 interface RuntimeTrajectoryHookEvent {
   id: string;
   runId: string;
@@ -1025,8 +983,6 @@ interface ProjectRuntimeTrajectoryEvidenceOptions<TRecord extends RuntimeTraject
   scenarioIdOf?: (record: TRecord, index: number) => string | undefined;
 }
 export declare function projectRuntimeTrajectoryEvidence<TRecord extends RuntimeTrajectoryRecord>(options: ProjectRuntimeTrajectoryEvidenceOptions<TRecord>): RuntimeTrajectoryEvidenceProjection;
-//#endregion
-//#region src/active-learning.d.ts
 type SynthesisReason = 'high-variance' | 'undersampled' | 'failure-cluster' | 'difficulty-gap';
 interface SynthesisTarget {
   reason: SynthesisReason;
@@ -1048,8 +1004,6 @@ interface ActiveLearningOptions {
   topK?: number;
 }
 export declare function proposeSynthesisTargets(dataset: Dataset, traceStore: TraceStore, options?: ActiveLearningOptions): Promise<SynthesisTarget[]>;
-//#endregion
-//#region src/bounded-process.d.ts
 /**
  * `runBoundedProcess` — run one command under a wall-clock bound and return
  * what it produced, leaving no descendant behind.
@@ -1246,8 +1200,6 @@ interface BoundedProcessResult {
   runnerError?: string;
 }
 export declare function runBoundedProcess(input: BoundedProcessInput): Promise<BoundedProcessResult>;
-//#endregion
-//#region src/command-runner.d.ts
 /**
  * CommandRunner — abstract subprocess execution surface.
  *
@@ -1320,8 +1272,6 @@ interface CommandRunner {
   readDir(path: string): Promise<DirEntry[]>;
 }
 export declare const localCommandRunner: CommandRunner;
-//#endregion
-//#region src/discover-personas.d.ts
 /**
  * Walk a personas directory and return every file matching the convention
  * `NN-slug.{yaml,yml,json,md}`. Sorted by filename so the numeric prefix
@@ -1356,8 +1306,6 @@ interface DiscoveredPersona {
   id: string;
 }
 export declare function discoverPersonas(dir: string, opts?: DiscoverPersonasOptions): Promise<DiscoveredPersona[]>;
-//#endregion
-//#region src/improvement-verdict.d.ts
 /** Verdict for one candidate relative to its parent. ITERATE is the neutral
  *  "keep collecting reps / no parent to compare against" state. */
 type ExperimentVerdict = 'KEEP' | 'ITERATE' | 'NOISE' | 'REGRESSION';
@@ -1416,8 +1364,6 @@ interface ImprovementVerdictResult {
   /** Human-readable reason for the verdict — for dashboards and logs. */
   reason: string;
 }
-//#endregion
-//#region src/propose-review.d.ts
 interface Verification {
   pass: boolean;
   score?: number;
@@ -1532,8 +1478,6 @@ interface LlmReviewerConfig<State, Summary = unknown> {
   systemPromptAddendum?: string;
 }
 export declare function createLlmReviewer<State, Summary = unknown>(cfg: LlmReviewerConfig<State, Summary>): ReviewFn<State, Summary>;
-//#endregion
-//#region src/propose-review-control.d.ts
 interface ProposeReviewControlState<State, Summary = unknown> {
   shot: number;
   state: State;
@@ -1576,8 +1520,6 @@ interface ProposeReviewControlConfig<State, Summary = unknown> {
   failureClassFromVerification?: (verification: Verification) => FailureClass | undefined;
   actionFailure?: ControlRuntimeConfig<ProposeReviewControlState<State, Summary>, ProposeReviewControlAction, ProposeReviewControlResult<State, Summary>>['actionFailure'];
 }
-//#endregion
-//#region src/reflective-mutation.d.ts
 interface TrialTrace {
   /** Stable id for the trial — surfaces in the prompt for grounding. */
   id: string;
@@ -1634,8 +1576,6 @@ interface ReflectionProposal {
  * fences and surrounding prose. Returns at most `maxProposals`.
  */
 export declare function parseReflectionResponse(raw: string, maxProposals?: number): ReflectionProposal[];
-//#endregion
-//#region src/eval-trace-store.d.ts
 /**
  * Backing persistence for `EvalTraceStore`. The in-memory store is the default;
  * the JSONL file store appends one validated `RunRecord` per line. Both keep an
@@ -1653,8 +1593,6 @@ interface RunRecordBackend {
 export declare function jsonlRunRecordBackend(path: string, opts?: {
   skipInvalid?: boolean;
 }): RunRecordBackend;
-//#endregion
-//#region src/integrity/preflight.d.ts
 interface ModelPreflight {
   /** The model id as supplied by the caller. */
   model: string;
@@ -1728,8 +1666,6 @@ interface PreflightOutcome {
  * unconfigured (a 401 `model_not_found` from the router) is caught.
  */
 export declare function preflightModels(opts: PreflightModelsOptions): Promise<PreflightOutcome>;
-//#endregion
-//#region src/model-seats.d.ts
 /**
  * ModelSeats — the program's model seating chart.
  *
@@ -1802,8 +1738,6 @@ export declare const seatPresets: Record<'economy' | 'frontier', ModelSeats>;
 export declare function resolveSeat(seats: ModelSeats, seat: 'judges', fallback?: string): string[];
 export declare function resolveSeat(seats: ModelSeats, seat: Exclude<SeatName, 'judges'>, fallback?: string): string;
 export declare function resolveSeat(seats: ModelSeats, seat: SeatName, fallback?: string): string | string[];
-//#endregion
-//#region src/prompt-registry.d.ts
 /**
  * Versioned prompt registry.
  *
@@ -1844,8 +1778,6 @@ export declare class PromptRegistry {
 }
 /** SHA-256(content) → first 12 hex chars. Stable across runtimes. */
 export declare function hashContent(content: string): Promise<string>;
-//#endregion
-//#region src/reference-replay.d.ts
 /**
  * Reference replay — score an agent against withheld historical outcomes.
  *
@@ -1930,8 +1862,6 @@ interface ReferenceReplayRun<Input = unknown> {
   score: ReferenceReplayScore;
   metadata?: Record<string, unknown>;
 }
-//#endregion
-//#region src/steering.d.ts
 interface SteeringRolePrompt {
   system?: string;
   append?: string;
@@ -1945,8 +1875,6 @@ interface SteeringBundle {
   rolePrompts?: Record<string, SteeringRolePrompt>;
   metadata?: Record<string, unknown>;
 }
-//#endregion
-//#region src/run-score.d.ts
 interface RunScore {
   success: number;
   goalProgress: number;
@@ -1976,8 +1904,6 @@ interface RunScoreWeights {
 }
 export declare function aggregateRunScore(score: RunScore, weights?: Partial<RunScoreWeights>): number;
 export declare function clamp01(value: number): number;
-//#endregion
-//#region src/steering-optimizer.d.ts
 type SteeringOptimizerBackend = 'pairwise';
 interface SteeringOptimizationRow {
   variantId: string;
@@ -2016,8 +1942,6 @@ interface SteeringOptimizerConfig {
 export declare class PairwiseSteeringOptimizer {
   optimize(rows: SteeringOptimizationRow[], config?: SteeringOptimizerConfig): SteeringOptimizationResult;
 }
-//#endregion
-//#region src/counterfactual.d.ts
 type CounterfactualMutation = {
   kind: 'swap-model';
   at: number;
@@ -2070,8 +1994,6 @@ interface CounterfactualRunner {
    */
   executeFrom: (ctx: CounterfactualContext, emitter: TraceEmitter) => Promise<void>;
 }
-//#endregion
-//#region src/scorecard.d.ts
 /** One commit's measurement of one (scenario, profile) cell. */
 interface ScorecardEntry {
   commitSha: string;
@@ -2184,8 +2106,6 @@ export declare function diffScorecard(scorecard: Scorecard, opts?: DiffScorecard
  * enumerated.
  */
 export declare function formatScorecardDiff(diff: ScorecardDiff): string;
-//#endregion
-//#region src/trace-contracts/types.d.ts
 /**
  * Minimal structural span the checker reads. The eval-side `Span`
  * (trace/schema) and the output of {@link contractSpansFromOtlp} satisfy it.
@@ -2262,11 +2182,11 @@ interface SpanPredicate {
    *  instead of silently dropping the check. */
   requiresCustom?: true;
 }
-declare const ORDER_MODES: readonly ['start-order', 'finish-before-start', 'all-occurrences'];
+declare const ORDER_MODES: readonly ["start-order", "finish-before-start", "all-occurrences"];
 type OrderMode = (typeof ORDER_MODES)[number];
-declare const ARGUMENT_OCCURRENCES: readonly ['first', 'last', 'any', 'all'];
+declare const ARGUMENT_OCCURRENCES: readonly ["first", "last", "any", "all"];
 type ArgumentOccurrence = (typeof ARGUMENT_OCCURRENCES)[number];
-declare const ARGUMENT_TYPES: readonly ['string', 'number', 'boolean', 'object', 'array', 'null'];
+declare const ARGUMENT_TYPES: readonly ["string", "number", "boolean", "object", "array", "null"];
 type ArgumentType = (typeof ARGUMENT_TYPES)[number];
 type ArgumentCheck = {
   op: 'exists';
@@ -2417,8 +2337,6 @@ interface EvaluateTraceContractOptions {
   /** The run `run` rules read. Ignored under `scope`, where the scope root is the run. */
   run?: ContractRun;
 }
-//#endregion
-//#region src/trace-contracts/builder.d.ts
 interface OrderOptions {
   order?: OrderMode;
 }
@@ -2470,8 +2388,6 @@ declare class TraceContractBuilder {
   private add;
 }
 export declare function traceContract(name: string): TraceContractBuilder;
-//#endregion
-//#region src/trace-contracts/evaluate.d.ts
 /**
  * Evaluate one contract over a span sequence. Pure and synchronous — works on
  * `Span[]` from a TraceStore, `contractSpansFromOtlp` output, or any array
@@ -2485,12 +2401,8 @@ export declare function evaluateTraceContract(contract: TraceContract, spans: re
  * contract list — `allValid: true` over zero contracts is a silent pass.
  */
 export declare function checkTraceContracts(spans: readonly ContractSpan[], contracts: readonly TraceContract[], options?: EvaluateTraceContractOptions): ContractCheckResult;
-//#endregion
-//#region src/trace-contracts/explain.d.ts
 /** Plain-language statement of everything a contract checks, one rule per line. */
 export declare function explainTraceContract(contract: TraceContract): string;
-//#endregion
-//#region src/trace-contracts/spec.d.ts
 interface ToolArgumentSpec {
   tool: string;
   /** RFC 6901 JSON Pointer into the call's arguments; `''` is the whole value. */
@@ -2583,8 +2495,6 @@ interface ContractLintFinding {
  * strictly first, so a malformed spec throws instead of linting.
  */
 export declare function lintTraceContractSpec(input: unknown): ContractLintFinding[];
-//#endregion
-//#region src/action-policy.d.ts
 interface ActionExecutionPolicy {
   allowedTypes?: string[];
   blockedTypes?: string[];
@@ -2607,8 +2517,6 @@ interface ActionPolicyDecision {
 export declare function evaluateActionPolicy(action: ProposedSideEffect, policy?: ActionExecutionPolicy, options?: {
   createdAt?: string;
 }): ActionPolicyDecision;
-//#endregion
-//#region src/contamination-guard.d.ts
 interface CanaryLeak {
   scenarioId: string;
   canary: string;
@@ -2616,8 +2524,6 @@ interface CanaryLeak {
   evidence: string;
 }
 export declare function checkCanaries(output: string, scenarios: DatasetScenario[]): CanaryLeak[];
-//#endregion
-//#region src/deploy-gate-layer.d.ts
 type DeployFamily = 'frontend-static' | 'nextjs' | 'remix' | 'fullstack-ts';
 interface DeployRunResult {
   ok: boolean;
@@ -2723,8 +2629,6 @@ interface WranglerDeployRunnerInput {
  * — the gate caller decides whether to require deploy validation).
  */
 export declare function wranglerDeployRunner(input: WranglerDeployRunnerInput): DeployRunner;
-//#endregion
-//#region src/hidden-criteria-grading.d.ts
 /**
  * Hidden-criteria grading firewall — grade an agent on criteria it never saw.
  *
@@ -2867,8 +2771,6 @@ interface JudgeScoreInput<TArtifact> {
  * to [0,1] — a judge on a non-unit scale must be normalized by the caller first.
  */
 export declare function blendHeldout(heldoutPassRate: number, judgeScore: number, weights?: BlendWeights): number;
-//#endregion
-//#region src/integrity/single-backend.d.ts
 /**
  * Minimal backend-config shape the assertion reads. Consumers may pass richer
  * types — only these five fields are inspected.
@@ -2915,8 +2817,6 @@ interface SingleBackendReport {
  * agent. Returns the report so callers can log it in either case.
  */
 export declare function assertSingleBackend(agent: BackendDescriptor, judge: BackendDescriptor, opts?: AssertSingleBackendOptions): SingleBackendReport;
-//#endregion
-//#region src/intent-match-judge.d.ts
 interface IntentMatchInput {
   /** The full natural-language prompt the agent was handed. */
   userRequest: string;
@@ -2963,8 +2863,6 @@ interface IntentMatchOptions {
  * Run the intent-match judge. Soft-fails to available=false on error.
  */
 export declare function runIntentMatchJudge(input: IntentMatchInput, options: IntentMatchOptions): Promise<IntentMatchResult>;
-//#endregion
-//#region src/judge-ensemble.d.ts
 /** One judge's verdict. `perDimension: null` ⇒ that judge call failed (threw or
  *  returned malformed output) — recorded, never folded into a zero. */
 interface JudgeVerdict<D extends string = string> {
@@ -3015,8 +2913,6 @@ interface EnsembleAggregate<D extends string = string> {
  * - Throws if `verdicts` or `dimensionKeys` is empty, or if every judge failed.
  */
 export declare function aggregateJudgeVerdicts<D extends string>(verdicts: readonly JudgeVerdict<D>[], dimensionKeys: readonly D[], weights?: Partial<Record<D, number>>): EnsembleAggregate<D>;
-//#endregion
-//#region src/judge-retry.d.ts
 /**
  * Wrap a single judge LLM call with retry, optional fallback-model
  * rotation, exponential backoff, and a typed `JudgeRetryOutcome`. Callers
@@ -3086,8 +2982,6 @@ interface JudgeRetryOutcome<T> {
  * synthetic zero is indistinguishable from a real low score downstream.
  */
 export declare function withJudgeRetry<T>(judgeFn: (model: string, signal: AbortSignal) => Promise<T>, policy?: JudgeRetryPolicy): Promise<JudgeRetryOutcome<T>>;
-//#endregion
-//#region src/judge-panel.d.ts
 interface EnsembleJudgeOptions<D extends string> {
   /** Judge name — becomes the returned `JudgeConfig.name`. */
   name: string;
@@ -3144,8 +3038,6 @@ interface EnsembleJudgeOptions<D extends string> {
  * failed cell instead of averaging a fabricated zero.
  */
 export declare function ensembleJudge<D extends string>(opts: EnsembleJudgeOptions<D>): JudgeConfig<unknown>;
-//#endregion
-//#region src/keyword-coverage-judge.d.ts
 /**
  * Keyword-coverage judge — baseline complement to the semantic concept
  * judge.
@@ -3221,8 +3113,6 @@ export declare function runKeywordCoverageJudge(html: string, expectedConcepts:
  * then score via {@link runKeywordCoverageJudge}.
  */
 export declare function runKeywordCoverageJudgeUrl(previewUrl: string, expectedConcepts: ReadonlyArray<KeywordConceptSpec>, options?: KeywordCoverageOptions): Promise<KeywordCoverageResult>;
-//#endregion
-//#region src/oracle.d.ts
 interface OracleObservation {
   /** Final observable text output from the agent (response, page snapshot, stdout). */
   text?: string;
@@ -3254,8 +3144,6 @@ interface OracleReport extends DefaultVerdict {
 }
 /** Run all oracles against one observation and aggregate. */
 export declare function evaluateOracles(obs: OracleObservation, oracles: Oracle[]): OracleReport;
-//#endregion
-//#region src/paired-promotion-power.d.ts
 /** A serializable law for one scalar draw. */
 type PairedValueLaw = {
   kind: 'point';
@@ -3411,8 +3299,6 @@ export declare function pairedPromotionPower(options: PairedPromotionPowerOption
  * recorded and the scan continues.
  */
 export declare function requiredPairsForPairedPromotion(options: RequiredPairsForPairedPromotionOptions): RequiredPairsForPairedPromotionResult;
-//#endregion
-//#region src/treatment-gate.d.ts
 /**
  * Treatment-applied gate — "did the treatment's tool actually fire this run?"
  *
@@ -3488,8 +3374,6 @@ interface TreatmentGate {
  * captured AND zero matching calls appear; an empty histogram fails open.
  */
 export declare function gateTreatmentApplied(input: TreatmentGateInput, opts?: TreatmentGateOptions): TreatmentGate;
-//#endregion
-//#region src/verdict-cache.d.ts
 /**
  * Canonical JSON (RFC 8785): object keys sorted recursively, so two
  * semantically equal values produce byte-identical output regardless of key
@@ -3515,8 +3399,6 @@ interface VerdictCacheStore {
  * a half-written file that needs operator attention.
  */
 export declare function fileVerdictCache(path: string): VerdictCacheStore;
-//#endregion
-//#region src/profile/baselines.d.ts
 /**
  * Strong, generically-useful baseline ROLES — the top zone of an `AgentProfile`
  * before any domain layer. A product composes one of these with its own
@@ -3640,8 +3522,6 @@ declare function applyDomainPatch(p: PromptProfile, sectionId: string, newBody:
  * content.
  */
 declare function sectionHash(section: AgentProfileSection): string;
-//#endregion
-//#region src/budget-guard.d.ts
 export declare class BudgetBreachError extends AgentEvalError {
   dimension: keyof BudgetSpec;
   limit: number;
@@ -3660,8 +3540,6 @@ export declare class BudgetGuard {
   tickWall(nowMs: number, spanId?: string): Promise<void>;
   get state(): Record<keyof BudgetSpec, number>;
 }
-//#endregion
-//#region src/client.d.ts
 /**
  * ProductClient — configurable HTTP client for exercising any agent's APIs.
  *
@@ -3739,15 +3617,9 @@ export declare class ProductClient {
   /** Generic PATCH for custom routes */
   patch(path: string, body: Record<string, unknown>): Promise<Record<string, unknown>>;
 }
-//#endregion
-//#region src/control-runtime.d.ts
 export declare function runAgentControlLoop<TState, TAction, TActionResult, TEval extends ControlEvalResult = ControlEvalResult>(config: ControlRuntimeConfig<TState, TAction, TActionResult, TEval>): Promise<ControlRunResult<TState, TAction, TActionResult, TEval>>;
-//#endregion
-//#region src/control-runtime/policy.d.ts
 export declare function objectiveEval(input: Omit<ControlEvalResult, 'objective'>): ControlEvalResult;
 export declare function subjectiveEval(input: Omit<ControlEvalResult, 'objective'>): ControlEvalResult;
-//#endregion
-//#region src/knowledge/readiness.d.ts
 interface ScoreKnowledgeReadinessOptions {
   taskId: string;
   requirements: KnowledgeRequirement[];
@@ -3770,8 +3642,6 @@ export declare function knowledgeReadinessTracePayload(report: KnowledgeReadines
 }): Record<string, unknown>;
 export declare function userQuestionsForKnowledgeGaps(gaps: KnowledgeRequirement[]): UserQuestion[];
 export declare function acquisitionPlansForKnowledgeGaps(gaps: KnowledgeRequirement[]): DataAcquisitionPlan[];
-//#endregion
-//#region src/metrics.d.ts
 interface TokenPrice {
   input: number;
   output: number;
@@ -3791,8 +3661,6 @@ export declare function estimateTokens(text: string): number;
  *  once (not a silent zero) and return 0 so callers that ignore pricing keep
  *  working; cost-sensitive callers should gate on {@link isModelPriced}. */
 export declare function estimateCost(inputTokens: number, outputTokens: number, model: string): number;
-//#endregion
-//#region src/anti-slop.d.ts
 interface AntiSlopConfig {
   /** Domain label — appears in the DimensionJudgeScore output */
   domain?: string;
@@ -3835,6 +3703,5 @@ interface AntiSlopReport {
 export declare function analyzeAntiSlop(outputs: string[], config: Omit<Required<AntiSlopConfig>, 'domain'> & {
   penaltyWeights: Record<SlopCategory, number>;
 }): AntiSlopReport;
-//#endregion
-export { AGENT_PROFILE_KINDS, ATTESTATION_ALGORITHM, type ActionExecutionPolicy, type ActionPolicyDecision, type ActionableSideInfo, type ActiveLearningOptions, type AgentEvalAgent, AgentEvalError, type AgentEvalErrorCode, type AgentEvalEvaluateOptions, type AgentEvalImproveOptions, type AgentProfile, type AgentProfileCell, type AgentProfileCellInput, type AgentProfileDimensionValue, type AgentProfileJson, type AgentProfileSourceInput, type Analyst, type AnalystContext, type AnalystFeedbackTrajectoryOptions, type AnalystFinding, type AnalystFindingDigest, AnalystRegistry, type AnalystRegistryOptions, type AnalystReviewDecision, type AnalystReviewRequest, type AnalystRunDigest, type AnalystRunEvent, type AnalystRunInputs, type AnalystRunResult, type AnalystRunSummary, type AnalystSeverity, type AnalystUsageReceipt, type AnalyzeRunsOptions, type AnalyzeTracesResult, type ArgumentCheck, type ArmAssignment, type ArmPopulationReading, type ArmPopulationSummary, type Artifact, type ArtifactCheck, type ArtifactEventLike, type ArtifactResult, type AssertCapabilityHeadroomOptions, type AssertCrossFamilyOptions, type AssertCrossFamilyServedOptions, type AssertServedModelOptions, type AssertSingleBackendOptions, type AttestationProvenance, type AttestationVerification, type AttestedReport, BOOTSTRAP_GATE_MIN_N, type BackendDescriptor, BackendIntegrityError, type BackendIntegrityReport, type BenchmarkEvaluation, type BlendWeights, type BootstrapOptions, type BootstrapResult, type BoundedProcessInput, type BoundedProcessResult, type BudgetLedgerEntry, type BudgetPolicy, type BudgetSpec, CODING_HARNESSES, type CalibrationExample, type CalibrationMerge, type CalibrationResult, type CampaignCellFailureReceipt, type CampaignCellRetryPolicy, type CampaignFactoryParams, type CampaignIntegrityPolicy, type CampaignResult, type CampaignRunContext, type CampaignRunOutcome, type CampaignScenario, type CampaignVariant, type CanaryAlert, type CanaryEvaluation, type CanaryKind, type CanaryLeak, type CanaryOptions, type CanaryReport, type CandidateScore, type CapabilityHeadroomOptions, type CapabilityHeadroomResult, type CaptureErrorHandler, type CaptureWrite, type CellVerdict, type ChannelRollup, type ChatCallOpts, type ChatClient, type ChatRequest, type ChatResponse, type ChatTraceEngineOptions, type ChatTransport, type CheckResult, type CheckerIdentity, type CheckerOutcome, type CliffsMagnitude, type CollectedArtifacts, type CommandRunner, type ComparePairedArmsOptions, type CompletionCriterion, type CompletionRequirement, type CompletionVerdict, type ConceptFinding, type ConceptSpec, ConfigError, type ContinuousAgreement, type ContinuousAgreementOptions, type ContinuousCalibrationResult, type ContractCheckResult, type ContractLintFinding, type ContractRule, type ContractRuleExecution, type ContractRun, type ContractSpan, type ContractStatus, type ContractVerdict, type ControlActionOutcome, type ControlBudget, type ControlContext, type ControlDecision, type ControlEvalResult, type ControlRunResult, type ControlRuntimeConfig, type ControlRuntimeError, type ControlStep, type CorpusAgreementOptions, type CorpusAgreementPerDimension, type CorpusAgreementReport, type CorpusScoreRecord, type CorrectnessChecker, type CostChannel, type CostEntry, CostLedger, type CostLedgerFilter, type CostLedgerHandle, type CostLedgerOptions, type CostLedgerPersistence, type CostLedgerSummary, type CostProvenance, type CostReceipt, type CostReceiptInput, type CostResult, type CostSummary, type CostUsage, type CounterfactualContext, type CounterfactualMutation, type CounterfactualResult, type CounterfactualRunner, type CreateChatClientOpts, type CreateTraceAnalystOptions, CrossFamilyError, type CustomTokenPricing, type CustomTransportOpts, DEFAULT_JUDGE_GATE_POLICY, DEFAULT_RED_TEAM_CORPUS, DEFAULT_TRACE_ANALYST_BUDGETS, DEFAULT_TRACE_ANALYST_KINDS, type DataAcquisitionPlan, type DatasetManifest, type DatasetOverview, type DatasetScenario, type DatasetSplit, type DefaultAnalystRegistryOptions, type DefaultVerdict, type DefineAgentEvalOptions, type DefinedAgentEval, type DeployGateLayerInput, type DeployRunResult, type DeployRunner, type DetectorEvent, type DetectorSignal, type DiffPolicy, type DiffScorecardOptions, type DimensionJudgeScore, type DirEntry, type DiscoverPersonasOptions, type DiscoveredPersona, type DriverState, type DspyRlmTraceEngineOptions, type EProcess, type EProcessOptions, type EProcessState, type EProcessStep, type EnsembleAggregate, type EnsembleJudgeOptions, type EquivalenceArm, type EquivalenceCheckDefinition, type EquivalenceCheckSpec, type EquivalenceChecker, type EquivalenceCheckerInput, type EquivalenceCheckerResult, type EquivalenceObligation, type EquivalenceObligationStatus, EquivalenceProtocolError, type EquivalenceRecord, type ErrorCluster, type ErrorCountPattern, type ErrorStreakOptions, type EvalCampaignOptions, type EvalCampaignResult, type EventFilter, type EvidenceRef, type ExactAnalystRunEvent, type ExactAnalystRunResult, type ExactCapableAnalyst, type ExactRegistryRunOpts, type ExactRiskDifferenceResult, type ExperimentPlan, type ExperimentRep, type ExperimentResult, type ExperimentStats, type ExtractOptions, type ExtractResult, type ExtractedUsage, FAILURE_CLASSES, FAILURE_MODE_KIND_SPEC, type FailedRun, type FailureBlame, type FailureClass, type FailureClassification, type FailureContext, type FailureMode, type FailureReasonRule, type FailureRule, type FeedbackArtifactType, type FeedbackAttempt, type FeedbackLabel, type FeedbackLabelKind, type FeedbackLabelSource, type FeedbackOptimizerRow, type FeedbackOutcome, type FeedbackSplitPolicy, type FeedbackTask, type FeedbackTrajectory, type FeedbackTrajectoryFilter, type FeedbackTrajectoryStore, type FieldDestination, FileSystemFeedbackTrajectoryStore, FileSystemRawProviderSink, FileSystemTraceStore, type Finding, type FindingSubject, type FindingSubjectKind, type FindingToPolicyEditOptions, type FindingsDiff, FindingsStore, type GainDistributionBin, type GainDistributionFigureSpec, type GainDistributionOptions, type GateDecision, type GateEvidence, type GenericSpan, type GoldenItem, HARNESS_NATIVE_MODEL, type HarnessType, type HeadroomInput, HeldOutGate, type HeldOutGateConfig, type HeldOutGateDecision, type HeldOutGateRejectionCode, type HeldOutPartition, type HiddenCriteriaGrader, type HiddenGradeResult, type HiddenLeak, type ImprovementThresholds, type ImprovementVerdictResult, InMemoryFeedbackTrajectoryStore, InMemoryRawProviderSink, InMemoryTraceStore, type InsightReport, type IntentMatchInput, type IntentMatchOptions, type IntentMatchResult, type InterimReleaseConfidence, type InterimReleaseConfidenceInput, type JudgeAgreement, type JudgeCalibrationSet, type JudgeConfusion, JudgeError, type JudgeExampleAgreement, type JudgeFamily, type JudgeFn, type JudgeGateDecision, type JudgeGateInput, type JudgeGatePolicy, JudgeGateRefusedError, type JudgeGateStatus, type JudgeInput, type JudgeRetryOutcome, type JudgeRetryPolicy, type JudgeRubric, type JudgeRun, type JudgeRunMeta, type JudgeScore, type JudgeScoreInput, type JudgeScoresRecord, type JudgeSpan, type JudgeVerdict, type KeywordConceptSpec, type KeywordCoverageFinding, type KeywordCoverageOptions, type KeywordCoverageResult, type KnowledgeAcquisitionMode, type KnowledgeBundle, type KnowledgeFreshness, type KnowledgeImportance, type KnowledgeReadinessReport, type KnowledgeRequirement, type KnowledgeRequirementCategory, type KnowledgeSensitivity, type Layer, type LayerResult, type LayerStatus, type LeaderboardOptions, type LeaderboardRow, type LinearRegressionResult, type LlmCallMetadata, type LlmCallRequest, type LlmCallResult, type LlmChargeBounds, type LlmCorrectnessCheckerOpts, type LlmJudgeOptions, type LlmMessage, type LlmReviewerConfig, type LlmSpan, type LlmTokenLogprob, type LlmToolCall, type LlmToolChoice, type LlmToolDefinition, type LlmUsage, type MannWhitneyResult, type MatchedPair, type MatchedRunRecordPair, type MaximumCharge, type McNemarResult, type MintRolloutOptions, type MintRolloutResult, type MintedRolloutLine, type MockTransportOpts, type ModelEndpointCheck, type ModelEndpointRequest, type ModelSeats, ModelSubstitutionError, MultiLayerVerifier, type NoLeakOptions, NoopRawProviderSink, NotFoundError, OUTPUT_VALUE, type Objective, type OpenAiCompatibleTransportOpts, type Oracle, type OracleObservation, type OracleReport, type OracleResult, type OrderMode, type OtlpExport, OtlpFileTraceStore, type OtlpFlatLine, type OtlpSpan, type OwnerVerdict, type PaidCallResult, type PairArmsOptions, type PairArmsResult, type PairRunRecordsResult, type PairedArmRow, type PairedArmsComparison, type PairedBootstrapOptions, type PairedBootstrapResult, type PairedCorrectness, type PairedDecisionMethod, type PairedDecisionShape, type PairedDecisionStatistic, type PairedDeltaTestOptions, type PairedDeltaTestResult, type PairedEvalueOptions, type PairedEvalueSequence, type PairedEvalueStep, type PairedMcNemarEvidence, type PairedMetricDelta, type PairedPromotionAlternative, type PairedPromotionAlternativeCell, type PairedPromotionCallRates, type PairedPromotionDecision, type PairedPromotionDecisionOptions, type PairedPromotionPowerCall, type PairedPromotionPowerOptions, type PairedPromotionPowerResult, type PairedSignTestResult, type PairedTTestResult, type PairedValueLaw, type ParetoPoint, type ParetoResult, type PartitionHeldOutOptions, type PendingCostCall, type PendingCostCallView, type PersistedFinding, type PersonaConfig, type PolicyEdit, type PolicyEditAdmission, type PolicyEditAdmissionOptions, type PolicyEditAxis, type PolicyEditCandidateRecord, type PolicyEditChange, type PolicyEditExpectedGain, type PolicyEditGainDirection, type PolicyEditGainUnit, type PolicyEditInit, type PolicyEditRisk, type PolicyEditSchemaVersion, type PolicyEditSource, type PolicyEditTarget, type PolicyEditTargetSurface, type PreferenceMemoryEntry, type ProducedProposal, type ProducedState, type ProductBenchmarkExportOptions, type ProductBenchmarkExportResult, type ProductBenchmarkManifest, type ProductBenchmarkSingleRunExportOptions, type ProductBenchmarkSplit, type ProductBenchmarkValidationReport, type ProductClientConfig, type ProductScenario, type ProfileAxisSpec, type ProjectRuntimeTrajectoryEvidenceOptions, type PromptHandle, type ProportionInterval, type ProposalEventLike, type ProposalFinding, type ProposeFn, type ProposeInput, type ProposeOutput, type ProposeReviewConfig, type ProposeReviewControlAction, type ProposeReviewControlConfig, type ProposeReviewControlResult, type ProposeReviewControlState, type ProposeReviewReport, type ProposedSideEffect, type QueryTracesPage, REDACTION_VERSION, type RankTestMethod, type RankTestMethodRequest, type RankTestOptions, type RawAnalystFinding, type RawProviderEvent, type RawProviderSink, type RecordRunsOptions, type RedTeamCase, type RedTeamCategory, type RedTeamFinding, type RedTeamReport, type RedactOptions, type RedactionProfile, type RedactionReport, type ReferenceReplayCaseRun, type ReferenceReplayRun, type ReferenceReplaySplit, type ReflectionContext, type ReflectionProposal, type RegistryRunOpts, type ReleaseConfidenceInput, type ReleaseConfidenceIssue, type ReleaseConfidenceMetrics, type ReleaseConfidenceScorecard, type ReleaseTraceEvidence, type RepeatedActionOptions, type RequiredPairsForPairedPromotionOptions, type RequiredPairsForPairedPromotionResult, type RequirementCheck, type ResearchReport, type ResearchReportOptions, type Researcher, type RetryWrite, type Review, type ReviewFn, type ReviewInput, type ReviewMemoryEntry, type ReviewMemoryStore, type RiskDifferenceResult, type RolloutSearchLineage, type RouteMap, type RoutedField, type RubricDimension, type Run, type RunCampaignOptions, type RunCaptureReport, type RunCommandInput, type RunCommandResult, type RunCostProvenance, type RunFilter, RunIntegrityError, type RunIntegrityReport, type RunJudgeMetadata, type RunOutcome, type RunPaidCallInput, type RunRecord, type RunRecordBackend, RunRecordValidationError, type RunScore, type RunScoreWeights, type RunSearchCoordinates, type RunSplitTag, type RunStatus, type RunTaskFailure, type RunTerminalOutcome, type RunTokenUsage, type RunTraceRef, type RuntimeEventLike, type RuntimeTrajectoryEvidenceProjection, type RuntimeTrajectoryEvidenceSummary, type RuntimeTrajectoryHookEvent, type RuntimeTrajectoryRecord, type RuntimeTrajectoryRunRecord, SEMANTIC_CONCEPT_JUDGE_VERSION, type SampleSummary, type SampleSummaryOptions, type SandboxDriver, type SandboxSdkTransportOpts, type SatisfiedBy, type Scenario, type ScenarioCost, type ScenarioFile, type ScenarioResult, type ScorePreference, type ScoreRiskDifferenceResult, type Scorecard, type ScorecardCell, type ScorecardCellDiff, type ScorecardDiff, type ScorecardEntry, type ScorecardLogLine, type SearchSpanResult, type SearchTraceResult, type SelfImproveBudget, type SelfImproveMethodOptions, type SelfImproveMethodProvenance, type SelfImproveMethodResult, type SelfImproveOptions, type SelfImproveProgressEvent, type SelfImproveProposerOptions, type SelfImproveProposerResult, type SelfImproveResult, type SemanticConceptJudgeInput, type SemanticConceptJudgeOptions, type SemanticConceptJudgeResult, type SequentialDecision, type SeriesConvergenceOptions, type SeriesConvergenceResult, type SeriesDistribution, type ServedModelCheck, type ServedModelPolicy, type ServedModelVerdict, type Severity, type ShareSafetyStatus, type ShareSafetyVerdict, type SignTestAlternative, type SingleBackendDivergence, type SingleBackendReport, type Span, type SpanFilter, type SpanHandle, type SpanMatchRecord, type SpanPredicate, type SpanRecordsToTraceAnalysisStoreOptions, type SplitCoverage, type SteeringBundle, type SteeringChange, type SteeringOptimizationResult, type SteeringOptimizationRow, type StopDecision, type StrategyChecker, type StreamingDetector, type SummaryTable, type SummaryTableOptions, type SummaryTableRow, type SynthesisTarget, type TaskGold, type TaskHeadroom, type ToolCallEventLike, type ToolSpan, type ToolStats, type ToolUseMetrics, type ToolUseOptions, type TraceAnalysisEngine, type TraceAnalysisEngineResult, type TraceAnalysisStore, type TraceAnalystByteBudgets, type TraceAnalystDefinition, type TraceAnalystFilters, type TraceAnalystSpan, type TraceAnalystSpanKind, type TraceAnalystSpanStatus, type TraceAnalystTraceSummary, type TraceContract, type TraceContractSpec, TraceEmitter, type TraceEmitterOptions, type TraceEvent, type TraceInsightReadiness, type TraceInsightSuite, type TraceStore, type TrainSelectionTestOptions, type TrainSelectionTestPartition, type Trajectory, type TrajectoryStep, type TreatmentGate, type TreatmentGateInput, type TreatmentGateOptions, type TrialTrace, type Turn, type TurnResult, UNKNOWN_MODEL, type UserQuestion, VERIFICATION_STRATEGIES, VERIFICATION_STRATEGY_SOURCES, ValidationError, type VerbosityBiasResult, type Verdict, type VerdictCacheStore, type VerdictCertification, type Verification, type VerificationReport, type VerificationStrategyProfile, type VerificationStrategySource, type VerifyFn, type VerifyOptions, type ViewSpansResult, type ViewTraceOversized, type ViewTraceResult, type ViteDeployRunnerInput, type WeightedCompositeInput, type WeightedCompositeResult, type WelchTestResult, type WelchTestStatus, type WilcoxonSignedRankResult, type WranglerDeployRunnerInput, addCalibrationVerdicts, agentProfileCellKey, agentProfileHash, agentProfileId, analystRunDigest, analystRunToFeedbackTrajectory, analystRunToReviewRequests, analyzeRuns, analyzeSeries, analyzeTraces, argHash, assertCrossFamily, assertCrossFamilyServed, assertJudgeMayGate, assertRealBackend, assertRunCaptured, assertServedModel, assertServedModels, assessShareSafety, assignFeedbackSplit, attest, benjaminiHochberg, bonferroni, bootstrapCi, budgetBreachView, buildAgentProfileCell, buildDefaultAnalystRegistry, buildTraceInsightContext, buildTraceInsightPrompt, buildTrajectory, calibrateJudge, calibrateJudgeContinuous, captureFetchToRawSink, certificationEvidenceDigest, checkServedModel, classifyFailure, classifyFailureReason, comparePairedArms, completionVerdict, computeFindingId, computeToolUseMetrics, confidenceInterval, continuousAgreement, controlRunToFeedbackTrajectory, corpusInterRaterAgreement, corpusInterRaterAgreementFromJudgeScores, createBoundedTraceAnalysisStore, createChatClient, createChatTraceEngine, createDspyRlmTraceEngine, createFeedbackTrajectory, createLlmCorrectnessChecker, createTraceAnalyst, decidePairedPromotion, defineAgentEval, defineEquivalenceCheck, describeServedModelCheck, describeTraceInsightScope, diffFindings, domainEvidencePattern, dominates, eProcess, equivalenceVerdict, evaluateInterimReleaseConfidence, evaluateReleaseConfidence, expandProfileAxes, exportRunAsOtlp, extractProducedState, extractUsage, extractUsageFromSse, failureBlame, failureClusterView, feedbackTrajectoriesToDatasetScenarios, feedbackTrajectoriesToOptimizerRows, feedbackTrajectoryToOptimizerRow, gainHistogram, gradeSemanticStatus, groupRunsByAgentProfileCell, harnessAxisOf, hashJson, holm, inferDomainKeywords, interRaterReliability, iqr, isJudgeSpan, isLlmSpan, isRunRecord, isToolSpan, judgeAgreementView, judgeFamily, judgeGateDecision, judgeSpans, llmJudge, makeFinding, makeProposalFinding, manifestContentDigest, mannWhitneyU, mcnemar, mcnemarPower, mcnemarRequiredN, measureJudgeAgreement, mintRolloutRows, modelHasSnapshot, mulberry32, otlpTextToTraceAnalysisStore, pairArms, pairRunRecords, pairedBootstrap, pairedEvalueSequence, pairedMde, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, pairedSignTest, pairedTTest, paretoFrontier, parseRunRecordSafe, partialCredit, passAtK, pearsonR, index_d_exports as profile, ranks, redTeamDataset, redTeamReport, redact, redactForShare, redactText, registerCalibrationSet, renderPreferenceMemoryMarkdown, requiredPairedSampleSize, requiredSampleSize, runCampaign, runCanaries, runEquivalenceCheck, runEvalCampaign, runSemanticConceptJudge, runTaskScore, runsForScenario, scoreRedTeamOutput, scoreTraceInsightReadiness, selfImprove, shareAllowed, snapshotFromJudgeAgreement, spanRecordsToTraceAnalysisStore, spearmanR, summarizeBackendIntegrity, summarizeNumberSeries, summarizePreferenceMemory, summaryTable, toAgentProfileJson, tokenizeDomainWords, toolSpansToTraceAnalysisStore, toolWasteView, transientDispatchFailure, validateRunRecord, verbosityBias, verifyAgentProfileCell, verifyAttestation, verifyCompletion, weightedComposite, weightedMean, welchsTTest, wilcoxonSignedRank, wilson, withAssignedFeedbackSplit };
+export { AGENT_PROFILE_KINDS, ATTESTATION_ALGORITHM, type ActionExecutionPolicy, type ActionPolicyDecision, type ActionableSideInfo, type ActiveLearningOptions, type AgentEvalAgent, AgentEvalError, type AgentEvalErrorCode, type AgentEvalEvaluateOptions, type AgentEvalImproveOptions, type AgentProfile, type AgentProfileCell, type AgentProfileCellInput, type AgentProfileDimensionValue, type AgentProfileJson, type AgentProfileSourceInput, type Analyst, type AnalystContext, type AnalystFeedbackTrajectoryOptions, type AnalystFinding, type AnalystFindingDigest, AnalystRegistry, type AnalystRegistryOptions, type AnalystReviewDecision, type AnalystReviewRequest, type AnalystRunDigest, type AnalystRunEvent, type AnalystRunInputs, type AnalystRunResult, type AnalystRunSummary, type AnalystSeverity, type AnalystUsageReceipt, type AnalyzeRunsOptions, type AnalyzeTracesResult, type ArgumentCheck, type ArmAssignment, type ArmPopulationReading, type ArmPopulationSummary, type Artifact, type ArtifactCheck, type ArtifactEventLike, type ArtifactResult, type AssertCapabilityHeadroomOptions, type AssertCrossFamilyOptions, type AssertCrossFamilyServedOptions, type AssertServedModelOptions, type AssertSingleBackendOptions, type AttestationProvenance, type AttestationVerification, type AttestedReport, BOOTSTRAP_GATE_MIN_N, type BackendDescriptor, BackendIntegrityError, type BackendIntegrityReport, type BenchmarkEvaluation, type BlendWeights, type BootstrapOptions, type BootstrapResult, type BoundedProcessInput, type BoundedProcessResult, type BudgetLedgerEntry, type BudgetPolicy, type BudgetSpec, CODING_HARNESSES, type CalibrationExample, type CalibrationMerge, type CalibrationResult, type CampaignCellFailureReceipt, type CampaignCellRetryPolicy, type CampaignFactoryParams, type CampaignIntegrityPolicy, type CampaignResult, type CampaignRunContext, type CampaignRunOutcome, type CampaignScenario, type CampaignVariant, type CanaryAlert, type CanaryEvaluation, type CanaryKind, type CanaryLeak, type CanaryOptions, type CanaryReport, type CandidateScore, type CapabilityHeadroomOptions, type CapabilityHeadroomResult, type CaptureErrorHandler, type CaptureWrite, type CellVerdict, type ChannelRollup, type ChatCallOpts, type ChatClient, type ChatRequest, type ChatResponse, type ChatTraceEngineOptions, type ChatTransport, type CheckResult, type CheckerIdentity, type CheckerOutcome, type CliffsMagnitude, type CollectedArtifacts, type CommandRunner, type ComparePairedArmsOptions, type CompletionCriterion, type CompletionRequirement, type CompletionVerdict, type ConceptFinding, type ConceptSpec, ConfigError, type ContinuousAgreement, type ContinuousAgreementOptions, type ContinuousCalibrationResult, type ContractCheckResult, type ContractLintFinding, type ContractRule, type ContractRuleExecution, type ContractRun, type ContractSpan, type ContractStatus, type ContractVerdict, type ControlActionOutcome, type ControlBudget, type ControlContext, type ControlDecision, type ControlEvalResult, type ControlRunResult, type ControlRuntimeConfig, type ControlRuntimeError, type ControlStep, type CorpusAgreementOptions, type CorpusAgreementPerDimension, type CorpusAgreementReport, type CorpusScoreRecord, type CorrectnessChecker, type CostChannel, type CostEntry, CostLedger, type CostLedgerFilter, type CostLedgerHandle, type CostLedgerOptions, type CostLedgerPersistence, type CostLedgerSummary, type CostProvenance, type CostReceipt, type CostReceiptInput, type CostResult, type CostSummary, type CostUsage, type CounterfactualContext, type CounterfactualMutation, type CounterfactualResult, type CounterfactualRunner, type CreateChatClientOpts, type CreateTraceAnalystOptions, CrossFamilyError, type CustomTokenPricing, type CustomTransportOpts, DEFAULT_JUDGE_GATE_POLICY, DEFAULT_RED_TEAM_CORPUS, DEFAULT_TRACE_ANALYST_BUDGETS, DEFAULT_TRACE_ANALYST_KINDS, type DataAcquisitionPlan, type DatasetManifest, type DatasetOverview, type DatasetScenario, type DatasetSplit, type DefaultAnalystRegistryOptions, type DefaultVerdict, type DefineAgentEvalOptions, type DefinedAgentEval, type DeployGateLayerInput, type DeployRunResult, type DeployRunner, type DetectorEvent, type DetectorSignal, type DiffPolicy, type DiffScorecardOptions, type DimensionJudgeScore, type DirEntry, type DiscoverPersonasOptions, type DiscoveredPersona, type DriverState, type DspyRlmTraceEngineOptions, type EProcess, type EProcessOptions, type EProcessState, type EProcessStep, type EnsembleAggregate, type EnsembleJudgeOptions, type EquivalenceArm, type EquivalenceCheckDefinition, type EquivalenceCheckSpec, type EquivalenceChecker, type EquivalenceCheckerInput, type EquivalenceCheckerResult, type EquivalenceObligation, type EquivalenceObligationStatus, EquivalenceProtocolError, type EquivalenceRecord, type ErrorCluster, type ErrorCountPattern, type ErrorStreakOptions, type EvalCampaignOptions, type EvalCampaignResult, type EventFilter, type EvidenceRef, type ExactAnalystRunEvent, type ExactAnalystRunResult, type ExactCapableAnalyst, type ExactRegistryRunOpts, type ExactRiskDifferenceResult, type ExperimentPlan, type ExperimentRep, type ExperimentResult, type ExperimentStats, type ExtractOptions, type ExtractResult, type ExtractedUsage, FAILURE_CLASSES, FAILURE_MODE_KIND_SPEC, type FailedRun, type FailureBlame, type FailureClass, type FailureClassification, type FailureContext, type FailureMode, type FailureReasonRule, type FailureRule, type FeedbackArtifactType, type FeedbackAttempt, type FeedbackLabel, type FeedbackLabelKind, type FeedbackLabelSource, type FeedbackOptimizerRow, type FeedbackOutcome, type FeedbackSplitPolicy, type FeedbackTask, type FeedbackTrajectory, type FeedbackTrajectoryFilter, type FeedbackTrajectoryStore, type FieldDestination, FileSystemFeedbackTrajectoryStore, FileSystemRawProviderSink, FileSystemTraceStore, type Finding, type FindingSubject, type FindingSubjectKind, type FindingToPolicyEditOptions, type FindingsDiff, FindingsStore, type GainDistributionBin, type GainDistributionFigureSpec, type GainDistributionOptions, type GateDecision, type GateEvidence, type GenericSpan, type GoldenItem, HARNESS_NATIVE_MODEL, type HarnessType, type HeadroomInput, HeldOutGate, type HeldOutGateConfig, type HeldOutGateDecision, type HeldOutGateRejectionCode, type HeldOutPartition, type HiddenCriteriaGrader, type HiddenGradeResult, type HiddenLeak, type ImprovementThresholds, type ImprovementVerdictResult, InMemoryFeedbackTrajectoryStore, InMemoryRawProviderSink, InMemoryTraceStore, type InsightReport, type IntentMatchInput, type IntentMatchOptions, type IntentMatchResult, type InterimReleaseConfidence, type InterimReleaseConfidenceInput, type JudgeAgreement, type JudgeCalibrationSet, type JudgeConfusion, JudgeError, type JudgeExampleAgreement, type JudgeFamily, type JudgeFn, type JudgeGateDecision, type JudgeGateInput, type JudgeGatePolicy, JudgeGateRefusedError, type JudgeGateStatus, type JudgeInput, type JudgeRetryOutcome, type JudgeRetryPolicy, type JudgeRubric, type JudgeRun, type JudgeRunMeta, type JudgeScore, type JudgeScoreInput, type JudgeScoresRecord, type JudgeSpan, type JudgeVerdict, type KeywordConceptSpec, type KeywordCoverageFinding, type KeywordCoverageOptions, type KeywordCoverageResult, type KnowledgeAcquisitionMode, type KnowledgeBundle, type KnowledgeFreshness, type KnowledgeImportance, type KnowledgeReadinessReport, type KnowledgeRequirement, type KnowledgeRequirementCategory, type KnowledgeSensitivity, type Layer, type LayerResult, type LayerStatus, type LeaderboardOptions, type LeaderboardRow, type LinearRegressionResult, type LlmCallMetadata, type LlmCallRequest, type LlmCallResult, type LlmChargeBounds, type LlmCorrectnessCheckerOpts, type LlmJudgeOptions, type LlmMessage, type LlmReviewerConfig, type LlmSpan, type LlmTokenLogprob, type LlmToolCall, type LlmToolChoice, type LlmToolDefinition, type LlmUsage, type MannWhitneyResult, type MatchedPair, type MatchedRunRecordPair, type MaximumCharge, type McNemarResult, type MintRolloutOptions, type MintRolloutResult, type MintedRolloutLine, type MockTransportOpts, type ModelEndpointCheck, type ModelEndpointRequest, type ModelSeats, ModelSubstitutionError, MultiLayerVerifier, type NoLeakOptions, NoopRawProviderSink, NotFoundError, OUTPUT_VALUE, type Objective, type OpenAiCompatibleTransportOpts, type Oracle, type OracleObservation, type OracleReport, type OracleResult, type OrderMode, type OtlpExport, OtlpFileTraceStore, type OtlpFlatLine, type OtlpSpan, type OwnerVerdict, type PaidCallResult, type PairArmsOptions, type PairArmsResult, type PairRunRecordsResult, type PairedArmRow, type PairedArmsComparison, type PairedBootstrapOptions, type PairedBootstrapResult, type PairedCorrectness, type PairedDecisionMethod, type PairedDecisionShape, type PairedDecisionStatistic, type PairedDeltaTestOptions, type PairedDeltaTestResult, type PairedEvalueOptions, type PairedEvalueSequence, type PairedEvalueStep, type PairedMcNemarEvidence, type PairedMetricDelta, type PairedPromotionAlternative, type PairedPromotionAlternativeCell, type PairedPromotionCallRates, type PairedPromotionDecision, type PairedPromotionDecisionOptions, type PairedPromotionPowerCall, type PairedPromotionPowerOptions, type PairedPromotionPowerResult, type PairedSignTestResult, type PairedTTestResult, type PairedValueLaw, type ParetoPoint, type ParetoResult, type PartitionHeldOutOptions, type PendingCostCall, type PendingCostCallView, type PersistedFinding, type PersonaConfig, type PolicyEdit, type PolicyEditAdmission, type PolicyEditAdmissionOptions, type PolicyEditAxis, type PolicyEditCandidateRecord, type PolicyEditChange, type PolicyEditExpectedGain, type PolicyEditGainDirection, type PolicyEditGainUnit, type PolicyEditInit, type PolicyEditRisk, type PolicyEditSchemaVersion, type PolicyEditSource, type PolicyEditTarget, type PolicyEditTargetSurface, type PreferenceMemoryEntry, type ProducedProposal, type ProducedState, type ProductBenchmarkExportOptions, type ProductBenchmarkExportResult, type ProductBenchmarkManifest, type ProductBenchmarkSingleRunExportOptions, type ProductBenchmarkSplit, type ProductBenchmarkValidationReport, type ProductClientConfig, type ProductScenario, type ProfileAxisSpec, type ProjectRuntimeTrajectoryEvidenceOptions, type PromptHandle, type ProportionInterval, type ProposalEventLike, type ProposalFinding, type ProposeFn, type ProposeInput, type ProposeOutput, type ProposeReviewConfig, type ProposeReviewControlAction, type ProposeReviewControlConfig, type ProposeReviewControlResult, type ProposeReviewControlState, type ProposeReviewReport, type ProposedSideEffect, type QueryTracesPage, REDACTION_VERSION, type RankTestMethod, type RankTestMethodRequest, type RankTestOptions, type RawAnalystFinding, type RawProviderEvent, type RawProviderSink, type RecordRunsOptions, type RedTeamCase, type RedTeamCategory, type RedTeamFinding, type RedTeamReport, type RedactOptions, type RedactionProfile, type RedactionReport, type ReferenceReplayCaseRun, type ReferenceReplayRun, type ReferenceReplaySplit, type ReflectionContext, type ReflectionProposal, type RegistryRunOpts, type ReleaseConfidenceInput, type ReleaseConfidenceIssue, type ReleaseConfidenceMetrics, type ReleaseConfidenceScorecard, type ReleaseTraceEvidence, type RepeatedActionOptions, type RequiredPairsForPairedPromotionOptions, type RequiredPairsForPairedPromotionResult, type RequirementCheck, type ResearchReport, type ResearchReportOptions, type Researcher, type RetryWrite, type Review, type ReviewFn, type ReviewInput, type ReviewMemoryEntry, type ReviewMemoryStore, type RiskDifferenceResult, type RolloutSearchLineage, type RouteMap, type RoutedField, type RubricDimension, type Run, type RunCampaignOptions, type RunCaptureReport, type RunCommandInput, type RunCommandResult, type RunCostProvenance, type RunFilter, RunIntegrityError, type RunIntegrityReport, type RunJudgeMetadata, type RunOutcome, type RunPaidCallInput, type RunRecord, type RunRecordBackend, RunRecordValidationError, type RunScore, type RunScoreWeights, type RunSearchCoordinates, type RunSplitTag, type RunStatus, type RunTaskFailure, type RunTerminalOutcome, type RunTokenUsage, type RunTraceRef, type RuntimeEventLike, type RuntimeTrajectoryEvidenceProjection, type RuntimeTrajectoryEvidenceSummary, type RuntimeTrajectoryHookEvent, type RuntimeTrajectoryRecord, type RuntimeTrajectoryRunRecord, SEMANTIC_CONCEPT_JUDGE_VERSION, type SampleSummary, type SampleSummaryOptions, type SandboxDriver, type SandboxSdkTransportOpts, type SatisfiedBy, type Scenario, type ScenarioCost, type ScenarioFile, type ScenarioResult, type ScorePreference, type ScoreRiskDifferenceResult, type Scorecard, type ScorecardCell, type ScorecardCellDiff, type ScorecardDiff, type ScorecardEntry, type ScorecardLogLine, type SearchSpanResult, type SearchTraceResult, type SelfImproveBudget, type SelfImproveMethodOptions, type SelfImproveMethodProvenance, type SelfImproveMethodResult, type SelfImproveOptions, type SelfImproveProgressEvent, type SelfImproveProposerOptions, type SelfImproveProposerResult, type SelfImproveResult, type SemanticConceptJudgeInput, type SemanticConceptJudgeOptions, type SemanticConceptJudgeResult, type SequentialDecision, type SeriesConvergenceOptions, type SeriesConvergenceResult, type SeriesDistribution, type SeriesDistributionOptions, type ServedModelCheck, type ServedModelPolicy, type ServedModelVerdict, type Severity, type ShareSafetyStatus, type ShareSafetyVerdict, type SignTestAlternative, type SingleBackendDivergence, type SingleBackendReport, type Span, type SpanFilter, type SpanHandle, type SpanMatchRecord, type SpanPredicate, type SpanRecordsToTraceAnalysisStoreOptions, type SplitCoverage, type SteeringBundle, type SteeringChange, type SteeringOptimizationResult, type SteeringOptimizationRow, type StopDecision, type StrategyChecker, type StreamingDetector, type SummaryTable, type SummaryTableOptions, type SummaryTableRow, type SynthesisTarget, type TaskGold, type TaskHeadroom, type ToolCallEventLike, type ToolSpan, type ToolStats, type ToolUseMetrics, type ToolUseOptions, type TraceAnalysisEngine, type TraceAnalysisEngineResult, type TraceAnalysisStore, type TraceAnalystByteBudgets, type TraceAnalystDefinition, type TraceAnalystFilters, type TraceAnalystSpan, type TraceAnalystSpanKind, type TraceAnalystSpanStatus, type TraceAnalystTraceSummary, type TraceContract, type TraceContractSpec, TraceEmitter, type TraceEmitterOptions, type TraceEvent, type TraceInsightReadiness, type TraceInsightSuite, type TraceStore, type TrainSelectionTestOptions, type TrainSelectionTestPartition, type Trajectory, type TrajectoryStep, type TreatmentGate, type TreatmentGateInput, type TreatmentGateOptions, type TrialTrace, type Turn, type TurnResult, UNKNOWN_MODEL, type UserQuestion, VERIFICATION_STRATEGIES, VERIFICATION_STRATEGY_SOURCES, ValidationError, type VerbosityBiasResult, type Verdict, type VerdictCacheStore, type VerdictCertification, type Verification, type VerificationReport, type VerificationStrategyProfile, type VerificationStrategySource, type VerifyFn, type VerifyOptions, type ViewSpansResult, type ViewTraceOversized, type ViewTraceResult, type ViteDeployRunnerInput, type WeightedCompositeInput, type WeightedCompositeResult, type WelchTestResult, type WelchTestStatus, type WilcoxonSignedRankResult, type WranglerDeployRunnerInput, addCalibrationVerdicts, agentProfileCellKey, agentProfileHash, agentProfileId, analystRunDigest, analystRunToFeedbackTrajectory, analystRunToReviewRequests, analyzeRuns, analyzeSeries, analyzeTraces, argHash, assertCrossFamily, assertCrossFamilyServed, assertJudgeMayGate, assertRealBackend, assertRunCaptured, assertServedModel, assertServedModels, assessShareSafety, assignFeedbackSplit, attest, benjaminiHochberg, bonferroni, bootstrapCi, budgetBreachView, buildAgentProfileCell, buildDefaultAnalystRegistry, buildTraceInsightContext, buildTraceInsightPrompt, buildTrajectory, calibrateJudge, calibrateJudgeContinuous, captureFetchToRawSink, certificationEvidenceDigest, checkServedModel, classifyFailure, classifyFailureReason, comparePairedArms, completionVerdict, computeFindingId, computeToolUseMetrics, confidenceInterval, continuousAgreement, controlRunToFeedbackTrajectory, corpusInterRaterAgreement, corpusInterRaterAgreementFromJudgeScores, createBoundedTraceAnalysisStore, createChatClient, createChatTraceEngine, createDspyRlmTraceEngine, createFeedbackTrajectory, createLlmCorrectnessChecker, createTraceAnalyst, decidePairedPromotion, defineAgentEval, defineEquivalenceCheck, describeServedModelCheck, describeTraceInsightScope, diffFindings, domainEvidencePattern, dominates, eProcess, equivalenceVerdict, evaluateInterimReleaseConfidence, evaluateReleaseConfidence, expandProfileAxes, exportRunAsOtlp, extractProducedState, extractUsage, extractUsageFromSse, failureBlame, failureClusterView, feedbackTrajectoriesToDatasetScenarios, feedbackTrajectoriesToOptimizerRows, feedbackTrajectoryToOptimizerRow, gainHistogram, gradeSemanticStatus, groupRunsByAgentProfileCell, harnessAxisOf, hashJson, holm, inferDomainKeywords, interRaterReliability, iqr, isJudgeSpan, isLlmSpan, isRunRecord, isToolSpan, judgeAgreementView, judgeFamily, judgeGateDecision, judgeSpans, llmJudge, makeFinding, makeProposalFinding, manifestContentDigest, mannWhitneyU, mcnemar, mcnemarPower, mcnemarRequiredN, measureJudgeAgreement, mintRolloutRows, modelHasSnapshot, mulberry32, otlpTextToTraceAnalysisStore, pairArms, pairRunRecords, pairedBootstrap, pairedEvalueSequence, pairedMde, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, pairedSignTest, pairedTTest, paretoFrontier, parseRunRecordSafe, partialCredit, passAtK, pearsonR, index_d_exports as profile, ranks, redTeamDataset, redTeamReport, redact, redactForShare, redactText, registerCalibrationSet, renderPreferenceMemoryMarkdown, requiredPairedSampleSize, requiredSampleSize, runCampaign, runCanaries, runEquivalenceCheck, runEvalCampaign, runSemanticConceptJudge, runTaskScore, runsForScenario, scoreRedTeamOutput, scoreTraceInsightReadiness, selfImprove, shareAllowed, snapshotFromJudgeAgreement, spanRecordsToTraceAnalysisStore, spearmanR, summarizeBackendIntegrity, summarizeNumberSeries, summarizePreferenceMemory, summaryTable, toAgentProfileJson, tokenizeDomainWords, toolSpansToTraceAnalysisStore, toolWasteView, transientDispatchFailure, validateRunRecord, verbosityBias, verifyAgentProfileCell, verifyAttestation, verifyCompletion, weightedComposite, weightedMean, welchsTTest, wilcoxonSignedRank, wilson, withAssignedFeedbackSplit };
 //# sourceMappingURL=index.d.ts.map