@tangle-network/agent-eval 0.19.0 → 0.19.1

package/README.md

@@ -83,11 +83,49 @@ The recipe for a code-generator eval is in [`SKILL.md` §Minimal working path](.
 | `evaluateActionPolicy` | Generic action preflight for approval, budget, expected-outcome, and kill-criteria checks. | [feature-guide.md](./docs/feature-guide.md) |
 | `ExperimentTracker`, steering optimizers, `bisector` | A/B prompts, optimize steering, bisect regressions. | SKILL.md |
 | `runMultiShotOptimization`, `trialTraceFromMultiShotTrial` | GEPA-style optimization for variable-length agent trajectories with ASI, paired seeds, and optional held-out promotion gating. | [multi-shot-optimization.md](./docs/multi-shot-optimization.md) |
+| `evaluateReleaseConfidence`, `assertReleaseConfidence` | Release scorecard that composes corpus coverage, search/holdout run evidence, ASI diagnostics, overfit checks, and cost/latency budgets. | §Release confidence |
 | `runPromptEvolution`, `createCompositeMutator`, `createSandboxPool`, `createSandboxCodeMutator`, `MutationTelemetry`, `LineageRecorder`, `CostLedger`, `JsonlTrialCache` | Prompt + code evolution loops with bounded sandbox pools, durable JSONL telemetry, plateau-detecting composite mutators, crash-resumable trial cache. | §Evolution loop |
 | `reflective-mutation` (`buildReflectionPrompt`, `parseReflectionResponse`, `DEFAULT_MUTATION_PRIMITIVES`) | Trace-conditioned LLM mutator that reasons over top/bottom trials instead of blind rewrites. | inline JSDoc |
 | `correlationStudy`, `OutcomeStore`, `ProductRegistry` | Meta-eval: do our scores predict deployment outcomes (revenue, retention)? | inline JSDoc |
 | Telemetry (`telemetry/`, `telemetry/file`) | OTLP export, trace replay, file sinks. | inline JSDoc |
 
+## Release confidence
+
+Use `evaluateReleaseConfidence` at the release boundary for every consuming
+agent surface. It fails closed unless the release has a versioned corpus,
+search and holdout run evidence, score/pass-rate evidence, ASI for failures,
+and budget/overfit checks. Single-shot and multi-shot apps use the same path:
+single-shot traces are just trace evidence with `turnCount: 1`.
+
+```ts
+import {
+  evaluateReleaseConfidence,
+  releaseTraceEvidenceFromMultiShotTrials,
+} from '@tangle-network/agent-eval'
+
+const scorecard = evaluateReleaseConfidence({
+  target: 'blueprint-agent/autoresearch',
+  candidateId: 'candidate-v3',
+  baselineId: 'baseline',
+  dataset: await dataset.manifest(),
+  runs: [...candidateRuns, ...baselineRuns],
+  traces: releaseTraceEvidenceFromMultiShotTrials(result.evolution.generations.flatMap((g) => g.trials)),
+  gateDecision: result.gate?.decision,
+  thresholds: {
+    minScenarioCount: 50,
+    minSearchRuns: 50,
+    minHoldoutRuns: 20,
+    minPassRate: 0.9,
+    minMeanScore: 0.8,
+    maxOverfitGap: 0.1,
+    maxMeanCostUsd: 0.05,
+    maxP95WallMs: 120_000,
+  },
+})
+
+if (!scorecard.promote) throw new Error(scorecard.summary)
+```
+
 ## Evolution loop
 
 For agent tasks that run across many chat turns or tool calls, start with

package/dist/index.d.ts

@@ -7760,6 +7760,114 @@ declare function runMultiShotOptimization<P>(config: MultiShotOptimizationConfig
 declare function defaultMultiShotObjectives(): Objective<VariantAggregate>[];
 declare function trialTraceFromMultiShotTrial(trial: MultiShotTrialResult): TrialTrace;
 
+/**
+ * Release confidence gate.
+ *
+ * This is the production-facing composition layer over the lower-level
+ * primitives:
+ *   - Dataset manifests prove corpus/version coverage.
+ *   - RunRecord rows prove reproducible search/holdout outcomes.
+ *   - Multi-shot trace evidence carries turn counts and ASI diagnostics.
+ *   - HeldOutGate decisions remain the paired promotion authority.
+ *
+ * The gate is intentionally pure and conservative. Missing declared evidence
+ * fails closed instead of being treated as a neutral zero.
+ */
+
+type ReleaseConfidenceStatus = 'pass' | 'warn' | 'fail';
+type ReleaseConfidenceAxisName = 'corpus' | 'quality' | 'generalization' | 'diagnostics' | 'efficiency';
+interface ReleaseTraceEvidence {
+    scenarioId: string;
+    candidateId?: string;
+    split?: RunSplitTag;
+    score?: number;
+    ok?: boolean;
+    turnCount?: number;
+    costUsd?: number;
+    durationMs?: number;
+    failureMode?: string;
+    asi?: ActionableSideInfo[];
+    metadata?: Record<string, unknown>;
+}
+interface ReleaseConfidenceThresholds {
+    /** Require a Dataset manifest or explicit scenarios. Default true. */
+    requireCorpus?: boolean;
+    minScenarioCount?: number;
+    minSearchRuns?: number;
+    minHoldoutRuns?: number;
+    /** Require at least one holdout scenario/run. Default true. */
+    requireHoldout?: boolean;
+    minPassRate?: number;
+    minMeanScore?: number;
+    /** Search mean may exceed holdout mean by at most this much. */
+    maxOverfitGap?: number;
+    maxMeanCostUsd?: number;
+    maxP95WallMs?: number;
+    /** Low-score/failed rows must carry ASI. Default true. */
+    requireAsiForFailures?: boolean;
+    /** Score below this is considered a failure for ASI coverage. Default 0.5. */
+    failureScoreThreshold?: number;
+}
+interface ReleaseConfidenceInput {
+    target: string;
+    candidateId?: string;
+    baselineId?: string;
+    dataset?: DatasetManifest;
+    scenarios?: readonly DatasetScenario[];
+    runs?: readonly RunRecord[];
+    traces?: readonly ReleaseTraceEvidence[];
+    gateDecision?: GateDecision | null;
+    thresholds?: ReleaseConfidenceThresholds;
+}
+interface ReleaseConfidenceAxis {
+    name: ReleaseConfidenceAxisName;
+    status: ReleaseConfidenceStatus;
+    score: number;
+    detail: string;
+}
+interface ReleaseConfidenceIssue {
+    axis: ReleaseConfidenceAxisName;
+    severity: 'critical' | 'warning';
+    code: string;
+    detail: string;
+}
+interface ReleaseConfidenceMetrics {
+    scenarioCount: number;
+    searchRuns: number;
+    holdoutRuns: number;
+    passRate: number;
+    meanScore: number;
+    searchMeanScore: number;
+    holdoutMeanScore: number;
+    overfitGap: number;
+    meanCostUsd: number;
+    p95WallMs: number;
+    failedRows: number;
+    failuresWithAsi: number;
+    singleShotTraces: number;
+    multiShotTraces: number;
+    splitCounts: Record<DatasetSplit, number>;
+    domainCounts: Record<string, number>;
+    failureModeCounts: Record<string, number>;
+    responsibleSurfaceCounts: Record<string, number>;
+}
+interface ReleaseConfidenceScorecard {
+    target: string;
+    candidateId: string | null;
+    baselineId: string | null;
+    status: ReleaseConfidenceStatus;
+    promote: boolean;
+    axes: ReleaseConfidenceAxis[];
+    issues: ReleaseConfidenceIssue[];
+    metrics: ReleaseConfidenceMetrics;
+    dataset: DatasetManifest | null;
+    gateDecision: GateDecision | null;
+    summary: string;
+}
+declare function releaseTraceEvidenceFromMultiShotTrials(trials: readonly MultiShotTrialResult[]): ReleaseTraceEvidence[];
+declare function evaluateReleaseConfidence(input: ReleaseConfidenceInput): ReleaseConfidenceScorecard;
+declare function assertReleaseConfidence(input: ReleaseConfidenceInput): ReleaseConfidenceScorecard;
+
 /**
  * concurrency — small primitives the evolution loop needs.
  *
@@ -8407,4 +8515,4 @@ declare function judgeReplayGate<TOutput>(args: JudgeReplayGateArgs<TOutput>): P
     candidateSamples: number;
 }>;
 
-export { type ActionExecutionPolicy, type ActionPolicyDecision, type ActionableSideInfo, type ActiveLearningOptions, type AdapterRun, AgentDriver, type AgentDriverConfig, type AlignmentOp, type AntiSlopConfig, type AntiSlopIssue, type AntiSlopReport, type Artifact$1 as Artifact, type ArtifactCheck, type Artifact as ArtifactCheckArtifact, type ArtifactResult, type ArtifactValidator, type AsiSeverity, AxGepaSteeringOptimizer, type AxSteeringOptimizerConfig, BENCHMARK_SPLIT_SEED, type BaselineOptions, type BaselineReport, BehaviorAssertion, type BenchmarkAdapter, type BenchmarkDatasetItem, type BenchmarkEvaluation, type BenchmarkReport, BenchmarkRunner, type BenchmarkRunnerConfig, type BestOfNResult, type BisectOptions, type BisectResult, type BisectStep, type BootstrapOptions, type BootstrapResult, BudgetBreachError, type BudgetBreachFinding, type BudgetBreachReport, BudgetGuard, type BudgetLedgerEntry, type BudgetSpec, BuilderSession, type BuilderSessionInit, type CalibrationBin, type CalibrationOptions, type CalibrationReport, type CalibrationResult, CallExpectation, type CanaryAlert, type CanaryKind, type CanaryLeak, type CanaryOptions, type CanaryReport, type CanarySeverity, type CandidateScenario, type CandidateScore, type CausalAttributionReport, type ChatSummary, type CheckResult, type CodeMutationOutcome, type CodeMutationRunner, type CollectedArtifacts, type CommandRunner, type CompletionCriterion, type CompositePolicy, type ConceptComplexity, type ConceptFinding, type ConceptSpec, type ConceptWeightStrategy, type ContinuityCheck, type ContinuityCheckResult, type ContinuityReport, type ContinuitySnapshotPair, type ContractMetric, type ContractReport, type ControlActionFailureMode, type ControlActionOutcome, type ControlBudget, type ControlContext, type ControlDecision, type ControlEvalResult, type ControlRunResult, type ControlRuntimeConfig, type ControlRuntimeError, type ControlSeverity, type ControlStep, type ControlStopPolicies, ConvergenceTracker, type CorrelationReport, type CorrelationResult, type CorrelationStudyOptions, type CorrelationStudyResult, type CostEntry, CostLedger, type CostLedgerGeneration, type CostLedgerSnapshot, type CostSummary, CostTracker, type CounterfactualContext, type CounterfactualMutation, type CounterfactualResult, type CounterfactualRunner, type CreateCompositeMutatorOpts, type CreateDefaultReviewerOptions, type CreateSandboxCodeMutatorOpts, type CreateSandboxPoolOpts, type CrossTraceDiff, type CrossTraceDiffOptions, D1ExperimentStore, type D1ExperimentStoreOptions, type D1Like, type D1PreparedStatementLike, DEFAULT_AGENT_SLOS, DEFAULT_COMPLEXITY_WEIGHTS, DEFAULT_RULES as DEFAULT_FAILURE_RULES, DEFAULT_FINDERS, DEFAULT_HARNESS_OBJECTIVES, DEFAULT_MUTATION_PRIMITIVES, DEFAULT_MUTATORS, DEFAULT_REDACTION_RULES, DEFAULT_RED_TEAM_CORPUS, DEFAULT_RUN_SCORE_WEIGHTS, DEFAULT_SEVERITY_WEIGHTS, Dataset, type DatasetDifficulty, type DatasetManifest, type DatasetProvenance, type DatasetScenario, type DatasetSplit, type DeployFamily, type DeployGateLayerInput, type DeployRunResult, type DeployRunner, type DeploymentOutcome, type DirEntry, type Direction, type DivergenceOptions, type DivergenceReport, DockerSandboxDriver, type DriverResult, type DriverState, DualAgentBench, type DualAgentBenchConfig, type DualAgentReport, type DualAgentRound, type DualAgentScenario, type DualAgentScenarioResult, ERROR_COUNT_PATTERNS, type ErrorCountPattern, type EuRiskClass, type EvalMetricSpec, type EvalResult, type EventFilter, type EventKind, type EvolutionRound, type EvolvableVariant, type ExecutorConfig, type Expectation, type Experiment, type ExperimentPlan, type ExperimentResult, type Run as ExperimentRun, type ExperimentStore, ExperimentTracker, type ExportedRewardModel, type ExtractOptions, type ExtractResult, FAILURE_CLASSES, type FactorContribution, type FactorialCell, type FailureClass, type FailureClassification, type FailureCluster, type FailureClusterReport, type FailureContext, type FailureMode, type FailureRule, type FeedbackArtifactType, type FeedbackAttempt, type FeedbackLabel, type FeedbackLabelKind, type FeedbackLabelSource, type FeedbackOptimizerRow, type FeedbackOutcome, type FeedbackPattern, type FeedbackReplayAdapter, type FeedbackReplayResult, type FeedbackSeverity, type FeedbackSplitPolicy, type FeedbackTask, type FeedbackTrajectory, type FeedbackTrajectoryFilter, type FeedbackTrajectoryStore, FileSystemExperimentStore, type FileSystemExperimentStoreOptions, FileSystemFeedbackTrajectoryStore, FileSystemOutcomeStore, type FileSystemOutcomeStoreOptions, FileSystemTraceStore, type FileSystemTraceStoreOptions, type Finding, type FlowAction, type FlowLayerEnv, type FlowLayerFactoryInput, type FlowRunner, type FlowRunnerStepResult, type FlowSpec, type FlowStep, type GainDistributionBin, type GainDistributionFigureSpec, type GainDistributionOptions, type GateDecision, type GateEvidence, type GenerationReport, type GenericSpan, type GoldenItem, type GoldenSeverity, type GoldenSpec, type GovernanceContext, type GovernanceFinding, type GovernanceReport, type GradedStep, type HarnessAdapter, type HarnessConfig, type HarnessExperimentConfig, type HarnessExperimentResult, type HarnessIntervention, type HarnessRunRequest, type HarnessRunResult, type HarnessScenario, type HarnessSelection, type HarnessVariant, type HarnessVariantReport, HeldOutGate, type HeldOutGateConfig, type HeldOutGateRejectionCode, HoldoutAuditor, HoldoutLockedError, type HostedJudgeConfig, type HostedJudgeDimension, type HostedJudgeRequest, type HostedJudgeResponse, type HostedRunCriticConfig, type HostedRunScoreRequest, type HostedRunScoreResponse, type HypothesisManifest, type HypothesisResult, INTENT_MATCH_JUDGE_VERSION, type ImageData, InMemoryExperimentStore, InMemoryFeedbackTrajectoryStore, InMemoryOutcomeStore, InMemoryTraceStore, InMemoryTrialCache, InMemoryWorkspaceInspector, type InferenceScorer, type InspectorContext, type IntentMatchInput, type IntentMatchOptions, type IntentMatchResult, type InteractionContribution, JsonlTrialCache, type JudgeAgreementReport, type JudgeConfig, type JudgeFleetOptions, type JudgeFn, type JudgeInput, type JudgePair, type JudgeReplayGateArgs, type JudgeReplayResult, type JudgeRubric, JudgeRunner, type JudgeScore, type JudgeSpan, type KeywordConceptSpec, type KeywordCoverageFinding, type KeywordCoverageOptions, type KeywordCoverageResult, type LangfuseEnvelope, type LangfuseGeneration, type LangfuseScore, type Layer, type LayerCorrelation, type LayerResult, type LayerStatus, type LineageKind, type LineageKindResolver, type LineageNode, LineageRecorder, LlmCallError, type LlmCallRequest, type LlmCallResult, LlmClient, type LlmClientOptions, type LlmJsonCall, type LlmMessage, type LlmReviewerConfig, type LlmSpan, type LlmUsage, LockedJsonlAppender, MODEL_PRICING, type MatchResult, type MatcherResult, type MeasurementPolicy, type MergeOptions, type Message, type MetricSamples, type MetricVerdict, MetricsCollector, type MuffledFinder, type MuffledFinding, MultiLayerVerifier, type MultiShotGateConfig, type MultiShotGateResult, type MultiShotMutateAdapter, type MultiShotOptimizationConfig, type MultiShotOptimizationResult, type MultiShotRun, type MultiShotRunInput, type MultiShotRunner, type MultiShotScore, type MultiShotScorer, type MultiShotSplit, type MultiShotTrace, type MultiShotTrialResult, type MultiShotVariant, type MultiToolchainLayerConfig, type MutateAdapter, type MutationAttempt, type MutationChannel, MutationTelemetry, type Mutator, Mutex, NoopResearcher, OTEL_AGENT_EVAL_SCOPE, type Objective, type Oracle, type OracleObservation, type OracleReport, type OracleResult, type OrthogonalityInput, type OrthogonalityResult, type OtlpExport, type OtlpResourceSpans, type OtlpSpan, type OutcomeFilter, type OutcomePair, type OutcomeStore, type PairedBootstrapOptions, type PairedBootstrapResult, PairwiseSteeringOptimizer, type ParetoFigureSpec, type ParetoPoint, type ParetoResult, type PersonaConfig, type Playbook, type PlaybookEntry, type PoolSlot, type PositionalBiasResult, type PreferenceMemoryEntry, type PrmGradedTrace, PrmGrader, type PrmTrainingSample, ProductClient, type ProductClientConfig, type ProjectKind, ProjectRegistry, type ProjectSummary, type ProjectTimelineEntry, type PromptEvolutionConfig, type PromptEvolutionEvent, type PromptEvolutionResult, type PromptHandle, PromptRegistry, type TrialResult as PromptTrialResult, type ProposeFn, type ProposeInput, type ProposeOutput, type ProposeReviewConfig, type ProposeReviewControlAction, type ProposeReviewControlConfig, type ProposeReviewControlResult, type ProposeReviewControlState, type ProposeReviewReport, type ProposeReviewShot, type ProposedSideEffect, REDACTION_VERSION, type RedTeamCase, type RedTeamCategory, type RedTeamFinding, type RedTeamPayload, type RedTeamReport, type RedactionReport, type RedactionRule, type ReferenceMatchResult, type ReferenceReplayAdapter, type ReferenceReplayAdapterFn, type ReferenceReplayAdapterLike, type ReferenceReplayAggregate, type ReferenceReplayCandidate, type ReferenceReplayCase, type ReferenceReplayCaseRun, type ReferenceReplayExecutionScenario, type ReferenceReplayItem, type ReferenceReplayMatch, type ReferenceReplayMatchStrategy, type ReferenceReplayMatcher, type ReferenceReplayPromotionDecision, type ReferenceReplayPromotionPolicy, type ReferenceReplayRun, type ReferenceReplayRunContext, type ReferenceReplayRunOptions, type ReferenceReplayRunStore, type ReferenceReplayScenario, type ReferenceReplayScenarioScore, type ReferenceReplayScore, type ReferenceReplayScoreOptions, type ReferenceReplaySplit, type ReferenceReplaySplitComparison, type ReferenceReplaySteeringRowsOptions, type ReflectionContext, type ReflectionProposal, type RegressionOptions, type RegressionSpec, type Researcher, type RetrievalSpan, type Review, type ReviewFn, type ReviewInput, type ReviewMemoryEntry, type ReviewMemoryStore, type ReviewerMemoryEntry, type ReviewerOutput, type ReviewerPromptInput, type ReviewerSoftFailDefaults, type ReviewerVerificationSummary, type RobustnessResult, type RouteMap, type RubricDimension, type Run$1 as Run, type RunAppScenarioOptions, type RunCommandInput, type RunCommandResult, type RunConfig, RunCritic, type RunCriticOptions, type RunDiff, type RunFilter, type RunJudgeMetadata, type RunLayer, type RunOutcome, type RunRecord, RunRecordValidationError, type RunScore, type RunScoreWeights, type RunSplitTag, type RunStatus, type RunTokenUsage, type RunTrace, SEMANTIC_CONCEPT_JUDGE_VERSION, type SandboxDriver, SandboxHarness, type SandboxHarnessResult, type SandboxJudgeKind, type SandboxJudgeResult, type SandboxJudgeSpec, type SandboxPool, type SandboxResult, type SandboxSpan, type ScanOptions, type Scenario, type ScenarioAggregate, type ScenarioCost, type ScenarioFile, ScenarioRegistry, type ScenarioResult, type ScoreAdapter, type ScoredTarget, type SelfPlayOptions, type SelfPlayProposer, type SelfPlayScorer, type SelfPreferenceResult, type SemanticConceptJudgeInput, type SemanticConceptJudgeOptions, type SemanticConceptJudgeResult, type SeriesConvergenceOptions, type SeriesConvergenceResult, type Severity, type ShipOptions, type SignedManifest, type SliceOptions, type Slo, type SloCheckResult, type SloComparator, type SloReport, type SloSeverity, type SlopCategory, type SlotFactory, type Span, type SpanBase, type SpanFilter, type SpanHandle, type SpanKind, type SpanStatus, type SteeringBundle, type SteeringChange, type SteeringDelta, type SteeringOptimizationResult, type SteeringOptimizationRow, type SteeringOptimizationSelector, type SteeringOptimizerBackend, type SteeringOptimizerConfig, type SteeringRolePrompt, type StepAttribution, type StepContext, type StepRubric, type StopDecision, type StuckLoopFinding, type StuckLoopOptions, type StuckLoopReport, SubprocessSandboxDriver, type SubprocessSandboxDriverOptions, type SummaryTable, type SummaryTableOptions, type SummaryTableRow, type SynthesisReason, type SynthesisTarget, TRACE_SCHEMA_VERSION, type TestGradedRunOptions, type TestGradedRunResult, type TestGradedScenario, type TestOutputParser, type TestResult, type ThreeLayerProjectReport, type ThresholdContract, TokenCounter, type TokenSpec, type ToolSpan, type ToolStats, type ToolUseMetrics, type ToolUseOptions, type ToolWasteFinding, type ToolWasteOptions, type ToolWasteReport, TraceEmitter, type TraceEmitterOptions, type TraceEvent, type TraceStore, type Trajectory, type TrajectoryStep, type TrialAttempt, type TrialCache, TrialTelemetry, type TrialTrace, type Turn, type TurnMetrics, type TurnResult, UNIVERSAL_FINDERS, type UseCaseSignals, type ValidationContext, type ValidationIssue, type ValidationResult, type VariantAggregate, type VerbosityBiasResult, type Verdict, type Verification, type VerificationReport, type VerifyContext, type VerifyFn, type VerifyOptions, type VisualDiffOptions, type VisualDiffResult, type ViteDeployRunnerInput, type WorkflowTopology, type WorkspaceAssertion, type WorkspaceAssertionResult, type WorkspaceInspector, type WorkspaceSnapshot, type WranglerDeployRunnerInput, adversarialJudge, aggregateLlm, aggregateRunScore, allCriticalPassed, analyzeAntiSlop, analyzeSeries, argHash, assignFeedbackSplit, attributeCounterfactuals, deterministicSplit as benchmarkDeterministicSplit, index as benchmarks, benjaminiHochberg, bhAdjust, bisect, bonferroni, bootstrapCi, budgetBreachView, buildReflectionPrompt, buildReviewerPrompt, buildTrajectory, byteLengthRange, calibrateJudge, calibrationCurve, callLlm, callLlmJson, canaryLeakView, causalAttribution, checkCanaries, checkSlos, clamp01, classifyEuAiRisk, classifyFailure, codeExecutionJudge, cohensD, coherenceJudge, collectionPreserved, commitBisect, compareReferenceReplay, compareToBaseline, compilerJudge, composeParsers, composeValidators, computeToolUseMetrics, confidenceInterval, containsAll, controlFailureClassFromVerification, controlRunToFeedbackTrajectory, correlateLayers, correlationStudy, createAntiSlopJudge, createCompositeMutator, createCustomJudge, createDefaultReviewer, createDomainExpertJudge, createFeedbackTrajectory, createIntentMatchJudge, createLlmReviewer, createSandboxCodeMutator, createSandboxPool, createSemanticConceptJudge, crossTraceDiff, crowdingDistance, decideReferenceReplayPromotion, decideReferenceReplayRunPromotion, defaultJudges, defaultMultiShotObjectives, defaultReferenceReplayMatcher, deployGateLayer, distillPlaybook, dominates, estimateCost, estimateTokens, euAiActReport, evaluateActionPolicy, evaluateContract, evaluateHypothesis, evaluateOracles, executeScenario, expectAgent, exportRewardModel, exportRunAsOtlp, exportTrainingData, extractAssetUrls, extractErrorCount, failureClusterView, feedbackTrajectoriesToDatasetScenarios, feedbackTrajectoriesToOptimizerRows, feedbackTrajectoryToDatasetScenario, feedbackTrajectoryToOptimizerRow, fileContains, fileExists, findAutoMatchNoExpectation, findConstructorCwdDropped, findFallbackToPass, findLiteralTruePass, findSkipCountsAsPass, firstDivergenceView, flowLayer, formatBenchmarkReport, formatDriverReport, formatFindings, gainHistogram, precision as goldenPrecision, gradeSemanticStatus, groupBy, hashContent, hashScenarios, htmlContainsElement, inMemoryReferenceReplayStore, inMemoryReviewStore, interRaterReliability, iqr, isJudgeSpan, isLlmSpan, isPrmVerdict, isRetrievalSpan, isRunRecord, isSandboxSpan, isToolSpan, jestTestParser, jsonHasKeys, jsonShape, jsonlReferenceReplayStore, jsonlReviewStore, judgeAgreementView, judgeReplayGate, judgeSpans, keyPreserved, linterJudge, llmSpanFromProvider, llmSpans, loadScorerFromGrader, localCommandRunner, lowercaseMutator, mannWhitneyU, matchGoldens, mergeLayerResults, mergeSteeringBundle, multiToolchainLayer, nistAiRmfReport, nonRefusalRubric, normalizeScores, notBlocked, objectiveEval, outputLengthRubric, pairedBootstrap, pairedTTest, pairedWilcoxon, paraphraseRobustness, paretoChart, paretoFrontier, paretoFrontierWithCrowding, parseFeedbackTrajectoriesJsonl, parseReflectionResponse, parseRunRecordSafe, partialCredit, passOrthogonality, pixelDeltaRatio, politenessPrefixMutator, positionalBias, printDriverSummary, prmBestOfN, prmEnsembleBestOfN, probeLlm, promptBisect, proposeSynthesisTargets, pytestTestParser, redTeamDataset, redTeamReport, redactString, redactValue, referenceReplayRunsToSteeringRows, referenceReplayScenarioToRunScore, regexMatch, regexMatches, regressionView, renderMarkdown, renderMarkdownReport, renderPlaybookMarkdown, renderPreferenceMemoryMarkdown, renderSteeringText, replayFeedbackTrajectories, replayFeedbackTrajectory, replayScorerOverCorpus, replayTraceThroughJudge, requiredSampleSize, resetLockedAppendersForTesting, resumeBuilderSession, roundTripRunRecord, rowCount, rowWhere, runAgentControlLoop, runAssertions, runCanaries, runCounterfactual, runE2EWorkflow, runExpectations, runFailureClass, runHarnessExperiment, runIntentMatchJudge, runJudgeFleet, runKeywordCoverageJudge, runKeywordCoverageJudgeUrl, runMultiShotOptimization, runPromptEvolution, runProposeReview, runProposeReviewAsControlLoop, runReferenceReplay, runSelfPlay, runSemanticConceptJudge, runTestGradedScenario, runsForScenario, scalarScore, scanForMuffledGates, scoreAllProjects, scoreContinuity, scoreProject, scoreRedTeamOutput, scoreReferenceReplay, securityJudge, selectHarnessVariant, selfPreference, sentenceReorderMutator, serializeFeedbackTrajectoriesJsonl, signManifest, soc2Report, statusAdvanced, stopOnNoProgress, stopOnRepeatedAction, stripFencedJson, stuckLoopView, subjectiveEval, summarize, summarizeHarnessResults, summarizePreferenceMemory, summaryTable, testJudge, textInSnapshot, toLangfuseEnvelope, toNdjson, toPrometheusText, toolIntentAlignmentRubric, toolNamesForRun, toolNonRedundantRubric, toolSpans, toolSuccessRubric, toolWasteView, trialTraceFromMultiShotTrial, typoMutator, urlContains, validateRunRecord, verbosityBias, verifyManifest, visualDiff, viteDeployRunner, vitestTestParser, weightedMean, weightedRecall, welchsTTest, whitespaceCollapseMutator, wilcoxonSignedRank, withAssignedFeedbackSplit, wranglerDeployRunner };
+export { type ActionExecutionPolicy, type ActionPolicyDecision, type ActionableSideInfo, type ActiveLearningOptions, type AdapterRun, AgentDriver, type AgentDriverConfig, type AlignmentOp, type AntiSlopConfig, type AntiSlopIssue, type AntiSlopReport, type Artifact$1 as Artifact, type ArtifactCheck, type Artifact as ArtifactCheckArtifact, type ArtifactResult, type ArtifactValidator, type AsiSeverity, AxGepaSteeringOptimizer, type AxSteeringOptimizerConfig, BENCHMARK_SPLIT_SEED, type BaselineOptions, type BaselineReport, BehaviorAssertion, type BenchmarkAdapter, type BenchmarkDatasetItem, type BenchmarkEvaluation, type BenchmarkReport, BenchmarkRunner, type BenchmarkRunnerConfig, type BestOfNResult, type BisectOptions, type BisectResult, type BisectStep, type BootstrapOptions, type BootstrapResult, BudgetBreachError, type BudgetBreachFinding, type BudgetBreachReport, BudgetGuard, type BudgetLedgerEntry, type BudgetSpec, BuilderSession, type BuilderSessionInit, type CalibrationBin, type CalibrationOptions, type CalibrationReport, type CalibrationResult, CallExpectation, type CanaryAlert, type CanaryKind, type CanaryLeak, type CanaryOptions, type CanaryReport, type CanarySeverity, type CandidateScenario, type CandidateScore, type CausalAttributionReport, type ChatSummary, type CheckResult, type CodeMutationOutcome, type CodeMutationRunner, type CollectedArtifacts, type CommandRunner, type CompletionCriterion, type CompositePolicy, type ConceptComplexity, type ConceptFinding, type ConceptSpec, type ConceptWeightStrategy, type ContinuityCheck, type ContinuityCheckResult, type ContinuityReport, type ContinuitySnapshotPair, type ContractMetric, type ContractReport, type ControlActionFailureMode, type ControlActionOutcome, type ControlBudget, type ControlContext, type ControlDecision, type ControlEvalResult, type ControlRunResult, type ControlRuntimeConfig, type ControlRuntimeError, type ControlSeverity, type ControlStep, type ControlStopPolicies, ConvergenceTracker, type CorrelationReport, type CorrelationResult, type CorrelationStudyOptions, type CorrelationStudyResult, type CostEntry, CostLedger, type CostLedgerGeneration, type CostLedgerSnapshot, type CostSummary, CostTracker, type CounterfactualContext, type CounterfactualMutation, type CounterfactualResult, type CounterfactualRunner, type CreateCompositeMutatorOpts, type CreateDefaultReviewerOptions, type CreateSandboxCodeMutatorOpts, type CreateSandboxPoolOpts, type CrossTraceDiff, type CrossTraceDiffOptions, D1ExperimentStore, type D1ExperimentStoreOptions, type D1Like, type D1PreparedStatementLike, DEFAULT_AGENT_SLOS, DEFAULT_COMPLEXITY_WEIGHTS, DEFAULT_RULES as DEFAULT_FAILURE_RULES, DEFAULT_FINDERS, DEFAULT_HARNESS_OBJECTIVES, DEFAULT_MUTATION_PRIMITIVES, DEFAULT_MUTATORS, DEFAULT_REDACTION_RULES, DEFAULT_RED_TEAM_CORPUS, DEFAULT_RUN_SCORE_WEIGHTS, DEFAULT_SEVERITY_WEIGHTS, Dataset, type DatasetDifficulty, type DatasetManifest, type DatasetProvenance, type DatasetScenario, type DatasetSplit, type DeployFamily, type DeployGateLayerInput, type DeployRunResult, type DeployRunner, type DeploymentOutcome, type DirEntry, type Direction, type DivergenceOptions, type DivergenceReport, DockerSandboxDriver, type DriverResult, type DriverState, DualAgentBench, type DualAgentBenchConfig, type DualAgentReport, type DualAgentRound, type DualAgentScenario, type DualAgentScenarioResult, ERROR_COUNT_PATTERNS, type ErrorCountPattern, type EuRiskClass, type EvalMetricSpec, type EvalResult, type EventFilter, type EventKind, type EvolutionRound, type EvolvableVariant, type ExecutorConfig, type Expectation, type Experiment, type ExperimentPlan, type ExperimentResult, type Run as ExperimentRun, type ExperimentStore, ExperimentTracker, type ExportedRewardModel, type ExtractOptions, type ExtractResult, FAILURE_CLASSES, type FactorContribution, type FactorialCell, type FailureClass, type FailureClassification, type FailureCluster, type FailureClusterReport, type FailureContext, type FailureMode, type FailureRule, type FeedbackArtifactType, type FeedbackAttempt, type FeedbackLabel, type FeedbackLabelKind, type FeedbackLabelSource, type FeedbackOptimizerRow, type FeedbackOutcome, type FeedbackPattern, type FeedbackReplayAdapter, type FeedbackReplayResult, type FeedbackSeverity, type FeedbackSplitPolicy, type FeedbackTask, type FeedbackTrajectory, type FeedbackTrajectoryFilter, type FeedbackTrajectoryStore, FileSystemExperimentStore, type FileSystemExperimentStoreOptions, FileSystemFeedbackTrajectoryStore, FileSystemOutcomeStore, type FileSystemOutcomeStoreOptions, FileSystemTraceStore, type FileSystemTraceStoreOptions, type Finding, type FlowAction, type FlowLayerEnv, type FlowLayerFactoryInput, type FlowRunner, type FlowRunnerStepResult, type FlowSpec, type FlowStep, type GainDistributionBin, type GainDistributionFigureSpec, type GainDistributionOptions, type GateDecision, type GateEvidence, type GenerationReport, type GenericSpan, type GoldenItem, type GoldenSeverity, type GoldenSpec, type GovernanceContext, type GovernanceFinding, type GovernanceReport, type GradedStep, type HarnessAdapter, type HarnessConfig, type HarnessExperimentConfig, type HarnessExperimentResult, type HarnessIntervention, type HarnessRunRequest, type HarnessRunResult, type HarnessScenario, type HarnessSelection, type HarnessVariant, type HarnessVariantReport, HeldOutGate, type HeldOutGateConfig, type HeldOutGateRejectionCode, HoldoutAuditor, HoldoutLockedError, type HostedJudgeConfig, type HostedJudgeDimension, type HostedJudgeRequest, type HostedJudgeResponse, type HostedRunCriticConfig, type HostedRunScoreRequest, type HostedRunScoreResponse, type HypothesisManifest, type HypothesisResult, INTENT_MATCH_JUDGE_VERSION, type ImageData, InMemoryExperimentStore, InMemoryFeedbackTrajectoryStore, InMemoryOutcomeStore, InMemoryTraceStore, InMemoryTrialCache, InMemoryWorkspaceInspector, type InferenceScorer, type InspectorContext, type IntentMatchInput, type IntentMatchOptions, type IntentMatchResult, type InteractionContribution, JsonlTrialCache, type JudgeAgreementReport, type JudgeConfig, type JudgeFleetOptions, type JudgeFn, type JudgeInput, type JudgePair, type JudgeReplayGateArgs, type JudgeReplayResult, type JudgeRubric, JudgeRunner, type JudgeScore, type JudgeSpan, type KeywordConceptSpec, type KeywordCoverageFinding, type KeywordCoverageOptions, type KeywordCoverageResult, type LangfuseEnvelope, type LangfuseGeneration, type LangfuseScore, type Layer, type LayerCorrelation, type LayerResult, type LayerStatus, type LineageKind, type LineageKindResolver, type LineageNode, LineageRecorder, LlmCallError, type LlmCallRequest, type LlmCallResult, LlmClient, type LlmClientOptions, type LlmJsonCall, type LlmMessage, type LlmReviewerConfig, type LlmSpan, type LlmUsage, LockedJsonlAppender, MODEL_PRICING, type MatchResult, type MatcherResult, type MeasurementPolicy, type MergeOptions, type Message, type MetricSamples, type MetricVerdict, MetricsCollector, type MuffledFinder, type MuffledFinding, MultiLayerVerifier, type MultiShotGateConfig, type MultiShotGateResult, type MultiShotMutateAdapter, type MultiShotOptimizationConfig, type MultiShotOptimizationResult, type MultiShotRun, type MultiShotRunInput, type MultiShotRunner, type MultiShotScore, type MultiShotScorer, type MultiShotSplit, type MultiShotTrace, type MultiShotTrialResult, type MultiShotVariant, type MultiToolchainLayerConfig, type MutateAdapter, type MutationAttempt, type MutationChannel, MutationTelemetry, type Mutator, Mutex, NoopResearcher, OTEL_AGENT_EVAL_SCOPE, type Objective, type Oracle, type OracleObservation, type OracleReport, type OracleResult, type OrthogonalityInput, type OrthogonalityResult, type OtlpExport, type OtlpResourceSpans, type OtlpSpan, type OutcomeFilter, type OutcomePair, type OutcomeStore, type PairedBootstrapOptions, type PairedBootstrapResult, PairwiseSteeringOptimizer, type ParetoFigureSpec, type ParetoPoint, type ParetoResult, type PersonaConfig, type Playbook, type PlaybookEntry, type PoolSlot, type PositionalBiasResult, type PreferenceMemoryEntry, type PrmGradedTrace, PrmGrader, type PrmTrainingSample, ProductClient, type ProductClientConfig, type ProjectKind, ProjectRegistry, type ProjectSummary, type ProjectTimelineEntry, type PromptEvolutionConfig, type PromptEvolutionEvent, type PromptEvolutionResult, type PromptHandle, PromptRegistry, type TrialResult as PromptTrialResult, type ProposeFn, type ProposeInput, type ProposeOutput, type ProposeReviewConfig, type ProposeReviewControlAction, type ProposeReviewControlConfig, type ProposeReviewControlResult, type ProposeReviewControlState, type ProposeReviewReport, type ProposeReviewShot, type ProposedSideEffect, REDACTION_VERSION, type RedTeamCase, type RedTeamCategory, type RedTeamFinding, type RedTeamPayload, type RedTeamReport, type RedactionReport, type RedactionRule, type ReferenceMatchResult, type ReferenceReplayAdapter, type ReferenceReplayAdapterFn, type ReferenceReplayAdapterLike, type ReferenceReplayAggregate, type ReferenceReplayCandidate, type ReferenceReplayCase, type ReferenceReplayCaseRun, type ReferenceReplayExecutionScenario, type ReferenceReplayItem, type ReferenceReplayMatch, type ReferenceReplayMatchStrategy, type ReferenceReplayMatcher, type ReferenceReplayPromotionDecision, type ReferenceReplayPromotionPolicy, type ReferenceReplayRun, type ReferenceReplayRunContext, type ReferenceReplayRunOptions, type ReferenceReplayRunStore, type ReferenceReplayScenario, type ReferenceReplayScenarioScore, type ReferenceReplayScore, type ReferenceReplayScoreOptions, type ReferenceReplaySplit, type ReferenceReplaySplitComparison, type ReferenceReplaySteeringRowsOptions, type ReflectionContext, type ReflectionProposal, type RegressionOptions, type RegressionSpec, type ReleaseConfidenceAxis, type ReleaseConfidenceAxisName, type ReleaseConfidenceInput, type ReleaseConfidenceIssue, type ReleaseConfidenceMetrics, type ReleaseConfidenceScorecard, type ReleaseConfidenceStatus, type ReleaseConfidenceThresholds, type ReleaseTraceEvidence, type Researcher, type RetrievalSpan, type Review, type ReviewFn, type ReviewInput, type ReviewMemoryEntry, type ReviewMemoryStore, type ReviewerMemoryEntry, type ReviewerOutput, type ReviewerPromptInput, type ReviewerSoftFailDefaults, type ReviewerVerificationSummary, type RobustnessResult, type RouteMap, type RubricDimension, type Run$1 as Run, type RunAppScenarioOptions, type RunCommandInput, type RunCommandResult, type RunConfig, RunCritic, type RunCriticOptions, type RunDiff, type RunFilter, type RunJudgeMetadata, type RunLayer, type RunOutcome, type RunRecord, RunRecordValidationError, type RunScore, type RunScoreWeights, type RunSplitTag, type RunStatus, type RunTokenUsage, type RunTrace, SEMANTIC_CONCEPT_JUDGE_VERSION, type SandboxDriver, SandboxHarness, type SandboxHarnessResult, type SandboxJudgeKind, type SandboxJudgeResult, type SandboxJudgeSpec, type SandboxPool, type SandboxResult, type SandboxSpan, type ScanOptions, type Scenario, type ScenarioAggregate, type ScenarioCost, type ScenarioFile, ScenarioRegistry, type ScenarioResult, type ScoreAdapter, type ScoredTarget, type SelfPlayOptions, type SelfPlayProposer, type SelfPlayScorer, type SelfPreferenceResult, type SemanticConceptJudgeInput, type SemanticConceptJudgeOptions, type SemanticConceptJudgeResult, type SeriesConvergenceOptions, type SeriesConvergenceResult, type Severity, type ShipOptions, type SignedManifest, type SliceOptions, type Slo, type SloCheckResult, type SloComparator, type SloReport, type SloSeverity, type SlopCategory, type SlotFactory, type Span, type SpanBase, type SpanFilter, type SpanHandle, type SpanKind, type SpanStatus, type SteeringBundle, type SteeringChange, type SteeringDelta, type SteeringOptimizationResult, type SteeringOptimizationRow, type SteeringOptimizationSelector, type SteeringOptimizerBackend, type SteeringOptimizerConfig, type SteeringRolePrompt, type StepAttribution, type StepContext, type StepRubric, type StopDecision, type StuckLoopFinding, type StuckLoopOptions, type StuckLoopReport, SubprocessSandboxDriver, type SubprocessSandboxDriverOptions, type SummaryTable, type SummaryTableOptions, type SummaryTableRow, type SynthesisReason, type SynthesisTarget, TRACE_SCHEMA_VERSION, type TestGradedRunOptions, type TestGradedRunResult, type TestGradedScenario, type TestOutputParser, type TestResult, type ThreeLayerProjectReport, type ThresholdContract, TokenCounter, type TokenSpec, type ToolSpan, type ToolStats, type ToolUseMetrics, type ToolUseOptions, type ToolWasteFinding, type ToolWasteOptions, type ToolWasteReport, TraceEmitter, type TraceEmitterOptions, type TraceEvent, type TraceStore, type Trajectory, type TrajectoryStep, type TrialAttempt, type TrialCache, TrialTelemetry, type TrialTrace, type Turn, type TurnMetrics, type TurnResult, UNIVERSAL_FINDERS, type UseCaseSignals, type ValidationContext, type ValidationIssue, type ValidationResult, type VariantAggregate, type VerbosityBiasResult, type Verdict, type Verification, type VerificationReport, type VerifyContext, type VerifyFn, type VerifyOptions, type VisualDiffOptions, type VisualDiffResult, type ViteDeployRunnerInput, type WorkflowTopology, type WorkspaceAssertion, type WorkspaceAssertionResult, type WorkspaceInspector, type WorkspaceSnapshot, type WranglerDeployRunnerInput, adversarialJudge, aggregateLlm, aggregateRunScore, allCriticalPassed, analyzeAntiSlop, analyzeSeries, argHash, assertReleaseConfidence, assignFeedbackSplit, attributeCounterfactuals, deterministicSplit as benchmarkDeterministicSplit, index as benchmarks, benjaminiHochberg, bhAdjust, bisect, bonferroni, bootstrapCi, budgetBreachView, buildReflectionPrompt, buildReviewerPrompt, buildTrajectory, byteLengthRange, calibrateJudge, calibrationCurve, callLlm, callLlmJson, canaryLeakView, causalAttribution, checkCanaries, checkSlos, clamp01, classifyEuAiRisk, classifyFailure, codeExecutionJudge, cohensD, coherenceJudge, collectionPreserved, commitBisect, compareReferenceReplay, compareToBaseline, compilerJudge, composeParsers, composeValidators, computeToolUseMetrics, confidenceInterval, containsAll, controlFailureClassFromVerification, controlRunToFeedbackTrajectory, correlateLayers, correlationStudy, createAntiSlopJudge, createCompositeMutator, createCustomJudge, createDefaultReviewer, createDomainExpertJudge, createFeedbackTrajectory, createIntentMatchJudge, createLlmReviewer, createSandboxCodeMutator, createSandboxPool, createSemanticConceptJudge, crossTraceDiff, crowdingDistance, decideReferenceReplayPromotion, decideReferenceReplayRunPromotion, defaultJudges, defaultMultiShotObjectives, defaultReferenceReplayMatcher, deployGateLayer, distillPlaybook, dominates, estimateCost, estimateTokens, euAiActReport, evaluateActionPolicy, evaluateContract, evaluateHypothesis, evaluateOracles, evaluateReleaseConfidence, executeScenario, expectAgent, exportRewardModel, exportRunAsOtlp, exportTrainingData, extractAssetUrls, extractErrorCount, failureClusterView, feedbackTrajectoriesToDatasetScenarios, feedbackTrajectoriesToOptimizerRows, feedbackTrajectoryToDatasetScenario, feedbackTrajectoryToOptimizerRow, fileContains, fileExists, findAutoMatchNoExpectation, findConstructorCwdDropped, findFallbackToPass, findLiteralTruePass, findSkipCountsAsPass, firstDivergenceView, flowLayer, formatBenchmarkReport, formatDriverReport, formatFindings, gainHistogram, precision as goldenPrecision, gradeSemanticStatus, groupBy, hashContent, hashScenarios, htmlContainsElement, inMemoryReferenceReplayStore, inMemoryReviewStore, interRaterReliability, iqr, isJudgeSpan, isLlmSpan, isPrmVerdict, isRetrievalSpan, isRunRecord, isSandboxSpan, isToolSpan, jestTestParser, jsonHasKeys, jsonShape, jsonlReferenceReplayStore, jsonlReviewStore, judgeAgreementView, judgeReplayGate, judgeSpans, keyPreserved, linterJudge, llmSpanFromProvider, llmSpans, loadScorerFromGrader, localCommandRunner, lowercaseMutator, mannWhitneyU, matchGoldens, mergeLayerResults, mergeSteeringBundle, multiToolchainLayer, nistAiRmfReport, nonRefusalRubric, normalizeScores, notBlocked, objectiveEval, outputLengthRubric, pairedBootstrap, pairedTTest, pairedWilcoxon, paraphraseRobustness, paretoChart, paretoFrontier, paretoFrontierWithCrowding, parseFeedbackTrajectoriesJsonl, parseReflectionResponse, parseRunRecordSafe, partialCredit, passOrthogonality, pixelDeltaRatio, politenessPrefixMutator, positionalBias, printDriverSummary, prmBestOfN, prmEnsembleBestOfN, probeLlm, promptBisect, proposeSynthesisTargets, pytestTestParser, redTeamDataset, redTeamReport, redactString, redactValue, referenceReplayRunsToSteeringRows, referenceReplayScenarioToRunScore, regexMatch, regexMatches, regressionView, releaseTraceEvidenceFromMultiShotTrials, renderMarkdown, renderMarkdownReport, renderPlaybookMarkdown, renderPreferenceMemoryMarkdown, renderSteeringText, replayFeedbackTrajectories, replayFeedbackTrajectory, replayScorerOverCorpus, replayTraceThroughJudge, requiredSampleSize, resetLockedAppendersForTesting, resumeBuilderSession, roundTripRunRecord, rowCount, rowWhere, runAgentControlLoop, runAssertions, runCanaries, runCounterfactual, runE2EWorkflow, runExpectations, runFailureClass, runHarnessExperiment, runIntentMatchJudge, runJudgeFleet, runKeywordCoverageJudge, runKeywordCoverageJudgeUrl, runMultiShotOptimization, runPromptEvolution, runProposeReview, runProposeReviewAsControlLoop, runReferenceReplay, runSelfPlay, runSemanticConceptJudge, runTestGradedScenario, runsForScenario, scalarScore, scanForMuffledGates, scoreAllProjects, scoreContinuity, scoreProject, scoreRedTeamOutput, scoreReferenceReplay, securityJudge, selectHarnessVariant, selfPreference, sentenceReorderMutator, serializeFeedbackTrajectoriesJsonl, signManifest, soc2Report, statusAdvanced, stopOnNoProgress, stopOnRepeatedAction, stripFencedJson, stuckLoopView, subjectiveEval, summarize, summarizeHarnessResults, summarizePreferenceMemory, summaryTable, testJudge, textInSnapshot, toLangfuseEnvelope, toNdjson, toPrometheusText, toolIntentAlignmentRubric, toolNamesForRun, toolNonRedundantRubric, toolSpans, toolSuccessRubric, toolWasteView, trialTraceFromMultiShotTrial, typoMutator, urlContains, validateRunRecord, verbosityBias, verifyManifest, visualDiff, viteDeployRunner, vitestTestParser, weightedMean, weightedRecall, welchsTTest, whitespaceCollapseMutator, wilcoxonSignedRank, withAssignedFeedbackSplit, wranglerDeployRunner };