@tangle-network/agent-eval 0.144.8 → 0.144.10

package/dist/index.d.ts

@@ -2,50 +2,50 @@ import { a as JudgeError, c as ReplayError, i as ConfigError, l as ValidationErr
 import { C as verifyAgentProfileCell, S as validateAgentProfileCell, _ as buildAgentInterfaceProfileCell, a as AgentProfileCellSchemaVersion, b as requireAgentProfileCell, c as AgentProfileHarness, d as AgentProfileKind, f as AgentProfileSource, g as assertRunAgentProfileCell, h as agentProfileCellKey, i as AgentProfileCellInput, l as AgentProfileJson, m as agentProfileCellHashMaterial, n as AgentInterfaceProfileLike, o as AgentProfileCellValidationError, p as AgentProfileSourceInput, r as AgentProfileCell, s as AgentProfileDimensionValue, t as AGENT_PROFILE_KINDS, u as AgentProfileJsonObject, v as buildAgentProfileCell, x as toAgentProfileJson, y as groupRunsByAgentProfileCell } from "./agent-profile-cell-BOP-iA9Q.js";
 import { a as StrategyChecker, c as VerificationStrategyProfile, i as CheckerOutcome, l as VerificationStrategySource, n as VerdictCertification, o as VERIFICATION_STRATEGIES, r as CheckerIdentity, s as VERIFICATION_STRATEGY_SOURCES, t as DefaultVerdict } from "./verdict-DExhxfgR.js";
 import { a as MultiLayerVerifier, c as VerifyContext, i as LayerStatus, l as VerifyOptions, n as Layer, o as Severity, r as LayerResult, s as VerificationReport, t as Finding, u as gradeSemanticStatus } from "./multi-layer-verifier-DnAqwl0h.js";
-import { $ as RunScore, B as ConceptSpec, E as FindingSubjectKind, G as SemanticConceptJudgeOptions, H as DEFAULT_COMPLEXITY_WEIGHTS, J as runSemanticConceptJudge, K as SemanticConceptJudgeResult, L as defineTraceAnalyst, M as DspyRlmTraceEngineOptions, N as createDspyRlmTraceEngine, Q as DEFAULT_RUN_SCORE_WEIGHTS, R as ConceptComplexity, T as FindingSubject, U as SEMANTIC_CONCEPT_JUDGE_VERSION, V as ConceptWeightStrategy, W as SemanticConceptJudgeInput, X as RunCriticOptions, Y as RunCritic, Z as RunTrace, _ as FindingsDiff, b as defaultIsMaterial, c as DEFAULT_TRACE_ANALYST_KINDS, d as IMPROVEMENT_KIND_SPEC, et as RunScoreWeights, f as FAILURE_MODE_KIND_SPEC, g as DiffPolicy, h as emitControlIntegrityFindings, l as KNOWLEDGE_POISONING_KIND_SPEC, m as ControlIntegrityAnalyst, n as SkillUsageAnalyst, nt as clamp01, p as CONTROL_INTEGRITY_ANALYST, q as createSemanticConceptJudge, t as SKILL_USAGE_ANALYST, tt as aggregateRunScore, u as KNOWLEDGE_GAP_KIND_SPEC, v as FindingsStore, x as diffFindings, y as PersistedFinding, z as ConceptFinding } from "./skill-usage-CJlWEUFt.js";
+import { $ as RunScore, B as ConceptSpec, E as FindingSubjectKind, G as SemanticConceptJudgeOptions, H as DEFAULT_COMPLEXITY_WEIGHTS, J as runSemanticConceptJudge, K as SemanticConceptJudgeResult, L as defineTraceAnalyst, M as DspyRlmTraceEngineOptions, N as createDspyRlmTraceEngine, Q as DEFAULT_RUN_SCORE_WEIGHTS, R as ConceptComplexity, T as FindingSubject, U as SEMANTIC_CONCEPT_JUDGE_VERSION, V as ConceptWeightStrategy, W as SemanticConceptJudgeInput, X as RunCriticOptions, Y as RunCritic, Z as RunTrace, _ as FindingsDiff, b as defaultIsMaterial, c as DEFAULT_TRACE_ANALYST_KINDS, d as IMPROVEMENT_KIND_SPEC, et as RunScoreWeights, f as FAILURE_MODE_KIND_SPEC, g as DiffPolicy, h as emitControlIntegrityFindings, l as KNOWLEDGE_POISONING_KIND_SPEC, m as ControlIntegrityAnalyst, n as SkillUsageAnalyst, nt as clamp01, p as CONTROL_INTEGRITY_ANALYST, q as createSemanticConceptJudge, t as SKILL_USAGE_ANALYST, tt as aggregateRunScore, u as KNOWLEDGE_GAP_KIND_SPEC, v as FindingsStore, x as diffFindings, y as PersistedFinding, z as ConceptFinding } from "./skill-usage-CO9OLRBx.js";
 import { C as PendingCostCall, D as costForUsage, E as costForTokenPricing, O as modelPriceKey, S as PaidCallResult, T as RunPaidCallInput, _ as CostReservationExceededError, a as CostChannel, b as CustomTokenPricing, c as CostLedgerHandle, d as CostLedgerPersistenceError, f as CostLedgerSummary, g as CostReceiptInput, h as CostReceiptCaptureError, i as CostCeilingReachedError, l as CostLedgerOptions, m as CostReceipt, n as CostAccountingIncompleteError, o as CostLedger, p as CostProvenance, r as CostCallConflictError, s as CostLedgerFilter, t as ChannelRollup, u as CostLedgerPersistence, v as CostResult, w as PendingCostCallView, x as MaximumCharge, y as CostUsage } from "./cost-ledger-Bv_e8XHY.js";
 import { C as TraceEvent, D as isSandboxSpan, E as isRetrievalSpan, O as isToolSpan, S as ToolSpan, T as isLlmSpan, _ as Span, a as FAILURE_CLASSES, b as SpanStatus, c as JudgeSpan, d as RetrievalSpan, f as Run, g as SandboxSpan, h as RunStatus, i as EventKind, l as LlmSpan, n as BudgetLedgerEntry, o as FailureClass, p as RunLayer, r as BudgetSpec, s as GenericSpan, t as Artifact, u as Message, v as SpanBase, w as isJudgeSpan, x as TRACE_SCHEMA_VERSION, y as SpanKind } from "./schema-BtVldJ3T.js";
-import { a as RunRecord, c as RunTaskFailure, d as isRunRecord, f as modelHasSnapshot, g as validateRunRecord, h as runTaskScore, i as RunOutcome, l as RunTerminalOutcome, m as roundTripRunRecord, n as RunCostProvenance, o as RunRecordValidationError, p as parseRunRecordSafe, r as RunJudgeMetadata, s as RunSplitTag, t as JudgeScoresRecord, u as RunTokenUsage } from "./run-record-DdSa93_W.js";
-import { a as ExtractedUsage, c as extractUsageFromResponse, i as ExtractUsageFromSseOptions, l as extractUsageFromSse, n as CaptureFetchOptions, o as SseUsageMode, r as captureFetchToRawSink, s as extractUsage, t as CaptureFetchContext } from "./index-CwDrUMe0.js";
+import { a as RunRecord, c as RunTaskFailure, d as isRunRecord, f as modelHasSnapshot, g as validateRunRecord, h as runTaskScore, i as RunOutcome, l as RunTerminalOutcome, m as roundTripRunRecord, n as RunCostProvenance, o as RunRecordValidationError, p as parseRunRecordSafe, r as RunJudgeMetadata, s as RunSplitTag, t as JudgeScoresRecord, u as RunTokenUsage } from "./run-record-CF4Dwpxr.js";
+import { a as ExtractedUsage, c as extractUsageFromResponse, i as ExtractUsageFromSseOptions, l as extractUsageFromSse, n as CaptureFetchOptions, o as SseUsageMode, r as captureFetchToRawSink, s as extractUsage, t as CaptureFetchContext } from "./index-CQTZ-4XN.js";
 import { $ as assertLlmRoute, A as ChatClient, At as InMemoryRawProviderSinkOptions, B as SandboxSdkTransportOpts, C as ScenarioFile, Ct as CrossFamilyError, D as TurnMetrics, Dt as FileSystemRawProviderSink, E as Turn, Et as judgeFamily, F as CreateChatClientOpts, Ft as RawProviderSink, G as LlmCallResult, H as LlmCallError, I as CustomTransportOpts, It as RawProviderSinkFilter, J as LlmMessage, K as LlmClient, L as DirectProviderTransportOpts, Lt as defaultProviderRedactor, M as ChatResponse, Mt as ProviderRedactor, N as ChatTransport, Nt as RawProviderDirection, O as TurnResult, Ot as FileSystemRawProviderSinkOptions, P as CliBridgeTransportOpts, Pt as RawProviderEvent, Q as LlmUsage, R as MockTransportOpts, Rt as providerFromBaseUrl, S as Scenario, St as AssertCrossFamilyOptions, T as TestResult, Tt as assertCrossFamily, U as LlmCallMetadata, V as createChatClient, W as LlmCallRequest, X as LlmRouteAssertionError, Y as LlmResponseError, Z as LlmRouteRequirements, _ as PersonaConfig, _t as assertServedModel, a as CheckResult, at as isTransientLlmError, b as RouteMap, bt as normalizeModelId, c as DriverResult, ct as stripFencedJson, d as FeedbackPattern, dt as ModelSubstitutionError, et as backoffMs, f as JudgeConfig, ft as PROBE_MAX_TOKENS, g as JudgeScore, gt as assertCrossFamilyServed, h as JudgeRubric, ht as ServedModelVerdict, i as BenchmarkRunnerConfig, it as costReceiptFromLlmError, j as ChatRequest, jt as NoopRawProviderSink, k as ChatCallOpts, kt as InMemoryRawProviderSink, l as DriverState, lt as AssertCrossFamilyServedOptions, m as JudgeInput, mt as ServedModelCheck, n as ArtifactResult, nt as callLlmJson, o as CollectedArtifacts, ot as maximumChargeForLlmRequest, p as JudgeFn, pt as ServedCrossFamilyError, q as LlmClientOptions, r as BenchmarkReport, rt as costReceiptFromLlm, s as CompletionCriterion, st as probeLlm, t as ArtifactCheck, tt as callLlm, u as EvalResult, ut as AssertServedModelOptions, v as PersonaRigor, vt as assertServedModels, w as ScenarioResult, wt as JudgeFamily, x as RubricDimension, xt as servedModelAcceptable, y as ProductClientConfig, yt as checkServedModel, z as RouterTransportOpts } from "./types-D216SgwM.js";
 import { a as RunFilter, i as InMemoryTraceStore, n as FileSystemTraceStore, o as SpanFilter, r as FileSystemTraceStoreOptions, s as TraceStore, t as EventFilter } from "./store-CT9YIIve.js";
 import { a as TraceEmitterOptions, i as TraceEmitter, n as RunCompleteHookContext, o as llmSpanFromProvider, r as SpanHandle, t as RunCompleteHook } from "./emitter-DGQGoLyj.js";
 import { a as RunIntegrityReport, i as RunIntegrityIssueCode, n as RunIntegrityExpectations, o as assertRunCaptured, r as RunIntegrityIssue, s as throwIfRunIncomplete, t as RunIntegrityError } from "./integrity-BuqEKu-x.js";
-import { $ as traceAnalystOnRunComplete, A as stringField, At as OtlpSpanRoleInput, B as TraceInsightReadiness, Bt as exportRunAsOtlp, C as ProjectedOtlpSpan, Ct as createOtelTracingStore, D as inferOtlpKind, Dt as OtelExporter, E as firstStringAttr, Et as OtelExportConfig, F as TraceInsightFinding, Ft as traceSpanKindToOpenInferenceKind, G as defaultTraceInsightPanel, H as TraceInsightTask, I as TraceInsightPanelRole, It as OTEL_AGENT_EVAL_SCOPE, J as inferDomainKeywords, K as describeTraceInsightScope, L as TraceInsightPromptInput, Lt as OtlpExport, M as flattenOtlpExportToNdjson, Mt as applyToolSpanOtlpAttributes, N as OtlpFlatLine, Nt as classifyOtlpSpanRole, O as projectOtlpFlatLine, Ot as createOtelExporter, P as TraceInsightContext, Pt as isOtlpModelCall, Q as TraceAnalystHookOptions, R as TraceInsightQualityGate, Rt as OtlpResourceSpans, S as otlpToTraceRunRecords, St as redactValue, T as extractOtlpAttributes, Tt as ExportableSpan, U as buildTraceInsightContext, V as TraceInsightSuite, W as buildTraceInsightPrompt, X as scoreTraceInsightReadiness, Y as planTraceInsightQuestions, Z as tokenizeDomainWords, _ as OtlpTraceRunRecord, _t as DEFAULT_REDACTION_RULES, a as ReplayFetchOptions, at as TraceFileMissingError, b as otlpRowsToTraceRunRecords, bt as RedactionRule, c as ToolTraceMissingError, ct as AnalyzeTracesInput, d as OtlpFileTraceStoreOptions, dt as analyzeTraces, et as SpanNotFoundError, f as ToolSpansToTraceAnalysisStoreOptions, ft as createBoundedTraceAnalysisStore, g as OtlpToRunRecordsOptions, gt as convertTraceStoresToOtlp, h as TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, ht as TracesToOtlpResult, i as ReplayCacheStats, it as TraceFileMalformedError, j as FlattenOtlpOptions, jt as ToolSpanOtlpInput, k as readOtlpStatus, kt as OtlpSpanRole, l as toolSpansToTraceAnalysisStore, lt as AnalyzeTracesOptions, m as TRACE_ANALYST_ACTOR_DESCRIPTION, mt as TraceStoreToOtlpOptions, n as ReplayCacheEntry, nt as TraceAnalysisStoreContractError, o as createReplayFetch, ot as TraceFileTooLargeError, p as otlpTextToTraceAnalysisStore, pt as TraceStoreSource, q as domainEvidencePattern, r as ReplayCacheMissError, rt as TraceAnalysisValidationError, s as iterateRawCalls, st as TraceNotFoundError, t as ReplayCache, tt as TraceAnalysisLimitError, u as OtlpFileTraceStore, ut as AnalyzeTracesResult, v as TraceAggregate, vt as REDACTION_VERSION, w as asString, wt as otelRunCompleteHook, x as otlpToRunRecords, xt as redactString, y as otlpRowsToRunRecords, yt as RedactionReport, z as TraceInsightQuestion, zt as OtlpSpan } from "./replay-DFf-teiC.js";
+import { $ as traceAnalystOnRunComplete, A as stringField, At as OtlpSpanRoleInput, B as TraceInsightReadiness, Bt as exportRunAsOtlp, C as ProjectedOtlpSpan, Ct as createOtelTracingStore, D as inferOtlpKind, Dt as OtelExporter, E as firstStringAttr, Et as OtelExportConfig, F as TraceInsightFinding, Ft as traceSpanKindToOpenInferenceKind, G as defaultTraceInsightPanel, H as TraceInsightTask, I as TraceInsightPanelRole, It as OTEL_AGENT_EVAL_SCOPE, J as inferDomainKeywords, K as describeTraceInsightScope, L as TraceInsightPromptInput, Lt as OtlpExport, M as flattenOtlpExportToNdjson, Mt as applyToolSpanOtlpAttributes, N as OtlpFlatLine, Nt as classifyOtlpSpanRole, O as projectOtlpFlatLine, Ot as createOtelExporter, P as TraceInsightContext, Pt as isOtlpModelCall, Q as TraceAnalystHookOptions, R as TraceInsightQualityGate, Rt as OtlpResourceSpans, S as otlpToTraceRunRecords, St as redactValue, T as extractOtlpAttributes, Tt as ExportableSpan, U as buildTraceInsightContext, V as TraceInsightSuite, W as buildTraceInsightPrompt, X as scoreTraceInsightReadiness, Y as planTraceInsightQuestions, Z as tokenizeDomainWords, _ as OtlpTraceRunRecord, _t as DEFAULT_REDACTION_RULES, a as ReplayFetchOptions, at as TraceFileMissingError, b as otlpRowsToTraceRunRecords, bt as RedactionRule, c as ToolTraceMissingError, ct as AnalyzeTracesInput, d as OtlpFileTraceStoreOptions, dt as analyzeTraces, et as SpanNotFoundError, f as ToolSpansToTraceAnalysisStoreOptions, ft as createBoundedTraceAnalysisStore, g as OtlpToRunRecordsOptions, gt as convertTraceStoresToOtlp, h as TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, ht as TracesToOtlpResult, i as ReplayCacheStats, it as TraceFileMalformedError, j as FlattenOtlpOptions, jt as ToolSpanOtlpInput, k as readOtlpStatus, kt as OtlpSpanRole, l as toolSpansToTraceAnalysisStore, lt as AnalyzeTracesOptions, m as TRACE_ANALYST_ACTOR_DESCRIPTION, mt as TraceStoreToOtlpOptions, n as ReplayCacheEntry, nt as TraceAnalysisStoreContractError, o as createReplayFetch, ot as TraceFileTooLargeError, p as otlpTextToTraceAnalysisStore, pt as TraceStoreSource, q as domainEvidencePattern, r as ReplayCacheMissError, rt as TraceAnalysisValidationError, s as iterateRawCalls, st as TraceNotFoundError, t as ReplayCache, tt as TraceAnalysisLimitError, u as OtlpFileTraceStore, ut as AnalyzeTracesResult, v as TraceAggregate, vt as REDACTION_VERSION, w as asString, wt as otelRunCompleteHook, x as otlpToRunRecords, xt as redactString, y as otlpRowsToRunRecords, yt as RedactionReport, z as TraceInsightQuestion, zt as OtlpSpan } from "./replay-DbhZ4Ked.js";
 import { C as TOOL_LATENCY_MS, D as asNumber, E as applyLlmSpanOtlpAttributes, O as contextInputTokens, S as TOOL_ARGS_CAPTURED, T as TOOL_NAME_ATTR_KEYS, _ as LlmSpanOtlpInput, a as LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, b as RUN_COST_ATTR_KEYS, c as LLM_COST_USD, d as LLM_MODEL_ATTR_KEYS, f as LLM_MODEL_NAME, g as LLM_REASONING_TOKEN_ATTR_KEYS, h as LLM_REASONING_TOKENS, i as LLM_CACHE_WRITE_TOKENS, k as firstNumberAttr, l as LLM_INPUT_TOKENS, m as LLM_OUTPUT_TOKEN_ATTR_KEYS, n as LLM_CACHED_TOKENS, o as LLM_CONTEXT_TOKENS, p as LLM_OUTPUT_TOKENS, r as LLM_CACHED_TOKEN_ATTR_KEYS, s as LLM_COST_ATTR_KEYS, t as INPUT_VALUE, u as LLM_INPUT_TOKEN_ATTR_KEYS, v as OPENINFERENCE_SPAN_KIND, w as TOOL_NAME, x as SPAN_KIND_ATTR_KEYS, y as OUTPUT_VALUE } from "./attribute-vocabulary-DLJ6303h.js";
 import { a as judgeSpans, c as runsForScenario, i as hasCapturedToolArgs, l as toolSpans, n as argHash, o as llmSpans, r as groupBy, s as runFailureClass, t as aggregateLlm } from "./query-CJ_DX8vl.js";
-import { A as SearchSpanResult, B as ViewSpansResult, C as TRACE_ANALYSIS_LIMITS, D as DatasetOverview, E as DEFAULT_TRACE_ANALYST_BUDGETS, F as TraceAnalystFilters, H as ViewTraceResult, I as TraceAnalystSpan, L as TraceAnalystSpanKind, M as SpanMatchRecord, N as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, O as ErrorCluster, P as TraceAnalystByteBudgets, R as TraceAnalystSpanStatus, S as BoundedTraceAnalysisStoreOptions, T as TraceAnalysisStoreContext, V as ViewTraceOversized, _ as ProposalFinding, a as AnalystInputKind, b as makeFinding, c as AnalystRunInputs, d as AnalystSeverity, f as AnalystUsageReceipt, g as ExecutionProbeRequest, h as ExecutionProbeOutcome, i as AnalystFinding, j as SearchTraceResult, k as QueryTracesPage, l as AnalystRunResult, m as ExecutionProbe, n as AnalystContext, o as AnalystRequirements, p as EvidenceRef, r as AnalystCost, s as AnalystRunEvent, t as Analyst, u as AnalystRunSummary, v as ProposalFindingOrigin, w as TraceAnalysisStore, x as makeProposalFinding, y as computeFindingId, z as TraceAnalystTraceSummary } from "./types-DF_Udrp-.js";
-import { a as createTraceAnalyst, c as runTraceAnalyst, f as BehavioralMetrics, g as computeTraceMetrics, h as SuboptimalSignal, i as TraceAnalystDefinition, m as SuboptimalCode, n as buildDefaultAnalystRegistry, o as renderPriorFindings, p as BehavioralTokenSequence, r as CreateTraceAnalystOptions, s as renderUpstreamFindings, t as DefaultAnalystRegistryOptions } from "./default-registry-BZhStdjl.js";
-import { a as ExactAnalystRunPolicySnapshot, c as ExactAnalystSnapshot, d as ExactExecutionComponentSnapshot, i as ExactAnalystRunEvent, l as ExactCapableAnalyst, n as ExactAnalystExecutionPlanSnapshot, o as ExactAnalystRunResult, r as ExactAnalystRunCompletion, s as ExactAnalystRunSummary, t as ExactAnalystBudgetSnapshot, u as ExactExecutionComponentIdentity } from "./exact-types-Djvzosly.js";
-import { A as ExactAnalystRunExecutionError, C as jsonHasKeys, D as AnalystRegistryOptions, E as AnalystRegistry, M as RegistryRunOpts, N as assertExactRegistryRunOpts, O as BudgetPolicy, S as containsAll, T as AnalystHooks, _ as ValidationContext, a as ProducedProposal, b as byteLengthRange, c as SatisfiedBy, d as createLlmCorrectnessChecker, f as createTokenRecallChecker, g as ArtifactValidator, h as Artifact$1, i as LlmCorrectnessCheckerOpts, j as ExactRegistryRunOpts, k as ExactAnalystBudgetPolicy, l as TaskGold, m as verifyCompletion, n as CompletionVerdict, o as ProducedState, p as parseCorrectnessResponse, r as CorrectnessChecker, s as RequirementCheck, t as CompletionRequirement, u as completionVerdict, v as ValidationIssue, w as regexMatch, x as composeValidators, y as ValidationResult } from "./completion-verifier-foUCLif_.js";
-import { a as ProfileAxisSpec, c as agentProfileModelId, i as HarnessType, l as expandProfileAxes, n as CODING_HARNESSES, o as agentProfileHash, r as HARNESS_NATIVE_MODEL, s as agentProfileId, t as AgentProfile, u as harnessAxisOf } from "./agent-profile-CgDTo40f.js";
-import { R as Scenario$1, S as JudgeConfig$1, w as JudgeScore$1 } from "./types-DYuNHo9R.js";
+import { A as SearchSpanResult, B as ViewSpansResult, C as TRACE_ANALYSIS_LIMITS, D as DatasetOverview, E as DEFAULT_TRACE_ANALYST_BUDGETS, F as TraceAnalystFilters, H as ViewTraceResult, I as TraceAnalystSpan, L as TraceAnalystSpanKind, M as SpanMatchRecord, N as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, O as ErrorCluster, P as TraceAnalystByteBudgets, R as TraceAnalystSpanStatus, S as BoundedTraceAnalysisStoreOptions, T as TraceAnalysisStoreContext, V as ViewTraceOversized, _ as ProposalFinding, a as AnalystInputKind, b as makeFinding, c as AnalystRunInputs, d as AnalystSeverity, f as AnalystUsageReceipt, g as ExecutionProbeRequest, h as ExecutionProbeOutcome, i as AnalystFinding, j as SearchTraceResult, k as QueryTracesPage, l as AnalystRunResult, m as ExecutionProbe, n as AnalystContext, o as AnalystRequirements, p as EvidenceRef, r as AnalystCost, s as AnalystRunEvent, t as Analyst, u as AnalystRunSummary, v as ProposalFindingOrigin, w as TraceAnalysisStore, x as makeProposalFinding, y as computeFindingId, z as TraceAnalystTraceSummary } from "./types-BvDKaULh.js";
+import { a as createTraceAnalyst, c as runTraceAnalyst, f as BehavioralMetrics, g as computeTraceMetrics, h as SuboptimalSignal, i as TraceAnalystDefinition, m as SuboptimalCode, n as buildDefaultAnalystRegistry, o as renderPriorFindings, p as BehavioralTokenSequence, r as CreateTraceAnalystOptions, s as renderUpstreamFindings, t as DefaultAnalystRegistryOptions } from "./default-registry-Bf8Woqmq.js";
+import { a as ExactAnalystRunPolicySnapshot, c as ExactAnalystSnapshot, d as ExactExecutionComponentSnapshot, i as ExactAnalystRunEvent, l as ExactCapableAnalyst, n as ExactAnalystExecutionPlanSnapshot, o as ExactAnalystRunResult, r as ExactAnalystRunCompletion, s as ExactAnalystRunSummary, t as ExactAnalystBudgetSnapshot, u as ExactExecutionComponentIdentity } from "./exact-types-DSFFpLLI.js";
+import { A as ExactAnalystRunExecutionError, C as jsonHasKeys, D as AnalystRegistryOptions, E as AnalystRegistry, M as RegistryRunOpts, N as assertExactRegistryRunOpts, O as BudgetPolicy, S as containsAll, T as AnalystHooks, _ as ValidationContext, a as ProducedProposal, b as byteLengthRange, c as SatisfiedBy, d as createLlmCorrectnessChecker, f as createTokenRecallChecker, g as ArtifactValidator, h as Artifact$1, i as LlmCorrectnessCheckerOpts, j as ExactRegistryRunOpts, k as ExactAnalystBudgetPolicy, l as TaskGold, m as verifyCompletion, n as CompletionVerdict, o as ProducedState, p as parseCorrectnessResponse, r as CorrectnessChecker, s as RequirementCheck, t as CompletionRequirement, u as completionVerdict, v as ValidationIssue, w as regexMatch, x as composeValidators, y as ValidationResult } from "./completion-verifier-DJA5BhPb.js";
+import { a as ProfileAxisSpec, c as agentProfileModelId, i as HarnessType, l as expandProfileAxes, n as CODING_HARNESSES, o as agentProfileHash, r as HARNESS_NATIVE_MODEL, s as agentProfileId, t as AgentProfile, u as harnessAxisOf } from "./agent-profile-DPi7IZg7.js";
+import { R as Scenario$1, S as JudgeConfig$1, w as JudgeScore$1 } from "./types-BnjdJ70P.js";
 import { a as DatasetScenario, c as SliceOptions, i as DatasetProvenance, l as hashScenarios, n as DatasetDifficulty, o as DatasetSplit, r as DatasetManifest, s as HoldoutLockedError, t as Dataset } from "./dataset-C8xaLXdY.js";
 import { $ as pairedCohensDz, A as WeightedCompositeInput, At as positionalBias, B as eProcess, C as RankTestMethod, Ct as GoldenItem, D as ScoreRiskDifferenceResult, Dt as calibrateJudge, E as RiskDifferenceResult, Et as VerbosityBiasResult, F as cliffsDelta, G as mannWhitneyU, H as interRaterReliability, I as cohensD, J as mcnemarRequiredN, K as mcnemar, L as confidenceInterval, M as WilcoxonSignedRankResult, Mt as verbosityBias, N as benjaminiHochberg, O as SignTestAlternative, Ot as calibrateJudgeContinuous, P as bonferroni, Q as pairedBootstrap, R as corpusInterRaterAgreement, S as ProportionInterval, St as ContinuousCalibrationResult, T as RankTestOptions, Tt as SelfPreferenceResult, U as interpretCliffs, V as holm, W as isBinaryOutcomeVector, X as normalizeScores, Y as mulberry32, Z as pairedBinaryScale, _ as McNemarResult, _t as wilson, a as CorpusAgreementReport, at as pairedSignTest, b as PairedSignTestResult, bt as ContinuousAgreement, c as DEFAULT_PERMUTATIONS, ct as passAtK, d as EProcessState, dt as requiredPairedSampleSize, et as pairedDeltaTieFraction, f as EProcessStep, ft as requiredSampleSize, g as MannWhitneyResult, gt as wilcoxonSignedRank, h as MANN_WHITNEY_EXACT_MAX_WORK, ht as weightedMean, i as CorpusAgreementPerDimension, it as pairedRiskDifferenceScore, j as WeightedCompositeResult, jt as selfPreference, k as WILCOXON_EXACT_MAX_N, kt as continuousAgreement, l as EProcess, lt as pearsonR, m as MANN_WHITNEY_EXACT_MAX_STATES, mt as weightedComposite, n as CliffsMagnitude, nt as pairedRiskDifference, o as CorpusScoreRecord, ot as pairedTTest, p as ExactRiskDifferenceResult, pt as spearmanR, q as mcnemarPower, r as CorpusAgreementOptions, rt as pairedRiskDifferenceExact, s as DECISION_PAIRED_DELTA_STATISTIC, st as partialCredit, t as BOOTSTRAP_GATE_MIN_N, tt as pairedMde, u as EProcessOptions, ut as ranks, v as PairedBootstrapOptions, vt as CalibrationResult, w as RankTestMethodRequest, wt as PositionalBiasResult, x as PairedTTestResult, xt as ContinuousAgreementOptions, y as PairedBootstrapResult, yt as CandidateScore, z as corpusInterRaterAgreementFromJudgeScores } from "./statistics-D6Uebe_4.js";
 import { a as PairedPromotionDecision, c as pairedDecisionShape, i as PairedMcNemarEvidence, n as PairedDecisionShape, o as PairedPromotionDecisionOptions, r as PairedDecisionStatistic, s as decidePairedPromotion, t as PairedDecisionMethod } from "./paired-promotion-decision-B6zJ3gYM.js";
-import { C as HeldOutGate, E as SplitCoverage, S as GateEvidence, T as HeldOutGateRejectionCode, _ as paretoChart, a as ParetoPoint, b as DeltaStatistic, c as ResearchReportCandidate, d as ResearchReportOptions, f as ResearchReportRecommendation, g as gainHistogram, h as SummaryTableRow, i as ParetoFigureSpec, l as ResearchReportDecision, m as SummaryTableOptions, n as GainDistributionFigureSpec, o as RESEARCH_REPORT_HARD_PAIR_FLOOR, p as SummaryTable, r as GainDistributionOptions, s as ResearchReport, t as GainDistributionBin, u as ResearchReportMethodology, v as researchReport, w as HeldOutGateConfig, x as GateDecision, y as summaryTable } from "./summary-report-DuUS_i7W.js";
+import { C as HeldOutGate, E as SplitCoverage, S as GateEvidence, T as HeldOutGateRejectionCode, _ as paretoChart, a as ParetoPoint, b as DeltaStatistic, c as ResearchReportCandidate, d as ResearchReportOptions, f as ResearchReportRecommendation, g as gainHistogram, h as SummaryTableRow, i as ParetoFigureSpec, l as ResearchReportDecision, m as SummaryTableOptions, n as GainDistributionFigureSpec, o as RESEARCH_REPORT_HARD_PAIR_FLOOR, p as SummaryTable, r as GainDistributionOptions, s as ResearchReport, t as GainDistributionBin, u as ResearchReportMethodology, v as researchReport, w as HeldOutGateConfig, x as GateDecision, y as summaryTable } from "./summary-report-BNR7DWTj.js";
 import { a as FailureClassification, c as classifyFailure, i as DEFAULT_RULES, o as FailureContext, s as FailureRule } from "./failure-cluster-CqcvCcdR.js";
-import { C as toOpenAiTool, S as makeEvalTools, _ as TraceAnalysisToolDescriptor, a as TraceAnalystLimits, b as EvalToolDef, d as RawAnalystFinding, g as TRACE_ANALYST_TOOL_NAMESPACE, h as BuildTraceAnalysisToolsOptions, i as TraceAnalysisEngineResult, l as RawAnalystEvidence, n as TraceAnalysisEngine, o as resolveTraceAnalystLimits, r as TraceAnalysisEngineRequest, t as DEFAULT_TRACE_ANALYST_LIMITS, v as buildTraceAnalysisToolDescriptors, x as MakeEvalToolsConfig, y as traceAnalystFunctionGroup } from "./engine-nB64f48I.js";
+import { C as toOpenAiTool, S as makeEvalTools, _ as TraceAnalysisToolDescriptor, a as TraceAnalystLimits, b as EvalToolDef, d as RawAnalystFinding, g as TRACE_ANALYST_TOOL_NAMESPACE, h as BuildTraceAnalysisToolsOptions, i as TraceAnalysisEngineResult, l as RawAnalystEvidence, n as TraceAnalysisEngine, o as resolveTraceAnalystLimits, r as TraceAnalysisEngineRequest, t as DEFAULT_TRACE_ANALYST_LIMITS, v as buildTraceAnalysisToolDescriptors, x as MakeEvalToolsConfig, y as traceAnalystFunctionGroup } from "./engine-3hL-XqwJ.js";
 import { A as isTrainableSplit, D as assertRolloutLine, E as assertMintedLines, T as assertMinted, b as RolloutSplit, h as RolloutLine, i as ChatToolCall, j as validateRolloutLine, k as isRolloutLine, n as ChatMessage, o as MintedRolloutLine, p as RolloutCapture, s as MintedRolloutOutcome, u as ROLLOUT_SCHEMA, w as ToolDef, x as RolloutStep, y as RolloutRole } from "./schema-Cef2cFmb2.js";
 import { C as Unavailable, D as showMeasured, E as isUnavailable, S as SupervisorRunTreeGapCode, _ as SupervisorRunReport, b as SupervisorRunTree, f as SUPERVISOR_RUN_SCHEMA, g as SupervisorRunReader, h as SupervisorRunNodeRole, p as SourceLimits, r as Measured, v as SupervisorRunRollup, x as SupervisorRunTreeGap, y as SupervisorRunSources } from "./types-D4mog56g.js";
-import { B as BenchmarkSource, F as BenchmarkDatasetItem, H as deterministicSplit, I as BenchmarkEvaluation, L as BenchmarkFamily, N as BENCHMARK_SPLIT_SEED, P as BenchmarkAdapter, R as BenchmarkResponder, V as BenchmarkTaskKind, t as index_d_exports, z as BenchmarkScenario } from "./index-C5HOo4ZF2.js";
-import { $ as redTeamDataset, J as RedTeamCase, Q as RedTeamReport, X as RedTeamFinding, Y as RedTeamCategory, Z as RedTeamPayload, an as ReferenceEquivalenceScenario, at as CanaryKind, cn as LlmJudgeDimension, ct as CanarySeverity, en as REFERENCE_EQUIVALENCE_INPUT_LIMITS, et as redTeamReport, in as ReferenceEquivalenceJudgeResult, it as CanaryEvaluation, ln as LlmJudgeOptions, lt as runCanaries, nn as ReferenceEquivalenceJudgeInput, nt as toolNamesForRun, on as createReferenceEquivalenceJudge, ot as CanaryOptions, q as DEFAULT_RED_TEAM_CORPUS, rn as ReferenceEquivalenceJudgeOptions, rt as CanaryAlert, sn as runReferenceEquivalenceJudge, st as CanaryReport, tn as REFERENCE_EQUIVALENCE_JUDGE_VERSION, tt as scoreRedTeamOutput, un as llmJudge } from "./skillopt-optimization-method-BO7NAl3b.js";
-import { $t as ToolCallEventLike, Gt as BackendIntegrityReport, Jt as summarizeAgentReceiptIntegrity, Kt as assertRealAgentReceipts, Qt as RuntimeEventLike, Wt as BackendIntegrityError, Xt as ArtifactEventLike, Yt as summarizeBackendIntegrity, Zt as ProposalEventLike, en as extractProducedState, qt as assertRealBackend } from "./index-Sh2I0DRc.js";
-import { A as PairArmsResult, C as hashJson, D as MatchedPair, E as ComparePairedArmsOptions, F as PairedMetricDelta, I as comparePairedArms, L as pairArms, M as PairedArmRow, N as PairedArmsComparison, O as MatchedRunRecordPair, P as PairedCorrectness, R as pairRunRecords, S as evaluateHypothesis, T as verifyManifest, _ as HypothesisManifest, b as SignedManifestAlgo, j as PairRunRecordsResult, k as PairArmsOptions, v as HypothesisResult, w as signManifest, x as canonicalize, y as SignedManifest } from "./statistical-heldout-Dn9ruizm.js";
-import { _ as paretoFrontier, f as Direction, g as dominates, h as crowdingDistance, m as ParetoResult, p as Objective, v as paretoFrontierWithCrowding, y as scalarScore } from "./promotion-policy-ChWhTDBH.js";
+import { B as BenchmarkSource, F as BenchmarkDatasetItem, H as deterministicSplit, I as BenchmarkEvaluation, L as BenchmarkFamily, N as BENCHMARK_SPLIT_SEED, P as BenchmarkAdapter, R as BenchmarkResponder, V as BenchmarkTaskKind, t as index_d_exports, z as BenchmarkScenario } from "./index-CwYZNzMD2.js";
+import { $ as redTeamDataset, J as RedTeamCase, Q as RedTeamReport, X as RedTeamFinding, Y as RedTeamCategory, Z as RedTeamPayload, an as REFERENCE_EQUIVALENCE_JUDGE_VERSION, at as CanaryKind, cn as ReferenceEquivalenceJudgeResult, ct as CanarySeverity, dn as runReferenceEquivalenceJudge, et as redTeamReport, fn as LlmJudgeDimension, in as REFERENCE_EQUIVALENCE_INPUT_LIMITS, it as CanaryEvaluation, ln as ReferenceEquivalenceScenario, lt as runCanaries, mn as llmJudge, nt as toolNamesForRun, on as ReferenceEquivalenceJudgeInput, ot as CanaryOptions, pn as LlmJudgeOptions, q as DEFAULT_RED_TEAM_CORPUS, rt as CanaryAlert, sn as ReferenceEquivalenceJudgeOptions, st as CanaryReport, tt as scoreRedTeamOutput, un as createReferenceEquivalenceJudge } from "./skillopt-optimization-method-9J8dJbrM.js";
+import { $t as ToolCallEventLike, Gt as BackendIntegrityReport, Jt as summarizeAgentReceiptIntegrity, Kt as assertRealAgentReceipts, Qt as RuntimeEventLike, Wt as BackendIntegrityError, Xt as ArtifactEventLike, Yt as summarizeBackendIntegrity, Zt as ProposalEventLike, en as extractProducedState, qt as assertRealBackend } from "./index-DvvvSho7.js";
+import { A as PairArmsResult, C as hashJson, D as MatchedPair, E as ComparePairedArmsOptions, F as PairedMetricDelta, I as comparePairedArms, L as pairArms, M as PairedArmRow, N as PairedArmsComparison, O as MatchedRunRecordPair, P as PairedCorrectness, R as pairRunRecords, S as evaluateHypothesis, T as verifyManifest, _ as HypothesisManifest, b as SignedManifestAlgo, j as PairRunRecordsResult, k as PairArmsOptions, v as HypothesisResult, w as signManifest, x as canonicalize, y as SignedManifest } from "./statistical-heldout-TQ-4CYiN.js";
+import { _ as paretoFrontier, f as Direction, g as dominates, h as crowdingDistance, m as ParetoResult, p as Objective, v as paretoFrontierWithCrowding, y as scalarScore } from "./promotion-policy-Ckjhzg_4.js";
 import { _ as vitestTestParser, a as DockerSandboxDriver, c as SandboxHarness, d as SubprocessSandboxDriver, f as SubprocessSandboxDriverOptions, g as pytestTestParser, h as jestTestParser, i as runTestGradedScenario, l as SandboxHarnessResult, m as composeParsers, n as TestGradedRunResult, o as HarnessConfig, p as TestOutputParser, r as TestGradedScenario, s as SandboxDriver, t as TestGradedRunOptions, u as SandboxResult } from "./test-graded-scenario-D1TaI2va.js";
-import { $ as ControlRunResult, A as feedbackTrajectoryToOptimizerRow, B as AnalystReviewCounts, C as analystRunToReviewRequests, D as feedbackTrajectoriesToDatasetScenarios, E as createFeedbackTrajectory, F as serializeFeedbackTrajectoriesJsonl, G as analystFindingDigest, H as AnalystReviewQuality, I as summarizePreferenceMemory, J as ControlActionOutcome, K as analystRunDigest, L as withAssignedFeedbackSplit, M as renderPreferenceMemoryMarkdown, N as replayFeedbackTrajectories, O as feedbackTrajectoriesToOptimizerRows, P as replayFeedbackTrajectory, Q as ControlEvalResult, R as AnalystFindingDigest, S as analystRunToFeedbackTrajectory, T as controlRunToFeedbackTrajectory, U as AnalystReviewSource, V as AnalystReviewDecision, W as AnalystRunDigest, X as ControlContext, Y as ControlBudget, Z as ControlDecision, _ as FeedbackTrajectoryStore, a as FeedbackLabel, at as StopDecision, b as PreferenceMemoryEntry, c as FeedbackOptimizerRow, ct as runAgentControlLoop, d as FeedbackReplayResult, dt as subjectiveEval, et as ControlRuntimeConfig, f as FeedbackSeverity, g as FeedbackTrajectoryFilter, h as FeedbackTrajectory, i as FeedbackAttempt, it as ControlStopPolicies, j as parseFeedbackTrajectoriesJsonl, k as feedbackTrajectoryToDatasetScenario, l as FeedbackOutcome, lt as stopOnNoProgress, m as FeedbackTask, n as AnalystReviewRequest, nt as ControlSeverity, o as FeedbackLabelKind, ot as allCriticalPassed, p as FeedbackSplitPolicy, q as ControlActionFailureMode, r as FeedbackArtifactType, rt as ControlStep, s as FeedbackLabelSource, st as objectiveEval, t as AnalystFeedbackTrajectoryOptions, tt as ControlRuntimeError, u as FeedbackReplayAdapter, ut as stopOnRepeatedAction, v as FileSystemFeedbackTrajectoryStore, w as assignFeedbackSplit, x as ProposedSideEffect, y as InMemoryFeedbackTrajectoryStore, z as AnalystMissedIssue } from "./feedback-trajectory-Rh280oXo.js";
-import { A as ActionExecutionPolicy, C as ReviewMemoryStore, D as inMemoryReviewStore, E as createLlmReviewer, M as evaluateActionPolicy, O as jsonlReviewStore, S as ReviewMemoryEntry, T as VerifyFn, _ as ProposeReviewReport, a as ProposeReviewControlAction, b as ReviewFn, c as ProposeReviewControlState, d as LlmJsonCall, f as LlmReviewerConfig, g as ProposeReviewConfig, h as ProposeOutput, i as scoreFromEvals, j as ActionPolicyDecision, k as runProposeReview, l as controlFailureClassFromVerification, m as ProposeInput, n as RunEvidenceMetadata, o as ProposeReviewControlConfig, p as ProposeFn, r as controlRunToRunRecord, s as ProposeReviewControlResult, t as ControlRunToRunRecordOptions, u as runProposeReviewAsControlLoop, v as ProposeReviewShot, w as Verification, x as ReviewInput, y as Review } from "./run-evidence-BDFFai9R.js";
+import { $ as ControlRunResult, A as feedbackTrajectoryToOptimizerRow, B as AnalystReviewCounts, C as analystRunToReviewRequests, D as feedbackTrajectoriesToDatasetScenarios, E as createFeedbackTrajectory, F as serializeFeedbackTrajectoriesJsonl, G as analystFindingDigest, H as AnalystReviewQuality, I as summarizePreferenceMemory, J as ControlActionOutcome, K as analystRunDigest, L as withAssignedFeedbackSplit, M as renderPreferenceMemoryMarkdown, N as replayFeedbackTrajectories, O as feedbackTrajectoriesToOptimizerRows, P as replayFeedbackTrajectory, Q as ControlEvalResult, R as AnalystFindingDigest, S as analystRunToFeedbackTrajectory, T as controlRunToFeedbackTrajectory, U as AnalystReviewSource, V as AnalystReviewDecision, W as AnalystRunDigest, X as ControlContext, Y as ControlBudget, Z as ControlDecision, _ as FeedbackTrajectoryStore, a as FeedbackLabel, at as StopDecision, b as PreferenceMemoryEntry, c as FeedbackOptimizerRow, ct as runAgentControlLoop, d as FeedbackReplayResult, dt as subjectiveEval, et as ControlRuntimeConfig, f as FeedbackSeverity, g as FeedbackTrajectoryFilter, h as FeedbackTrajectory, i as FeedbackAttempt, it as ControlStopPolicies, j as parseFeedbackTrajectoriesJsonl, k as feedbackTrajectoryToDatasetScenario, l as FeedbackOutcome, lt as stopOnNoProgress, m as FeedbackTask, n as AnalystReviewRequest, nt as ControlSeverity, o as FeedbackLabelKind, ot as allCriticalPassed, p as FeedbackSplitPolicy, q as ControlActionFailureMode, r as FeedbackArtifactType, rt as ControlStep, s as FeedbackLabelSource, st as objectiveEval, t as AnalystFeedbackTrajectoryOptions, tt as ControlRuntimeError, u as FeedbackReplayAdapter, ut as stopOnRepeatedAction, v as FileSystemFeedbackTrajectoryStore, w as assignFeedbackSplit, x as ProposedSideEffect, y as InMemoryFeedbackTrajectoryStore, z as AnalystMissedIssue } from "./feedback-trajectory-CacHpxsp.js";
+import { A as ActionExecutionPolicy, C as ReviewMemoryStore, D as inMemoryReviewStore, E as createLlmReviewer, M as evaluateActionPolicy, O as jsonlReviewStore, S as ReviewMemoryEntry, T as VerifyFn, _ as ProposeReviewReport, a as ProposeReviewControlAction, b as ReviewFn, c as ProposeReviewControlState, d as LlmJsonCall, f as LlmReviewerConfig, g as ProposeReviewConfig, h as ProposeOutput, i as scoreFromEvals, j as ActionPolicyDecision, k as runProposeReview, l as controlFailureClassFromVerification, m as ProposeInput, n as RunEvidenceMetadata, o as ProposeReviewControlConfig, p as ProposeFn, r as controlRunToRunRecord, s as ProposeReviewControlResult, t as ControlRunToRunRecordOptions, u as runProposeReviewAsControlLoop, v as ProposeReviewShot, w as Verification, x as ReviewInput, y as Review } from "./run-evidence-Dhi3C81V.js";
 import { C as ClusteredPairedBinaryOptions, E as clusteredPairedBinary, S as ClusteredMatchedPair, T as ClusteredPairedBinaryStatistics, _ as inMemoryExperimentStore, a as ExperimentStats, b as ClusterSignFlipResult, c as ExperimentTrackerOptions, d as ImprovementVerdictResult, f as ProvenanceReader, g as improvementVerdict, h as gitProvenanceReader, i as ExperimentRep, l as ExperimentVerdict, m as fileExperimentStore, n as Experiment, o as ExperimentStore, p as computeExperimentStats, r as ExperimentProvenance, s as ExperimentTracker, t as CreateExperimentInput, u as ImprovementThresholds, v as ClusterBootstrapInterval, w as ClusteredPairedBinaryResult, x as ClusteredBinaryCluster, y as ClusterSignFlipAlternative } from "./experiment-tracker-IMntXr6J.js";
 import { a as PairedEvalueStep, c as pairedEvalueSequence, i as PairedEvalueSequence, n as InterimReleaseConfidenceInput, o as SequentialDecision, r as PairedEvalueOptions, s as evaluateInterimReleaseConfidence, t as InterimReleaseConfidence } from "./sequential-CYwq6Ff_.js";
-import { _ as ReleaseConfidenceStatus, a as JudgeReplayGateArgs, b as assertReleaseConfidence, c as judgeReplayGate, d as ReleaseConfidenceAxis, f as ReleaseConfidenceAxisName, g as ReleaseConfidenceScorecard, h as ReleaseConfidenceMetrics, i as BootstrapResult, l as ActionableSideInfo, m as ReleaseConfidenceIssue, n as renderReleaseReport, o as Verdict, p as ReleaseConfidenceInput, r as BootstrapOptions, s as bootstrapCi, t as RenderReleaseReportOptions, u as AsiSeverity, v as ReleaseConfidenceThresholds, x as evaluateReleaseConfidence, y as ReleaseTraceEvidence } from "./release-report-CI8uisI1.js";
-import { $ as ScoreOrigin, $t as toSftRows, Ct as HarborSubagentTrajectoryRef, Dt as relabelImportedSplit, Et as fromHarborTrajectory, Gt as SftRow, Ht as RewardRow, J as MintRolloutOptions, Ot as toHarborTrajectories, Q as unmintableReasons, St as HarborStepSource, Tt as HarborTrajectory, Wt as SftExportOptions, X as RolloutScrubber, Xt as toRewardRows, Y as MintRolloutResult, Yt as toJsonl, Z as mintRolloutRows, _t as HarborImageSource, at as trainingReward, bt as HarborObservationResult, dt as ATIF_SCHEMA_VERSION, et as ScorePreference, ft as FromHarborOptions, gt as HarborFinalMetrics, ht as HarborContentPart, it as scoreOrigin, kt as toHarborTrajectory, mt as HarborAgent, nt as observedScore, ot as trainingScore, pt as HARBOR_IMPORT_GAP, rt as observedSplitScore, tt as isRealnessGated, vt as HarborMetrics, wt as HarborToolCall, xt as HarborStep, yt as HarborObservation } from "./index-CvXXlyz7.js";
+import { _ as ReleaseConfidenceStatus, a as JudgeReplayGateArgs, b as assertReleaseConfidence, c as judgeReplayGate, d as ReleaseConfidenceAxis, f as ReleaseConfidenceAxisName, g as ReleaseConfidenceScorecard, h as ReleaseConfidenceMetrics, i as BootstrapResult, l as ActionableSideInfo, m as ReleaseConfidenceIssue, n as renderReleaseReport, o as Verdict, p as ReleaseConfidenceInput, r as BootstrapOptions, s as bootstrapCi, t as RenderReleaseReportOptions, u as AsiSeverity, v as ReleaseConfidenceThresholds, x as evaluateReleaseConfidence, y as ReleaseTraceEvidence } from "./release-report-DA2BCu5a.js";
+import { $ as ScoreOrigin, $t as toSftRows, Ct as HarborSubagentTrajectoryRef, Dt as relabelImportedSplit, Et as fromHarborTrajectory, Gt as SftRow, Ht as RewardRow, J as MintRolloutOptions, Ot as toHarborTrajectories, Q as unmintableReasons, St as HarborStepSource, Tt as HarborTrajectory, Wt as SftExportOptions, X as RolloutScrubber, Xt as toRewardRows, Y as MintRolloutResult, Yt as toJsonl, Z as mintRolloutRows, _t as HarborImageSource, at as trainingReward, bt as HarborObservationResult, dt as ATIF_SCHEMA_VERSION, et as ScorePreference, ft as FromHarborOptions, gt as HarborFinalMetrics, ht as HarborContentPart, it as scoreOrigin, kt as toHarborTrajectory, mt as HarborAgent, nt as observedScore, ot as trainingScore, pt as HARBOR_IMPORT_GAP, rt as observedSplitScore, tt as isRealnessGated, vt as HarborMetrics, wt as HarborToolCall, xt as HarborStep, yt as HarborObservation } from "./index-BnEuDAK2.js";
 import { C as SupervisorRunIntegrityIssue, D as SupervisorRunIntegritySeverity, E as SupervisorRunIntegrityReport, F as analyzeSupervisorRunSources, L as rollupSupervisorRuns, N as claudeCodeSupervisorRunReader, O as SupervisorRunSourceOnlyCheckCode, P as readClaudeCodeSupervisorRun, S as SupervisorRunIntegrityEvidence, T as SupervisorRunIntegrityOptions, a as supervisorRunRolloutLines, b as analyzeSupervisorRunIntegrity, c as renderSupervisorRunMarkdown, d as analyzeSupervisorRun, n as readRuntimeSupervisorRun, r as runtimeSupervisorRunReader, s as renderSupervisorRunHeadline, t as isRuntimeSupervisorRunDir, v as writeSupervisorRunReport, w as SupervisorRunIntegrityIssueCode, x as SUPERVISOR_RUN_INTEGRITY_SCHEMA } from "./index-BZ3-y4YL.js";
 import { a as WelchTestResult, c as iqr, d as ToolUseMetrics, f as ToolUseOptions, i as MetricVerdict, l as welchsTTest, n as BaselineReport, o as WelchTestStatus, p as computeToolUseMetrics, r as MetricSamples, s as compareToBaseline, t as BaselineOptions, u as ToolStats } from "./baseline-CavEbRyH.js";
 import { n as TrajectoryStep, r as buildTrajectory, t as Trajectory } from "./trajectory-YC15QDYQ.js";
 import { n as SeriesConvergenceResult, r as analyzeSeries, t as SeriesConvergenceOptions } from "./series-convergence-ofsqPWhs.js";
 import { a as attributeCounterfactuals, i as CounterfactualRunner, n as CounterfactualMutation, o as runCounterfactual, r as CounterfactualResult, t as CounterfactualContext } from "./counterfactual-CxmxAONP.js";
-import { _ as EvalCampaignResult, a as FailureMode, c as SteeringChange, d as CampaignRunContext, f as CampaignRunOutcome, g as EvalCampaignOptions, h as CampaignVariant, i as ExperimentResult, l as CampaignFactoryParams, m as CampaignScenario, n as CallbackResearcherOptions, o as NoopResearcher, p as CampaignRunner, r as ExperimentPlan, s as Researcher, t as CallbackResearcher, u as CampaignIntegrityPolicy, v as FailedRun, y as runEvalCampaign } from "./researcher-BoaxeCzP.js";
+import { _ as EvalCampaignResult, a as FailureMode, c as SteeringChange, d as CampaignRunContext, f as CampaignRunOutcome, g as EvalCampaignOptions, h as CampaignVariant, i as ExperimentResult, l as CampaignFactoryParams, m as CampaignScenario, n as CallbackResearcherOptions, o as NoopResearcher, p as CampaignRunner, r as ExperimentPlan, s as Researcher, t as CallbackResearcher, u as CampaignIntegrityPolicy, v as FailedRun, y as runEvalCampaign } from "./researcher-BchpD55R.js";
 //#region src/auto-pr.d.ts
 /**
  * Automated pull-request transports for the production loop.

package/dist/index.js

@@ -1,5 +1,5 @@
 import { t as __exportAll } from "./rolldown-runtime-8H4AJuhK.js";
-import { _ as observeAll, a as jsonlReviewStore, c as scoreFromEvals, d as runAgentControlLoop, f as stopOnNoProgress, g as noProgressDetector, h as errorStreakDetector, i as inMemoryReviewStore, l as allCriticalPassed, m as subjectiveEval, n as runProposeReviewAsControlLoop, o as runProposeReview, p as stopOnRepeatedAction, r as createLlmReviewer, s as controlRunToRunRecord, t as controlFailureClassFromVerification, u as objectiveEval, v as repeatedActionDetector, y as evaluateActionPolicy } from "./propose-review-control-qgWLA6E9.js";
+import { _ as observeAll, a as jsonlReviewStore, c as scoreFromEvals, d as runAgentControlLoop, f as stopOnNoProgress, g as noProgressDetector, h as errorStreakDetector, i as inMemoryReviewStore, l as allCriticalPassed, m as subjectiveEval, n as runProposeReviewAsControlLoop, o as runProposeReview, p as stopOnRepeatedAction, r as createLlmReviewer, s as controlRunToRunRecord, t as controlFailureClassFromVerification, u as objectiveEval, v as repeatedActionDetector, y as evaluateActionPolicy } from "./propose-review-control-dSNPjFUH.js";
 import { a as LimitExceededError, c as ValidationError, i as JudgeError, l as VerificationError, n as CaptureIntegrityError, o as NotFoundError, r as ConfigError, s as ReplayError, t as AgentEvalError } from "./errors-D-LKuDhb.js";
 import { a as verifyManifest, i as signManifest, n as evaluateHypothesis, r as hashJson, t as canonicalize } from "./pre-registration-DakwTRXk.js";
 import { AGENT_PROFILE_KINDS, AgentProfileCellValidationError, agentProfileCellHashMaterial, agentProfileCellKey, assertRunAgentProfileCell, buildAgentInterfaceProfileCell, buildAgentProfileCell, groupRunsByAgentProfileCell, requireAgentProfileCell, toAgentProfileJson, validateAgentProfileCell, verifyAgentProfileCell } from "./profile-cell.js";
@@ -7,39 +7,39 @@ import { a as estimateTokens, i as estimateCost, n as MetricsCollector, o as isM
 import { a as CostLedgerPersistenceError, c as costForTokenPricing, i as CostLedger, l as costForUsage, n as CostCallConflictError, o as CostReceiptCaptureError, r as CostCeilingReachedError, s as CostReservationExceededError, t as CostAccountingIncompleteError, u as modelPriceKey } from "./cost-ledger-DMFxsLKr.js";
 import { C as servedModelAcceptable, E as judgeFamily, S as normalizeModelId, T as assertCrossFamily, _ as ServedCrossFamilyError, a as assertLlmRoute, b as assertServedModels, c as callLlmJson, d as isTransientLlmError, f as maximumChargeForLlmRequest, g as PROBE_MAX_TOKENS, h as ModelSubstitutionError, i as LlmRouteAssertionError, l as costReceiptFromLlm, m as stripFencedJson, n as LlmClient, o as backoffMs, p as probeLlm, r as LlmResponseError, s as callLlm, t as LlmCallError, u as costReceiptFromLlmError, v as assertCrossFamilyServed, w as CrossFamilyError, x as checkServedModel, y as assertServedModel } from "./llm-client-DzvMUsS_.js";
 import { a as providerFromBaseUrl, i as defaultProviderRedactor, n as InMemoryRawProviderSink, r as NoopRawProviderSink, t as FileSystemRawProviderSink } from "./raw-provider-sink-BQd7mzyT.js";
-import { C as createChatClient, S as computeTraceMetrics, _ as CONTROL_INTEGRITY_ANALYST, a as analystFindingDigest, c as completedAnalystReviewQuality, d as validateAnalystReviewDecisions, f as DEFAULT_TRACE_ANALYST_KINDS, g as FAILURE_MODE_KIND_SPEC, h as IMPROVEMENT_KIND_SPEC, i as assertExactRegistryRunOpts, l as readAnalystReview, m as KNOWLEDGE_GAP_KIND_SPEC, n as AnalystRegistry, o as analystRunDigest, p as KNOWLEDGE_POISONING_KIND_SPEC, r as ExactAnalystRunExecutionError, s as assertUniqueFindingIds, t as buildDefaultAnalystRegistry, u as snapshotAnalystRun, v as ControlIntegrityAnalyst, y as emitControlIntegrityFindings } from "./default-registry-BaQXW1Ow.js";
+import { C as createChatClient, S as computeTraceMetrics, _ as CONTROL_INTEGRITY_ANALYST, a as analystFindingDigest, c as completedAnalystReviewQuality, d as validateAnalystReviewDecisions, f as DEFAULT_TRACE_ANALYST_KINDS, g as FAILURE_MODE_KIND_SPEC, h as IMPROVEMENT_KIND_SPEC, i as assertExactRegistryRunOpts, l as readAnalystReview, m as KNOWLEDGE_GAP_KIND_SPEC, n as AnalystRegistry, o as analystRunDigest, p as KNOWLEDGE_POISONING_KIND_SPEC, r as ExactAnalystRunExecutionError, s as assertUniqueFindingIds, t as buildDefaultAnalystRegistry, u as snapshotAnalystRun, v as ControlIntegrityAnalyst, y as emitControlIntegrityFindings } from "./default-registry-BmktKy8r.js";
 import { INPUT_VALUE, LLM_CACHED_TOKENS, LLM_CACHED_TOKEN_ATTR_KEYS, LLM_CACHE_WRITE_TOKENS, LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, LLM_CONTEXT_TOKENS, LLM_COST_ATTR_KEYS, LLM_COST_USD, LLM_INPUT_TOKENS, LLM_INPUT_TOKEN_ATTR_KEYS, LLM_MODEL_ATTR_KEYS, LLM_MODEL_NAME, LLM_OUTPUT_TOKENS, LLM_OUTPUT_TOKEN_ATTR_KEYS, LLM_REASONING_TOKENS, LLM_REASONING_TOKEN_ATTR_KEYS, OPENINFERENCE_SPAN_KIND, OUTPUT_VALUE, RUN_COST_ATTR_KEYS, SPAN_KIND_ATTR_KEYS, TOOL_ARGS_CAPTURED, TOOL_LATENCY_MS, TOOL_NAME, TOOL_NAME_ATTR_KEYS, applyLlmSpanOtlpAttributes, asNumber, contextInputTokens, firstNumberAttr } from "./trace-attributes.js";
-import { C as TraceNotFoundError, G as resolveTraceAnalystLimits, J as extractOtlpAttributes, K as asString, Q as readOtlpStatus, S as TraceFileTooLargeError, W as DEFAULT_TRACE_ANALYST_LIMITS, X as inferOtlpKind, Y as firstStringAttr, Z as projectOtlpFlatLine, _ as TraceAnalysisLimitError, b as TraceFileMalformedError, c as traceAnalystFunctionGroup, et as stringField, g as SpanNotFoundError, h as TRACE_ANALYSIS_LIMITS, i as runTraceAnalyst, it as traceSpanKindToOpenInferenceKind, l as createBoundedTraceAnalysisStore, m as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, n as renderPriorFindings, nt as classifyOtlpSpanRole, o as TRACE_ANALYST_TOOL_NAMESPACE, p as DEFAULT_TRACE_ANALYST_BUDGETS, r as renderUpstreamFindings, rt as isOtlpModelCall, s as buildTraceAnalysisToolDescriptors, t as createTraceAnalyst, tt as applyToolSpanOtlpAttributes, v as TraceAnalysisStoreContractError, x as TraceFileMissingError, y as TraceAnalysisValidationError } from "./kind-factory-BHIgPmzS.js";
-import { a as computeFindingId, o as makeFinding, s as makeProposalFinding } from "./usage-receipt-EVI8B8Xu.js";
+import { $ as stringField, C as TraceNotFoundError, G as asString, J as firstStringAttr, S as TraceFileTooLargeError, U as DEFAULT_TRACE_ANALYST_LIMITS, W as resolveTraceAnalystLimits, X as projectOtlpFlatLine, Y as inferOtlpKind, Z as readOtlpStatus, _ as TraceAnalysisLimitError, b as TraceFileMalformedError, c as traceAnalystFunctionGroup, et as applyToolSpanOtlpAttributes, g as SpanNotFoundError, h as TRACE_ANALYSIS_LIMITS, i as runTraceAnalyst, l as createBoundedTraceAnalysisStore, m as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, n as renderPriorFindings, nt as isOtlpModelCall, o as TRACE_ANALYST_TOOL_NAMESPACE, p as DEFAULT_TRACE_ANALYST_BUDGETS, q as extractOtlpAttributes, r as renderUpstreamFindings, rt as traceSpanKindToOpenInferenceKind, s as buildTraceAnalysisToolDescriptors, t as createTraceAnalyst, tt as classifyOtlpSpanRole, v as TraceAnalysisStoreContractError, x as TraceFileMissingError, y as TraceAnalysisValidationError } from "./kind-factory-B8-r8-y8.js";
+import { c as makeProposalFinding, o as computeFindingId, s as makeFinding } from "./usage-receipt-t7vAzCRQ.js";
 import { c as SUPERVISOR_RUN_INTEGRITY_SCHEMA, n as supervisorRunRolloutLines, t as analyzeSupervisorRunIntegrity } from "./integrity-DY6tIbl0.js";
 import { c as assertMintedLines, f as isRolloutLine, i as ROLLOUT_SCHEMA, m as validateRolloutLine, p as isTrainableSplit, s as assertMinted, u as assertRolloutLine } from "./schema-C6DW4ZHR.js";
 import { a as scoreOrigin, n as observedScore, o as trainingReward, r as observedSplitScore, s as trainingScore, t as isRealnessGated } from "./reward-nw2xZGZG.js";
 import { a as clamp01, i as aggregateRunScore, r as DEFAULT_RUN_SCORE_WEIGHTS } from "./proposal-findings-2GIUo1et.js";
-import { a as RunCritic, d as defaultIsMaterial, f as diffFindings, h as defineTraceAnalyst, i as runSemanticConceptJudge, n as SEMANTIC_CONCEPT_JUDGE_VERSION, o as SKILL_USAGE_ANALYST, p as LockedJsonlAppender, r as createSemanticConceptJudge, s as SkillUsageAnalyst, t as DEFAULT_COMPLEXITY_WEIGHTS, u as FindingsStore } from "./semantic-concept-judge-Bmrq6yqU.js";
+import { a as RunCritic, d as defaultIsMaterial, f as diffFindings, h as defineTraceAnalyst, i as runSemanticConceptJudge, n as SEMANTIC_CONCEPT_JUDGE_VERSION, o as SKILL_USAGE_ANALYST, p as LockedJsonlAppender, r as createSemanticConceptJudge, s as SkillUsageAnalyst, t as DEFAULT_COMPLEXITY_WEIGHTS, u as FindingsStore } from "./semantic-concept-judge-D1z-KepS.js";
 import { a as inMemoryVerdictCache, i as fileVerdictCache, n as canonicalJson, r as contentHash, t as cachedJudge } from "./verdict-cache-BCcOh0kF.js";
-import { t as createDspyRlmTraceEngine } from "./dspy-rlm-engine-BiN49gK6.js";
+import { t as createDspyRlmTraceEngine } from "./dspy-rlm-engine-DbTk4JdR.js";
 import { f as Mutex, p as mapConcurrent } from "./ledger-core-DXZIqu17.js";
-import { $ as runReferenceEquivalenceJudge, L as DEFAULT_MUTATION_PRIMITIVES, M as surfaceContentHash, Q as createReferenceEquivalenceJudge, R as buildReflectionPrompt, X as REFERENCE_EQUIVALENCE_INPUT_LIMITS, Z as REFERENCE_EQUIVALENCE_JUDGE_VERSION, _t as assertRealBackend, at as Dataset, bt as JudgeParseError, ct as llmJudge, dt as paretoFrontier, et as DEFAULT_RED_TEAM_CORPUS, ft as paretoFrontierWithCrowding, gt as assertRealAgentReceipts, ht as BackendIntegrityError, it as toolNamesForRun, lt as crowdingDistance, nt as redTeamReport, ot as HoldoutLockedError, pt as scalarScore, rt as scoreRedTeamOutput, st as hashScenarios, tt as redTeamDataset, ut as dominates, vt as summarizeAgentReceiptIntegrity, y as runCanaries, yt as summarizeBackendIntegrity, z as parseReflectionResponse } from "./skillopt-optimization-method-CkaI2ly4.js";
+import { $ as REFERENCE_EQUIVALENCE_JUDGE_VERSION, B as parseReflectionResponse, N as surfaceContentHash, Q as REFERENCE_EQUIVALENCE_INPUT_LIMITS, R as DEFAULT_MUTATION_PRIMITIVES, St as JudgeParseError, _t as BackendIntegrityError, at as scoreRedTeamOutput, bt as summarizeAgentReceiptIntegrity, ct as HoldoutLockedError, dt as crowdingDistance, et as createReferenceEquivalenceJudge, ft as dominates, ht as scalarScore, it as redTeamReport, lt as hashScenarios, mt as paretoFrontierWithCrowding, nt as DEFAULT_RED_TEAM_CORPUS, ot as toolNamesForRun, pt as paretoFrontier, rt as redTeamDataset, st as Dataset, tt as runReferenceEquivalenceJudge, ut as llmJudge, vt as assertRealAgentReceipts, xt as summarizeBackendIntegrity, y as runCanaries, yt as assertRealBackend, z as buildReflectionPrompt } from "./skillopt-optimization-method-D6Q4dLbh.js";
 import { $ as selfPreference, A as pairedRiskDifference, B as requiredSampleSize, C as mulberry32, D as pairedCohensDz, E as pairedBootstrap, F as partialCredit, G as wilson, H as weightedComposite, I as passAtK, J as normalCdf, K as studentTCdf, L as pearsonR, M as pairedRiskDifferenceScore, N as pairedSignTest, O as pairedDeltaTieFraction, P as pairedTTest, Q as positionalBias, R as ranks, S as mcnemarRequiredN, T as pairedBinaryScale, U as weightedMean, V as spearmanR, W as wilcoxonSignedRank, X as calibrateJudgeContinuous, Y as calibrateJudge, Z as continuousAgreement, _ as interpretCliffs, a as MANN_WHITNEY_EXACT_MAX_WORK, b as mcnemar, c as bonferroni, d as confidenceInterval, et as verbosityBias, f as corpusInterRaterAgreement, g as interRaterReliability, h as holm, i as MANN_WHITNEY_EXACT_MAX_STATES, j as pairedRiskDifferenceExact, k as pairedMde, l as cliffsDelta, m as eProcess, n as DECISION_PAIRED_DELTA_STATISTIC, o as WILCOXON_EXACT_MAX_N, p as corpusInterRaterAgreementFromJudgeScores, r as DEFAULT_PERMUTATIONS, s as benjaminiHochberg, t as BOOTSTRAP_GATE_MIN_N, u as cohensD, v as isBinaryOutcomeVector, w as normalizeScores, x as mcnemarPower, y as mannWhitneyU, z as requiredPairedSampleSize } from "./statistics-ByxzSiOM.js";
 import { n as pairArms, r as pairRunRecords, t as comparePairedArms } from "./paired-arms-iZ08VFMN.js";
 import { a as improvementVerdict, i as gitProvenanceReader, n as computeExperimentStats, o as inMemoryExperimentStore, r as fileExperimentStore, s as clusteredPairedBinary, t as ExperimentTracker } from "./experiment-tracker-CnRICnMl.js";
 import { n as llmSpanFromProvider, t as TraceEmitter } from "./emitter-CPBAhxum.js";
 import { a as isRetrievalSpan, i as isLlmSpan, n as TRACE_SCHEMA_VERSION, o as isSandboxSpan, r as isJudgeSpan, s as isToolSpan, t as FAILURE_CLASSES } from "./schema-CRhEY1SO.js";
-import { a as roundTripRunRecord, i as parseRunRecordSafe, n as isRunRecord, o as runTaskScore, r as modelHasSnapshot, s as validateRunRecord, t as RunRecordValidationError } from "./run-record-DqOw5X6_.js";
-import { a as evaluateReleaseConfidence, i as assertReleaseConfidence, n as bootstrapCi, r as judgeReplayGate, t as renderReleaseReport } from "./release-report-Dy39cbFF.js";
+import { a as roundTripRunRecord, i as parseRunRecordSafe, n as isRunRecord, o as runTaskScore, r as modelHasSnapshot, s as validateRunRecord, t as RunRecordValidationError } from "./run-record-BmSPWXJR.js";
+import { a as evaluateReleaseConfidence, i as assertReleaseConfidence, n as bootstrapCi, r as judgeReplayGate, t as renderReleaseReport } from "./release-report-BUYmoKo2.js";
 import { d as pairedDecisionShape, f as minimumPairsForPairedDeltaTest, p as pairedDeltaTest, u as decidePairedPromotion } from "./promotion-policy-CrLrmys8.js";
 import { i as toRewardRows, r as toJsonl, s as toSftRows } from "./exporters-q9iL-2Jf.js";
-import { a as toHarborTrajectories, i as relabelImportedSplit, n as HARBOR_IMPORT_GAP, o as toHarborTrajectory, r as fromHarborTrajectory, t as ATIF_SCHEMA_VERSION } from "./rollout-2ECTXb2N.js";
+import { a as toHarborTrajectories, i as relabelImportedSplit, n as HARBOR_IMPORT_GAP, o as toHarborTrajectory, r as fromHarborTrajectory, t as ATIF_SCHEMA_VERSION } from "./rollout-DRrksrcV.js";
 import { t as buildTrajectory } from "./trajectory-D_7rLrvE.js";
-import { n as unmintableReasons, t as mintRolloutRows } from "./mint-CGEkzPLf.js";
+import { n as unmintableReasons, t as mintRolloutRows } from "./mint-B2O60ACG.js";
 import { C as rollupSupervisorRuns, D as isUnavailable, E as SUPERVISOR_RUN_SCHEMA, O as showMeasured, b as readClaudeCodeSupervisorRun, c as writeSupervisorRunReport, d as readRuntimeSupervisorRun, f as runtimeSupervisorRunReader, h as renderSupervisorRunMarkdown, m as renderSupervisorRunHeadline, t as analyzeSupervisorRun, u as isRuntimeSupervisorRunDir, x as analyzeSupervisorRunSources, y as claudeCodeSupervisorRunReader } from "./supervisor-run-D_sokXcO.js";
-import { A as TRACE_ANALYST_ACTOR_DESCRIPTION, C as domainEvidencePattern, D as tokenizeDomainWords, E as scoreTraceInsightReadiness, O as traceAnalystOnRunComplete, S as describeTraceInsightScope, T as planTraceInsightQuestions, _ as otlpToTraceRunRecords, a as convertTraceStoresToOtlp, b as buildTraceInsightPrompt, c as otelRunCompleteHook, d as captureFetchToRawSink, f as ToolTraceMissingError, g as otlpToRunRecords, h as otlpRowsToTraceRunRecords, i as iterateRawCalls, j as TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, k as analyzeTraces, l as OTEL_AGENT_EVAL_SCOPE, m as otlpRowsToRunRecords, n as ReplayCacheMissError, o as createOtelExporter, p as toolSpansToTraceAnalysisStore, r as createReplayFetch, s as createOtelTracingStore, t as ReplayCache, u as exportRunAsOtlp, v as flattenOtlpExportToNdjson, w as inferDomainKeywords, x as defaultTraceInsightPanel, y as buildTraceInsightContext } from "./replay-DyBLaKFc.js";
-import { i as otlpTextToTraceAnalysisStore, n as OtlpFileTraceStore } from "./store-otlp-CKtTpRhv.js";
-import { n as extractUsageFromResponse, r as extractUsageFromSse, t as extractUsage } from "./extract-usage-BW27f3XW.js";
-import { B as harnessAxisOf, F as HARNESS_NATIVE_MODEL, G as parseCorrectnessResponse, H as completionVerdict, I as agentProfileHash, K as verifyCompletion, L as agentProfileId, P as CODING_HARNESSES, R as agentProfileModelId, U as createLlmCorrectnessChecker, V as extractProducedState, W as createTokenRecallChecker, z as expandProfileAxes } from "./campaign--HVSuvV0.js";
+import { A as TRACE_ANALYST_ACTOR_DESCRIPTION, C as domainEvidencePattern, D as tokenizeDomainWords, E as scoreTraceInsightReadiness, O as traceAnalystOnRunComplete, S as describeTraceInsightScope, T as planTraceInsightQuestions, _ as otlpToTraceRunRecords, a as convertTraceStoresToOtlp, b as buildTraceInsightPrompt, c as otelRunCompleteHook, d as captureFetchToRawSink, f as ToolTraceMissingError, g as otlpToRunRecords, h as otlpRowsToTraceRunRecords, i as iterateRawCalls, j as TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, k as analyzeTraces, l as OTEL_AGENT_EVAL_SCOPE, m as otlpRowsToRunRecords, n as ReplayCacheMissError, o as createOtelExporter, p as toolSpansToTraceAnalysisStore, r as createReplayFetch, s as createOtelTracingStore, t as ReplayCache, u as exportRunAsOtlp, v as flattenOtlpExportToNdjson, w as inferDomainKeywords, x as defaultTraceInsightPanel, y as buildTraceInsightContext } from "./replay-CohS93nE.js";
+import { i as otlpTextToTraceAnalysisStore, n as OtlpFileTraceStore } from "./store-otlp-Dw8PPIlL.js";
+import { n as extractUsageFromResponse, r as extractUsageFromSse, t as extractUsage } from "./extract-usage-CdZdoj1s.js";
+import { B as harnessAxisOf, F as HARNESS_NATIVE_MODEL, G as parseCorrectnessResponse, H as completionVerdict, I as agentProfileHash, K as verifyCompletion, L as agentProfileId, P as CODING_HARNESSES, R as agentProfileModelId, U as createLlmCorrectnessChecker, V as extractProducedState, W as createTokenRecallChecker, z as expandProfileAxes } from "./campaign-ccOrVCuR.js";
 import { n as iqr, r as welchsTTest, t as compareToBaseline } from "./baseline-C-GocmIW.js";
 import { a as judgeSpans, c as runsForScenario, i as hasCapturedToolArgs, l as toolSpans, n as argHash, o as llmSpans, r as groupBy, s as runFailureClass, t as aggregateLlm } from "./query-Di7eEQ79.js";
-import { a as checkBehavioralCanary, i as canaryLeakView, o as checkCanaries, r as HoldoutAuditor, s as runBehavioralCanaries, t as analyzeRuns } from "./analyze-runs-BNNK7irB.js";
+import { a as checkBehavioralCanary, i as canaryLeakView, o as checkCanaries, r as HoldoutAuditor, s as runBehavioralCanaries, t as analyzeRuns } from "./analyze-runs-C30yljDJ.js";
 import { a as summaryTable, i as researchReport, n as gainHistogram, r as paretoChart, t as RESEARCH_REPORT_HARD_PAIR_FLOOR } from "./summary-report-Lf-5I7xh.js";
 import { a as composeParsers, c as vitestTestParser, i as SubprocessSandboxDriver, n as DockerSandboxDriver, o as jestTestParser, r as SandboxHarness, s as pytestTestParser, t as runTestGradedScenario } from "./test-graded-scenario-JHcKQNpq.js";
 import { n as InMemoryTraceStore, t as FileSystemTraceStore } from "./store-DNe_Uv1Q.js";
@@ -48,8 +48,8 @@ import { i as redactValue, n as REDACTION_VERSION, r as redactString, t as DEFAU
 import { n as DEFAULT_RULES, r as classifyFailure, t as computeToolUseMetrics } from "./tool-use-metrics-DEGMKycK.js";
 import { t as analyzeSeries } from "./series-convergence-CjO2QdRW.js";
 import { n as runCounterfactual, t as attributeCounterfactuals } from "./counterfactual-CWPTrMH7.js";
-import { _ as deterministicSplit, g as BENCHMARK_SPLIT_SEED, t as benchmarks_exports } from "./benchmarks-BlPmjd88.js";
-import { t as runEvalCampaign } from "./eval-campaign-DNjCvAm-.js";
+import { _ as deterministicSplit, g as BENCHMARK_SPLIT_SEED, t as benchmarks_exports } from "./benchmarks-BzqyTvAr.js";
+import { t as runEvalCampaign } from "./eval-campaign-B_7wcnav.js";
 import { n as pairedEvalueSequence, t as evaluateInterimReleaseConfidence } from "./sequential-Br0mAPHA.js";
 import { accessSync, appendFileSync, constants, cpSync, existsSync, mkdirSync, promises, readFileSync, readdirSync, statSync, writeFileSync } from "node:fs";
 import { basename, delimiter, dirname, extname, isAbsolute, join, relative, resolve } from "node:path";

package/dist/{insight-report-C6h6F_4L.d.ts → insight-report-CgX_s0Ez.d.ts}

@@ -1,7 +1,7 @@
 import { o as FailureClass } from "./schema-BtVldJ3T.js";
-import { l as RunTerminalOutcome } from "./run-record-DdSa93_W.js";
+import { l as RunTerminalOutcome } from "./run-record-CF4Dwpxr.js";
 import { bt as ContinuousAgreement } from "./statistics-D6Uebe_4.js";
-import { i as ParetoFigureSpec, t as GainDistributionBin } from "./summary-report-DuUS_i7W.js";
+import { i as ParetoFigureSpec, t as GainDistributionBin } from "./summary-report-BNR7DWTj.js";
 //#region src/contract/insight-report.d.ts
 interface InsightReport {
   /** Number of runs analyzed. */
@@ -403,4 +403,4 @@ interface Recommendation {
 }
 //#endregion
 export { FailureClusterInsight as a, JudgeInsight as c, Recommendation as d, ReleaseSummary as f, FailureClassTally as i, LiftInsight as l, TokenUsageInsight as m, ExecutionErrorOutcomeCell as n, InsightReport as o, ScalarDistribution as p, ExecutionInsight as r, InterRaterInsight as s, CostProvenanceSummary as t, OutcomeCorrelationInsight as u };
-//# sourceMappingURL=insight-report-C6h6F_4L.d.ts.map
+//# sourceMappingURL=insight-report-CgX_s0Ez.d.ts.map

package/dist/{insight-report-C6h6F_4L.d.ts.map → insight-report-CgX_s0Ez.d.ts.map}

@@ -1 +1 @@
-{"version":3,"file":"insight-report-C6h6F_4L.d.ts","names":[],"sources":["../src/contract/insight-report.ts"],"mappings":";;;;;UAqCiB;;EAEf;;;;EAKA,WAAW;;EAGX,WAAW;;;EAIX,cAAc,eAAe;;EAG7B;IACE,MAAM;IACN,QAAQ;;;;IAIR,aAAa;;;;;IAKb;MAAa;MAAe;;;;;;EAM9B,QAAQ,eAAe;;;;EAKvB,aAAa;;;;EAKb,OAAO;;;EAIP,kBAAkB;;;EAIlB,gBAAgB;;;;;EAMhB,qBAAqB;;;;;EAMrB,SAAS;;;;;;EAOT,wBAAwB;;;;EAKxB,iBAAiB;;;EAIjB,iBAAiB;;UAGF;EACf;IAAY;IAAW;;EACvB;IAAa;IAAW;;EACxB;IAAc;;EACd;;UAGe;;EAEf,YAAY;;EAEZ,SAAS;;;EAGT,YAAY;;;EAGZ;IACE;IACA,YAAY;IACZ,SAAS;IACT;;;EAGF,QAAQ;IAAQ;IAAe;;;;;EAI/B;IACE;IACA;IACA;;;;;EAKF;IACE;;;IAGA;;IAEA;;IAEA;;IAEA;;IAEA;;;;;IAKA,mBAAmB,OAAO,oBAAoB;;;;EAIhD;IACE;IACA;IACA;IACA;IACA;;;UAIa;;EAEf;;EAEA;;EAEA;;UAGe;EACf,OAAO;EACP,QAAQ;EACR,WAAW;EACX,QAAQ;EACR,YAAY;EACZ;IACE;IACA;IACA;IACA;IACA;;;;UAOa;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,WAAW;;;;;EAKX,WAAW;IAAQ;IAAe;;;UAGnB;;EAEf;;EAEA;;;EAGA,cAAc;;;EAGd;;;EAGA;;;EAGA;;UAGe;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,SAAS;;EAET,mBAAmB;IACjB;IACA,SAAS;MAAQ;MAAe;;IAChC;;;UAIa;EACf;EACA;;EAEA;;EAEA;;;EAGA;;EAEA;;EAEA;;;;;;;;;;;EAWA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;UAGe;;EAEf,UAAU;IACR;IACA;;IAEA;;IAEA;;IAEA;;EAEF;;;;;UAMe;;EAEf,cAAc;;EAEd;;EAEA;;UAGe;;EAEf;;;EAGA;EACA,UAAU;IAAQ;IAAe;IAAgB;;;UAGlC;;;EAGf;;EAEA;;EAEA;;EAEA;;EAEA;IACE;IACA;IACA;;;UAIa;;;EAGf;EACA,MAAM;IACJ;IACA;IACA;;;;EAIF;;UAGQ;;EAER;;EAEA;;;;EAIA;;EAEA;EACA;;KAGU,eACP;EACC;;EAEA;;EAEA;;EAEA;;EAEA;MAED;EACC;;EAEA;EACA;EACA;EACA;;UAGW;;EAEf;EACA;;EAEA;;;;EAIA,SAAS,eAAe;;;EAGxB;;EAEA;;EAEA;;UAGe;EACf;EACA;EACA;EACA;;EAEA"}
+{"version":3,"file":"insight-report-CgX_s0Ez.d.ts","names":[],"sources":["../src/contract/insight-report.ts"],"mappings":";;;;;UAqCiB;;EAEf;;;;EAKA,WAAW;;EAGX,WAAW;;;EAIX,cAAc,eAAe;;EAG7B;IACE,MAAM;IACN,QAAQ;;;;IAIR,aAAa;;;;;IAKb;MAAa;MAAe;;;;;;EAM9B,QAAQ,eAAe;;;;EAKvB,aAAa;;;;EAKb,OAAO;;;EAIP,kBAAkB;;;EAIlB,gBAAgB;;;;;EAMhB,qBAAqB;;;;;EAMrB,SAAS;;;;;;EAOT,wBAAwB;;;;EAKxB,iBAAiB;;;EAIjB,iBAAiB;;UAGF;EACf;IAAY;IAAW;;EACvB;IAAa;IAAW;;EACxB;IAAc;;EACd;;UAGe;;EAEf,YAAY;;EAEZ,SAAS;;;EAGT,YAAY;;;EAGZ;IACE;IACA,YAAY;IACZ,SAAS;IACT;;;EAGF,QAAQ;IAAQ;IAAe;;;;;EAI/B;IACE;IACA;IACA;;;;;EAKF;IACE;;;IAGA;;IAEA;;IAEA;;IAEA;;IAEA;;;;;IAKA,mBAAmB,OAAO,oBAAoB;;;;EAIhD;IACE;IACA;IACA;IACA;IACA;;;UAIa;;EAEf;;EAEA;;EAEA;;UAGe;EACf,OAAO;EACP,QAAQ;EACR,WAAW;EACX,QAAQ;EACR,YAAY;EACZ;IACE;IACA;IACA;IACA;IACA;;;;UAOa;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,WAAW;;;;;EAKX,WAAW;IAAQ;IAAe;;;UAGnB;;EAEf;;EAEA;;;EAGA,cAAc;;;EAGd;;;EAGA;;;EAGA;;UAGe;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,SAAS;;EAET,mBAAmB;IACjB;IACA,SAAS;MAAQ;MAAe;;IAChC;;;UAIa;EACf;EACA;;EAEA;;EAEA;;;EAGA;;EAEA;;EAEA;;;;;;;;;;;EAWA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;UAGe;;EAEf,UAAU;IACR;IACA;;IAEA;;IAEA;;IAEA;;EAEF;;;;;UAMe;;EAEf,cAAc;;EAEd;;EAEA;;UAGe;;EAEf;;;EAGA;EACA,UAAU;IAAQ;IAAe;IAAgB;;;UAGlC;;;EAGf;;EAEA;;EAEA;;EAEA;;EAEA;IACE;IACA;IACA;;;UAIa;;;EAGf;EACA,MAAM;IACJ;IACA;IACA;;;;EAIF;;UAGQ;;EAER;;EAEA;;;;EAIA;;EAEA;EACA;;KAGU,eACP;EACC;;EAEA;;EAEA;;EAEA;;EAEA;MAED;EACC;;EAEA;EACA;EACA;EACA;;UAGW;;EAEf;EACA;;EAEA;;;;EAIA,SAAS,eAAe;;;EAGxB;;EAEA;;EAEA;;UAGe;EACf;EACA;EACA;EACA;;EAEA"}

package/dist/{kind-factory-BHIgPmzS.js → kind-factory-B8-r8-y8.js}

@@ -1,7 +1,7 @@
 import { a as LimitExceededError, c as ValidationError, n as CaptureIntegrityError, o as NotFoundError, t as AgentEvalError } from "./errors-D-LKuDhb.js";
 import { i as CostLedger } from "./cost-ledger-DMFxsLKr.js";
 import { INPUT_VALUE, LLM_MODEL_ATTR_KEYS, OUTPUT_VALUE, SPAN_KIND_ATTR_KEYS, TOOL_ARGS_CAPTURED, TOOL_LATENCY_MS, TOOL_NAME, TOOL_NAME_ATTR_KEYS } from "./trace-attributes.js";
-import { i as validateUsageSettlementTimeout, n as settleUsageReceiptFromCostLedger, o as makeFinding } from "./usage-receipt-EVI8B8Xu.js";
+import { a as deepFreezeCanonicalJson, i as validateUsageSettlementTimeout, n as settleUsageReceiptFromCostLedger, s as makeFinding } from "./usage-receipt-t7vAzCRQ.js";
 import { i as hashCanonical, r as canonicalString } from "./canonical-D011XM8r.js";
 import { z } from "zod";
 import { RE2JS } from "re2js";
@@ -247,19 +247,6 @@ function resolveTraceAnalystLimits(limits) {
 	return resolved;
 }
 //#endregion
-//#region src/ledger-core/deep-freeze.ts
-/** Freeze a detached canonical-JSON graph. Canonicalization has already ruled out cycles.
-*
-* Lives outside canonical.ts so the analyst-benchmark implementation digest,
-* which covers canonical.ts, stays bound to the published benchmark evidence. */
-function deepFreezeCanonicalJson(value) {
-	if (value && typeof value === "object" && !Object.isFrozen(value)) {
-		Object.freeze(value);
-		for (const nested of Object.values(value)) deepFreezeCanonicalJson(nested);
-	}
-	return value;
-}
-//#endregion
 //#region src/analyst/exact-types.ts
 /** Canonical identity for any live component admitted to an exact run. */
 function snapshotExactExecutionComponentIdentity(value, context) {
@@ -2138,6 +2125,6 @@ function truncateForContext(value, max) {
 	return `${value.slice(0, max - 3).trimEnd()}...`;
 }
 //#endregion
-export { spanEpochMillis as $, coerceJson as A, renderFindingSubject as B, TraceNotFoundError as C, RawAnalystFindingSchema as D, RawAnalystEvidenceSchema as E, FINDING_SUBJECT_SYNTAX as F, resolveTraceAnalystLimits as G, snapshotExactExecutionPlan as H, FindingSubjectStringSchema as I, extractOtlpAttributes as J, asString as K, KIND_EXPECTED_SUBJECTS as L, stripCodeFences as M, FINDING_SUBJECT_GRAMMAR_PROMPT as N, evidenceRefsFromRawFinding as O, FINDING_SUBJECT_KINDS as P, readOtlpStatus as Q, findingSubjectGrammarPromptFor as R, TraceFileTooLargeError as S, RAW_FINDING_SCHEMA_PROMPT as T, deepFreezeCanonicalJson as U, snapshotExactExecutionComponentIdentity as V, DEFAULT_TRACE_ANALYST_LIMITS as W, inferOtlpKind as X, firstStringAttr as Y, projectOtlpFlatLine as Z, TraceAnalysisLimitError as _, buildTraceToolsForGroup as a, TraceFileMalformedError as b, traceAnalystFunctionGroup as c, truncateForBudget as d, stringField as et, validateInteger as f, SpanNotFoundError as g, TRACE_ANALYSIS_LIMITS as h, runTraceAnalyst as i, traceSpanKindToOpenInferenceKind as it, coerceToFindingRows as j, parseRawFinding as k, createBoundedTraceAnalysisStore as l, TRACE_ANALYST_TRUNCATION_MARKER_PREFIX as m, renderPriorFindings as n, classifyOtlpSpanRole as nt, TRACE_ANALYST_TOOL_NAMESPACE as o, DEFAULT_TRACE_ANALYST_BUDGETS as p, compareSpanTime as q, renderUpstreamFindings as r, isOtlpModelCall as rt, buildTraceAnalysisToolDescriptors as s, createTraceAnalyst as t, applyToolSpanOtlpAttributes as tt, compileSearchRegex as u, TraceAnalysisStoreContractError as v, ANALYST_SEVERITIES as w, TraceFileMissingError as x, TraceAnalysisValidationError as y, parseFindingSubject as z };
+export { stringField as $, coerceJson as A, renderFindingSubject as B, TraceNotFoundError as C, RawAnalystFindingSchema as D, RawAnalystEvidenceSchema as E, FINDING_SUBJECT_SYNTAX as F, asString as G, snapshotExactExecutionPlan as H, FindingSubjectStringSchema as I, firstStringAttr as J, compareSpanTime as K, KIND_EXPECTED_SUBJECTS as L, stripCodeFences as M, FINDING_SUBJECT_GRAMMAR_PROMPT as N, evidenceRefsFromRawFinding as O, FINDING_SUBJECT_KINDS as P, spanEpochMillis as Q, findingSubjectGrammarPromptFor as R, TraceFileTooLargeError as S, RAW_FINDING_SCHEMA_PROMPT as T, DEFAULT_TRACE_ANALYST_LIMITS as U, snapshotExactExecutionComponentIdentity as V, resolveTraceAnalystLimits as W, projectOtlpFlatLine as X, inferOtlpKind as Y, readOtlpStatus as Z, TraceAnalysisLimitError as _, buildTraceToolsForGroup as a, TraceFileMalformedError as b, traceAnalystFunctionGroup as c, truncateForBudget as d, applyToolSpanOtlpAttributes as et, validateInteger as f, SpanNotFoundError as g, TRACE_ANALYSIS_LIMITS as h, runTraceAnalyst as i, coerceToFindingRows as j, parseRawFinding as k, createBoundedTraceAnalysisStore as l, TRACE_ANALYST_TRUNCATION_MARKER_PREFIX as m, renderPriorFindings as n, isOtlpModelCall as nt, TRACE_ANALYST_TOOL_NAMESPACE as o, DEFAULT_TRACE_ANALYST_BUDGETS as p, extractOtlpAttributes as q, renderUpstreamFindings as r, traceSpanKindToOpenInferenceKind as rt, buildTraceAnalysisToolDescriptors as s, createTraceAnalyst as t, classifyOtlpSpanRole as tt, compileSearchRegex as u, TraceAnalysisStoreContractError as v, ANALYST_SEVERITIES as w, TraceFileMissingError as x, TraceAnalysisValidationError as y, parseFindingSubject as z };
 
-//# sourceMappingURL=kind-factory-BHIgPmzS.js.map
+//# sourceMappingURL=kind-factory-B8-r8-y8.js.map