@tangle-network/agent-eval 0.112.0 → 0.114.0

package/dist/index.d.ts

@@ -16,24 +16,24 @@ import { F as FailureClass, T as ToolSpan, h as BudgetSpec, B as BudgetLedgerEnt
 export { A as Artifact, E as EventKind, i as FAILURE_CLASSES, G as GenericSpan, J as JudgeSpan, M as Message, c as RetrievalSpan, g as RunLayer, f as RunStatus, d as SandboxSpan, j as SpanBase, b as SpanKind, k as SpanStatus, l as TRACE_SCHEMA_VERSION, e as TraceEvent, m as isJudgeSpan, n as isLlmSpan, o as isRetrievalSpan, p as isSandboxSpan, q as isToolSpan } from './schema-SGWcK9wa.js';
 import { A as AgentEvalError, J as JudgeError, a as ConfigError } from './errors-oeQrLqXC.js';
 export { b as AgentEvalErrorCode, C as CaptureIntegrityError, N as NotFoundError, R as ReplayError, V as ValidationError, c as VerificationError } from './errors-oeQrLqXC.js';
-import { b as CorrectnessChecker } from './pre-registration-BepVVa6P.js';
-export { A as ArtifactCheckArtifact, d as ArtifactEventLike, e as ArtifactValidator, f as BackendIntegrityError, B as BackendIntegrityReport, C as CompletionRequirement, a as CompletionVerdict, H as HypothesisManifest, g as HypothesisResult, h as LlmCorrectnessCheckerOpts, L as LlmJudgeDimension, c as LlmJudgeOptions, i as ProducedProposal, P as ProducedState, j as ProposalEventLike, k as RequirementCheck, R as RuntimeEventLike, m as SatisfiedBy, S as SignedManifest, n as SignedManifestAlgo, T as TaskGold, o as ToolCallEventLike, V as ValidationContext, p as ValidationIssue, q as ValidationResult, r as assertRealBackend, s as byteLengthRange, t as canonicalize, u as completionVerdict, v as composeValidators, w as containsAll, x as createLlmCorrectnessChecker, y as createTokenRecallChecker, z as evaluateHypothesis, D as extractProducedState, E as hashJson, F as jsonHasKeys, l as llmJudge, G as parseCorrectnessResponse, I as regexMatch, J as signManifest, K as summarizeBackendIntegrity, M as verifyCompletion, N as verifyManifest } from './pre-registration-BepVVa6P.js';
+import { b as CorrectnessChecker } from './pre-registration-CTQbZbpX.js';
+export { A as ArtifactCheckArtifact, d as ArtifactEventLike, e as ArtifactValidator, f as BackendIntegrityError, B as BackendIntegrityReport, C as CompletionRequirement, a as CompletionVerdict, H as HypothesisManifest, g as HypothesisResult, h as LlmCorrectnessCheckerOpts, L as LlmJudgeDimension, c as LlmJudgeOptions, i as ProducedProposal, P as ProducedState, j as ProposalEventLike, k as RequirementCheck, R as RuntimeEventLike, m as SatisfiedBy, S as SignedManifest, n as SignedManifestAlgo, T as TaskGold, o as ToolCallEventLike, V as ValidationContext, p as ValidationIssue, q as ValidationResult, r as assertRealBackend, s as byteLengthRange, t as canonicalize, u as completionVerdict, v as composeValidators, w as containsAll, x as createLlmCorrectnessChecker, y as createTokenRecallChecker, z as evaluateHypothesis, D as extractProducedState, E as hashJson, F as jsonHasKeys, l as llmJudge, G as parseCorrectnessResponse, I as regexMatch, J as signManifest, K as summarizeBackendIntegrity, M as verifyCompletion, N as verifyManifest } from './pre-registration-CTQbZbpX.js';
 import { T as TraceEmitter } from './emitter-BRchAAAx.js';
 export { R as RunCompleteHook, a as RunCompleteHookContext, S as SpanHandle, b as TraceEmitterOptions, l as llmSpanFromProvider } from './emitter-BRchAAAx.js';
 import { h as ReleaseConfidenceThresholds, f as ReleaseConfidenceScorecard } from './release-report-oBfOz8ku.js';
 export { A as ActionableSideInfo, o as AsiSeverity, B as BootstrapOptions, a as BootstrapResult, J as JudgeReplayGateArgs, R as ReleaseConfidenceAxis, b as ReleaseConfidenceAxisName, c as ReleaseConfidenceInput, d as ReleaseConfidenceIssue, e as ReleaseConfidenceMetrics, g as ReleaseConfidenceStatus, i as ReleaseTraceEvidence, j as RenderReleaseReportOptions, V as Verdict, k as assertReleaseConfidence, l as bootstrapCi, m as evaluateReleaseConfidence, n as judgeReplayGate, r as renderReleaseReport } from './release-report-oBfOz8ku.js';
-import { P as PairedBootstrapOptions, M as McNemarResult, R as RiskDifferenceResult, a as PairedBootstrapResult } from './statistics-CDfpwIdX.js';
-export { c as CliffsMagnitude, d as CorpusAgreementOptions, e as CorpusAgreementPerDimension, C as CorpusAgreementReport, f as CorpusScoreRecord, g as EProcess, h as EProcessOptions, E as EProcessState, i as EProcessStep, j as ProportionInterval, W as WeightedCompositeInput, k as WeightedCompositeResult, b as benjaminiHochberg, l as bonferroni, m as cliffsDelta, n as cohensD, o as confidenceInterval, q as corpusInterRaterAgreement, r as corpusInterRaterAgreementFromJudgeScores, s as eProcess, t as holm, u as interRaterReliability, v as interpretCliffs, x as mannWhitneyU, y as mcnemar, z as mcnemarPower, A as mcnemarRequiredN, B as mulberry32, D as normalizeScores, p as pairedBootstrap, F as pairedMde, G as pairedRiskDifference, H as pairedTTest, I as partialCredit, J as passAtK, K as pearsonR, L as ranks, N as requiredSampleSize, O as spearmanR, Q as weightedComposite, S as weightedMean, w as wilcoxonSignedRank, T as wilson } from './statistics-CDfpwIdX.js';
+import { P as PairedBootstrapOptions, M as McNemarResult, R as RiskDifferenceResult, a as PairedBootstrapResult } from './statistics-oUbOJe-S.js';
+export { c as CliffsMagnitude, d as CorpusAgreementOptions, e as CorpusAgreementPerDimension, C as CorpusAgreementReport, f as CorpusScoreRecord, g as EProcess, h as EProcessOptions, E as EProcessState, i as EProcessStep, j as PairedSignTestResult, k as ProportionInterval, S as SignTestAlternative, W as WeightedCompositeInput, l as WeightedCompositeResult, b as benjaminiHochberg, m as bonferroni, n as cliffsDelta, o as cohensD, q as confidenceInterval, r as corpusInterRaterAgreement, s as corpusInterRaterAgreementFromJudgeScores, t as eProcess, u as holm, v as interRaterReliability, x as interpretCliffs, y as mannWhitneyU, z as mcnemar, A as mcnemarPower, B as mcnemarRequiredN, D as mulberry32, F as normalizeScores, p as pairedBootstrap, G as pairedMde, H as pairedRiskDifference, I as pairedSignTest, J as pairedTTest, K as partialCredit, L as passAtK, N as pearsonR, O as ranks, Q as requiredSampleSize, T as spearmanR, U as weightedComposite, V as weightedMean, w as wilcoxonSignedRank, X as wilson } from './statistics-oUbOJe-S.js';
 import { OtelExporter, OtelExportConfig } from './traces.js';
 export { CaptureFetchContext, CaptureFetchOptions, DEFAULT_REDACTION_RULES, ExportableSpan, ExtractedUsage, FlattenOtlpOptions, LLM_CACHED_TOKENS, LLM_CACHED_TOKEN_ATTR_KEYS, LLM_COST_ATTR_KEYS, LLM_COST_USD, LLM_INPUT_TOKENS, LLM_INPUT_TOKEN_ATTR_KEYS, LLM_MODEL_ATTR_KEYS, LLM_MODEL_NAME, LLM_OUTPUT_TOKENS, LLM_OUTPUT_TOKEN_ATTR_KEYS, OPENINFERENCE_SPAN_KIND, OTEL_AGENT_EVAL_SCOPE, OtlpExport, OtlpFileTraceStore, OtlpFileTraceStoreOptions, OtlpFlatLine, OtlpResourceSpans, OtlpSpan, OtlpToRunRecordsOptions, OtlpTraceRunRecord, ProjectedOtlpSpan, REDACTION_VERSION, RedactionReport, RedactionRule, ReplayCache, ReplayCacheEntry, ReplayCacheMissError, ReplayCacheStats, ReplayFetchOptions, SPAN_KIND_ATTR_KEYS, SpanNotFoundError, TOOL_NAME, TOOL_NAME_ATTR_KEYS, TRACE_ANALYST_ACTOR_DESCRIPTION, TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, TRACE_ANALYST_SUBAGENT_DESCRIPTION, TraceAggregate, TraceAnalystHookOptions, TraceFileMissingError, TraceInsightContext, TraceInsightFinding, TraceInsightPanelRole, TraceInsightPromptInput, TraceInsightQualityGate, TraceInsightQuestion, TraceInsightReadiness, TraceInsightSuite, TraceInsightTask, TraceNotFoundError, TraceStoreSource, TraceStoreToOtlpOptions, TracesToOtlpResult, asNumber, asString, buildTraceAnalystTools, buildTraceInsightContext, buildTraceInsightPrompt, captureFetchToRawSink, convertTraceStoresToOtlp, createOtelExporter, createOtelTracingStore, createReplayFetch, defaultTraceInsightPanel, describeTraceInsightScope, domainEvidencePattern, exportRunAsOtlp, extractOtlpAttributes, extractUsage, extractUsageFromResponse, extractUsageFromSse, firstNumberAttr, firstStringAttr, flattenOtlpExportToNdjson, inferDomainKeywords, inferOtlpKind, iterateRawCalls, otelRunCompleteHook, otlpToRunRecords, otlpToTraceRunRecords, planTraceInsightQuestions, projectOtlpFlatLine, readOtlpStatus, redactString, redactValue, scoreTraceInsightReadiness, stringField, tokenizeDomainWords, traceAnalystFunctionGroup, traceAnalystOnRunComplete, traceSpanKindToOpenInferenceKind } from './traces.js';
 import { a as AnalyzeTracesInput, A as AnalyzeTracesOptions, b as AnalyzeTracesResult } from './analyst-C8HHvfJp.js';
 export { c as AnalyzeTracesTurnSnapshot, d as analyzeTraces } from './analyst-C8HHvfJp.js';
 import { a as TraceAnalystSpan } from './store-C1YxJDEK.js';
 export { D as DEFAULT_TRACE_ANALYST_BUDGETS, b as DatasetOverview, E as ErrorCluster, Q as QueryTracesPage, S as SearchSpanResult, c as SearchTraceResult, d as SpanMatchRecord, e as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, T as TraceAnalysisStore, f as TraceAnalystByteBudgets, g as TraceAnalystFilters, h as TraceAnalystSpanKind, i as TraceAnalystSpanStatus, j as TraceAnalystTraceSummary, V as ViewSpansResult, k as ViewTraceOversized, l as ViewTraceResult } from './store-C1YxJDEK.js';
-import { b as JudgeConfig, J as JudgeScore, S as Scenario$1, g as Gate } from './types-v--ctu-b.js';
+import { b as JudgeConfig, J as JudgeScore, S as Scenario$1, g as Gate } from './types-CgSlO6wT.js';
 import { A as AnalyzeRunsOptions } from './analyze-runs-Dmz6LA9e.js';
-import { q as Objective, t as ParetoResult, h as GepaProposerConstraints, a as RunImprovementLoopResult } from './gepa-T8T215nw.js';
-export { u as DEFAULT_RED_TEAM_CORPUS, D as Direction, e as RedTeamCase, v as RedTeamCategory, w as RedTeamFinding, x as RedTeamPayload, y as RedTeamReport, z as crowdingDistance, A as dominates, B as paretoFrontier, E as paretoFrontierWithCrowding, F as redTeamDataset, H as redTeamReport, I as scalarScore, J as scoreRedTeamOutput, K as toolNamesForRun } from './gepa-T8T215nw.js';
+import { q as Objective, s as ParetoResult, h as GepaProposerConstraints, a as RunImprovementLoopResult } from './gepa-DolL_Fko.js';
+export { t as DEFAULT_RED_TEAM_CORPUS, D as Direction, e as RedTeamCase, u as RedTeamCategory, v as RedTeamFinding, w as RedTeamPayload, x as RedTeamReport, y as crowdingDistance, z as dominates, A as paretoFrontier, B as paretoFrontierWithCrowding, E as redTeamDataset, F as redTeamReport, H as scalarScore, I as scoreRedTeamOutput, J as toolNamesForRun } from './gepa-DolL_Fko.js';
 import { S as SandboxDriver, H as HarnessConfig, a as SandboxHarnessResult } from './test-graded-scenario-mzYBKspu.js';
 export { D as DockerSandboxDriver, c as SandboxHarness, d as SandboxResult, e as SubprocessSandboxDriver, f as SubprocessSandboxDriverOptions, g as TestGradedRunOptions, b as TestGradedRunResult, T as TestGradedScenario, h as TestOutputParser, i as composeParsers, j as jestTestParser, p as pytestTestParser, r as runTestGradedScenario, v as vitestTestParser } from './test-graded-scenario-mzYBKspu.js';
 export { b as RunIntegrityError, R as RunIntegrityExpectations, c as RunIntegrityIssue, d as RunIntegrityIssueCode, a as RunIntegrityReport, e as assertRunCaptured, t as throwIfRunIncomplete } from './integrity-qemeBAyx.js';
@@ -58,7 +58,7 @@ import { L as Layer, S as Severity, b as LayerResult, c as VerifyContext } from
 export { F as Finding, d as LayerStatus, M as MultiLayerVerifier, a as VerificationReport, V as VerifyOptions, g as gradeSemanticStatus } from './multi-layer-verifier-BsqKuLyN.js';
 import { L as LlmClientOptions } from './llm-client-DyqEH4jH.js';
 export { d as LlmCallError, b as LlmCallRequest, c as LlmCallResult, e as LlmClient, f as LlmMessage, g as LlmRouteAssertionError, a as LlmRouteRequirements, h as LlmUsage, i as assertLlmRoute, j as backoffMs, k as callLlm, l as callLlmJson, m as isTransientLlmError, p as probeLlm, s as stripFencedJson } from './llm-client-DyqEH4jH.js';
-export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, c as BenchmarkEvaluation, d as BenchmarkFamily, e as BenchmarkResponder, f as BenchmarkScenario, g as BenchmarkSource, h as BenchmarkTaskKind, i as benchmarkDeterministicSplit, j as benchmarks } from './index-Dc3VLGhp.js';
+export { B as BENCHMARK_SPLIT_SEED, a as BenchmarkAdapter, b as BenchmarkDatasetItem, c as BenchmarkEvaluation, d as BenchmarkFamily, e as BenchmarkResponder, f as BenchmarkScenario, g as BenchmarkSource, h as BenchmarkTaskKind, i as benchmarkDeterministicSplit, j as benchmarks } from './index-CWr5SIG-.js';
 export { C as CallbackResearcher, d as CallbackResearcherOptions, e as CampaignFactoryParams, f as CampaignIntegrityPolicy, g as CampaignRunContext, h as CampaignRunOutcome, i as CampaignRunner, j as CampaignScenario, k as CampaignVariant, c as EvalCampaignOptions, b as EvalCampaignResult, E as ExperimentPlan, a as ExperimentResult, l as FailedRun, F as FailureMode, N as NoopResearcher, R as Researcher, S as SteeringChange, r as runEvalCampaign } from './researcher-CaH0CwFC.js';
 export { G as GainDistributionBin, a as GainDistributionFigureSpec, b as GainDistributionOptions, m as GateDecision, n as GateEvidence, H as HeldOutGate, o as HeldOutGateConfig, q as HeldOutGateRejectionCode, P as ParetoFigureSpec, c as ParetoPoint, R as RESEARCH_REPORT_HARD_PAIR_FLOOR, d as ResearchReport, e as ResearchReportCandidate, f as ResearchReportDecision, g as ResearchReportMethodology, h as ResearchReportOptions, i as ResearchReportRecommendation, S as SummaryTable, j as SummaryTableOptions, k as SummaryTableRow, l as gainHistogram, p as paretoChart, r as researchReport, s as summaryTable } from './summary-report-Bz-0-t8v.js';
 export { I as InterimReleaseConfidence, a as InterimReleaseConfidenceInput, P as PairedEvalueOptions, b as PairedEvalueSequence, c as PairedEvalueStep, S as SequentialDecision, e as evaluateInterimReleaseConfidence, p as pairedEvalueSequence } from './sequential-5iSVfzl2.js';

package/dist/index.js

@@ -9,12 +9,12 @@ import {
   checkBehavioralCanary,
   checkCanaries,
   runBehavioralCanaries
-} from "./chunk-BOLYPSAE.js";
+} from "./chunk-WBOGKYM4.js";
 import {
   BENCHMARK_SPLIT_SEED,
   benchmarks_exports,
   deterministicSplit
-} from "./chunk-VLNGPUJJ.js";
+} from "./chunk-3LXTCTWL.js";
 import {
   DEFAULT_RULES,
   buildTrajectory,
@@ -23,7 +23,7 @@ import {
   computeToolUseMetrics,
   iqr,
   welchsTTest
-} from "./chunk-YWGKQARF.js";
+} from "./chunk-NYFUT3B3.js";
 import {
   analyzeSeries
 } from "./chunk-BOD4O7OF.js";
@@ -59,7 +59,7 @@ import {
   llmJudge,
   parseCorrectnessResponse,
   verifyCompletion
-} from "./chunk-QQVYMBCZ.js";
+} from "./chunk-VHPD6AXX.js";
 import {
   DEFAULT_MUTATION_PRIMITIVES,
   DEFAULT_RED_TEAM_CORPUS,
@@ -83,7 +83,7 @@ import {
   scoreRedTeamOutput,
   surfaceContentHash,
   toolNamesForRun
-} from "./chunk-RCXMWGRY.js";
+} from "./chunk-N6MTC3GK.js";
 import {
   MODEL_PRICING,
   MetricsCollector,
@@ -102,7 +102,7 @@ import {
   fileVerdictCache,
   inMemoryVerdictCache,
   summarizeBackendIntegrity
-} from "./chunk-XNCF3JU3.js";
+} from "./chunk-FAOEFFRT.js";
 import {
   DEFAULT_COMPLEXITY_WEIGHTS,
   FindingsStore,
@@ -179,11 +179,11 @@ import {
   evaluateReleaseConfidence,
   judgeReplayGate,
   renderReleaseReport
-} from "./chunk-RT2AFUXM.js";
+} from "./chunk-MOXWMGPC.js";
 import {
   runEvalCampaign
-} from "./chunk-WZOIFFG2.js";
-import "./chunk-7QPK6W7R.js";
+} from "./chunk-LOBMT6SB.js";
+import "./chunk-ARU2PZFM.js";
 import {
   evaluateInterimReleaseConfidence,
   pairedEvalueSequence
@@ -194,7 +194,7 @@ import {
   paretoChart,
   researchReport,
   summaryTable
-} from "./chunk-EMKORATZ.js";
+} from "./chunk-DPZAEKA6.js";
 import {
   benjaminiHochberg,
   bonferroni,
@@ -219,6 +219,7 @@ import {
   pairedBootstrap,
   pairedMde,
   pairedRiskDifference,
+  pairedSignTest,
   pairedTTest,
   partialCredit,
   passAtK,
@@ -233,7 +234,7 @@ import {
   weightedMean,
   wilcoxonSignedRank,
   wilson
-} from "./chunk-VC43KQCK.js";
+} from "./chunk-PJQFMIOX.js";
 import {
   FileSystemTraceStore,
   InMemoryTraceStore,
@@ -12003,6 +12004,7 @@ export {
   pairedEvalueSequence,
   pairedMde,
   pairedRiskDifference,
+  pairedSignTest,
   pairedTTest,
   paraphraseRobustness,
   paraphraseRobustnessScenarios,