@tangle-network/agent-eval 0.204.0 → 0.205.0

package/dist/index.js

@@ -1,36 +1,38 @@
 import { t as __exportAll } from "./rolldown-runtime-8H4AJuhK.js";
 import { i as JudgeError, n as CaptureIntegrityError, o as NotFoundError, r as ConfigError, s as ValidationError, t as AgentEvalError } from "./errors-Dngq5h35.js";
 import { a as hashCanonical, r as canonicalString } from "./canonical-BsZVgBhW.js";
-import { a as CODING_HARNESSES, c as agentProfileId, d as harnessAxisOf, i as verifyCompletion, l as agentProfileModelId, n as completionVerdict, o as HARNESS_NATIVE_MODEL, r as createLlmCorrectnessChecker, s as agentProfileHash, t as extractProducedState, u as expandProfileAxes } from "./produced-state-DLToUXrD.js";
+import { a as CODING_HARNESSES, c as agentProfileId, d as harnessAxisOf, i as verifyCompletion, l as agentProfileModelId, n as completionVerdict, o as HARNESS_NATIVE_MODEL, r as createLlmCorrectnessChecker, s as agentProfileHash, t as extractProducedState, u as expandProfileAxes } from "./produced-state-FdbE-88t.js";
 import { c as groupRunsByAgentProfileCell, d as validateAgentProfileCell, f as verifyAgentProfileCell, h as manifestContentDigest, i as agentProfileCellKey, m as hashJson, r as agentProfileCellHashMaterial, s as buildAgentProfileCell, t as AGENT_PROFILE_KINDS, u as toAgentProfileJson } from "./agent-profile-cell-DcFW4vOc.js";
-import { t as mulberry32 } from "./random-Dn5fPWkt.js";
-import { a as spearmanR, c as weightedMean, i as ranks, n as partialCredit, o as summarizeNumberSeries, r as pearsonR, s as weightedComposite, t as confidenceInterval } from "./descriptive-D9LUFk1-.js";
-import { n as calibrateJudgeContinuous, o as verbosityBias, r as continuousAgreement, t as calibrateJudge } from "./judge-calibration-C5zgtiJT.js";
-import { a as budgetBreachView, c as interRaterReliability, i as failureClusterView, n as computeToolUseMetrics, o as corpusInterRaterAgreement, r as judgeAgreementView, s as corpusInterRaterAgreementFromJudgeScores, t as toolWasteView } from "./tool-waste-B-u3rzCz.js";
+import { t as interpolatedQuantile } from "./quantile--0yb07pD.js";
+import { a as medianInPlace, i as makeRng, t as assertFiniteSample } from "./internal-CRHwS9iz.js";
 import { i as pairedCohensDz, n as cohensD, r as interpretCliffs, t as cliffsDelta } from "./effect-sizes-DiH8MGOH.js";
+import { t as mulberry32 } from "./random-Dn5fPWkt.js";
+import { a as chiSquaredPValue, c as normalTwoSidedPValue, i as chiSquared, l as studentTTwoSidedPValue, n as iqr, o as fisherExactTwoSided, r as welchsTTest, s as linearRegressionSlope, u as twoByTwoPValue } from "./baseline-Dm8966du.js";
+import { a as spearmanR, c as weightedMean, i as ranks, n as partialCredit, o as summarizeNumberSeries, r as pearsonR, s as weightedComposite, t as confidenceInterval } from "./descriptive-CnU8Yq29.js";
+import { n as calibrateJudgeContinuous, o as verbosityBias, r as continuousAgreement, t as calibrateJudge } from "./judge-calibration-N44V04Se.js";
+import { a as budgetBreachView, c as interRaterReliability, i as failureClusterView, n as computeToolUseMetrics, o as corpusInterRaterAgreement, r as judgeAgreementView, s as corpusInterRaterAgreementFromJudgeScores, t as toolWasteView } from "./tool-waste-x_E8h9Ih.js";
 import { n as bonferroni, r as holm, t as benjaminiHochberg } from "./multiplicity-D9cjmStq.js";
-import { a as pairedRiskDifferenceExact, c as wilson, i as pairedRiskDifference, n as mcnemar, o as pairedRiskDifferenceScore, r as pairedBinaryScale, s as passAtK, t as isBinaryOutcomeVector } from "./paired-binary-OMBVRtc-.js";
-import { a as pairedSignTest, i as pairedDeltaTieFraction, n as DECISION_PAIRED_DELTA_STATISTIC, o as pairedTTest, r as pairedBootstrap, t as BOOTSTRAP_GATE_MIN_N } from "./paired-tests-Ccr16gcW.js";
-import { a as MANN_WHITNEY_EXACT_MAX_STATES, c as mannWhitneyU, i as DEFAULT_PERMUTATIONS, l as wilcoxonSignedRank, n as pairArms, o as MANN_WHITNEY_EXACT_MAX_WORK, r as pairRunRecords, s as WILCOXON_EXACT_MAX_N, t as comparePairedArms } from "./paired-arms-DpS_h8ON.js";
-import { a as requiredSampleSize, i as requiredPairedSampleSize, n as mcnemarRequiredN, r as pairedMde, t as mcnemarPower } from "./power-and-mde-bie6i965.js";
+import { a as pairedRiskDifferenceExact, c as wilson, i as pairedRiskDifference, n as mcnemar, o as pairedRiskDifferenceScore, r as pairedBinaryScale, s as passAtK, t as isBinaryOutcomeVector } from "./paired-binary-DpwwK1fU.js";
+import { a as pairedSignTest, i as pairedDeltaTieFraction, n as DECISION_PAIRED_DELTA_STATISTIC, o as pairedTTest, r as pairedBootstrap, t as BOOTSTRAP_GATE_MIN_N } from "./paired-tests-jbwPdSz5.js";
+import { a as requiredSampleSize, i as requiredPairedSampleSize, n as mcnemarRequiredN, r as pairedMde, t as mcnemarPower } from "./power-and-mde-fK5Qocwo.js";
+import { a as MANN_WHITNEY_EXACT_MAX_STATES, c as mannWhitneyU, i as DEFAULT_PERMUTATIONS, l as wilcoxonSignedRank, n as pairArms, o as MANN_WHITNEY_EXACT_MAX_WORK, r as pairRunRecords, s as WILCOXON_EXACT_MAX_N, t as comparePairedArms } from "./paired-arms-Dsu-r6xd.js";
 import { t as eProcess } from "./sequential-eprocess-D1jKoihe.js";
-import { n as iqr, r as welchsTTest } from "./baseline-CVKsfWTg.js";
-import { i as pairedDeltaTest, n as pairedDecisionShape, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-DhrmjdV8.js";
+import { i as pairedDeltaTest, n as pairedDecisionShape, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-CO0kA0bq.js";
 import { a as isToolSpan, i as isLlmSpan, r as isJudgeSpan, t as FAILURE_CLASSES } from "./schema-CSf6qWgZ.js";
 import { d as runsForScenario, r as argHash, s as judgeSpans } from "./query-DlV0pdx4.js";
-import { r as checkCanaries, t as analyzeRuns } from "./analyze-runs-BrIibNd0.js";
+import { r as checkCanaries, t as analyzeRuns } from "./analyze-runs-d5dcMlxi.js";
 import { r as observedSplitScore, t as isRealnessGated } from "./reward-nw2xZGZG.js";
 import { a as parseRunRecordSafe, c as runTaskScore, i as modelHasSnapshot, l as searchCellRunId, n as UNKNOWN_MODEL, o as roundTripRunRecord, r as isRunRecord, s as runCostFloorUsd, t as RunRecordValidationError, u as validateRunRecord } from "./run-record-BRlM99rX.js";
-import { a as summaryTable, n as gainHistogram, r as paretoChart } from "./summary-report-DLzptLSA.js";
+import { a as summaryTable, n as gainHistogram, r as paretoChart } from "./summary-report-BG6-hOkm.js";
 import { n as contentHash, r as fileVerdictCache, t as canonicalJson } from "./verdict-cache-BQEsVy9n.js";
-import { J as redTeamDataset, X as scoreRedTeamOutput, Y as redTeamReport, Z as runCanaries, a as transientDispatchFailure, i as quotaExhaustedUntil, o as aggregateRunScore, q as DEFAULT_RED_TEAM_CORPUS, rt as runCampaign, s as clamp01, t as llmJudge } from "./llm-judge-DyZdrp6I.js";
+import { J as redTeamDataset, X as scoreRedTeamOutput, Y as redTeamReport, Z as runCanaries, a as transientDispatchFailure, i as quotaExhaustedUntil, o as aggregateRunScore, q as DEFAULT_RED_TEAM_CORPUS, rt as runCampaign, s as clamp01, t as llmJudge } from "./llm-judge-Bka5q-ds.js";
 import { a as resolveModelPricing, i as isModelPriced, n as estimateCost, r as estimateTokens, t as MODEL_PRICING } from "./metrics-DOMpkoeb.js";
 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-_IbkejCF.js";
-import { B as assertRealBackend, F as surfaceContentHash, T as parseReflectionResponse, V as summarizeBackendIntegrity, c as ATTESTATION_ALGORITHM, l as attest, u as verifyAttestation, w as buildReflectionPrompt, z as BackendIntegrityError } from "./campaign-evidence-Dt_rWPk2.js";
+import { B as assertRealBackend, F as surfaceContentHash, T as parseReflectionResponse, V as summarizeBackendIntegrity, c as ATTESTATION_ALGORITHM, l as attest, u as verifyAttestation, w as buildReflectionPrompt, z as BackendIntegrityError } from "./campaign-evidence-C__cUpCt.js";
 import { d as redactForShare, f as redactText, n as REDACTION_VERSION, p as shareAllowed, r as assessShareSafety, u as redact } from "./redact-Dk1ZdJEm.js";
-import { D as paretoFrontier, E as dominates, d as pairedPromotionPower, f as requiredPairsForPairedPromotion } from "./search-kernel-Dpniloks.js";
+import { D as paretoFrontier, E as dominates, d as pairedPromotionPower, f as requiredPairsForPairedPromotion } from "./search-kernel-CsISc5t8.js";
 import { S as judgeFamily, _ as assertServedModels, b as CrossFamilyError, d as stripFencedJson, f as ModelSubstitutionError, h as assertServedModel, i as backoffMs, l as isTransientLlmError, m as assertCrossFamilyServed, o as costReceiptFromLlm, p as ServedCrossFamilyError, r as LlmResponseError, s as costReceiptFromLlmError, t as LlmCallError, u as maximumChargeForLlmRequest, v as checkServedModel, x as assertCrossFamily, y as servedModelAcceptable } from "./llm-client-DmCNzrc9.js";
-import { r as selfImprove, t as defineAgentEval } from "./define-agent-eval-B25Cc3ki.js";
+import { r as selfImprove, t as defineAgentEval } from "./define-agent-eval-CeYrXpl7.js";
 import { n as equivalenceVerdict, t as certificationEvidenceDigest } from "./verdict-DhPoXApk.js";
 import { t as TraceEmitter } from "./emitter-DbMXUIVP.js";
 import { t as buildTrajectory } from "./trajectory-D_7rLrvE.js";
@@ -46,7 +48,7 @@ import { n as InMemoryTraceStore, t as FileSystemTraceStore } from "./store-BhNu
 import { t as packageVersion$1 } from "./package-version-D7lQHt_-.js";
 import { a as classifyFailure, i as INFRA_FAILURE_BLAMES, o as classifyFailureReason, r as FAILURE_BLAME, s as failureBlame, t as DEFAULT_FAILURE_REASON_RULES } from "./failure-taxonomy-CjabTATj.js";
 import { n as runBoundedProcess, t as DEFAULT_BOUNDED_PROCESS_TIMEOUT_MS } from "./bounded-process-Cj9YjXbE.js";
-import { t as runEvalCampaign } from "./eval-campaign-ha_aWtk7.js";
+import { t as runEvalCampaign } from "./eval-campaign-DsRmYsQI.js";
 import "./rollout-BdCoWV9n.js";
 import { t as mintRolloutRows } from "./mint-CWG-s2mI.js";
 import { n as pairedEvalueSequence, t as evaluateInterimReleaseConfidence } from "./sequential-CzK5DarL.js";
@@ -63,6 +65,81 @@ import { accessSync, appendFileSync, constants, cpSync, existsSync, mkdirSync, p
 import { basename, delimiter, dirname, extname, isAbsolute, join, relative, resolve } from "node:path";
 import { spawnSync } from "node:child_process";
 import { INPUT_TOKEN_ATTR_KEYS, MODEL_ATTR_KEYS, OUTPUT_TOKEN_ATTR_KEYS, SPAN_KINDS, TOOL_NAME_ATTR_KEYS, firstNumberAttr, firstStringAttr, resolveSpanKind } from "@tangle-network/agent-trace-contract";
+//#region src/statistics/samples.ts
+/** Interpolated sample median; an empty population has no measured median. */
+function sampleMedian(values) {
+	assertFiniteSample("sampleMedian", "values", values);
+	return values.length ? medianInPlace([...values]) : null;
+}
+/** Descriptive facts, not stability policy or a significance test. */
+function summarizeSample(values, options = {}) {
+	assertFiniteSample("summarizeSample", "values", values);
+	const confidence = options.confidence ?? .95;
+	const resamples = options.resamples ?? 1e3;
+	if (!Number.isFinite(confidence) || confidence <= 0 || confidence >= 1) throw new ValidationError("summarizeSample: confidence must be between zero and one");
+	if (!Number.isSafeInteger(resamples) || resamples < 1) throw new ValidationError("summarizeSample: resamples must be a positive safe integer");
+	const rng = makeRng(options.seed, [...values]);
+	if (!values.length) return null;
+	const sorted = [...values].sort((a, b) => a - b);
+	const n = sorted.length;
+	const mean = sorted.reduce((sum, value) => sum + value / n, 0);
+	const centered = sorted.map((value) => value - mean);
+	const sumSquares = centered.reduce((sum, value) => sum + value ** 2, 0);
+	const stddev = n > 1 ? Math.sqrt(sumSquares / (n - 1)) : 0;
+	const median = interpolatedQuantile(sorted, .5);
+	const q1 = interpolatedQuantile(sorted, .25);
+	const q3 = interpolatedQuantile(sorted, .75);
+	const iqr = q3 - q1;
+	const deviations = sorted.map((value) => Math.abs(value - median)).sort((a, b) => a - b);
+	const skewness = n >= 3 && stddev > 0 ? n * centered.reduce((sum, value) => sum + (value / stddev) ** 3, 0) / ((n - 1) * (n - 2)) : null;
+	const kurtosis = n >= 4 && stddev > 0 ? n * (n + 1) * centered.reduce((sum, value) => sum + (value / stddev) ** 4, 0) / ((n - 1) * (n - 2) * (n - 3)) - 3 * (n - 1) ** 2 / ((n - 2) * (n - 3)) : null;
+	const bootstrap = [];
+	if (n >= 3) {
+		for (let i = 0; i < resamples; i++) {
+			const sample = Array.from({ length: n }, () => sorted[Math.floor(rng() * n)]);
+			bootstrap.push(medianInPlace(sample));
+		}
+		bootstrap.sort((a, b) => a - b);
+	}
+	const alpha = (1 - confidence) / 2;
+	const summary = {
+		n,
+		median,
+		mean,
+		stddev,
+		q1,
+		q3,
+		iqr,
+		min: sorted[0],
+		max: sorted[n - 1],
+		range: sorted[n - 1] - sorted[0],
+		cv: mean === 0 ? null : stddev / Math.abs(mean),
+		skewness,
+		kurtosis,
+		mad: interpolatedQuantile(deviations, .5),
+		outliers: sorted.filter((value) => value < q1 - 1.5 * iqr || value > q3 + 1.5 * iqr),
+		medianInterval: n < 3 ? null : {
+			low: interpolatedQuantile(bootstrap, alpha),
+			high: interpolatedQuantile(bootstrap, 1 - alpha),
+			confidence,
+			resamples
+		}
+	};
+	if ([
+		mean,
+		stddev,
+		iqr,
+		summary.range,
+		summary.mad,
+		summary.cv,
+		skewness,
+		kurtosis,
+		summary.medianInterval?.low ?? null,
+		summary.medianInterval?.high ?? null
+	].some((value) => value !== null && !Number.isFinite(value))) throw new ValidationError("summarizeSample: sample arithmetic exceeded finite range");
+	return summary;
+}
+//#endregion
 //#region src/profile/baselines.ts
 /**
 * Strong, generically-useful baseline ROLES — the top zone of an `AgentProfile`
@@ -9694,6 +9771,6 @@ function rankRows(rows, weights) {
 	})).sort((a, b) => b.mean - a.mean);
 }
 //#endregion
-export { AGENT_PROFILE_KINDS, ATTESTATION_ALGORITHM, AgentEvalError, AnalystRegistry, BOOTSTRAP_GATE_MIN_N, BackendIntegrityError, BudgetBreachError, BudgetGuard, CODING_HARNESSES, ConfigError, CostAccountingIncompleteError, CostCallConflictError, CostCeilingReachedError, CostLedger, CostLedgerPersistenceError, CostReceiptCaptureError, CostReservationExceededError, CostTracker, CrossFamilyError, DECISION_PAIRED_DELTA_STATISTIC, DEFAULT_BOUNDED_PROCESS_TIMEOUT_MS, DEFAULT_FAILURE_REASON_RULES, DEFAULT_PERMUTATIONS, DEFAULT_RED_TEAM_CORPUS, DEFAULT_TRACE_ANALYST_BUDGETS, DEFAULT_TRACE_ANALYST_KINDS, DEFAULT_TRACE_ANALYST_OUTPUT_TOKENS, ERROR_COUNT_PATTERNS, EquivalenceProtocolError, FAILURE_BLAME, FAILURE_CLASSES, FAILURE_MODE_KIND_SPEC, FileSystemFeedbackTrajectoryStore, FileSystemRawProviderSink, FileSystemTraceStore, FindingsStore, HARNESS_NATIVE_MODEL, HeldOutGate, INFRA_FAILURE_BLAMES, InMemoryFeedbackTrajectoryStore, InMemoryRawProviderSink, InMemoryTraceStore, JudgeError, LOCAL_COMMAND_RUNNER_DEFAULT_CAP_MS, LlmCallError, LlmResponseError, MANN_WHITNEY_EXACT_MAX_STATES, MANN_WHITNEY_EXACT_MAX_WORK, MODEL_PRICING, ModelSubstitutionError, MultiLayerVerifier, NoopRawProviderSink, NotFoundError, OUTPUT_VALUE, OtlpFileTraceStore, POLICY_EDIT_AXES, POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, POLICY_EDIT_TARGET_SURFACES, PairwiseSteeringOptimizer, PolicyEditValidationError, ProductClient, PromptRegistry, REDACTION_VERSION, RunIntegrityError, RunRecordValidationError, SEMANTIC_CONCEPT_JUDGE_VERSION, ServedCrossFamilyError, TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, TraceEmitter, UNKNOWN_MODEL, VERIFICATION_STRATEGIES, VERIFICATION_STRATEGY_SOURCES, ValidationError, WILCOXON_EXACT_MAX_N, acquisitionPlansForKnowledgeGaps, admitPolicyEdit, agentProfileCellHashMaterial, agentProfileCellKey, agentProfileHash, agentProfileId, aggregateJudgeVerdicts, aggregateRunScore, analystFindingDigest, analystRunDigest, analystRunToFeedbackTrajectory, analystRunToReviewRequests, analyzeAntiSlop, analyzeRuns, analyzeSeries, analyzeTraces, applyPolicyEditToSurface, argHash, assertCapabilityHeadroom, assertCrossFamily, assertCrossFamilyServed, assertNoHiddenLeak, assertNoJudgeVerdict, assertProductBenchmarkRun, assertRealBackend, assertRunCaptured, assertServedModel, assertServedModels, assertSingleBackend, assessShareSafety, assignFeedbackSplit, attest, benjaminiHochberg, blendHeldout, blockingKnowledgeEval, bonferroni, bootstrapCi, budgetBreachView, buildAgentProfileCell, buildDefaultAnalystRegistry, buildEquivalenceRecord, buildReflectionPrompt, buildTraceInsightContext, buildTraceInsightPrompt, buildTrajectory, calibrateJudge, calibrateJudgeContinuous, canonicalJson, capabilityHeadroom, captureFetchToRawSink, certificationEvidenceDigest, checkCanaries, checkServedModel, checkTraceContracts, clamp01, classifyFailure, classifyFailureReason, cliffsDelta, cohensD, comparePairedArms, compileTraceContractSpec, completionVerdict, computeExperimentStats, computeFindingId, computePolicyEditId, computeToolUseMetrics, confidenceInterval, contentHash, continuousAgreement, controlRunToFeedbackTrajectory, corpusInterRaterAgreement, corpusInterRaterAgreementFromJudgeScores, costForTokenPricing, costForUsage, costReceiptFromLlm, costReceiptFromLlmError, createAntiSlopJudge, createBoundedTraceAnalysisStore, createChatClient, createChatTraceEngine, createDspyRlmTraceEngine, createFeedbackTrajectory, createLlmCorrectnessChecker, createLlmReviewer, createTraceAnalyst, decidePairedPromotion, defaultBlendWeights, defineAgentEval, defineEquivalenceCheck, deployGateLayer, describeTraceInsightScope, diffFindings, diffScorecard, discoverPersonas, domainEvidencePattern, dominates, eProcess, ensembleJudge, equivalenceVerdict, errorStreakDetector, estimateCost, estimateTokens, evaluateActionPolicy, evaluateInterimReleaseConfidence, evaluateOracles, evaluateReleaseConfidence, evaluateTraceContract, expandProfileAxes, explainTraceContract, exportProductBenchmark, exportProductBenchmarkRuns, exportRunAsOtlp, extractErrorCount, extractProducedState, extractUsage, extractUsageFromSse, failureBlame, failureClusterView, feedbackTrajectoriesToDatasetScenarios, feedbackTrajectoriesToOptimizerRows, feedbackTrajectoryToOptimizerRow, fileVerdictCache, formatScorecardDiff, gainHistogram, gateTreatmentApplied, gradeOnHidden, gradeSemanticStatus, groupRunsByAgentProfileCell, harnessAxisOf, hashContent, hashJson, hiddenGrade, holm, improvementVerdict, inMemoryReviewStore, inferDomainKeywords, interRaterReliability, interpretCliffs, iqr, isBinaryOutcomeVector, isJudgeSpan, isLlmSpan, isModelPriced, isPolicyEdit, isRunRecord, isToolSpan, isTransientLlmError, jsonShape, jsonlReviewStore, jsonlRunRecordBackend, judgeAgreementView, judgeFamily, judgeSpans, knowledgeReadinessTracePayload, leaderboard, lintTraceContractSpec, llmJudge, loadScorecard, localCommandRunner, makeFinding, makePolicyEdit, makePolicyEditCandidateRecord, makeProposalFinding, manifestContentDigest, mannWhitneyU, maximumChargeForLlmRequest, mcnemar, mcnemarPower, mcnemarRequiredN, minimumPairsForPairedDeltaTest, mintRolloutRows, modelHasSnapshot, modelPriceKey, mulberry32, notBlocked, objectiveEval, observeAll, otlpTextToTraceAnalysisStore, pairArms, pairRunRecords, pairedBinaryScale, pairedBootstrap, pairedCohensDz, pairedDeltaTest, pairedDeltaTieFraction, pairedEvalueSequence, pairedMde, pairedPromotionPower, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, pairedSignTest, pairedTTest, paretoChart, paretoFrontier, parseReflectionResponse, parseRunRecordSafe, partialCredit, partitionHeldOut, partitionTrainSelectionTest, passAtK, pearsonR, policyEditFromFinding, policyEditsFromFindings, preflightModels, productBenchmarkRepoIdentity, profile_exports as profile, projectRuntimeTrajectoryEvidence, proposeSynthesisTargets, quotaExhaustedUntil, ranks, readProductBenchmarkManifest, recordRuns, recordRunsToScorecard, redTeamDataset, redTeamReport, redact, redactForShare, redactText, regexMatches, renderPreferenceMemoryMarkdown, repeatedActionDetector, requiredPairedSampleSize, requiredPairsForPairedPromotion, requiredSampleSize, resolveModelPricing, resolveSeat, roundTripRunRecord, routeFields, runAgentControlLoop, runBoundedProcess, runCampaign, runCanaries, runCostFloorUsd, runCounterfactual, runEquivalenceCheck, runEvalCampaign, runIntentMatchJudge, runKeywordCoverageJudge, runKeywordCoverageJudgeUrl, runProposeReview, runProposeReviewAsControlLoop, runSemanticConceptJudge, runTaskScore, runsForScenario, scoreKnowledgeReadiness, scorePolicyEditReadiness, scoreRedTeamOutput, scoreTraceInsightReadiness, searchCellRunId, seatPresets, selfImprove, servedModelAcceptable, shareAllowed, spanRecordsToTraceAnalysisStore, spearmanR, stripFencedJson, subjectiveEval, summarizeBackendIntegrity, summarizeNumberSeries, summarizePreferenceMemory, summaryTable, textInSnapshot, toAgentProfileJson, tokenizeDomainWords, toolSpansToTraceAnalysisStore, toolWasteView, traceContract, transientDispatchFailure, urlContains, userQuestionsForKnowledgeGaps, validatePolicyEdit, validatePolicyEditCandidateRecord, validateRunRecord, verbosityBias, verifyAgentProfileCell, verifyAttestation, verifyCompletion, viteDeployRunner, weightedComposite, weightedMean, wilcoxonSignedRank, wilson, withAssignedFeedbackSplit, withHeldoutBlend, withJudgeRetry, wranglerDeployRunner };
+export { AGENT_PROFILE_KINDS, ATTESTATION_ALGORITHM, AgentEvalError, AnalystRegistry, BOOTSTRAP_GATE_MIN_N, BackendIntegrityError, BudgetBreachError, BudgetGuard, CODING_HARNESSES, ConfigError, CostAccountingIncompleteError, CostCallConflictError, CostCeilingReachedError, CostLedger, CostLedgerPersistenceError, CostReceiptCaptureError, CostReservationExceededError, CostTracker, CrossFamilyError, DECISION_PAIRED_DELTA_STATISTIC, DEFAULT_BOUNDED_PROCESS_TIMEOUT_MS, DEFAULT_FAILURE_REASON_RULES, DEFAULT_PERMUTATIONS, DEFAULT_RED_TEAM_CORPUS, DEFAULT_TRACE_ANALYST_BUDGETS, DEFAULT_TRACE_ANALYST_KINDS, DEFAULT_TRACE_ANALYST_OUTPUT_TOKENS, ERROR_COUNT_PATTERNS, EquivalenceProtocolError, FAILURE_BLAME, FAILURE_CLASSES, FAILURE_MODE_KIND_SPEC, FileSystemFeedbackTrajectoryStore, FileSystemRawProviderSink, FileSystemTraceStore, FindingsStore, HARNESS_NATIVE_MODEL, HeldOutGate, INFRA_FAILURE_BLAMES, InMemoryFeedbackTrajectoryStore, InMemoryRawProviderSink, InMemoryTraceStore, JudgeError, LOCAL_COMMAND_RUNNER_DEFAULT_CAP_MS, LlmCallError, LlmResponseError, MANN_WHITNEY_EXACT_MAX_STATES, MANN_WHITNEY_EXACT_MAX_WORK, MODEL_PRICING, ModelSubstitutionError, MultiLayerVerifier, NoopRawProviderSink, NotFoundError, OUTPUT_VALUE, OtlpFileTraceStore, POLICY_EDIT_AXES, POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, POLICY_EDIT_TARGET_SURFACES, PairwiseSteeringOptimizer, PolicyEditValidationError, ProductClient, PromptRegistry, REDACTION_VERSION, RunIntegrityError, RunRecordValidationError, SEMANTIC_CONCEPT_JUDGE_VERSION, ServedCrossFamilyError, TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, TraceEmitter, UNKNOWN_MODEL, VERIFICATION_STRATEGIES, VERIFICATION_STRATEGY_SOURCES, ValidationError, WILCOXON_EXACT_MAX_N, acquisitionPlansForKnowledgeGaps, admitPolicyEdit, agentProfileCellHashMaterial, agentProfileCellKey, agentProfileHash, agentProfileId, aggregateJudgeVerdicts, aggregateRunScore, analystFindingDigest, analystRunDigest, analystRunToFeedbackTrajectory, analystRunToReviewRequests, analyzeAntiSlop, analyzeRuns, analyzeSeries, analyzeTraces, applyPolicyEditToSurface, argHash, assertCapabilityHeadroom, assertCrossFamily, assertCrossFamilyServed, assertNoHiddenLeak, assertNoJudgeVerdict, assertProductBenchmarkRun, assertRealBackend, assertRunCaptured, assertServedModel, assertServedModels, assertSingleBackend, assessShareSafety, assignFeedbackSplit, attest, benjaminiHochberg, blendHeldout, blockingKnowledgeEval, bonferroni, bootstrapCi, budgetBreachView, buildAgentProfileCell, buildDefaultAnalystRegistry, buildEquivalenceRecord, buildReflectionPrompt, buildTraceInsightContext, buildTraceInsightPrompt, buildTrajectory, calibrateJudge, calibrateJudgeContinuous, canonicalJson, capabilityHeadroom, captureFetchToRawSink, certificationEvidenceDigest, checkCanaries, checkServedModel, checkTraceContracts, chiSquared, chiSquaredPValue, clamp01, classifyFailure, classifyFailureReason, cliffsDelta, cohensD, comparePairedArms, compileTraceContractSpec, completionVerdict, computeExperimentStats, computeFindingId, computePolicyEditId, computeToolUseMetrics, confidenceInterval, contentHash, continuousAgreement, controlRunToFeedbackTrajectory, corpusInterRaterAgreement, corpusInterRaterAgreementFromJudgeScores, costForTokenPricing, costForUsage, costReceiptFromLlm, costReceiptFromLlmError, createAntiSlopJudge, createBoundedTraceAnalysisStore, createChatClient, createChatTraceEngine, createDspyRlmTraceEngine, createFeedbackTrajectory, createLlmCorrectnessChecker, createLlmReviewer, createTraceAnalyst, decidePairedPromotion, defaultBlendWeights, defineAgentEval, defineEquivalenceCheck, deployGateLayer, describeTraceInsightScope, diffFindings, diffScorecard, discoverPersonas, domainEvidencePattern, dominates, eProcess, ensembleJudge, equivalenceVerdict, errorStreakDetector, estimateCost, estimateTokens, evaluateActionPolicy, evaluateInterimReleaseConfidence, evaluateOracles, evaluateReleaseConfidence, evaluateTraceContract, expandProfileAxes, explainTraceContract, exportProductBenchmark, exportProductBenchmarkRuns, exportRunAsOtlp, extractErrorCount, extractProducedState, extractUsage, extractUsageFromSse, failureBlame, failureClusterView, feedbackTrajectoriesToDatasetScenarios, feedbackTrajectoriesToOptimizerRows, feedbackTrajectoryToOptimizerRow, fileVerdictCache, fisherExactTwoSided, formatScorecardDiff, gainHistogram, gateTreatmentApplied, gradeOnHidden, gradeSemanticStatus, groupRunsByAgentProfileCell, harnessAxisOf, hashContent, hashJson, hiddenGrade, holm, improvementVerdict, inMemoryReviewStore, inferDomainKeywords, interRaterReliability, interpretCliffs, iqr, isBinaryOutcomeVector, isJudgeSpan, isLlmSpan, isModelPriced, isPolicyEdit, isRunRecord, isToolSpan, isTransientLlmError, jsonShape, jsonlReviewStore, jsonlRunRecordBackend, judgeAgreementView, judgeFamily, judgeSpans, knowledgeReadinessTracePayload, leaderboard, linearRegressionSlope, lintTraceContractSpec, llmJudge, loadScorecard, localCommandRunner, makeFinding, makePolicyEdit, makePolicyEditCandidateRecord, makeProposalFinding, manifestContentDigest, mannWhitneyU, maximumChargeForLlmRequest, mcnemar, mcnemarPower, mcnemarRequiredN, minimumPairsForPairedDeltaTest, mintRolloutRows, modelHasSnapshot, modelPriceKey, mulberry32, normalTwoSidedPValue, notBlocked, objectiveEval, observeAll, otlpTextToTraceAnalysisStore, pairArms, pairRunRecords, pairedBinaryScale, pairedBootstrap, pairedCohensDz, pairedDeltaTest, pairedDeltaTieFraction, pairedEvalueSequence, pairedMde, pairedPromotionPower, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, pairedSignTest, pairedTTest, paretoChart, paretoFrontier, parseReflectionResponse, parseRunRecordSafe, partialCredit, partitionHeldOut, partitionTrainSelectionTest, passAtK, pearsonR, policyEditFromFinding, policyEditsFromFindings, preflightModels, productBenchmarkRepoIdentity, profile_exports as profile, projectRuntimeTrajectoryEvidence, proposeSynthesisTargets, quotaExhaustedUntil, ranks, readProductBenchmarkManifest, recordRuns, recordRunsToScorecard, redTeamDataset, redTeamReport, redact, redactForShare, redactText, regexMatches, renderPreferenceMemoryMarkdown, repeatedActionDetector, requiredPairedSampleSize, requiredPairsForPairedPromotion, requiredSampleSize, resolveModelPricing, resolveSeat, roundTripRunRecord, routeFields, runAgentControlLoop, runBoundedProcess, runCampaign, runCanaries, runCostFloorUsd, runCounterfactual, runEquivalenceCheck, runEvalCampaign, runIntentMatchJudge, runKeywordCoverageJudge, runKeywordCoverageJudgeUrl, runProposeReview, runProposeReviewAsControlLoop, runSemanticConceptJudge, runTaskScore, runsForScenario, sampleMedian, scoreKnowledgeReadiness, scorePolicyEditReadiness, scoreRedTeamOutput, scoreTraceInsightReadiness, searchCellRunId, seatPresets, selfImprove, servedModelAcceptable, shareAllowed, spanRecordsToTraceAnalysisStore, spearmanR, stripFencedJson, studentTTwoSidedPValue, subjectiveEval, summarizeBackendIntegrity, summarizeNumberSeries, summarizePreferenceMemory, summarizeSample, summaryTable, textInSnapshot, toAgentProfileJson, tokenizeDomainWords, toolSpansToTraceAnalysisStore, toolWasteView, traceContract, transientDispatchFailure, twoByTwoPValue, urlContains, userQuestionsForKnowledgeGaps, validatePolicyEdit, validatePolicyEditCandidateRecord, validateRunRecord, verbosityBias, verifyAgentProfileCell, verifyAttestation, verifyCompletion, viteDeployRunner, weightedComposite, weightedMean, welchsTTest, wilcoxonSignedRank, wilson, withAssignedFeedbackSplit, withHeldoutBlend, withJudgeRetry, wranglerDeployRunner };
 
 //# sourceMappingURL=index.js.map