@tangle-network/agent-eval 1.0.19 → 1.0.21

package/dist/index.js

@@ -6,41 +6,32 @@ import { a as groupRunsByAgentProfileCell, c as verifyAgentProfileCell, i as bui
 import { r as observedSplitScore, t as isRealnessGated } from "./reward-DqMfaEax.js";
 import { i as isToolSpan, n as isJudgeSpan, r as isLlmSpan, t as FAILURE_CLASSES } from "./schema-D-xu3MU_.js";
 import { a as parseRunRecordSafe, c as validateRunRecord, i as modelHasSnapshot, n as UNKNOWN_MODEL, o as runTaskScore, r as isRunRecord, t as RunRecordValidationError } from "./run-record-CpLS4Uid.js";
-import { a as cliffsDelta, c as pairedCohensDz, i as summaryTable, n as gainHistogram, o as cohensD, s as interpretCliffs } from "./summary-report-C3IJvp0i.js";
+import { i as summaryTable, n as gainHistogram } from "./summary-report-BEhqoxrA.js";
 import { t as interpolatedQuantile } from "./quantile-CncXP-Ss.js";
-import { a as wilcoxonSignedRank, i as mannWhitneyU, n as pairArms, r as pairRunRecords, t as comparePairedArms } from "./paired-arms-3-tsXvvC.js";
-import { a as assertFiniteSample, l as makeRng, t as mean$1, u as medianInPlace } from "./mean-D1d6gedE.js";
+import { n as pairArms, r as pairRunRecords, t as comparePairedArms } from "./paired-arms-DjMEtYI9.js";
+import { $ as meanConfidenceInterval, A as fisherExactTwoSided, B as summarizeDistribution, C as passAtK, D as holm, E as bonferroni, F as cliffsDelta, H as corpusInterRaterAgreement, I as cohensD, J as continuousAgreement, K as calibrateJudge, L as interpretCliffs, M as normalTwoSidedPValue, N as studentTTwoSidedPValue, O as chiSquared, P as twoByTwoPValue, Q as confidenceInterval, R as pairedCohensDz, S as pairedRiskDifferenceScore, T as benjaminiHochberg, U as corpusInterRaterAgreementFromJudgeScores, V as rocAuc, W as interRaterReliability, Z as verbosityBias, a as wilcoxonSignedRank, at as summarizeNumberSeries, b as pairedRiskDifference, c as pairedMde, ct as mean$1, d as BOOTSTRAP_GATE_MIN_N, et as partialCredit, ft as medianInPlace, g as pairedTTest, h as pairedSignTest, i as mannWhitneyU, it as studentTCritical, j as linearRegressionSlope, k as chiSquaredPValue, l as requiredPairedSampleSize, n as sampleMedian, nt as ranks, o as mcnemarPower, ot as weightedComposite, p as pairedBootstrap, q as calibrateJudgeContinuous, r as summarizeSample, rt as spearmanR, s as mcnemarRequiredN, st as weightedMean, t as eProcess, tt as pearsonR, u as requiredSampleSize, v as mcnemar, w as wilson, x as pairedRiskDifferenceExact, z as percentileNearestRank } from "./statistics-CvvC-MPv.js";
 import { t as mulberry32 } from "./random-DrzcgC7F.js";
-import { a as ranks, c as summarizeNumberSeries, i as pearsonR, l as weightedComposite, n as meanConfidenceInterval, o as spearmanR, r as partialCredit, s as studentTCritical, t as confidenceInterval, u as weightedMean } from "./descriptive-B3Eb7tX0.js";
-import { n as calibrateJudgeContinuous, o as verbosityBias, r as continuousAgreement, t as calibrateJudge } from "./judge-calibration-DKR6Xi-w.js";
-import { n as corpusInterRaterAgreementFromJudgeScores, r as interRaterReliability, t as corpusInterRaterAgreement } from "./agreement-irr-DXcXEt4I.js";
-import { i as rocAuc, n as diffFindings, r as createChatTraceEngine, t as FindingsStore } from "./findings-store-BnYN57-h.js";
-import { a as welchsTTest, c as fisherExactTwoSided, d as studentTTwoSidedPValue, f as twoByTwoPValue, i as iqr, l as linearRegressionSlope, n as checkCanaries, o as chiSquared, s as chiSquaredPValue, t as analyzeRuns, u as normalTwoSidedPValue } from "./analyze-runs-C7vWBDhl.js";
-import { n as bonferroni, r as holm, t as benjaminiHochberg } from "./multiplicity-DAd4xaDC.js";
-import { a as pairedRiskDifferenceScore, i as pairedRiskDifferenceExact, o as passAtK, r as pairedRiskDifference, s as wilson, t as mcnemar } from "./paired-binary-CMzmSSd5.js";
-import { a as pairedSignTest, o as pairedTTest, r as pairedBootstrap, t as BOOTSTRAP_GATE_MIN_N } from "./paired-tests-Buh6feJ4.js";
-import { a as requiredSampleSize, i as requiredPairedSampleSize, n as mcnemarRequiredN, r as pairedMde, t as mcnemarPower } from "./power-and-mde-g3vOn0XK.js";
-import { t as eProcess } from "./sequential-eprocess-C1YLwFaz.js";
-import { i as pairedDeltaTest, n as pairedDecisionShape, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-CiueaTTj.js";
+import { a as welchsTTest, i as iqr, n as checkCanaries, t as analyzeRuns } from "./analyze-runs-jLTRKuoK.js";
+import { i as pairedDeltaTest, n as pairedDecisionShape, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-Bi9__RQS.js";
 import { c as runsForScenario, i as judgeSpans, t as argHash } from "./query-CGXzO9d4.js";
 import { n as contentHash, r as fileVerdictCache, t as canonicalJson } from "./verdict-cache-BVH-fl18.js";
-import { B as DEFAULT_RED_TEAM_CORPUS, H as redTeamReport, U as scoreRedTeamOutput, V as redTeamDataset, W as runCanaries, Z as runCampaign, a as clamp01, i as aggregateRunScore, r as transientDispatchFailure, t as llmJudge } from "./llm-judge-VD_MI6kd.js";
-import { a as resolveModelPricing, i as isModelPriced, n as estimateCost, r as estimateTokens, t as MODEL_PRICING } from "./metrics-DNCAUtuB.js";
-import { c as costForTokenPricing, i as CostCeilingReachedError, l as costForUsage, n as CostAccountingIncompleteError, t as CostLedger } from "./cost-ledger-CS8R3IZj.js";
-import { C as assertRealBackend, S as BackendIntegrityError, _ as parseReflectionResponse, a as ATTESTATION_ALGORITHM, g as buildReflectionPrompt, o as attest, s as verifyAttestation, w as summarizeBackendIntegrity } from "./campaign-evidence-DONEWvVY.js";
+import { B as DEFAULT_RED_TEAM_CORPUS, H as redTeamReport, U as scoreRedTeamOutput, V as redTeamDataset, W as runCanaries, Z as runCampaign, a as clamp01, i as aggregateRunScore, r as transientDispatchFailure, t as llmJudge } from "./llm-judge-C813vmpc.js";
+import { a as resolveModelPricing, i as isModelPriced, n as estimateCost, r as estimateTokens, t as MODEL_PRICING } from "./metrics-zALlR17s.js";
+import { c as costForTokenPricing, i as CostCeilingReachedError, l as costForUsage, n as CostAccountingIncompleteError, t as CostLedger } from "./cost-ledger-DfhStGcZ.js";
+import { C as assertRealBackend, S as BackendIntegrityError, _ as parseReflectionResponse, a as ATTESTATION_ALGORITHM, g as buildReflectionPrompt, o as attest, s as verifyAttestation, w as summarizeBackendIntegrity } from "./campaign-evidence-NtH29ctt.js";
 import { s as surfaceContentHash } from "./json-recovery-DK4MDm5r.js";
 import { c as redactText, l as shareAllowed, n as REDACTION_VERSION, o as redact, r as assessShareSafety, s as redactForShare } from "./redact-zp_ILC4l.js";
 import { a as makeProposalFinding, i as makeFinding, r as computeFindingId } from "./types-CL1_N7AZ.js";
-import { S as requiredPairsForPairedPromotion, n as paretoFrontier, t as dominates, x as pairedPromotionPower } from "./pareto-a3EOmjfe.js";
+import { S as requiredPairsForPairedPromotion, n as paretoFrontier, t as dominates, x as pairedPromotionPower } from "./pareto-fJZg8pXm.js";
 import { a as LlmCallError, i as stripFencedJson, n as isRecord, o as LlmResponseError } from "./json-DtvZNAU9.js";
-import { r as selfImprove, t as defineAgentEval } from "./define-agent-eval-Q4FdXGj7.js";
+import { r as selfImprove, t as defineAgentEval } from "./define-agent-eval-BROmIwn1.js";
 import { n as equivalenceVerdict, t as certificationEvidenceDigest } from "./verdict-C8TcdwfD.js";
-import { c as createBoundedTraceAnalysisStore, p as DEFAULT_TRACE_ANALYST_BUDGETS, t as createTraceAnalyst } from "./kind-factory-Cfz8DUCc.js";
-import { C as assertUniqueFindingIds, D as validateAnalystReviewDecisions, E as snapshotAnalystRun, S as analystRunDigest, T as readAnalystReview, c as FAILURE_MODE_KIND_SPEC, i as DEFAULT_TRACE_ANALYST_KINDS, n as AnalystRegistry, t as buildDefaultAnalystRegistry, w as completedAnalystReviewQuality, x as analystFindingDigest } from "./default-registry-BB9_jzyH.js";
+import { c as createBoundedTraceAnalysisStore, p as DEFAULT_TRACE_ANALYST_BUDGETS, t as createTraceAnalyst } from "./kind-factory-CV0FNCMM.js";
+import { C as assertUniqueFindingIds, D as validateAnalystReviewDecisions, E as snapshotAnalystRun, S as analystRunDigest, T as readAnalystReview, c as FAILURE_MODE_KIND_SPEC, i as DEFAULT_TRACE_ANALYST_KINDS, n as AnalystRegistry, t as buildDefaultAnalystRegistry, w as completedAnalystReviewQuality, x as analystFindingDigest } from "./default-registry-CdFbxPnr.js";
 import { OUTPUT_VALUE } from "./trace-attributes.js";
 import { n as extractUsageFromSse, t as extractUsage } from "./extract-usage-ISFqhfDz.js";
 import { n as InMemoryRawProviderSink, r as NoopRawProviderSink, t as FileSystemRawProviderSink } from "./raw-provider-sink-RbaEv2vB.js";
-import { _ as exportRunAsOtlp, c as describeTraceInsightScope, g as analyzeTraces, h as tokenizeDomainWords, l as domainEvidencePattern, m as scoreTraceInsightReadiness, n as toolSpansToTraceAnalysisStore, o as buildTraceInsightContext, s as buildTraceInsightPrompt, u as inferDomainKeywords, v as captureFetchToRawSink } from "./store-tool-spans-BrSWHB9O.js";
+import { _ as exportRunAsOtlp, c as describeTraceInsightScope, g as analyzeTraces, h as tokenizeDomainWords, l as domainEvidencePattern, m as scoreTraceInsightReadiness, n as toolSpansToTraceAnalysisStore, o as buildTraceInsightContext, s as buildTraceInsightPrompt, u as inferDomainKeywords, v as captureFetchToRawSink } from "./store-tool-spans-DalWlKY8.js";
 import { t as TraceEmitter } from "./emitter-DhQ65mza.js";
 import { n as assertRunCaptured, t as RunIntegrityError } from "./integrity-DOHt1mwU.js";
 import { n as InMemoryTraceStore, t as FileSystemTraceStore } from "./store-ZykIyQmy.js";
@@ -48,152 +39,25 @@ import { t as packageVersion$1 } from "./package-version-CJl_pY9W.js";
 import { t as buildTrajectory } from "./trajectory-O3GDHmev.js";
 import { i as failureBlame, n as classifyFailure, r as classifyFailureReason } from "./failure-taxonomy-9eQfXKoW.js";
 import { n as runBoundedProcess, t as DEFAULT_BOUNDED_PROCESS_TIMEOUT_MS } from "./bounded-process-BZrWXIKW.js";
-import { t as runEvalCampaign } from "./eval-campaign-CZSCjxVb.js";
-import { _ as assertCrossFamily, a as maximumChargeForLlmRequest, d as assertServedModel, g as CrossFamilyError, h as describeServedModelCheck, i as costReceiptFromLlmError, l as ModelSubstitutionError, m as checkServedModel, p as assertServedModels, r as costReceiptFromLlm, s as isTransientLlmError, u as assertCrossFamilyServed, v as judgeFamily } from "./llm-client-BdSiZ4qz.js";
-import { n as createLlmCorrectnessChecker, r as verifyCompletion, t as completionVerdict } from "./completion-verifier-ClmJV_Yc.js";
-import "./rollout-Cu7kHGIe.js";
+import { t as runEvalCampaign } from "./eval-campaign-Dq8DEKmy.js";
+import { _ as assertCrossFamily, a as maximumChargeForLlmRequest, d as assertServedModel, g as CrossFamilyError, h as describeServedModelCheck, i as costReceiptFromLlmError, l as ModelSubstitutionError, m as checkServedModel, p as assertServedModels, r as costReceiptFromLlm, s as isTransientLlmError, u as assertCrossFamilyServed, v as judgeFamily } from "./llm-client-DMMSg-Ro.js";
+import { n as createLlmCorrectnessChecker, r as verifyCompletion, t as completionVerdict } from "./completion-verifier-DFyYFezn.js";
+import "./rollout-DKa54qer.js";
 import { t as mintRolloutRows } from "./mint-yF51iBgY.js";
-import { n as pairedEvalueSequence, t as evaluateInterimReleaseConfidence } from "./sequential-Bjs5roiK.js";
-import { n as paidJsonChat } from "./chat-json-call-C473CwTt.js";
-import { t as createDspyRlmTraceEngine } from "./dspy-rlm-engine-CqQc0R2l.js";
-import { a as budgetBreachView, i as failureClusterView, n as computeToolUseMetrics, r as judgeAgreementView, t as toolWasteView } from "./tool-waste-Czt3aWID.js";
-import { _ as analyzeSeries, a as judgeGateDecision, c as snapshotFromJudgeAgreement, g as withJudgeRetry, i as assertJudgeMayGate, n as JudgeGateRefusedError, o as measureJudgeAgreement, r as addCalibrationVerdicts, s as registerCalibrationSet, t as DEFAULT_JUDGE_GATE_POLICY } from "./judge-gate-CIlBrn6q.js";
-import { a as spanRecordsToTraceAnalysisStore, i as otlpTextToTraceAnalysisStore, r as OtlpFileTraceStore } from "./store-otlp-BfPRa2p8.js";
-import { n as bootstrapCi, t as evaluateReleaseConfidence } from "./release-confidence-nfQNh3eb.js";
-import { t as createChatClient } from "./chat-client-D0me8iwq.js";
+import { n as pairedEvalueSequence, r as sequentialCrossingHorizon, t as evaluateInterimReleaseConfidence } from "./sequential-CAksw6EU.js";
+import { n as paidJsonChat } from "./chat-json-call-Dgi7-RH-.js";
+import { t as createDspyRlmTraceEngine } from "./dspy-rlm-engine-cWyn8G1a.js";
+import { n as diffFindings, r as createChatTraceEngine, t as FindingsStore } from "./findings-store-BSuaV3dN.js";
+import { a as budgetBreachView, i as failureClusterView, n as computeToolUseMetrics, r as judgeAgreementView, t as toolWasteView } from "./tool-waste-BqSmmPUz.js";
+import { _ as analyzeSeries, a as judgeGateDecision, c as snapshotFromJudgeAgreement, g as withJudgeRetry, i as assertJudgeMayGate, n as JudgeGateRefusedError, o as measureJudgeAgreement, r as addCalibrationVerdicts, s as registerCalibrationSet, t as DEFAULT_JUDGE_GATE_POLICY } from "./judge-gate-CrN0PaAL.js";
+import { a as spanRecordsToTraceAnalysisStore, i as otlpTextToTraceAnalysisStore, r as OtlpFileTraceStore } from "./store-otlp-C-jyj3WX.js";
+import { n as bootstrapCi, t as evaluateReleaseConfidence } from "./release-confidence-DGtuPDxJ.js";
+import { t as createChatClient } from "./chat-client-Zz97O3NB.js";
 import { createHash } from "node:crypto";
 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";
 import { spawnSync } from "node:child_process";
 import { ATTR, 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";
-// src/statistics/distribution-summary.ts
-/** Nearest-rank percentile on a sorted, non-empty sample (0-100 scale). */
-function nearestRank(sorted, percentile) {
-	return sorted[Math.min(sorted.length - 1, Math.max(0, Math.ceil(percentile / 100 * sorted.length) - 1))];
-}
-/**
-* Nearest-rank percentile on an arbitrary sample, `percentile` on a 0-100
-* scale — the same rule `summarizeDistribution` applies, exposed for a single
-* percentile the summary struct does not carry (e.g. p99.9). Returns NaN for
-* an empty sample; callers that need a numeric floor guard it.
-*/
-function percentileNearestRank(values, percentile) {
-	if (!Array.isArray(values)) throw new ValidationError("percentileNearestRank: values must be an array");
-	if (values.length === 0) return NaN;
-	return nearestRank([...values].sort((a, b) => a - b), percentile);
-}
-function summarizeDistribution(values, options = {}) {
-	if (values.some((value) => !Number.isFinite(value))) throw new ValidationError("summarizeDistribution: samples must be finite");
-	const finite = options.lowerBound === void 0 ? values : values.filter((value) => value >= options.lowerBound);
-	if (finite.length === 0) return {
-		min: NaN,
-		p50: NaN,
-		p90: NaN,
-		p95: NaN,
-		p99: NaN,
-		max: NaN,
-		mean: NaN,
-		n: 0,
-		values: []
-	};
-	const sorted = [...finite].sort((a, b) => a - b);
-	const n = sorted.length;
-	const mean = sorted.reduce((sum, value) => sum + value / n, 0);
-	const sumSquares = sorted.map((value) => value - mean).reduce((sum, value) => sum + value ** 2, 0);
-	const summary = {
-		min: sorted[0],
-		p50: nearestRank(sorted, 50),
-		p90: nearestRank(sorted, 90),
-		p95: nearestRank(sorted, 95),
-		p99: nearestRank(sorted, 99),
-		max: sorted[n - 1],
-		mean,
-		n,
-		values: sorted
-	};
-	if (n >= 2) {
-		const stddev = Math.sqrt(sumSquares / (n - 1));
-		const margin = studentTCritical(n - 1, .95) * stddev / Math.sqrt(n);
-		summary.stddev = stddev;
-		summary.ci95Low = mean - margin;
-		summary.ci95High = mean + margin;
-	}
-	return summary;
-}
-// 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;
-}
 // src/profile/baselines.ts
 /**
 * Strong, generically-useful baseline ROLES — the top zone of an `AgentProfile`
@@ -8916,6 +8780,6 @@ function assertHasField(name, obj, field) {
 		actual: `keys: ${Object.keys(obj).join(", ")}`
 	};
 }
-export { AGENT_PROFILE_KINDS, ATTESTATION_ALGORITHM, AgentEvalError, AnalystRegistry, BOOTSTRAP_GATE_MIN_N, BackendIntegrityError, BudgetBreachError, BudgetGuard, CODING_HARNESSES, ConfigError, CostAccountingIncompleteError, CostCeilingReachedError, CostLedger, CostTracker, CrossFamilyError, DEFAULT_BOUNDED_PROCESS_TIMEOUT_MS, DEFAULT_JUDGE_GATE_POLICY, DEFAULT_RED_TEAM_CORPUS, DEFAULT_TRACE_ANALYST_BUDGETS, DEFAULT_TRACE_ANALYST_KINDS, ERROR_COUNT_PATTERNS, EquivalenceProtocolError, FAILURE_CLASSES, FAILURE_MODE_KIND_SPEC, FileSystemFeedbackTrajectoryStore, FileSystemRawProviderSink, FileSystemTraceStore, FindingsStore, HARNESS_NATIVE_MODEL, HeldOutGate, InMemoryFeedbackTrajectoryStore, InMemoryRawProviderSink, InMemoryTraceStore, JudgeError, JudgeGateRefusedError, LlmCallError, LlmResponseError, MODEL_PRICING, ModelSubstitutionError, MultiLayerVerifier, NoopRawProviderSink, NotFoundError, OUTPUT_VALUE, OtlpFileTraceStore, PairwiseSteeringOptimizer, ProductClient, PromptRegistry, REDACTION_VERSION, RunIntegrityError, RunRecordValidationError, SEMANTIC_CONCEPT_JUDGE_VERSION, TraceEmitter, UNKNOWN_MODEL, VERIFICATION_STRATEGIES, VERIFICATION_STRATEGY_SOURCES, ValidationError, acquisitionPlansForKnowledgeGaps, addCalibrationVerdicts, agentProfileCellKey, agentProfileHash, agentProfileId, aggregateJudgeVerdicts, aggregateRunScore, analystRunDigest, analystRunToFeedbackTrajectory, analystRunToReviewRequests, analyzeAntiSlop, analyzeRuns, analyzeSeries, analyzeTraces, argHash, assertCapabilityHeadroom, assertContains, assertCrossFamily, assertCrossFamilyServed, assertEventCount, assertExists, assertHasField, assertHttpOk, assertJudgeMayGate, assertLatencyUnder, assertNoHiddenLeak, assertNotContains, assertProductBenchmarkRun, assertRealBackend, assertRunCaptured, assertServedModel, assertServedModels, assertSingleBackend, assertStatus, assessShareSafety, assignFeedbackSplit, attest, benjaminiHochberg, blendHeldout, blockingKnowledgeEval, bonferroni, bootstrapCi, budgetBreachView, buildAgentProfileCell, buildDefaultAnalystRegistry, buildReflectionPrompt, buildTraceInsightContext, buildTraceInsightPrompt, buildTrajectory, calibrateJudge, calibrateJudgeContinuous, canonicalJson, capabilityHeadroom, captureFetchToRawSink, certificationEvidenceDigest, checkCanaries, checkServedModel, checkTraceContracts, chiSquared, chiSquaredPValue, clamp01, classifyFailure, classifyFailureReason, cliffsDelta, cohensD, comparePairedArms, compileTraceContractSpec, completionVerdict, computeFindingId, computeToolUseMetrics, confidenceInterval, contentHash, continuousAgreement, controlRunToFeedbackTrajectory, corpusInterRaterAgreement, corpusInterRaterAgreementFromJudgeScores, costForTokenPricing, costForUsage, costReceiptFromLlm, costReceiptFromLlmError, createAntiSlopJudge, createBoundedTraceAnalysisStore, createChatClient, createChatTraceEngine, createDspyRlmTraceEngine, createFeedbackTrajectory, createLlmCorrectnessChecker, createLlmReviewer, createTraceAnalyst, decidePairedPromotion, defineAgentEval, defineEquivalenceCheck, deployGateLayer, describeServedModelCheck, 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, gradeSemanticStatus, groupRunsByAgentProfileCell, harnessAxisOf, hashContent, hashJson, holm, inMemoryReviewStore, inferDomainKeywords, interRaterReliability, interpretCliffs, iqr, isJudgeSpan, isLlmSpan, isModelPriced, isRunRecord, isToolSpan, isTransientLlmError, jsonShape, jsonlReviewStore, jsonlRunRecordBackend, judgeAgreementView, judgeFamily, judgeGateDecision, judgeSpans, knowledgeReadinessTracePayload, leaderboard, linearRegressionSlope, lintTraceContractSpec, llmJudge, loadScorecard, localCommandRunner, makeFinding, makePolicyEditCandidateRecord, makeProposalFinding, manifestContentDigest, mannWhitneyU, maximumChargeForLlmRequest, mcnemar, mcnemarPower, mcnemarRequiredN, meanConfidenceInterval, measureJudgeAgreement, minimumPairsForPairedDeltaTest, mintRolloutRows, modelHasSnapshot, mulberry32, normalTwoSidedPValue, objectiveEval, observeAll, otlpTextToTraceAnalysisStore, pairArms, pairRunRecords, pairedBootstrap, pairedCohensDz, pairedDeltaTest, pairedEvalueSequence, pairedMde, pairedPromotionPower, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, pairedSignTest, pairedTTest, paretoFrontier, parseReflectionResponse, parseRunRecordSafe, partialCredit, partitionHeldOut, passAtK, pearsonR, percentileNearestRank, preflightModels, productBenchmarkRepoIdentity, profile_exports as profile, projectRuntimeTrajectoryEvidence, proposeSynthesisTargets, ranks, readProductBenchmarkManifest, recordRuns, recordRunsToScorecard, redTeamDataset, redTeamReport, redact, redactForShare, redactText, registerCalibrationSet, renderPreferenceMemoryMarkdown, repeatedActionDetector, requiredPairedSampleSize, requiredPairsForPairedPromotion, requiredSampleSize, resolveModelPricing, resolveSeat, rocAuc, runAgentControlLoop, runBoundedProcess, runCampaign, runCanaries, runEquivalenceCheck, runEvalCampaign, runIntentMatchJudge, runKeywordCoverageJudge, runKeywordCoverageJudgeUrl, runProposeReview, runSemanticConceptJudge, runTaskScore, runsForScenario, sampleMedian, scoreKnowledgeReadiness, scoreRedTeamOutput, scoreTraceInsightReadiness, seatPresets, selfImprove, shareAllowed, snapshotFromJudgeAgreement, spanRecordsToTraceAnalysisStore, spearmanR, stripFencedJson, studentTCritical, studentTTwoSidedPValue, subjectiveEval, summarizeBackendIntegrity, summarizeDistribution, summarizeNumberSeries, summarizePreferenceMemory, summarizeSample, summaryTable, toAgentProfileJson, tokenizeDomainWords, toolSpansToTraceAnalysisStore, toolWasteView, traceContract, transientDispatchFailure, twoByTwoPValue, userQuestionsForKnowledgeGaps, validateRunRecord, verbosityBias, verifyAgentProfileCell, verifyAttestation, verifyCompletion, viteDeployRunner, weightedComposite, weightedMean, welchsTTest, wilcoxonSignedRank, wilson, withAssignedFeedbackSplit, withJudgeRetry, wranglerDeployRunner };
+export { AGENT_PROFILE_KINDS, ATTESTATION_ALGORITHM, AgentEvalError, AnalystRegistry, BOOTSTRAP_GATE_MIN_N, BackendIntegrityError, BudgetBreachError, BudgetGuard, CODING_HARNESSES, ConfigError, CostAccountingIncompleteError, CostCeilingReachedError, CostLedger, CostTracker, CrossFamilyError, DEFAULT_BOUNDED_PROCESS_TIMEOUT_MS, DEFAULT_JUDGE_GATE_POLICY, DEFAULT_RED_TEAM_CORPUS, DEFAULT_TRACE_ANALYST_BUDGETS, DEFAULT_TRACE_ANALYST_KINDS, ERROR_COUNT_PATTERNS, EquivalenceProtocolError, FAILURE_CLASSES, FAILURE_MODE_KIND_SPEC, FileSystemFeedbackTrajectoryStore, FileSystemRawProviderSink, FileSystemTraceStore, FindingsStore, HARNESS_NATIVE_MODEL, HeldOutGate, InMemoryFeedbackTrajectoryStore, InMemoryRawProviderSink, InMemoryTraceStore, JudgeError, JudgeGateRefusedError, LlmCallError, LlmResponseError, MODEL_PRICING, ModelSubstitutionError, MultiLayerVerifier, NoopRawProviderSink, NotFoundError, OUTPUT_VALUE, OtlpFileTraceStore, PairwiseSteeringOptimizer, ProductClient, PromptRegistry, REDACTION_VERSION, RunIntegrityError, RunRecordValidationError, SEMANTIC_CONCEPT_JUDGE_VERSION, TraceEmitter, UNKNOWN_MODEL, VERIFICATION_STRATEGIES, VERIFICATION_STRATEGY_SOURCES, ValidationError, acquisitionPlansForKnowledgeGaps, addCalibrationVerdicts, agentProfileCellKey, agentProfileHash, agentProfileId, aggregateJudgeVerdicts, aggregateRunScore, analystRunDigest, analystRunToFeedbackTrajectory, analystRunToReviewRequests, analyzeAntiSlop, analyzeRuns, analyzeSeries, analyzeTraces, argHash, assertCapabilityHeadroom, assertContains, assertCrossFamily, assertCrossFamilyServed, assertEventCount, assertExists, assertHasField, assertHttpOk, assertJudgeMayGate, assertLatencyUnder, assertNoHiddenLeak, assertNotContains, assertProductBenchmarkRun, assertRealBackend, assertRunCaptured, assertServedModel, assertServedModels, assertSingleBackend, assertStatus, assessShareSafety, assignFeedbackSplit, attest, benjaminiHochberg, blendHeldout, blockingKnowledgeEval, bonferroni, bootstrapCi, budgetBreachView, buildAgentProfileCell, buildDefaultAnalystRegistry, buildReflectionPrompt, buildTraceInsightContext, buildTraceInsightPrompt, buildTrajectory, calibrateJudge, calibrateJudgeContinuous, canonicalJson, capabilityHeadroom, captureFetchToRawSink, certificationEvidenceDigest, checkCanaries, checkServedModel, checkTraceContracts, chiSquared, chiSquaredPValue, clamp01, classifyFailure, classifyFailureReason, cliffsDelta, cohensD, comparePairedArms, compileTraceContractSpec, completionVerdict, computeFindingId, computeToolUseMetrics, confidenceInterval, contentHash, continuousAgreement, controlRunToFeedbackTrajectory, corpusInterRaterAgreement, corpusInterRaterAgreementFromJudgeScores, costForTokenPricing, costForUsage, costReceiptFromLlm, costReceiptFromLlmError, createAntiSlopJudge, createBoundedTraceAnalysisStore, createChatClient, createChatTraceEngine, createDspyRlmTraceEngine, createFeedbackTrajectory, createLlmCorrectnessChecker, createLlmReviewer, createTraceAnalyst, decidePairedPromotion, defineAgentEval, defineEquivalenceCheck, deployGateLayer, describeServedModelCheck, 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, gradeSemanticStatus, groupRunsByAgentProfileCell, harnessAxisOf, hashContent, hashJson, holm, inMemoryReviewStore, inferDomainKeywords, interRaterReliability, interpretCliffs, iqr, isJudgeSpan, isLlmSpan, isModelPriced, isRunRecord, isToolSpan, isTransientLlmError, jsonShape, jsonlReviewStore, jsonlRunRecordBackend, judgeAgreementView, judgeFamily, judgeGateDecision, judgeSpans, knowledgeReadinessTracePayload, leaderboard, linearRegressionSlope, lintTraceContractSpec, llmJudge, loadScorecard, localCommandRunner, makeFinding, makePolicyEditCandidateRecord, makeProposalFinding, manifestContentDigest, mannWhitneyU, maximumChargeForLlmRequest, mcnemar, mcnemarPower, mcnemarRequiredN, meanConfidenceInterval, measureJudgeAgreement, minimumPairsForPairedDeltaTest, mintRolloutRows, modelHasSnapshot, mulberry32, normalTwoSidedPValue, objectiveEval, observeAll, otlpTextToTraceAnalysisStore, pairArms, pairRunRecords, pairedBootstrap, pairedCohensDz, pairedDeltaTest, pairedEvalueSequence, pairedMde, pairedPromotionPower, pairedRiskDifference, pairedRiskDifferenceExact, pairedRiskDifferenceScore, pairedSignTest, pairedTTest, paretoFrontier, parseReflectionResponse, parseRunRecordSafe, partialCredit, partitionHeldOut, passAtK, pearsonR, percentileNearestRank, preflightModels, productBenchmarkRepoIdentity, profile_exports as profile, projectRuntimeTrajectoryEvidence, proposeSynthesisTargets, ranks, readProductBenchmarkManifest, recordRuns, recordRunsToScorecard, redTeamDataset, redTeamReport, redact, redactForShare, redactText, registerCalibrationSet, renderPreferenceMemoryMarkdown, repeatedActionDetector, requiredPairedSampleSize, requiredPairsForPairedPromotion, requiredSampleSize, resolveModelPricing, resolveSeat, rocAuc, runAgentControlLoop, runBoundedProcess, runCampaign, runCanaries, runEquivalenceCheck, runEvalCampaign, runIntentMatchJudge, runKeywordCoverageJudge, runKeywordCoverageJudgeUrl, runProposeReview, runSemanticConceptJudge, runTaskScore, runsForScenario, sampleMedian, scoreKnowledgeReadiness, scoreRedTeamOutput, scoreTraceInsightReadiness, seatPresets, selfImprove, sequentialCrossingHorizon, shareAllowed, snapshotFromJudgeAgreement, spanRecordsToTraceAnalysisStore, spearmanR, stripFencedJson, studentTCritical, studentTTwoSidedPValue, subjectiveEval, summarizeBackendIntegrity, summarizeDistribution, summarizeNumberSeries, summarizePreferenceMemory, summarizeSample, summaryTable, toAgentProfileJson, tokenizeDomainWords, toolSpansToTraceAnalysisStore, toolWasteView, traceContract, transientDispatchFailure, twoByTwoPValue, userQuestionsForKnowledgeGaps, validateRunRecord, verbosityBias, verifyAgentProfileCell, verifyAttestation, verifyCompletion, viteDeployRunner, weightedComposite, weightedMean, welchsTTest, wilcoxonSignedRank, wilson, withAssignedFeedbackSplit, withJudgeRetry, wranglerDeployRunner };
 
 //# sourceMappingURL=index.js.map