@tangle-network/agent-eval 0.205.0 → 0.205.2

package/dist/index.js

@@ -1,7 +1,7 @@
 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-FdbE-88t.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-CywjstcW.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 interpolatedQuantile } from "./quantile--0yb07pD.js";
 import { a as medianInPlace, i as makeRng, t as assertFiniteSample } from "./internal-CRHwS9iz.js";
@@ -12,27 +12,27 @@ import { a as spearmanR, c as weightedMean, i as ranks, n as partialCredit, o as
 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-DpwwK1fU.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-b2L3WKqr.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 { 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-CgWT7sHy.js";
 import { t as eProcess } from "./sequential-eprocess-D1jKoihe.js";
-import { i as pairedDeltaTest, n as pairedDecisionShape, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-CO0kA0bq.js";
+import { i as pairedDeltaTest, n as pairedDecisionShape, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-2deJ7sDX.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-d5dcMlxi.js";
+import { r as checkCanaries, t as analyzeRuns } from "./analyze-runs-GZ0VSwui.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-BG6-hOkm.js";
+import { a as summaryTable, n as gainHistogram, r as paretoChart } from "./summary-report-fXurE2TR.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-Bka5q-ds.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-Fc3UIcPq.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-C__cUpCt.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-DW3Krxov.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-CsISc5t8.js";
+import { D as paretoFrontier, E as dominates, d as pairedPromotionPower, f as requiredPairsForPairedPromotion } from "./search-kernel-mzeNuzr8.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-CeYrXpl7.js";
+import { r as selfImprove, t as defineAgentEval } from "./define-agent-eval-CmItPfIq.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";
@@ -48,8 +48,8 @@ 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-DsRmYsQI.js";
-import "./rollout-BdCoWV9n.js";
+import { t as runEvalCampaign } from "./eval-campaign-DU_p8qVt.js";
+import "./rollout-l9w3Gdvh.js";
 import { t as mintRolloutRows } from "./mint-CWG-s2mI.js";
 import { n as pairedEvalueSequence, t as evaluateInterimReleaseConfidence } from "./sequential-CzK5DarL.js";
 import { n as paidJsonChat } from "./chat-json-call-Digk-pfF.js";

package/dist/{llm-judge-Bka5q-ds.js → llm-judge-Fc3UIcPq.js}

@@ -2,13 +2,13 @@ import { i as JudgeError, s as ValidationError } from "./errors-Dngq5h35.js";
 import { a as hashCanonical, i as compareCodeUnits, r as canonicalString, t as LEDGER_HASH_PATTERN } from "./canonical-BsZVgBhW.js";
 import { t as mulberry32 } from "./random-Dn5fPWkt.js";
 import { o as summarizeNumberSeries, s as weightedComposite, t as confidenceInterval } from "./descriptive-CnU8Yq29.js";
-import { r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-CO0kA0bq.js";
-import { a as heldoutSignificance, i as dimensionRegressions, o as pairHoldout } from "./statistical-heldout-C26hOFWD.js";
+import { r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-2deJ7sDX.js";
+import { a as heldoutSignificance, i as dimensionRegressions, o as pairHoldout } from "./statistical-heldout-BP6iEJvR.js";
 import { i as modelHasSnapshot, l as searchCellRunId } from "./run-record-BRlM99rX.js";
 import { n as contentHash } from "./verdict-cache-BQEsVy9n.js";
 import { r as projectCampaignCellQuality, t as campaignCellCostProvenance } from "./run-record-BB290Y9A.js";
 import { i as CostLedger, t as CostAccountingIncompleteError } from "./cost-ledger-_IbkejCF.js";
-import { F as surfaceContentHash, G as campaignCoverage, H as assertCampaignDesign, I as surfaceDispatchRef, K as campaignScenarioIdentity, L as surfaceHash, O as FinalEvidenceConflictError, P as renderSurfaceDiff, R as surfaceHashMatches, S as campaignMeanCompositeOrNull, U as assertCampaignSplitIdentity, W as assertCompleteCampaign, Y as formatCoverageFailures, b as campaignBreakdown, k as FinalEvidenceError, q as campaignSplitDigest, t as createCampaignEvidenceReceipt, x as campaignMeanComposite, z as BackendIntegrityError } from "./campaign-evidence-C__cUpCt.js";
+import { F as surfaceContentHash, G as campaignCoverage, H as assertCampaignDesign, I as surfaceDispatchRef, K as campaignScenarioIdentity, L as surfaceHash, O as FinalEvidenceConflictError, P as renderSurfaceDiff, R as surfaceHashMatches, S as campaignMeanCompositeOrNull, U as assertCampaignSplitIdentity, W as assertCompleteCampaign, Y as formatCoverageFailures, b as campaignBreakdown, k as FinalEvidenceError, q as campaignSplitDigest, t as createCampaignEvidenceReceipt, x as campaignMeanComposite, z as BackendIntegrityError } from "./campaign-evidence-DW3Krxov.js";
 import { m as mapConcurrent } from "./journal-DvKCvjty.js";
 import { t as deepFreezeCanonicalJson } from "./deep-freeze-CurcOQg3.js";
 import { S as fsCampaignStorage, g as isRecord, h as isExternalTextCandidate, x as createRunCostLedger } from "./external-optimizer-subprocess-BmKtH3T6.js";
@@ -16,9 +16,9 @@ import { r as summarizeEvaluationUnits, t as defineEvaluationClaim } from "./cla
 import { f as redactText, n as REDACTION_VERSION, o as detectCredential, s as detectPersonalData } from "./redact-Dk1ZdJEm.js";
 import { t as detectRewardHacking } from "./reward-hacking-BFcQgkux.js";
 import { t as assertProposalFindings } from "./proposal-findings-CXblsskH.js";
-import { D as paretoFrontier, O as paretoFrontierWithCrowding, a as searchPolicyView, b as estimateNode, i as searchExpansionIndex, n as runSearch, p as EXPANSION_OPERATORS, t as SEARCH_KERNEL_SOURCE } from "./search-kernel-CsISc5t8.js";
+import { D as paretoFrontier, O as paretoFrontierWithCrowding, a as searchPolicyView, b as estimateNode, i as searchExpansionIndex, n as runSearch, p as EXPANSION_OPERATORS, t as SEARCH_KERNEL_SOURCE } from "./search-kernel-mzeNuzr8.js";
 import { a as validateSearchLedgerEvent, d as searchTaskSetDigest, i as replaySearchLedgerText, l as searchCellId, n as openSearchLedger, u as searchNodeId } from "./search-ledger-CXQPwdWz.js";
-import { n as searchProposerView, t as renderSearchSummary } from "./search-summary-BZkoLwbB.js";
+import { n as searchProposerView, t as renderSearchSummary } from "./search-summary-CDURlgZR.js";
 import { C as combineAbortSignals, d as stripFencedJson, o as costReceiptFromLlm, s as costReceiptFromLlmError, u as maximumChargeForLlmRequest } from "./llm-client-DmCNzrc9.js";
 import { createHash } from "node:crypto";
 import { writeFileSync } from "node:fs";
@@ -6174,4 +6174,4 @@ function firstString(value) {
 //#endregion
 export { captureFinalEvidencePolicy as $, assertCompleteSearchHistory as A, searchReceiptAccounting as B, incumbentWithOperatorBandit as C, surfaceDiff as D, searchEdgeId as E, verifySearchHistoryArtifact as F, captureJudge as G, uniform as H, verifySearchHistoryReceipt as I, redTeamDataset as J, defaultProductionGate as K, campaignCellSearchResult as L, assertSearchHistoryMatchesState as M, createSearchHistoryReceipt as N, surfaceNode as O, searchHistoryCoverageRow as P, assertIndependentEvaluationSplit as Q, searchModelIdentity as R, incumbent as S, developmentClaim as T, runFinalComparison as U, asha as V, openAutoPr as W, scoreRedTeamOutput as X, redTeamReport as Y, runCanaries as Z, isProposedCandidate as _, transientDispatchFailure as a, resolveRunDir as at, beam as b, compareOptimizationMethods as c, buildCellSchedule as ct, combineComparisonCosts as d, evaluationUnitMap as et, costFromLedgerSummary as f, runOptimization as g, runImprovementLoop as h, quotaExhaustedUntil as i, planCampaignRun as it, assertSearchHistoryAdmissionOptions as j, SearchHistoryRequiredError as k, optimizationTokenUsageFromSummary as l, cellCachePath as lt, readGepaCandidatePopulationArtifact as m, JudgeParseError as n, runEval as nt, aggregateRunScore as o, tangleTracesRoot as ot, assertGepaCandidatePopulationSummary as p, DEFAULT_RED_TEAM_CORPUS as q, isTransientTransportFailure as r, runCampaign as rt, clamp01 as s, readCachedCell as st, llmJudge as t, reserveFinalEvidence as tt, executeOptimizationMethod as u, computeManifestHash as ut, labelTrustRank as v, SearchRecorder as w, crowdedFrontierParent as x, aide as y, searchProposalExecution as z };
 
-//# sourceMappingURL=llm-judge-Bka5q-ds.js.map
+//# sourceMappingURL=llm-judge-Fc3UIcPq.js.map