@tangle-network/agent-eval 0.181.0 → 0.182.0

package/dist/benchmarks/index.js

@@ -1,7 +1,7 @@
 import { t as __exportAll } from "../rolldown-runtime-8H4AJuhK.js";
-import { X as runCampaign } from "../llm-judge-DEFZeSiu.js";
+import { X as runCampaign } from "../llm-judge-BPFR12VJ.js";
 import { S as fsCampaignStorage } from "../external-optimizer-subprocess-D4dzUBZI.js";
-import "../campaign-4_ppJW5X.js";
+import "../campaign-B2U7WcFz.js";
 import { join } from "node:path";
 //#region src/benchmarks/calibration.ts
 async function calibrateBenchmarkMetric(options) {

package/dist/campaign/index.js

@@ -1,10 +1,10 @@
-import { B as surfaceDispatchRef, F as assertCodeSurfaceIdentity, I as codeSurfaceIdentityMaterial, K as assertCampaignDesign, L as componentSurfaceIdentityMaterial, Q as campaignSplitDigestFromIdentities, R as renderSurfaceDiff, S as campaignMeanComposite, V as surfaceHash, X as campaignScenarioIdentity, Z as campaignSplitDigest, _ as provenanceRecordPath, at as pairHoldout, d as buildLoopProvenanceRecord, f as campaignMeasurementDigest, g as loopProvenanceSpans, h as loopProvenanceArgsFromResult, it as heldoutSignificance, k as powerPreflight, m as emitLoopProvenance, nt as detectScale, p as canonicalDigest, q as assertCampaignSplitIdentity, rt as dimensionRegressions, v as provenanceSpansPath, w as compareRankKeys, x as campaignBreakdown, y as verifyLoopProvenanceRecord, z as surfaceContentHash } from "../campaign-evidence-B8oF9xQ6.js";
-import { $ as tangleTracesRoot, A as openSearchLedger, D as verifySearchHistoryArtifact, E as searchHistoryCoverageRow, L as openAutoPr, M as validateSearchLedgerEvent, O as verifySearchHistoryReceipt, Q as resolveRunDir, S as assertCompleteSearchHistory, T as createSearchHistoryReceipt, X as runCampaign, Y as runEval, Z as planCampaignRun, _ as isProposedCandidate, a as transientDispatchFailure, b as recordCandidatePopulationSearch, c as compareOptimizationMethods, d as combineComparisonCosts, et as readCachedCell, f as costFromLedgerSummary, g as runOptimization, h as runImprovementLoop, i as quotaExhaustedUntil, j as replaySearchLedgerText, k as FileSearchLedger, l as optimizationTokenUsageFromSummary, m as readGepaCandidatePopulationArtifact, nt as cellCachePath, r as isTransientTransportFailure, t as llmJudge, tt as buildCellSchedule, v as labelTrustRank, w as assertSearchHistoryMatchesReplay, x as SearchHistoryRequiredError, y as SearchRecorder, z as defaultProductionGate } from "../llm-judge-DEFZeSiu.js";
+import { B as surfaceDispatchRef, F as assertCodeSurfaceIdentity, I as codeSurfaceIdentityMaterial, K as assertCampaignDesign, L as componentSurfaceIdentityMaterial, Q as campaignSplitDigestFromIdentities, R as renderSurfaceDiff, S as campaignMeanComposite, V as surfaceHash, X as campaignScenarioIdentity, Z as campaignSplitDigest, _ as provenanceRecordPath, at as pairHoldout, d as buildLoopProvenanceRecord, f as campaignMeasurementDigest, g as loopProvenanceSpans, h as loopProvenanceArgsFromResult, it as heldoutSignificance, k as powerPreflight, m as emitLoopProvenance, nt as detectScale, p as canonicalDigest, q as assertCampaignSplitIdentity, rt as dimensionRegressions, v as provenanceSpansPath, w as compareRankKeys, x as campaignBreakdown, y as verifyLoopProvenanceRecord, z as surfaceContentHash } from "../campaign-evidence-PYPUftRy.js";
+import { $ as tangleTracesRoot, A as openSearchLedger, D as verifySearchHistoryArtifact, E as searchHistoryCoverageRow, L as openAutoPr, M as validateSearchLedgerEvent, O as verifySearchHistoryReceipt, Q as resolveRunDir, S as assertCompleteSearchHistory, T as createSearchHistoryReceipt, X as runCampaign, Y as runEval, Z as planCampaignRun, _ as isProposedCandidate, a as transientDispatchFailure, b as recordCandidatePopulationSearch, c as compareOptimizationMethods, d as combineComparisonCosts, et as readCachedCell, f as costFromLedgerSummary, g as runOptimization, h as runImprovementLoop, i as quotaExhaustedUntil, j as replaySearchLedgerText, k as FileSearchLedger, l as optimizationTokenUsageFromSummary, m as readGepaCandidatePopulationArtifact, nt as cellCachePath, r as isTransientTransportFailure, t as llmJudge, tt as buildCellSchedule, v as labelTrustRank, w as assertSearchHistoryMatchesReplay, x as SearchHistoryRequiredError, y as SearchRecorder, z as defaultProductionGate } from "../llm-judge-BPFR12VJ.js";
 import { C as inMemoryCampaignStorage, D as SearchLedgerIntegrityError, E as SearchLedgerError, S as fsCampaignStorage, T as SearchLedgerConflictError, l as DEFAULT_EXTERNAL_OPTIMIZER_CALLBACK_LIMITS, u as DEFAULT_EXTERNAL_OPTIMIZER_PROCESS_LIMITS, v as resolveExternalOptimizerCallbackLimits, x as createRunCostLedger, y as resolveExternalOptimizerProcessLimits } from "../external-optimizer-subprocess-D4dzUBZI.js";
 import { i as makeProposalFinding } from "../types-DQ0e2E7y.js";
 import { n as acquireSingleRunLock } from "../external-optimizer-process-QDRURJAM.js";
-import { a as externalTextOptimizationMethod, i as composeGate, n as gepaOptimizationMethod, o as decodeExternalTextCandidate, r as heldOutGate, s as readExternalOptimizerObservationArtifact, t as skillOptOptimizationMethod, u as createReferenceEquivalenceJudge } from "../skillopt-optimization-method-CL_0aArC.js";
-import { A as loadEvalFixtureScenarios, C as FsLabeledScenarioStore, D as neutralizationGate, E as rolloutArgumentDiff, M as analyzeCrossSurfaceInteractions, N as buildTraceAnalystSurfaceDispatch, O as discoverEvalFixtures, P as traceAnalystQualityJudge, S as neutralizeText, T as classifyUngroundedLiterals, _ as scoreboardSummary, a as autoevalsScorerJudge, b as scopedOptimizationMethod, c as selectDiscriminative, d as runProfileMatrixSegment, f as ProfileMatrixError, g as scoreUserStory, h as renderScoreboardMarkdown, i as verifyCodeSurface, j as planEvalFixtureRun, k as loadEvalFixture, l as createProfileMatrixPlan, m as makePlaybackDispatch, n as gitWorktreeAdapter, o as phoenixEvaluatorJudge, p as runProfileMatrix, r as resolveWorktreePath, s as scoreDiscrimination, t as WorktreeAdapterError, u as finalizeProfileMatrix, v as userStoryScoreboard, w as LabeledScenarioStoreError, x as sequentialOptimizationMethod, y as crowdedFrontierParent } from "../campaign-4_ppJW5X.js";
-import { n as paretoPolicy, r as paretoSignificanceGate, t as buildEvidenceVector } from "../promotion-policy-CDMMxzb6.js";
-import { n as sequentialPairedGate, t as sequentialDecide } from "../sequential-DAsyV2T9.js";
+import { a as externalTextOptimizationMethod, i as composeGate, n as gepaOptimizationMethod, o as decodeExternalTextCandidate, r as heldOutGate, s as readExternalOptimizerObservationArtifact, t as skillOptOptimizationMethod, u as createReferenceEquivalenceJudge } from "../skillopt-optimization-method-B2IEaCeF.js";
+import { A as loadEvalFixtureScenarios, C as FsLabeledScenarioStore, D as neutralizationGate, E as rolloutArgumentDiff, M as analyzeCrossSurfaceInteractions, N as buildTraceAnalystSurfaceDispatch, O as discoverEvalFixtures, P as traceAnalystQualityJudge, S as neutralizeText, T as classifyUngroundedLiterals, _ as scoreboardSummary, a as autoevalsScorerJudge, b as scopedOptimizationMethod, c as selectDiscriminative, d as runProfileMatrixSegment, f as ProfileMatrixError, g as scoreUserStory, h as renderScoreboardMarkdown, i as verifyCodeSurface, j as planEvalFixtureRun, k as loadEvalFixture, l as createProfileMatrixPlan, m as makePlaybackDispatch, n as gitWorktreeAdapter, o as phoenixEvaluatorJudge, p as runProfileMatrix, r as resolveWorktreePath, s as scoreDiscrimination, t as WorktreeAdapterError, u as finalizeProfileMatrix, v as userStoryScoreboard, w as LabeledScenarioStoreError, x as sequentialOptimizationMethod, y as crowdedFrontierParent } from "../campaign-B2U7WcFz.js";
+import { n as paretoPolicy, r as paretoSignificanceGate, t as buildEvidenceVector } from "../promotion-policy-BKFrl7CC.js";
+import { n as sequentialPairedGate, t as sequentialDecide } from "../sequential-D_xHrBZe.js";
 export { DEFAULT_EXTERNAL_OPTIMIZER_CALLBACK_LIMITS, DEFAULT_EXTERNAL_OPTIMIZER_PROCESS_LIMITS, FileSearchLedger, FsLabeledScenarioStore, LabeledScenarioStoreError, ProfileMatrixError, SearchHistoryRequiredError, SearchLedgerConflictError, SearchLedgerError, SearchLedgerIntegrityError, SearchRecorder, WorktreeAdapterError, acquireSingleRunLock, analyzeCrossSurfaceInteractions, assertCampaignDesign, assertCampaignSplitIdentity, assertCodeSurfaceIdentity, assertCompleteSearchHistory, assertSearchHistoryMatchesReplay, autoevalsScorerJudge, buildCellSchedule, buildEvidenceVector, buildLoopProvenanceRecord, buildTraceAnalystSurfaceDispatch, campaignBreakdown, campaignMeanComposite, campaignMeasurementDigest, campaignScenarioIdentity, campaignSplitDigest, campaignSplitDigestFromIdentities, canonicalDigest, cellCachePath, classifyUngroundedLiterals, codeSurfaceIdentityMaterial, combineComparisonCosts, compareOptimizationMethods, compareRankKeys, componentSurfaceIdentityMaterial, composeGate, costFromLedgerSummary, createProfileMatrixPlan, createReferenceEquivalenceJudge, createRunCostLedger, createSearchHistoryReceipt, crowdedFrontierParent, decodeExternalTextCandidate, defaultProductionGate, detectScale, dimensionRegressions, discoverEvalFixtures, emitLoopProvenance, externalTextOptimizationMethod, finalizeProfileMatrix, fsCampaignStorage, gepaOptimizationMethod, gitWorktreeAdapter, heldOutGate, heldoutSignificance, inMemoryCampaignStorage, isProposedCandidate, isTransientTransportFailure, labelTrustRank, llmJudge, loadEvalFixture, loadEvalFixtureScenarios, loopProvenanceArgsFromResult, loopProvenanceSpans, makePlaybackDispatch, makeProposalFinding, neutralizationGate, neutralizeText, openAutoPr, openSearchLedger, optimizationTokenUsageFromSummary, pairHoldout, paretoPolicy, paretoSignificanceGate, phoenixEvaluatorJudge, planCampaignRun, planEvalFixtureRun, powerPreflight, provenanceRecordPath, provenanceSpansPath, quotaExhaustedUntil, readCachedCell, readExternalOptimizerObservationArtifact, readGepaCandidatePopulationArtifact, recordCandidatePopulationSearch, renderScoreboardMarkdown, renderSurfaceDiff, replaySearchLedgerText, resolveExternalOptimizerCallbackLimits, resolveExternalOptimizerProcessLimits, resolveRunDir, resolveWorktreePath, rolloutArgumentDiff, runCampaign, runEval, runImprovementLoop, runOptimization, runProfileMatrix, runProfileMatrixSegment, scopedOptimizationMethod, scoreDiscrimination, scoreUserStory, scoreboardSummary, searchHistoryCoverageRow, selectDiscriminative, sequentialDecide, sequentialOptimizationMethod, sequentialPairedGate, skillOptOptimizationMethod, surfaceContentHash, surfaceDispatchRef, surfaceHash, tangleTracesRoot, traceAnalystQualityJudge, transientDispatchFailure, userStoryScoreboard, validateSearchLedgerEvent, verifyCodeSurface, verifyLoopProvenanceRecord, verifySearchHistoryArtifact, verifySearchHistoryReceipt };

package/dist/{campaign-4_ppJW5X.js → campaign-B2U7WcFz.js}

@@ -1,22 +1,22 @@
 import { s as ValidationError, t as AgentEvalError } from "./errors-Dngq5h35.js";
 import { i as compareCodeUnits } from "./canonical-DPyQ_rpt.js";
-import { c as agentProfileId, d as harnessAxisOf, i as verifyCompletion, l as agentProfileModelId, s as agentProfileHash, t as extractProducedState } from "./produced-state-BHboMaab.js";
+import { c as agentProfileId, d as harnessAxisOf, i as verifyCompletion, l as agentProfileModelId, s as agentProfileHash, t as extractProducedState } from "./produced-state-CkoG43gI.js";
 import { s as buildAgentProfileCell } from "./agent-profile-cell-Cv6UA-W_.js";
 import { t as mulberry32 } from "./random-Dn5fPWkt.js";
-import { t as comparePairedArms } from "./paired-arms-D4aeIHUy.js";
+import { t as comparePairedArms } from "./paired-arms-JSDcs2zx.js";
 import { r as pairedBootstrap } from "./paired-tests-C8iCsioC.js";
-import { a as campaignCellToRunRecord, o as projectCampaignCellQuality, t as campaignCellCostProvenance } from "./run-record-Br-Yzt_k.js";
-import { F as assertCodeSurfaceIdentity, G as summarizeBackendIntegrity, K as assertCampaignDesign, W as assertRealBackend, X as campaignScenarioIdentity, at as pairHoldout, z as surfaceContentHash } from "./campaign-evidence-B8oF9xQ6.js";
+import { a as campaignCellToRunRecord, o as projectCampaignCellQuality, t as campaignCellCostProvenance } from "./run-record-DPa2PeHx.js";
+import { F as assertCodeSurfaceIdentity, G as summarizeBackendIntegrity, K as assertCampaignDesign, W as assertRealBackend, X as campaignScenarioIdentity, at as pairHoldout, z as surfaceContentHash } from "./campaign-evidence-PYPUftRy.js";
 import { c as validateRunRecord, i as modelHasSnapshot, n as UNKNOWN_MODEL, s as runTaskScore } from "./run-record-DualPTn2.js";
 import { n as contentHash, t as canonicalJson } from "./verdict-cache-B3eCVQtY.js";
-import { F as paretoFrontierWithCrowding, Q as resolveRunDir, X as runCampaign, Z as planCampaignRun, d as combineComparisonCosts, nt as cellCachePath, rt as computeManifestHash, u as executeOptimizationMethod, v as labelTrustRank } from "./llm-judge-DEFZeSiu.js";
+import { F as paretoFrontierWithCrowding, Q as resolveRunDir, X as runCampaign, Z as planCampaignRun, d as combineComparisonCosts, nt as cellCachePath, rt as computeManifestHash, u as executeOptimizationMethod, v as labelTrustRank } from "./llm-judge-BPFR12VJ.js";
 import { t as CostAccountingIncompleteError } from "./cost-ledger-B1qx30B4.js";
 import { d as mapConcurrent } from "./journal-Cs9f7385.js";
 import { S as fsCampaignStorage } from "./external-optimizer-subprocess-D4dzUBZI.js";
 import { a as scoreAnalystFindings } from "./benchmark-DQKzykkO.js";
 import "./external-optimizer-process-QDRURJAM.js";
-import "./skillopt-optimization-method-CL_0aArC.js";
-import "./sequential-DAsyV2T9.js";
+import "./skillopt-optimization-method-B2IEaCeF.js";
+import "./sequential-D_xHrBZe.js";
 import { createHash } from "node:crypto";
 import { closeSync, constants, existsSync, fstatSync, lstatSync, mkdirSync, mkdtempSync, openSync, readFileSync, readSync, readdirSync, readlinkSync, realpathSync, rmSync, statSync, writeFileSync } from "node:fs";
 import { basename, dirname, isAbsolute, join, relative, resolve, sep } from "node:path";
@@ -3152,4 +3152,4 @@ function resolveWorktreePath(surface, worktreeDir) {
 //#endregion
 export { loadEvalFixtureScenarios as A, FsLabeledScenarioStore as C, neutralizationGate as D, rolloutArgumentDiff as E, analyzeCrossSurfaceInteractions as M, buildTraceAnalystSurfaceDispatch as N, discoverEvalFixtures as O, traceAnalystQualityJudge as P, neutralizeText as S, classifyUngroundedLiterals as T, scoreboardSummary as _, autoevalsScorerJudge as a, scopedOptimizationMethod as b, selectDiscriminative as c, runProfileMatrixSegment as d, ProfileMatrixError as f, scoreUserStory as g, renderScoreboardMarkdown as h, verifyCodeSurface as i, planEvalFixtureRun as j, loadEvalFixture as k, createProfileMatrixPlan as l, makePlaybackDispatch as m, gitWorktreeAdapter as n, phoenixEvaluatorJudge as o, runProfileMatrix as p, resolveWorktreePath as r, scoreDiscrimination as s, WorktreeAdapterError as t, finalizeProfileMatrix as u, userStoryScoreboard as v, LabeledScenarioStoreError as w, sequentialOptimizationMethod as x, crowdedFrontierParent as y };
 
-//# sourceMappingURL=campaign-4_ppJW5X.js.map
+//# sourceMappingURL=campaign-B2U7WcFz.js.map