@tangle-network/agent-eval 0.145.8 → 0.145.10

package/dist/benchmarks/index.js

@@ -1,7 +1,7 @@
 import { t as __exportAll } from "../rolldown-runtime-8H4AJuhK.js";
 import { K as runCampaign } from "../llm-judge-DhiJqSjB.js";
 import { x as fsCampaignStorage } from "../external-optimizer-subprocess-BhKYK0Jv.js";
-import "../campaign-Dg2B4W6d.js";
+import "../campaign-B3GJpJ4h.js";
 import { join } from "node:path";
 //#region src/benchmarks/calibration.ts
 async function calibrateBenchmarkMetric(options) {

package/dist/campaign/index.js

@@ -4,7 +4,7 @@ import { a as heldoutSignificance, i as dimensionRegressions, o as pairHoldout,
 import { r as makeProposalFinding } from "../types-BI4fT3HN.js";
 import { n as acquireSingleRunLock } from "../external-optimizer-process-BRE56woM.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-C6JSfYxT.js";
-import { A as neutralizationGate, C as scoreboardSummary, D as LabeledScenarioStoreError, E as FsLabeledScenarioStore, F as analyzeCrossSurfaceInteractions, I as buildTraceAnalystSurfaceDispatch, L as traceAnalystQualityJudge, M as loadEvalFixture, N as loadEvalFixtureScenarios, O as classifyUngroundedLiterals, P as planEvalFixtureRun, S as scoreUserStory, T as neutralizeText, _ as runProfileMatrixSegment, a as autoevalsScorerJudge, b as makePlaybackDispatch, c as FileSearchLedger, d as validateSearchLedgerEvent, f as scoreDiscrimination, g as finalizeProfileMatrix, h as createProfileMatrixPlan, i as verifyCodeSurface, j as discoverEvalFixtures, k as rolloutArgumentDiff, l as SEARCH_LEDGER_SCHEMA, m as SegmentedProfileMatrixError, n as gitWorktreeAdapter, o as phoenixEvaluatorJudge, p as selectDiscriminative, r as resolveWorktreePath, s as isTransientTransportFailure, t as WorktreeAdapterError, u as openSearchLedger, v as ProfileMatrixError, w as userStoryScoreboard, x as renderScoreboardMarkdown, y as runProfileMatrix } from "../campaign-Dg2B4W6d.js";
+import { A as neutralizationGate, C as scoreboardSummary, D as LabeledScenarioStoreError, E as FsLabeledScenarioStore, F as analyzeCrossSurfaceInteractions, I as buildTraceAnalystSurfaceDispatch, L as traceAnalystQualityJudge, M as loadEvalFixture, N as loadEvalFixtureScenarios, O as classifyUngroundedLiterals, P as planEvalFixtureRun, S as scoreUserStory, T as neutralizeText, _ as runProfileMatrixSegment, a as autoevalsScorerJudge, b as makePlaybackDispatch, c as FileSearchLedger, d as validateSearchLedgerEvent, f as scoreDiscrimination, g as finalizeProfileMatrix, h as createProfileMatrixPlan, i as verifyCodeSurface, j as discoverEvalFixtures, k as rolloutArgumentDiff, l as SEARCH_LEDGER_SCHEMA, m as SegmentedProfileMatrixError, n as gitWorktreeAdapter, o as phoenixEvaluatorJudge, p as selectDiscriminative, r as resolveWorktreePath, s as isTransientTransportFailure, t as WorktreeAdapterError, u as openSearchLedger, v as ProfileMatrixError, w as userStoryScoreboard, x as renderScoreboardMarkdown, y as runProfileMatrix } from "../campaign-B3GJpJ4h.js";
 import { n as paretoPolicy, r as paretoSignificanceGate, t as buildEvidenceVector } from "../promotion-policy-xzA40Evo.js";
 import { n as sequentialPairedGate, t as sequentialDecide } from "../sequential-C458DXNf.js";
 export { DEFAULT_EXTERNAL_OPTIMIZER_CALLBACK_LIMITS, DEFAULT_EXTERNAL_OPTIMIZER_PROCESS_LIMITS, FileSearchLedger, FsLabeledScenarioStore, LabeledScenarioStoreError, ProfileMatrixError, SEARCH_LEDGER_SCHEMA, SearchLedgerConflictError, SearchLedgerError, SearchLedgerIntegrityError, SegmentedProfileMatrixError, WorktreeAdapterError, acquireSingleRunLock, analyzeCrossSurfaceInteractions, assertCampaignDesign, assertCampaignSplitIdentity, assertCodeSurfaceIdentity, assertComponentSurface, autoevalsScorerJudge, buildEvidenceVector, buildLoopProvenanceRecord, buildTraceAnalystSurfaceDispatch, campaignBreakdown, campaignMeanComposite, campaignMeasurementDigest, campaignScenarioIdentity, campaignSplitDigest, campaignSplitDigestFromIdentities, canonicalDigest, cellCachePath, classifyUngroundedLiterals, codeSurfaceIdentityMaterial, combineComparisonCosts, compareOptimizationMethods, compareRankKeys, componentSurfaceIdentityMaterial, composeGate, costFromLedgerSummary, createProfileMatrixPlan, createReferenceEquivalenceJudge, createRunCostLedger, 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, readCachedCell, readExternalOptimizerObservationArtifact, readGepaCandidatePopulationArtifact, renderScoreboardMarkdown, renderSurfaceDiff, resolveExternalOptimizerCallbackLimits, resolveExternalOptimizerProcessLimits, resolveRunDir, resolveWorktreePath, rolloutArgumentDiff, runCampaign, runEval, runImprovementLoop, runOptimization, runProfileMatrix, runProfileMatrixSegment, scoreDiscrimination, scoreUserStory, scoreboardSummary, selectDiscriminative, sequentialDecide, sequentialPairedGate, skillOptOptimizationMethod, surfaceContentHash, surfaceHash, tangleTracesRoot, traceAnalystQualityJudge, userStoryScoreboard, validateSearchLedgerEvent, verifyCodeSurface, verifyLoopProvenanceRecord };

package/dist/{campaign-Dg2B4W6d.js → campaign-B3GJpJ4h.js}

@@ -1542,10 +1542,13 @@ function receiptModels(cell) {
 	const reported = cell.resolvedModels ?? (cell.resolvedModel ? [cell.resolvedModel] : []);
 	return [...new Set(reported.map((model) => model.trim()).filter(Boolean))];
 }
+function isUnknownModelOnly(reported) {
+	return reported.length > 0 && reported.every((model) => model === "unknown");
+}
 /** Resolve and validate every paid-call model used by one cell. */
 function recordModel(cell, profileId, declaredModel) {
 	const reported = receiptModels(cell);
-	if (reported.length === 0 && cell.error !== void 0) return UNKNOWN_MODEL;
+	if (cell.error !== void 0 && (reported.length === 0 || isUnknownModelOnly(reported))) return UNKNOWN_MODEL;
 	if (modelHasSnapshot(declaredModel)) {
 		for (const model of reported) if (model !== declaredModel) throw new ProfileMatrixError(`profile '${profileId}' paid-call model '${model}' for cell '${cell.cellId}' does not match its declared exact model '${declaredModel}'`);
 		return declaredModel;
@@ -1595,7 +1598,7 @@ function buildRunRecord(args) {
 */
 function durableCellForRecord(cell) {
 	if (cell.error === void 0) return cell;
-	const hasSettledAgentCalls = receiptModels(cell).length > 0;
+	const hasSettledAgentCalls = receiptModels(cell).some((model) => model !== UNKNOWN_MODEL);
 	return {
 		...cell,
 		tokenUsage: cell.tokenUsage.tokensKnown === false || hasSettledAgentCalls ? cell.tokenUsage : {
@@ -3682,4 +3685,4 @@ function resolveWorktreePath(surface, worktreeDir) {
 //#endregion
 export { neutralizationGate as A, scoreboardSummary as C, LabeledScenarioStoreError as D, FsLabeledScenarioStore as E, analyzeCrossSurfaceInteractions as F, buildTraceAnalystSurfaceDispatch as I, traceAnalystQualityJudge as L, loadEvalFixture as M, loadEvalFixtureScenarios as N, classifyUngroundedLiterals as O, planEvalFixtureRun as P, scoreUserStory as S, neutralizeText as T, runProfileMatrixSegment as _, autoevalsScorerJudge as a, makePlaybackDispatch as b, FileSearchLedger as c, validateSearchLedgerEvent as d, scoreDiscrimination as f, finalizeProfileMatrix as g, createProfileMatrixPlan as h, verifyCodeSurface as i, discoverEvalFixtures as j, rolloutArgumentDiff as k, SEARCH_LEDGER_SCHEMA as l, SegmentedProfileMatrixError as m, gitWorktreeAdapter as n, phoenixEvaluatorJudge as o, selectDiscriminative as p, resolveWorktreePath as r, isTransientTransportFailure as s, WorktreeAdapterError as t, openSearchLedger as u, ProfileMatrixError as v, userStoryScoreboard as w, renderScoreboardMarkdown as x, runProfileMatrix as y };
 
-//# sourceMappingURL=campaign-Dg2B4W6d.js.map
+//# sourceMappingURL=campaign-B3GJpJ4h.js.map