@tangle-network/agent-eval 0.102.1 → 0.103.0

package/dist/index.d.ts

@@ -18,8 +18,8 @@ import { A as AgentEvalError, J as JudgeError, a as ConfigError } from './errors
 export { b as AgentEvalErrorCode, C as CaptureIntegrityError, N as NotFoundError, R as ReplayError, V as ValidationError, c as VerificationError } from './errors-CzMUYo7b.js';
 import { b as FeedbackLabel, F as FeedbackTrajectoryStore, a as FeedbackTrajectory } from './feedback-trajectory-BxY0cKfs.js';
 export { c as FeedbackArtifactType, d as FeedbackAttempt, e as FeedbackLabelKind, f as FeedbackLabelSource, g as FeedbackOptimizerRow, h as FeedbackOutcome, i as FeedbackReplayAdapter, j as FeedbackReplayResult, k as FeedbackSeverity, l as FeedbackSplitPolicy, m as FeedbackTask, n as FeedbackTrajectoryFilter, o as FileSystemFeedbackTrajectoryStore, I as InMemoryFeedbackTrajectoryStore, P as PreferenceMemoryEntry, p as ProposedSideEffect, q as assignFeedbackSplit, r as controlRunToFeedbackTrajectory, s as createFeedbackTrajectory, t as feedbackTrajectoriesToDatasetScenarios, u as feedbackTrajectoriesToOptimizerRows, v as feedbackTrajectoryToDatasetScenario, w as feedbackTrajectoryToOptimizerRow, x as parseFeedbackTrajectoriesJsonl, y as renderPreferenceMemoryMarkdown, z as replayFeedbackTrajectories, A as replayFeedbackTrajectory, B as serializeFeedbackTrajectoriesJsonl, C as summarizePreferenceMemory, D as withAssignedFeedbackSplit } from './feedback-trajectory-BxY0cKfs.js';
-import { b as CorrectnessChecker } from './pre-registration-Dzg61IQA.js';
-export { A as ArtifactCheckArtifact, d as ArtifactEventLike, e as ArtifactValidator, f as BackendIntegrityError, B as BackendIntegrityReport, C as CompletionRequirement, a as CompletionVerdict, H as HypothesisManifest, g as HypothesisResult, h as LlmCorrectnessCheckerOpts, L as LlmJudgeDimension, c as LlmJudgeOptions, i as ProducedProposal, P as ProducedState, j as ProposalEventLike, k as RequirementCheck, R as RuntimeEventLike, m as SatisfiedBy, S as SignedManifest, n as SignedManifestAlgo, T as TaskGold, o as ToolCallEventLike, V as ValidationContext, p as ValidationIssue, q as ValidationResult, r as assertRealBackend, s as byteLengthRange, t as canonicalize, u as completionVerdict, v as composeValidators, w as containsAll, x as createLlmCorrectnessChecker, y as createTokenRecallChecker, z as evaluateHypothesis, D as extractProducedState, E as hashJson, F as jsonHasKeys, l as llmJudge, G as parseCorrectnessResponse, I as regexMatch, J as signManifest, K as summarizeBackendIntegrity, M as verifyCompletion, N as verifyManifest } from './pre-registration-Dzg61IQA.js';
+import { b as CorrectnessChecker } from './pre-registration-mWG2w8d-.js';
+export { A as ArtifactCheckArtifact, d as ArtifactEventLike, e as ArtifactValidator, f as BackendIntegrityError, B as BackendIntegrityReport, C as CompletionRequirement, a as CompletionVerdict, H as HypothesisManifest, g as HypothesisResult, h as LlmCorrectnessCheckerOpts, L as LlmJudgeDimension, c as LlmJudgeOptions, i as ProducedProposal, P as ProducedState, j as ProposalEventLike, k as RequirementCheck, R as RuntimeEventLike, m as SatisfiedBy, S as SignedManifest, n as SignedManifestAlgo, T as TaskGold, o as ToolCallEventLike, V as ValidationContext, p as ValidationIssue, q as ValidationResult, r as assertRealBackend, s as byteLengthRange, t as canonicalize, u as completionVerdict, v as composeValidators, w as containsAll, x as createLlmCorrectnessChecker, y as createTokenRecallChecker, z as evaluateHypothesis, D as extractProducedState, E as hashJson, F as jsonHasKeys, l as llmJudge, G as parseCorrectnessResponse, I as regexMatch, J as signManifest, K as summarizeBackendIntegrity, M as verifyCompletion, N as verifyManifest } from './pre-registration-mWG2w8d-.js';
 export { DataAcquisitionPlan, KnowledgeAcquisitionMode, KnowledgeBundle, KnowledgeFallbackPolicy, KnowledgeFreshness, KnowledgeImportance, KnowledgeReadinessReport, KnowledgeRecommendedAction, KnowledgeRequirement, KnowledgeRequirementCategory, KnowledgeResponsibleSurface, KnowledgeSensitivity, ScoreKnowledgeReadinessOptions, UserQuestion, acquisitionPlansForKnowledgeGaps, blockingKnowledgeEval, knowledgeReadinessTracePayload, scoreKnowledgeReadiness, userQuestionsForKnowledgeGaps } from './knowledge/index.js';
 import { h as ReleaseConfidenceThresholds, f as ReleaseConfidenceScorecard } from './release-report-BQ1Ziyu-.js';
 export { A as ActionableSideInfo, o as AsiSeverity, B as BootstrapOptions, a as BootstrapResult, J as JudgeReplayGateArgs, R as ReleaseConfidenceAxis, b as ReleaseConfidenceAxisName, c as ReleaseConfidenceInput, d as ReleaseConfidenceIssue, e as ReleaseConfidenceMetrics, g as ReleaseConfidenceStatus, i as ReleaseTraceEvidence, j as RenderReleaseReportOptions, V as Verdict, k as assertReleaseConfidence, l as bootstrapCi, m as evaluateReleaseConfidence, n as judgeReplayGate, r as renderReleaseReport } from './release-report-BQ1Ziyu-.js';
@@ -77,7 +77,7 @@ export { L as LockedJsonlAppender } from './testing-C21CHsq2.js';
 export { I as InterimReleaseConfidence, a as InterimReleaseConfidenceInput, P as PairedEvalueOptions, b as PairedEvalueSequence, c as PairedEvalueStep, S as SequentialDecision, e as evaluateInterimReleaseConfidence, p as pairedEvalueSequence } from './sequential-5iSVfzl2.js';
 import { j as GepaProposerConstraints, b as RunImprovementLoopResult } from './gepa-DeyPTlvx.js';
 export { IntegrityResult, IntegrityViolation, JourneySpec, PerfBaseline, PerfGateResult, PerfRegression, PerfScenario, PerfStat, ScenarioAxes, assertRecordIntegrity, checkRecordIntegrity, expandMatrix, gatePerf, scenarioKey, summarizeRecords } from './perf/index.js';
-export { AgentProfileRuntimeReceipt, ProductBenchmarkArm, ProductBenchmarkArtifactPaths, ProductBenchmarkBudgets, ProductBenchmarkManifest, ProductBenchmarkProfileRef, ProductBenchmarkRecord, ProductBenchmarkRepoRef, ProductBenchmarkRunInput, ProductBenchmarkScenario, ProductBenchmarkSplit, ProductBenchmarkSubstrateVersions, ProductBenchmarkValidationReport, RuntimeResolution, findProductBenchmarkArtifacts, productBenchmarkIntegrityFailures, productBenchmarkSplits, readProductBenchmarkManifest, readProductBenchmarkRecords, validateProductBenchmarkManifest, validateProductBenchmarkRecord, validateProductBenchmarkRun } from './product-benchmark/index.js';
+export { AgentProfileRuntimeReceipt, ProductBenchmarkArm, ProductBenchmarkArtifactPaths, ProductBenchmarkBudgets, ProductBenchmarkExportOptions, ProductBenchmarkExportResult, ProductBenchmarkManifest, ProductBenchmarkProfileRef, ProductBenchmarkRecord, ProductBenchmarkRepoRef, ProductBenchmarkRunInput, ProductBenchmarkScenario, ProductBenchmarkSingleRunExportOptions, ProductBenchmarkSplit, ProductBenchmarkSubstrateVersions, ProductBenchmarkValidationReport, RuntimeResolution, assertProductBenchmarkRun, buildProductBenchmarkManifest, exportProductBenchmark, exportProductBenchmarkRuns, findProductBenchmarkArtifacts, productBenchmarkIntegrityFailures, productBenchmarkMutableSurfaces, productBenchmarkRepoIdentity, productBenchmarkSplits, readProductBenchmarkManifest, readProductBenchmarkRecords, runRecordToProductBenchmarkRecord, validateProductBenchmarkManifest, validateProductBenchmarkRecord, validateProductBenchmarkRun } from './product-benchmark/index.js';
 import '@ax-llm/ax';
 import 'zod';
 import './insight-report-k0sRTzKg.js';
@@ -5271,6 +5271,10 @@ interface ReflectionProposal {
     rationale: string;
     payload: unknown;
 }
+/**
+ * Parse the model's JSON response back into proposals. Tolerates markdown
+ * fences and surrounding prose. Returns at most `maxProposals`.
+ */
 declare function parseReflectionResponse(raw: string, maxProposals?: number): ReflectionProposal[];
 
 /**

package/dist/index.js

@@ -9,7 +9,7 @@ import {
   checkBehavioralCanary,
   checkCanaries,
   runBehavioralCanaries
-} from "./chunk-6Q2DYRWV.js";
+} from "./chunk-U3IDYATS.js";
 import {
   classifyEuAiRisk,
   euAiActReport,
@@ -34,15 +34,22 @@ import {
   summarizeRecords
 } from "./chunk-STGVSCDH.js";
 import {
+  assertProductBenchmarkRun,
+  buildProductBenchmarkManifest,
+  exportProductBenchmark,
+  exportProductBenchmarkRuns,
   findProductBenchmarkArtifacts,
   productBenchmarkIntegrityFailures,
+  productBenchmarkMutableSurfaces,
+  productBenchmarkRepoIdentity,
   productBenchmarkSplits,
   readProductBenchmarkManifest,
   readProductBenchmarkRecords,
+  runRecordToProductBenchmarkRecord,
   validateProductBenchmarkManifest,
   validateProductBenchmarkRecord,
   validateProductBenchmarkRun
-} from "./chunk-63MBSQTX.js";
+} from "./chunk-RQNOLV3I.js";
 import {
   CODING_HARNESSES,
   HARNESS_NATIVE_MODEL,
@@ -65,7 +72,7 @@ import {
   llmJudge,
   parseCorrectnessResponse,
   verifyCompletion
-} from "./chunk-CMJSTXUR.js";
+} from "./chunk-IXOV77YF.js";
 import {
   DEFAULT_MUTATION_PRIMITIVES,
   DEFAULT_RED_TEAM_CORPUS,
@@ -89,7 +96,7 @@ import {
   scoreRedTeamOutput,
   surfaceContentHash,
   toolNamesForRun
-} from "./chunk-Z3FLN24V.js";
+} from "./chunk-NTVWIH24.js";
 import {
   BackendIntegrityError,
   assertRealBackend,
@@ -99,7 +106,7 @@ import {
   fileVerdictCache,
   inMemoryVerdictCache,
   summarizeBackendIntegrity
-} from "./chunk-QIT2XZ4E.js";
+} from "./chunk-XKA6ZGEY.js";
 import {
   MODEL_PRICING,
   MetricsCollector,
@@ -5490,7 +5497,7 @@ function matchText(actual, matcher) {
 }
 function resolveToolName(span) {
   if (typeof span.toolName === "string") return span.toolName;
-  const fromAttr = span.attributes?.["tool.name"] ?? span.attributes?.["toolName"];
+  const fromAttr = span.attributes?.["tool.name"] ?? span.attributes?.toolName;
   if (typeof fromAttr === "string") return fromAttr;
   if (span.kind === "tool" && typeof span.name === "string") return span.name;
   return void 0;
@@ -10144,6 +10151,7 @@ export {
   assertLlmRoute,
   assertModelsServed,
   assertNoHiddenLeak,
+  assertProductBenchmarkRun,
   assertRealBackend,
   assertRecordIntegrity,
   assertReleaseConfidence,
@@ -10169,6 +10177,7 @@ export {
   buildAgreementJudge,
   buildDefaultAnalystRegistry,
   buildDriverSystemPrompt,
+  buildProductBenchmarkManifest,
   buildReflectionPrompt,
   buildReviewerPrompt,
   buildTraceAnalystTools,
@@ -10283,6 +10292,8 @@ export {
   expandMatrix,
   expandProfileAxes,
   expectAgent,
+  exportProductBenchmark,
+  exportProductBenchmarkRuns,
   exportRewardModel,
   exportRunAsOtlp,
   extractAssetUrls,
@@ -10437,6 +10448,8 @@ export {
   printDriverSummary,
   probeLlm,
   productBenchmarkIntegrityFailures,
+  productBenchmarkMutableSurfaces,
+  productBenchmarkRepoIdentity,
   productBenchmarkSplits,
   profile_exports as profile,
   projectOtlpFlatLine,
@@ -10498,6 +10511,7 @@ export {
   runLiveProof,
   runProposeReview,
   runProposeReviewAsControlLoop,
+  runRecordToProductBenchmarkRecord,
   runReferenceReplay,
   runScore,
   runSelfPlay,