@tangle-network/agent-eval 0.144.0 → 0.144.1

package/dist/cli.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import { o as runRolloutReleaseCli } from "./hf-dataset-XggBupCr.js";
-import { n as runAnalystBenchmarkCommand } from "./benchmark-command-4c7N_rlw.js";
-import { a as runRpcBatch, o as runRpcOnce, p as handleVersion, r as startServerAsync, s as buildOpenApi } from "./server-Cu4M3NSO.js";
+import { n as runAnalystBenchmarkCommand } from "./benchmark-command-95Zqd1e8.js";
+import { a as runRpcBatch, o as runRpcOnce, p as handleVersion, r as startServerAsync, s as buildOpenApi } from "./server-Df00sdwz.js";
 import { writeFileSync } from "node:fs";
 //#region src/cli-config.ts
 function resolveCliLlmConfig(env = process.env) {

package/dist/contract/index.d.ts

@@ -6,7 +6,7 @@ import { n as buildDefaultAnalystRegistry, t as DefaultAnalystRegistryOptions }
 import { C as JudgeDimension, H as SurfaceProposer, M as OptimizerConfig, R as Scenario, S as JudgeConfig, V as SessionScript, _ as GateDecision, a as CampaignResult, b as GenerationRecord, c as CampaignTraceWriter, d as DispatchContext, f as DispatchFn, g as GateContribution, h as GateContext, i as CampaignCostMeter, j as MutableSurface, k as LabeledScenarioStore, l as CodeSurface, m as GateCheckStatus, n as CampaignArtifactWriter, p as Gate, r as CampaignCellResult, t as CampaignAggregates, v as GateResult, w as JudgeScore, y as GenerationCandidate } from "../types-KEqL1pZc.js";
 import { a as FailureClusterInsight, c as JudgeInsight, d as Recommendation, f as ReleaseSummary, i as FailureClassTally, l as LiftInsight, m as TokenUsageInsight, n as ExecutionErrorOutcomeCell, o as InsightReport, p as ScalarDistribution, r as ExecutionInsight, s as InterRaterInsight, t as CostProvenanceSummary, u as OutcomeCorrelationInsight } from "../insight-report-Bh_8ksel.js";
 import { A as summarizeExecution, D as ExecutionReport, E as AnalyzeRunsOptions, O as SummarizeExecutionOptions, k as analyzeRuns } from "../tool-groups-Bqy4A3QB.js";
-import { A as RunEvalOptions, An as LlmJudgeDimension, B as OptimizerModelBudget, Bt as ComparisonCost, C as RunImprovementLoopOptions, Cn as REFERENCE_EQUIVALENCE_JUDGE_VERSION, D as RunOptimizationOptions, Dn as ReferenceEquivalenceScenario, E as PremeasuredOptimizationBaseline, En as ReferenceEquivalenceJudgeResult, F as GepaOptimizationMethodConfig, Ft as ExternalTextOptimizerContext, G as ObjectiveSource, Gt as OptimizationMethodProvenance, H as AxisVerdict, Ht as OptimizationMethodComparison, I as GepaOptimizationRecipe, It as ExternalTextOptimizerResult, J as PromotionPolicy, K as ParetoSignificanceGateOptions, Kt as OptimizationMethodResult, L as GepaRunnerCommand, Lt as ExternalOptimizationExample, M as GepaAdaptiveEngineRun, Mn as llmJudge, Mt as composeGate, N as GepaEngineOptions, Nt as externalTextOptimizationMethod, On as createReferenceEquivalenceJudge, P as GepaEngineRun, Pt as ExternalTextOptimizationMethodConfig, Qt as compareOptimizationMethods, R as gepaOptimizationMethod, Rt as ExternalTextEvaluationResponse, Sn as REFERENCE_EQUIVALENCE_INPUT_LIMITS, T as runImprovementLoop, Tn as ReferenceEquivalenceJudgeOptions, U as BuildEvidenceVectorOptions, Ut as OptimizationMethodInput, V as AxisEvidence, Vt as OptimizationMethod, W as EvidenceVector, X as paretoPolicy, Xt as OptimizationTokenUsage, Y as buildEvidenceVector, Yt as OptimizationPackageSource, Z as paretoSignificanceGate, _n as campaignSplitDigest, an as RunCampaignOptions, dt as DefaultProductionGateCheck, fn as fsCampaignStorage, ft as DefaultProductionGateOptions, i as skillOptOptimizationMethod, j as runEval, jn as LlmJudgeOptions, kn as runReferenceEquivalenceJudge, lt as HeldOutGateOptions, mt as defaultProductionGate, n as SkillOptRunnerCommand, ot as PowerPreflight, p as LoopProvenanceRecord, pn as inMemoryCampaignStorage, pt as DefaultProductionRewardHackingOptions, q as PromotionObjective, r as SkillOptTrainerConfig, sn as runCampaign, t as SkillOptOptimizationMethodConfig, tn as CampaignCellFailureReceipt, un as CampaignStorage, ut as heldOutGate, w as RunImprovementLoopResult, wn as ReferenceEquivalenceJudgeInput, z as OpenAICompatibleOptimizerModel, zt as CompareOptimizationMethodsOptions } from "../skillopt-optimization-method-B7o01OdX.js";
+import { A as RunEvalOptions, An as LlmJudgeDimension, B as OptimizerModelBudget, Bt as ComparisonCost, C as RunImprovementLoopOptions, Cn as REFERENCE_EQUIVALENCE_JUDGE_VERSION, D as RunOptimizationOptions, Dn as ReferenceEquivalenceScenario, E as PremeasuredOptimizationBaseline, En as ReferenceEquivalenceJudgeResult, F as GepaOptimizationMethodConfig, Ft as ExternalTextOptimizerContext, G as ObjectiveSource, Gt as OptimizationMethodProvenance, H as AxisVerdict, Ht as OptimizationMethodComparison, I as GepaOptimizationRecipe, It as ExternalTextOptimizerResult, J as PromotionPolicy, K as ParetoSignificanceGateOptions, Kt as OptimizationMethodResult, L as GepaRunnerCommand, Lt as ExternalOptimizationExample, M as GepaAdaptiveEngineRun, Mn as llmJudge, Mt as composeGate, N as GepaEngineOptions, Nt as externalTextOptimizationMethod, On as createReferenceEquivalenceJudge, P as GepaEngineRun, Pt as ExternalTextOptimizationMethodConfig, Qt as compareOptimizationMethods, R as gepaOptimizationMethod, Rt as ExternalTextEvaluationResponse, Sn as REFERENCE_EQUIVALENCE_INPUT_LIMITS, T as runImprovementLoop, Tn as ReferenceEquivalenceJudgeOptions, U as BuildEvidenceVectorOptions, Ut as OptimizationMethodInput, V as AxisEvidence, Vt as OptimizationMethod, W as EvidenceVector, X as paretoPolicy, Xt as OptimizationTokenUsage, Y as buildEvidenceVector, Yt as OptimizationPackageSource, Z as paretoSignificanceGate, _n as campaignSplitDigest, an as RunCampaignOptions, dt as DefaultProductionGateCheck, fn as fsCampaignStorage, ft as DefaultProductionGateOptions, i as skillOptOptimizationMethod, j as runEval, jn as LlmJudgeOptions, kn as runReferenceEquivalenceJudge, lt as HeldOutGateOptions, mt as defaultProductionGate, n as SkillOptRunnerCommand, ot as PowerPreflight, p as LoopProvenanceRecord, pn as inMemoryCampaignStorage, pt as DefaultProductionRewardHackingOptions, q as PromotionObjective, r as SkillOptTrainerConfig, sn as runCampaign, t as SkillOptOptimizationMethodConfig, tn as CampaignCellFailureReceipt, un as CampaignStorage, ut as heldOutGate, w as RunImprovementLoopResult, wn as ReferenceEquivalenceJudgeInput, z as OpenAICompatibleOptimizerModel, zt as CompareOptimizationMethodsOptions } from "../skillopt-optimization-method-DNniakI5.js";
 import { c as EvalRunGenerationSnapshot, g as TraceSpanEvent, n as HostedTenant, o as EvalRunCellScore, s as EvalRunEvent } from "../client-Cgl6KasJ.js";
 import { i as InMemoryOutcomeStore, n as FileSystemOutcomeStore, o as OutcomeStore, r as FileSystemOutcomeStoreOptions, t as DeploymentOutcome } from "../outcome-store-BYHIuO0e.js";
 import { AgentCandidateBenchmarkCellRef, AgentCandidateBenchmarkSuiteInputs, AgentCandidateBenchmarkTask, AgentCandidateBenchmarkTaskMaterial, AgentCandidateBundle, AgentCandidateEvaluationPolicy, AgentCandidateExperiment, AgentCandidateExperimentMaterial, AgentCandidateExperimentMeasurement, AgentImprovementCost, AgentImprovementMeasuredComparison, AgentProfileImprovementExperiment, AgentProfileImprovementExperimentMaterial, AgentProfileImprovementMeasuredComparison, AgentProfileImprovementMeasurement, AgentProfileImprovementRunCell, AgentProfileImprovementRunReceipt, AgentProfileImprovementSuiteInputs, AgentProfileImprovementTask, AgentProfileImprovementTaskMaterial, CandidateExecutionEvidence, Sha256Digest } from "@tangle-network/agent-interface";

package/dist/contract/index.js

@@ -1,13 +1,13 @@
 import { c as ValidationError } from "../errors-D-LKuDhb.js";
 import { i as estimateCost, o as isModelPriced } from "../metrics-C9YY1OcL.js";
 import { i as CostLedger } from "../cost-ledger-DMFxsLKr.js";
-import { C as createChatClient, t as buildDefaultAnalystRegistry } from "../default-registry-DfHJEwYh.js";
+import { C as createChatClient, t as buildDefaultAnalystRegistry } from "../default-registry-SOyHB6qG.js";
 import { LLM_MODEL_ATTR_KEYS, SPAN_KIND_ATTR_KEYS } from "../trace-attributes.js";
 import { nt as classifyOtlpSpanRole, rt as isOtlpModelCall } from "../kind-factory-Bvwe3pup.js";
 import { s as makeProposalFinding } from "../usage-receipt-CgxMEBZq.js";
-import { i as inMemoryCampaignStorage, n as createRunCostLedger, r as fsCampaignStorage } from "../single-run-lock-t1si1ob7.js";
+import { i as inMemoryCampaignStorage, n as createRunCostLedger, r as fsCampaignStorage } from "../single-run-lock-B-cFnnNn.js";
 import { m as mapConcurrentRange } from "../ledger-core-Dxz0Rkwa.js";
-import { A as assertOptimizationResult, B as surfaceContentHash, Ct as decidePairedPromotion, M as compareOptimizationMethods, O as composeGate, S as defaultProductionGate, T as heldoutSignificance, V as surfaceHash, X as resolveRunDir, Y as runCampaign, _ as buildEvidenceVector, a as emitLoopProvenance, at as REFERENCE_EQUIVALENCE_JUDGE_VERSION, b as powerPreflight, d as runImprovementLoop, g as gepaOptimizationMethod, gt as llmJudge, it as REFERENCE_EQUIVALENCE_INPUT_LIMITS, k as externalTextOptimizationMethod, o as loopProvenanceArgsFromResult, ot as createReferenceEquivalenceJudge, p as runEval, st as runReferenceEquivalenceJudge, t as skillOptOptimizationMethod, tt as campaignSplitDigest, v as paretoPolicy, x as heldOutGate, y as paretoSignificanceGate } from "../skillopt-optimization-method-Ds8J1_K8.js";
+import { A as assertOptimizationResult, B as surfaceContentHash, Ct as decidePairedPromotion, M as compareOptimizationMethods, O as composeGate, S as defaultProductionGate, T as heldoutSignificance, V as surfaceHash, X as resolveRunDir, Y as runCampaign, _ as buildEvidenceVector, a as emitLoopProvenance, at as REFERENCE_EQUIVALENCE_JUDGE_VERSION, b as powerPreflight, d as runImprovementLoop, g as gepaOptimizationMethod, gt as llmJudge, it as REFERENCE_EQUIVALENCE_INPUT_LIMITS, k as externalTextOptimizationMethod, o as loopProvenanceArgsFromResult, ot as createReferenceEquivalenceJudge, p as runEval, st as runReferenceEquivalenceJudge, t as skillOptOptimizationMethod, tt as campaignSplitDigest, v as paretoPolicy, x as heldOutGate, y as paretoSignificanceGate } from "../skillopt-optimization-method-DA5mJmFE.js";
 import { E as pairedBootstrap } from "../statistics-ByxzSiOM.js";
 import { i as parseRunRecordSafe, r as modelHasSnapshot } from "../run-record-vRgqWmJw.js";
 import { a as recordAggregateMeasurements, i as readTaskFailureLabels, o as summarizeExecutionMeasurements, s as summarizeTraceErrors, t as extractUsage } from "../extract-usage-7l1Xq5ti.js";

package/dist/{default-registry-DfHJEwYh.js → default-registry-SOyHB6qG.js}

@@ -1,4 +1,4 @@
-import { n as LlmClient } from "./llm-client-B3WXSH5Y.js";
+import { n as LlmClient } from "./llm-client-D3EoChAU.js";
 import { LLM_CONTEXT_TOKENS, LLM_INPUT_TOKEN_ATTR_KEYS, LLM_OUTPUT_TOKEN_ATTR_KEYS, TOOL_NAME_ATTR_KEYS } from "./trace-attributes.js";
 import { t as executionTrackByLane } from "./execution-tracks-CpgFPpS5.js";
 import { $ as spanEpochMillis, H as snapshotExactExecutionPlan, R as findingSubjectGrammarPromptFor, U as deepFreezeCanonicalJson, V as snapshotExactExecutionComponentIdentity, t as createTraceAnalyst } from "./kind-factory-Bvwe3pup.js";
@@ -2449,4 +2449,4 @@ function buildDefaultAnalystRegistry(options = {}) {
 //#endregion
 export { createChatClient as C, computeTraceMetrics as S, CONTROL_INTEGRITY_ANALYST as _, analystFindingDigest as a, behavioralAnalyst as b, completedAnalystReviewQuality as c, validateAnalystReviewDecisions as d, DEFAULT_TRACE_ANALYST_KINDS as f, FAILURE_MODE_KIND_SPEC as g, IMPROVEMENT_KIND_SPEC as h, assertExactRegistryRunOpts as i, readAnalystReview as l, KNOWLEDGE_GAP_KIND_SPEC as m, AnalystRegistry as n, analystRunDigest as o, KNOWLEDGE_POISONING_KIND_SPEC as p, ExactAnalystRunExecutionError as r, assertUniqueFindingIds as s, buildDefaultAnalystRegistry as t, snapshotAnalystRun as u, ControlIntegrityAnalyst as v, deriveEfficiencyFindings as x, emitControlIntegrityFindings as y };
 
-//# sourceMappingURL=default-registry-DfHJEwYh.js.map
+//# sourceMappingURL=default-registry-SOyHB6qG.js.map