@tangle-network/agent-eval 0.27.0 → 0.28.0

package/dist/{integrity-DK2EBVZC.d.ts → integrity-BAxLGJ9I.d.ts}

@@ -1,5 +1,5 @@
 import { C as CaptureIntegrityError } from './errors-BZ9sTdz7.js';
-import { T as TraceStore } from './store-Db2Bv8Cf.js';
+import { T as TraceStore } from './store-BP5be6s7.js';
 
 /**
  * RawProviderSink — first-class persistence for the actual HTTP-level
@@ -208,4 +208,4 @@ declare function assertRunCaptured(store: TraceStore, runId: string, expectation
 /** Strict mode: throws `RunIntegrityError` when the report isn't ok. */
 declare function throwIfRunIncomplete(report: RunIntegrityReport): void;
 
-export { FileSystemRawProviderSink as F, InMemoryRawProviderSink as I, NoopRawProviderSink as N, type ProviderRedactor as P, type RawProviderSink as R, type RunIntegrityExpectations as a, type RunIntegrityReport as b, type RawProviderEvent as c, type FileSystemRawProviderSinkOptions as d, type InMemoryRawProviderSinkOptions as e, type RawProviderDirection as f, type RawProviderSinkFilter as g, RunIntegrityError as h, type RunIntegrityIssue as i, type RunIntegrityIssueCode as j, assertRunCaptured as k, defaultProviderRedactor as l, providerFromBaseUrl as p, throwIfRunIncomplete as t };
+export { FileSystemRawProviderSink as F, InMemoryRawProviderSink as I, NoopRawProviderSink as N, type ProviderRedactor as P, type RawProviderSink as R, type RunIntegrityExpectations as a, type RunIntegrityReport as b, type FileSystemRawProviderSinkOptions as c, type InMemoryRawProviderSinkOptions as d, type RawProviderDirection as e, type RawProviderEvent as f, type RawProviderSinkFilter as g, RunIntegrityError as h, type RunIntegrityIssue as i, type RunIntegrityIssueCode as j, assertRunCaptured as k, defaultProviderRedactor as l, providerFromBaseUrl as p, throwIfRunIncomplete as t };

package/dist/knowledge/index.d.ts

@@ -1,6 +1,6 @@
-import { j as ControlSeverity, C as ControlEvalResult } from '../control-runtime-BuJHoLg0.js';
-import { T as TraceEmitter } from '../emitter-DP_cSSiw.js';
-import '../store-Db2Bv8Cf.js';
+import { j as ControlSeverity, C as ControlEvalResult } from '../control-runtime-BRdQ0wrx.js';
+import { T as TraceEmitter } from '../emitter-BqjeOvJh.js';
+import '../store-BP5be6s7.js';
 
 type KnowledgeRequirementCategory = 'user_specific' | 'company_specific' | 'domain_specific' | 'codebase_specific' | 'market_specific' | 'regulatory' | 'tool_api' | 'credential_or_secret' | 'runtime_environment' | 'preference' | 'historical_context';
 type KnowledgeAcquisitionMode = 'ask_user' | 'search_web' | 'query_connector' | 'inspect_repo' | 'run_command' | 'infer_low_confidence' | 'not_available';

package/dist/knowledge/index.js

@@ -4,8 +4,8 @@ import {
   knowledgeReadinessTracePayload,
   scoreKnowledgeReadiness,
   userQuestionsForKnowledgeGaps
-} from "../chunk-WWYCWKUM.js";
-import "../chunk-LSH4MMOZ.js";
+} from "../chunk-3CKU6VGU.js";
+import "../chunk-NCRFYPS3.js";
 import "../chunk-TVVP3ZZQ.js";
 import "../chunk-PZ5AY32C.js";
 export {

package/dist/meta-eval/index.d.ts

@@ -1,4 +1,4 @@
-import { R as Run, T as TraceStore } from '../store-Db2Bv8Cf.js';
+import { R as Run, T as TraceStore } from '../store-BP5be6s7.js';
 import { a as OutcomeFilter, O as OutcomeStore } from '../outcome-store-D6KWmYvj.js';
 export { D as DeploymentOutcome, F as FileSystemOutcomeStore, b as FileSystemOutcomeStoreOptions, I as InMemoryOutcomeStore } from '../outcome-store-D6KWmYvj.js';
 export { R as RubricOutcomePair, a as RubricPredictiveValidityInput, b as RubricPredictiveValidityReport, c as RubricRanking, r as rubricPredictiveValidity } from '../rubric-predictive-validity-C0uDYwG6.js';

package/dist/{multi-layer-verifier-LkP3LVKj.d.ts → multi-layer-verifier-BNi4-8lR.d.ts}

@@ -51,7 +51,7 @@ interface LayerResult {
      * diagnostic name; null = "diagnostic not applicable / not measured."
      * Renderers that know the keys can display them; ones that don't,
      * ignore. Free-form on purpose — consumers type the value shape in
-     * their own namespace. Added in 0.10.
+     * their own namespace.
      */
     diagnostics?: Record<string, number | null>;
     /** Any rich per-layer detail — rendered as-is by consumers that know the layer. */
@@ -138,4 +138,4 @@ declare class MultiLayerVerifier<Env = unknown> {
     run(opts: VerifyOptions<Env>): Promise<VerificationReport>;
 }
 
-export { type Finding as F, type Layer as L, MultiLayerVerifier as M, type Severity as S, type VerificationReport as V, type LayerResult as a, type VerifyContext as b, type LayerStatus as c, type VerifyOptions as d, gradeSemanticStatus as g };
+export { type Finding as F, type Layer as L, MultiLayerVerifier as M, type Severity as S, type VerificationReport as V, type VerifyOptions as a, type LayerResult as b, type VerifyContext as c, type LayerStatus as d, gradeSemanticStatus as g };

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.1.0",
   "info": {
     "title": "@tangle-network/agent-eval — wire protocol",
-    "version": "0.27.0",
+    "version": "0.28.0",
     "description": "HTTP and stdio RPC interface to agent-eval. The TypeScript runtime is the source of truth; this spec is the contract that cross-language clients (Python, Rust, Go) generate from.\n\nWire-protocol version: 1.0.0. Bumps on breaking changes to request/response schemas.",
     "contact": {
       "name": "Tangle Network",

package/dist/optimization.d.ts

@@ -1,11 +1,11 @@
-export { C as CallbackResearcher, a as CallbackResearcherOptions, b as CampaignFactoryParams, c as CampaignIntegrityPolicy, d as CampaignRunContext, e as CampaignRunOutcome, f as CampaignRunner, g as CampaignScenario, h as CampaignVariant, E as EvalCampaignOptions, i as EvalCampaignResult, j as ExperimentPlan, k as ExperimentResult, F as FailedRun, l as FailureMode, N as NoopResearcher, R as Researcher, S as SteeringChange, r as runEvalCampaign } from './researcher-bGkI7vCl.js';
-export { F as FeedbackArtifactType, a as FeedbackAttempt, b as FeedbackLabel, c as FeedbackLabelKind, d as FeedbackLabelSource, e as FeedbackOptimizerRow, f as FeedbackOutcome, g as FeedbackReplayAdapter, h as FeedbackReplayResult, i as FeedbackSeverity, j as FeedbackSplitPolicy, k as FeedbackTask, l as FeedbackTrajectory, m as FeedbackTrajectoryFilter, n as FeedbackTrajectoryStore, 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-DfFdrraJ.js';
-export { A as ActionableSideInfo, a as AsiSeverity, D as DEFAULT_MUTATION_PRIMITIVES, E as EvolvableVariant, G as GenerationReport, I as InMemoryTrialCache, M as MultiShotGateConfig, b as MultiShotGateResult, c as MultiShotMutateAdapter, d as MultiShotOptimizationConfig, e as MultiShotOptimizationResult, f as MultiShotRun, g as MultiShotRunInput, h as MultiShotRunner, i as MultiShotScore, j as MultiShotScorer, k as MultiShotSplit, l as MultiShotTrace, m as MultiShotTrialResult, n as MultiShotVariant, o as MutateAdapter, P as PromptEvolutionConfig, p as PromptEvolutionEvent, q as PromptEvolutionResult, R as ReflectionContext, r as ReflectionProposal, S as ScenarioAggregate, s as ScoreAdapter, T as TrialCache, t as TrialResult, u as TrialTrace, V as VariantAggregate, v as buildReflectionPrompt, w as defaultMultiShotObjectives, x as parseReflectionResponse, y as runMultiShotOptimization, z as runPromptEvolution, B as trialTraceFromMultiShotTrial } from './summary-report-DZVXOCK_.js';
+export { C as CallbackResearcher, a as CallbackResearcherOptions, b as CampaignFactoryParams, c as CampaignIntegrityPolicy, d as CampaignRunContext, e as CampaignRunOutcome, f as CampaignRunner, g as CampaignScenario, h as CampaignVariant, E as EvalCampaignOptions, i as EvalCampaignResult, j as ExperimentPlan, k as ExperimentResult, F as FailedRun, l as FailureMode, N as NoopResearcher, R as Researcher, S as SteeringChange, r as runEvalCampaign } from './researcher-ClDX3KZx.js';
+export { F as FeedbackArtifactType, a as FeedbackAttempt, b as FeedbackLabel, c as FeedbackLabelKind, d as FeedbackLabelSource, e as FeedbackOptimizerRow, f as FeedbackOutcome, g as FeedbackReplayAdapter, h as FeedbackReplayResult, i as FeedbackSeverity, j as FeedbackSplitPolicy, k as FeedbackTask, l as FeedbackTrajectory, m as FeedbackTrajectoryFilter, n as FeedbackTrajectoryStore, 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-j0nJFgC6.js';
+export { A as ActionableSideInfo, a as AsiSeverity, D as DEFAULT_MUTATION_PRIMITIVES, E as EvolvableVariant, G as GenerationReport, I as InMemoryTrialCache, M as MultiShotGateConfig, b as MultiShotGateResult, c as MultiShotMutateAdapter, d as MultiShotOptimizationConfig, e as MultiShotOptimizationResult, f as MultiShotRun, g as MultiShotRunInput, h as MultiShotRunner, i as MultiShotScore, j as MultiShotScorer, k as MultiShotSplit, l as MultiShotTrace, m as MultiShotTrialResult, n as MultiShotVariant, o as MutateAdapter, P as PromptEvolutionConfig, p as PromptEvolutionEvent, q as PromptEvolutionResult, R as ReflectionContext, r as ReflectionProposal, S as ScenarioAggregate, s as ScoreAdapter, T as TrialCache, t as TrialResult, u as TrialTrace, V as VariantAggregate, v as buildReflectionPrompt, w as defaultMultiShotObjectives, x as parseReflectionResponse, y as runMultiShotOptimization, z as runPromptEvolution, B as trialTraceFromMultiShotTrial } from './summary-report-jrSGb2xZ.js';
 import './errors-BZ9sTdz7.js';
-import './integrity-DK2EBVZC.js';
-import './store-Db2Bv8Cf.js';
+import './integrity-BAxLGJ9I.js';
+import './store-BP5be6s7.js';
 import './run-record-CqzahIbx.js';
-import './emitter-DP_cSSiw.js';
-import './control-runtime-BuJHoLg0.js';
+import './emitter-BqjeOvJh.js';
+import './control-runtime-BRdQ0wrx.js';
 import './dataset-CiK_3LDr.js';
-import './failure-cluster-C2EGSDiT.js';
+import './failure-cluster-D1NZKqYu.js';

package/dist/optimization.js

@@ -25,18 +25,18 @@ import {
   summarizePreferenceMemory,
   trialTraceFromMultiShotTrial,
   withAssignedFeedbackSplit
-} from "./chunk-WHZMVFUV.js";
+} from "./chunk-SZSBQUIJ.js";
 import "./chunk-NLMNWKVM.js";
 import {
   runEvalCampaign
-} from "./chunk-SESZDQPX.js";
+} from "./chunk-RUI6SIHY.js";
 import "./chunk-4S4BM3QQ.js";
-import "./chunk-2A5XJB43.js";
-import "./chunk-I4MBDTY5.js";
+import "./chunk-5AKPEK5L.js";
+import "./chunk-R5UQJNKC.js";
 import "./chunk-KTGTIOFD.js";
 import "./chunk-PC4UYEBM.js";
 import "./chunk-TVVP3ZZQ.js";
-import "./chunk-4F5DQN55.js";
+import "./chunk-VSMTAMNK.js";
 import "./chunk-NG236HPC.js";
 import "./chunk-PZ5AY32C.js";
 export {

package/dist/pipelines/index.d.ts

@@ -1,9 +1,9 @@
-import { g as BudgetSpec, T as TraceStore, h as RunFilter, R as Run, a as ToolSpan } from '../store-Db2Bv8Cf.js';
-export { a as FailureCluster, F as FailureClusterReport, f as failureClusterView } from '../failure-cluster-C2EGSDiT.js';
-import { a as TrajectoryStep } from '../trajectory-CnoBo-JY.js';
-import { B as BaselineOptions, a as BaselineReport } from '../baseline-4R5deP0N.js';
-export { c as computeToolUseMetrics } from '../baseline-4R5deP0N.js';
-import { l as llmSpans } from '../query-DODUYdPg.js';
+import { g as BudgetSpec, T as TraceStore, l as RunFilter, R as Run, a as ToolSpan } from '../store-BP5be6s7.js';
+export { a as FailureCluster, F as FailureClusterReport, f as failureClusterView } from '../failure-cluster-D1NZKqYu.js';
+import { a as TrajectoryStep } from '../trajectory-BFmveYZt.js';
+import { B as BaselineOptions, a as BaselineReport } from '../baseline-BwdCXUS8.js';
+export { c as computeToolUseMetrics } from '../baseline-BwdCXUS8.js';
+import { l as llmSpans } from '../query-BFDT0kX_.js';
 
 /**
  * BudgetBreachView — aggregates breach events across the corpus.

package/dist/pipelines/index.js

@@ -2,13 +2,13 @@ import {
   compareToBaseline,
   computeToolUseMetrics,
   failureClusterView
-} from "../chunk-JLZQWFV3.js";
+} from "../chunk-K33INZHH.js";
 import {
   buildTrajectory
 } from "../chunk-RZTMDUO7.js";
 import {
   interRaterReliability
-} from "../chunk-I4MBDTY5.js";
+} from "../chunk-R5UQJNKC.js";
 import {
   aggregateLlm,
   argHash,

package/dist/prm/index.d.ts

@@ -1,7 +1,7 @@
-import { P as PrmGradedTrace, S as StepRubric, a as PrmGrader } from '../rubric-D5tjHNJQ.js';
-export { G as GradedStep, b as StepContext, i as isPrmVerdict } from '../rubric-D5tjHNJQ.js';
-import { S as Span, T as TraceStore } from '../store-Db2Bv8Cf.js';
-import '../trajectory-CnoBo-JY.js';
+import { P as PrmGradedTrace, S as StepRubric, a as PrmGrader } from '../rubric-DgSqjqqj.js';
+export { G as GradedStep, b as StepContext, i as isPrmVerdict } from '../rubric-DgSqjqqj.js';
+import { S as Span, T as TraceStore } from '../store-BP5be6s7.js';
+import '../trajectory-BFmveYZt.js';
 
 /**
  * Export PRM-graded traces as training data for downstream reward-model

package/dist/{query-DODUYdPg.d.ts → query-BFDT0kX_.d.ts}

@@ -1,4 +1,4 @@
-import { L as LlmSpan, T as TraceStore, J as JudgeSpan, R as Run, F as FailureClass, a as ToolSpan } from './store-Db2Bv8Cf.js';
+import { L as LlmSpan, T as TraceStore, J as JudgeSpan, R as Run, F as FailureClass, a as ToolSpan } from './store-BP5be6s7.js';
 
 /**
  * Typed query helpers over TraceStore.

package/dist/{release-report-wfUySN5F.d.ts → release-report-PWhGlpfO.d.ts}

@@ -1,5 +1,5 @@
 import { D as DatasetSplit, b as DatasetManifest, a as DatasetScenario } from './dataset-CiK_3LDr.js';
-import { a3 as GateDecision, A as ActionableSideInfo, m as MultiShotTrialResult } from './summary-report-DZVXOCK_.js';
+import { a3 as GateDecision, A as ActionableSideInfo, m as MultiShotTrialResult } from './summary-report-jrSGb2xZ.js';
 import { R as RunRecord, a as RunSplitTag } from './run-record-CqzahIbx.js';
 
 /**