@tangle-network/agent-eval 0.27.0 → 0.28.0

package/dist/{control-runtime-BuJHoLg0.d.ts → control-runtime-BRdQ0wrx.d.ts}

@@ -1,5 +1,5 @@
-import { T as TraceEmitter } from './emitter-DP_cSSiw.js';
-import { F as FailureClass, T as TraceStore } from './store-Db2Bv8Cf.js';
+import { T as TraceEmitter } from './emitter-BqjeOvJh.js';
+import { F as FailureClass, T as TraceStore } from './store-BP5be6s7.js';
 
 /**
  * Policy-based agent control runtime.
@@ -120,6 +120,7 @@ interface ControlRunResult<TState, TAction, TActionResult, TEval extends Control
     finalEvals: TEval[];
     wallMs: number;
     spentCostUsd: number;
+    /** null when the run executed without a TraceEmitter wired (no run record was persisted). */
     runId: string | null;
     failureClass?: FailureClass;
     runtimeErrors: ControlRuntimeError[];

package/dist/control.d.ts

@@ -1,8 +1,8 @@
-export { A as ActionExecutionPolicy, a as ActionPolicyDecision, C as ControlRunToRunRecordOptions, e as ProposeReviewConfig, f as ProposeReviewControlAction, g as ProposeReviewControlConfig, h as ProposeReviewControlResult, i as ProposeReviewControlState, j as ProposeReviewReport, p as RunEvidenceMetadata, s as controlRunToRunRecord, u as evaluateActionPolicy, x as runProposeReview, y as runProposeReviewAsControlLoop, z as scoreFromEvals } from './control-CBShYYA6.js';
-export { c as ControlActionFailureMode, d as ControlActionOutcome, e as ControlBudget, f as ControlContext, g as ControlDecision, C as ControlEvalResult, a as ControlRunResult, h as ControlRuntimeConfig, i as ControlRuntimeError, j as ControlSeverity, b as ControlStep, k as ControlStopPolicies, S as StopDecision, l as allCriticalPassed, o as objectiveEval, r as runAgentControlLoop, s as stopOnNoProgress, m as stopOnRepeatedAction, n as subjectiveEval } from './control-runtime-BuJHoLg0.js';
-import './feedback-trajectory-DfFdrraJ.js';
+export { A as ActionExecutionPolicy, a as ActionPolicyDecision, C as ControlRunToRunRecordOptions, e as ProposeReviewConfig, f as ProposeReviewControlAction, g as ProposeReviewControlConfig, h as ProposeReviewControlResult, i as ProposeReviewControlState, j as ProposeReviewReport, p as RunEvidenceMetadata, s as controlRunToRunRecord, u as evaluateActionPolicy, x as runProposeReview, y as runProposeReviewAsControlLoop, z as scoreFromEvals } from './control-rJhEDdpy.js';
+export { c as ControlActionFailureMode, d as ControlActionOutcome, e as ControlBudget, f as ControlContext, g as ControlDecision, C as ControlEvalResult, a as ControlRunResult, h as ControlRuntimeConfig, i as ControlRuntimeError, j as ControlSeverity, b as ControlStep, k as ControlStopPolicies, S as StopDecision, l as allCriticalPassed, o as objectiveEval, r as runAgentControlLoop, s as stopOnNoProgress, m as stopOnRepeatedAction, n as subjectiveEval } from './control-runtime-BRdQ0wrx.js';
+import './feedback-trajectory-j0nJFgC6.js';
 import './dataset-CiK_3LDr.js';
 import './errors-BZ9sTdz7.js';
-import './emitter-DP_cSSiw.js';
-import './store-Db2Bv8Cf.js';
+import './emitter-BqjeOvJh.js';
+import './store-BP5be6s7.js';
 import './run-record-CqzahIbx.js';

package/dist/control.js

@@ -4,7 +4,7 @@ import {
   runProposeReview,
   runProposeReviewAsControlLoop,
   scoreFromEvals
-} from "./chunk-ZN274SWR.js";
+} from "./chunk-PALJO75S.js";
 import {
   allCriticalPassed,
   objectiveEval,
@@ -12,7 +12,7 @@ import {
   stopOnNoProgress,
   stopOnRepeatedAction,
   subjectiveEval
-} from "./chunk-LSH4MMOZ.js";
+} from "./chunk-NCRFYPS3.js";
 import "./chunk-NLMNWKVM.js";
 import "./chunk-TVVP3ZZQ.js";
 import "./chunk-NG236HPC.js";

package/dist/{emitter-DP_cSSiw.d.ts → emitter-BqjeOvJh.d.ts}

@@ -1,4 +1,4 @@
-import { T as TraceStore, c as RunOutcome, R as Run, S as Span, d as SpanKind, L as LlmSpan, a as ToolSpan, e as RetrievalSpan, J as JudgeSpan, f as SandboxSpan, E as EventKind, b as TraceEvent, B as BudgetLedgerEntry, A as Artifact, M as Message } from './store-Db2Bv8Cf.js';
+import { T as TraceStore, c as RunOutcome, R as Run, S as Span, d as SpanKind, L as LlmSpan, a as ToolSpan, e as RetrievalSpan, J as JudgeSpan, f as SandboxSpan, E as EventKind, b as TraceEvent, B as BudgetLedgerEntry, A as Artifact, M as Message } from './store-BP5be6s7.js';
 
 /**
  * TraceEmitter — hierarchical span builder that auto-parents using an

package/dist/{failure-cluster-C2EGSDiT.d.ts → failure-cluster-D1NZKqYu.d.ts}

@@ -1,4 +1,4 @@
-import { R as Run, S as Span, b as TraceEvent, F as FailureClass, T as TraceStore } from './store-Db2Bv8Cf.js';
+import { R as Run, S as Span, b as TraceEvent, F as FailureClass, T as TraceStore } from './store-BP5be6s7.js';
 
 /**
  * Failure taxonomy — canonical classes + a default classifier.
@@ -54,8 +54,7 @@ interface FailureCluster {
      * Source dimension when the trigger was a judge span (e.g. `'format'`,
      * `'safety'`, `'correctness'`). Lets cross-template aggregators
      * group failures by the dimension that fired without overloading
-     * `argPrefix`. Optional — legacy clusters without this field
-     * deserialize cleanly.
+     * `argPrefix`. Optional — clusters without this field deserialize cleanly.
      */
     dimension?: string;
     runCount: number;

package/dist/{feedback-trajectory-DfFdrraJ.d.ts → feedback-trajectory-j0nJFgC6.d.ts}

@@ -1,4 +1,4 @@
-import { C as ControlEvalResult, a as ControlRunResult, b as ControlStep } from './control-runtime-BuJHoLg0.js';
+import { C as ControlEvalResult, a as ControlRunResult, b as ControlStep } from './control-runtime-BRdQ0wrx.js';
 import { D as DatasetSplit, a as DatasetScenario } from './dataset-CiK_3LDr.js';
 
 type FeedbackArtifactType = 'text' | 'code' | 'plan' | 'research' | 'action' | 'ui' | 'decision' | 'data' | 'other';

package/dist/governance/index.d.ts

@@ -1,5 +1,5 @@
-export { E as EuRiskClass, e as GovernanceContext, f as GovernanceFinding, g as GovernanceReport, U as UseCaseSignals, n as classifyEuAiRisk, p as euAiActReport, q as nistAiRmfReport, u as renderMarkdown, x as soc2Report, y as summarize } from '../index-D3iBCjdF.js';
+export { E as EuRiskClass, e as GovernanceContext, f as GovernanceFinding, g as GovernanceReport, U as UseCaseSignals, n as classifyEuAiRisk, p as euAiActReport, q as nistAiRmfReport, u as renderMarkdown, x as soc2Report, y as summarize } from '../index-Cgt3DKXr.js';
 import '../dataset-CiK_3LDr.js';
 import '../errors-BZ9sTdz7.js';
 import '../outcome-store-D6KWmYvj.js';
-import '../store-Db2Bv8Cf.js';
+import '../store-BP5be6s7.js';

package/dist/{index-D3iBCjdF.d.ts → index-Cgt3DKXr.d.ts}

@@ -1,6 +1,6 @@
 import { a as DatasetScenario, c as Dataset, b as DatasetManifest } from './dataset-CiK_3LDr.js';
 import { O as OutcomeStore } from './outcome-store-D6KWmYvj.js';
-import { T as TraceStore } from './store-Db2Bv8Cf.js';
+import { T as TraceStore } from './store-BP5be6s7.js';
 
 /**
  * Judge calibration — measure judge quality against human gold + bias.
@@ -328,4 +328,4 @@ declare function nistAiRmfReport(ctx: GovernanceContext): Promise<GovernanceRepo
 
 declare function soc2Report(ctx: GovernanceContext): Promise<GovernanceReport>;
 
-export { verbosityBias as A, type CalibrationResult as C, DEFAULT_RED_TEAM_CORPUS as D, type EuRiskClass as E, type GoldenItem as G, type PositionalBiasResult as P, type RedTeamCase as R, type SelfPreferenceResult as S, type UseCaseSignals as U, type VerbosityBiasResult as V, type CandidateScore as a, type ContinuousAgreement as b, type ContinuousAgreementOptions as c, type ContinuousCalibrationResult as d, type GovernanceContext as e, type GovernanceFinding as f, type GovernanceReport as g, type RedTeamCategory as h, type RedTeamFinding as i, type RedTeamPayload as j, type RedTeamReport as k, calibrateJudge as l, calibrateJudgeContinuous as m, classifyEuAiRisk as n, continuousAgreement as o, euAiActReport as p, nistAiRmfReport as q, positionalBias as r, redTeamDataset as s, redTeamReport as t, renderMarkdown as u, scoreRedTeamOutput as v, selfPreference as w, soc2Report as x, summarize as y, toolNamesForRun as z };
+export { verbosityBias as A, type ContinuousAgreementOptions as C, DEFAULT_RED_TEAM_CORPUS as D, type EuRiskClass as E, type GoldenItem as G, type PositionalBiasResult as P, type RedTeamCase as R, type SelfPreferenceResult as S, type UseCaseSignals as U, type VerbosityBiasResult as V, type ContinuousAgreement as a, type CalibrationResult as b, type CandidateScore as c, type ContinuousCalibrationResult as d, type GovernanceContext as e, type GovernanceFinding as f, type GovernanceReport as g, type RedTeamCategory as h, type RedTeamFinding as i, type RedTeamPayload as j, type RedTeamReport as k, calibrateJudge as l, calibrateJudgeContinuous as m, classifyEuAiRisk as n, continuousAgreement as o, euAiActReport as p, nistAiRmfReport as q, positionalBias as r, redTeamDataset as s, redTeamReport as t, renderMarkdown as u, scoreRedTeamOutput as v, selfPreference as w, soc2Report as x, summarize as y, toolNamesForRun as z };