@tangle-network/agent-eval 0.86.0 → 0.90.0

package/dist/contract/index.d.ts

@@ -1,32 +1,36 @@
-import { S as Scenario, M as MutableSurface, b as DispatchContext, a as JudgeConfig, I as ImprovementDriver, G as Gate, L as LabeledScenarioStore, c as GateDecision } from '../types-D7lLRYe9.js';
-export { C as CampaignAggregates, d as CampaignArtifactWriter, e as CampaignCellResult, f as CampaignCostMeter, g as CampaignResult, h as CampaignTraceWriter, i as CodeSurface, D as Dispatch, j as GateContext, k as GateResult, l as GenerationCandidate, m as GenerationRecord, n as JudgeDimension, J as JudgeScore, o as Mutator, O as OptimizerConfig, p as SessionScript } from '../types-D7lLRYe9.js';
-import { L as LoopProvenanceRecord } from '../provenance-DPpNIOJD.js';
-export { A as AxisEvidence, a as AxisVerdict, B as BuildEvidenceVectorOptions, D as DefaultProductionGateOptions, E as EvidenceVector, b as EvolutionaryDriverOptions, H as HeldOutGateOptions, O as ObjectiveSource, P as ParetoSignificanceGateOptions, c as PromotionObjective, d as PromotionPolicy, R as RunEvalOptions, e as buildEvidenceVector, f as composeGate, g as defaultProductionGate, h as evolutionaryDriver, i as heldOutGate, p as paretoPolicy, j as paretoSignificanceGate, r as runEval } from '../provenance-DPpNIOJD.js';
-import { C as CampaignStorage, R as RunOptimizationOptions, a as RunImprovementLoopResult } from '../run-improvement-loop-CNqQckTj.js';
-export { G as GepaDriverOptions, b as RunCampaignOptions, c as RunImprovementLoopOptions, f as fsCampaignStorage, g as gepaDriver, i as inMemoryCampaignStorage, r as runCampaign, d as runImprovementLoop } from '../run-improvement-loop-CNqQckTj.js';
+import { S as Scenario, M as MutableSurface, b as DispatchContext, a as JudgeConfig, I as ImprovementDriver, G as Gate, L as LabeledScenarioStore, c as GateDecision } from '../types-BU-7W85F.js';
+export { C as CampaignAggregates, d as CampaignArtifactWriter, e as CampaignCellResult, f as CampaignCostMeter, g as CampaignResult, h as CampaignTraceWriter, i as CodeSurface, D as Dispatch, j as GateContext, k as GateResult, l as GenerationCandidate, m as GenerationRecord, n as JudgeDimension, J as JudgeScore, o as Mutator, O as OptimizerConfig, p as SessionScript } from '../types-BU-7W85F.js';
+import { L as LoopProvenanceRecord } from '../provenance-LnqRT0sS.js';
+export { A as AxisEvidence, a as AxisVerdict, B as BuildEvidenceVectorOptions, D as DefaultProductionGateOptions, E as EvidenceVector, b as EvolutionaryDriverOptions, H as HeldOutGateOptions, O as ObjectiveSource, P as ParetoSignificanceGateOptions, c as PromotionObjective, d as PromotionPolicy, R as RunEvalOptions, e as buildEvidenceVector, f as composeGate, g as defaultProductionGate, h as evolutionaryDriver, i as heldOutGate, p as paretoPolicy, j as paretoSignificanceGate, r as runEval } from '../provenance-LnqRT0sS.js';
+import { C as CampaignStorage, R as RunOptimizationOptions, a as RunImprovementLoopResult } from '../run-improvement-loop-5z_l5zDz.js';
+export { G as GepaDriverOptions, b as RunCampaignOptions, c as RunImprovementLoopOptions, f as fsCampaignStorage, g as gepaDriver, i as inMemoryCampaignStorage, r as runCampaign, d as runImprovementLoop } from '../run-improvement-loop-5z_l5zDz.js';
 export { D as DeploymentOutcome, F as FileSystemOutcomeStore, a as FileSystemOutcomeStoreOptions, I as InMemoryOutcomeStore, b as OutcomeStore } from '../outcome-store-rnXLEqSn.js';
 import { HostedTenant, EvalRunCellScore, EvalRunGenerationSnapshot, EvalRunEvent, TraceSpanEvent } from '../hosted/index.js';
-import { R as RunRecord, a as RunSplitTag } from '../run-record-De9VarXR.js';
-import { I as InsightReport } from '../insight-report-3ADTfClO.js';
-export { F as FailureClusterInsight, a as InterRaterInsight, J as JudgeInsight, L as LiftInsight, O as OutcomeCorrelationInsight, R as Recommendation, b as ReleaseSummary, S as ScalarDistribution } from '../insight-report-3ADTfClO.js';
-import { a as AnalystRegistry } from '../registry-DrEQ3Luj.js';
-import { a as DatasetScenario } from '../dataset-B2kL-fSM.js';
-export { C as CodeAgentSessionDiagnostic, a as CodeAgentSessionIntakeOptions, b as CodeAgentSessionIntakeResult, c as CodeAgentSessionMetrics, d as CodeAgentSessionSource, P as ParsedCodeAgentJsonl, f as fromClaudeCodeSession, e as fromCodexSession, g as fromKimiCodeSession, h as fromOpenCodeSession, i as fromPiSession, j as fromPigraphSession, p as parseCodeAgentJsonl } from '../code-agent-session-BRXmavYv.js';
-import '../red-team-DW9Ca_tj.js';
+import { R as RunRecord, a as RunSplitTag } from '../run-record-e7vj1uZQ.js';
+import { I as InsightReport } from '../insight-report-BBwvOh6x.js';
+export { F as FailureClusterInsight, a as InterRaterInsight, J as JudgeInsight, L as LiftInsight, O as OutcomeCorrelationInsight, R as Recommendation, b as ReleaseSummary, S as ScalarDistribution } from '../insight-report-BBwvOh6x.js';
+export { D as DefaultAnalystRegistryOptions, b as buildDefaultAnalystRegistry } from '../default-registry-zoGHUQEH.js';
+export { c as AnalystFinding } from '../types-2VVIL04s.js';
+export { A as AnalyzeRunsOptions, a as analyzeRuns } from '../analyze-runs-DwCEkpO_.js';
+export { C as CodeAgentSessionDiagnostic, a as CodeAgentSessionIntakeOptions, b as CodeAgentSessionIntakeResult, c as CodeAgentSessionMetrics, d as CodeAgentSessionSource, P as ParsedCodeAgentJsonl, f as fromClaudeCodeSession, e as fromCodexSession, g as fromKimiCodeSession, h as fromOpenCodeSession, i as fromPiSession, j as fromPigraphSession, p as parseCodeAgentJsonl } from '../code-agent-session-BO8nCnv3.js';
+import '../red-team-BXHil6c8.js';
+import '../dataset-BbGkaN2I.js';
+import '../errors-CzMUYo7b.js';
 import '../store-CKUAgsJz.js';
 import '../schema-m0gsnbt3.js';
 import '../pareto-E-pembql.js';
-import '../statistics-CnC1FMbx.js';
-import '../judge-calibration-DilmB3Ml.js';
+import '../statistics-C7PozGrZ.js';
+import '../judge-calibration-0p2QcWNE.js';
 import '../types-Croy5h7V.js';
 import '@tangle-network/tcloud';
-import '../llm-client-CuUg2Mn3.js';
-import '../errors-Dwqw-T_m.js';
+import '../llm-client-BeEcAokY.js';
 import '../raw-provider-sink-C46HDghv.js';
-import '../summary-report-Db0dDSWP.js';
+import '../summary-report-DGmUucwQ.js';
 import '../failure-cluster-CL7IVgkJ.js';
-import '../types-Cu3u_x59.js';
+import '@ax-llm/ax';
+import '../kind-factory-5b7xXXOr.js';
 import '../store-C1YxJDEK.js';
+import 'zod';
 
 /**
  * # `selfImprove()` — the LAND-tier one-shot.
@@ -306,81 +310,6 @@ interface SelfImproveResult<TScenario extends Scenario, TArtifact> {
  */
 declare function selfImprove<TScenario extends Scenario, TArtifact>(opts: SelfImproveOptions<TScenario, TArtifact>): Promise<SelfImproveResult<TScenario, TArtifact>>;
 
-/**
- * # `analyzeRuns()` — turn a set of agent runs into an actionable decision packet.
- *
- * Wires the substrate's statistical, calibration, clustering, Pareto, and
- * release-confidence primitives into one `InsightReport`. Two top-level
- * entry points use this function:
- *
- *   - `selfImprove()` calls it on the campaign output to attach a packet
- *     to every run.
- *   - Consumers with observed `RunRecord[]` (production traces, gold
- *     corpora, approve/reject tables) call it directly via `analyzeRuns()`
- *     for analysis without a closed loop.
- *
- * Every section is opt-in based on what the input data supports — the
- * function never invents signal. If runs carry no judge scores, `judges`
- * is empty. If there's no baseline/candidate split, `lift` is undefined.
- * If no `analyst` is wired, `failureClusters` is undefined.
- *
- * The `recommendations` array is the human-readable layer; everything
- * else is the evidence backing each recommendation.
- */
-
-interface AnalyzeRunsOptions {
-    /** The runs to analyze. */
-    runs: RunRecord[];
-    /** Which split to score against when reading composite from RunOutcome.
-     *  Default: holdout when ANY run has a `holdoutScore`, else search. */
-    split?: 'search' | 'holdout' | 'auto';
-    /** Pairwise analysis configuration. When both `baselineCandidateId` and
-     *  `candidateCandidateId` are present, lift is computed on paired
-     *  (experimentId, seed) tuples shared between the two sides. */
-    baselineCandidateId?: string;
-    candidateCandidateId?: string;
-    /** Canary scenarios — checked against every run's raw output for
-     *  holdout contamination. */
-    canaryScenarios?: DatasetScenario[];
-    /** Analyst registry for failure clustering. When omitted, the
-     *  `failureClusters` section is left undefined. */
-    analyst?: AnalystRegistry;
-    /** Downstream outcome metric per run (e.g. engagement rate, approval
-     *  rate, downstream pass rate). When present, the report includes
-     *  `outcomeCorrelation` + a simple linear reward model fit. */
-    outcomeSignal?: {
-        metric: string;
-        valueByRunId: Record<string, number>;
-    };
-    /** Multi-rater feedback for inter-rater agreement. Each entry is one
-     *  rater's score for one run. Two or more raters → kappa + disagreement
-     *  triage list. */
-    raterScores?: Array<{
-        runId: string;
-        rater: string;
-        score: number;
-    }>;
-    /** Number of histogram bins for distributional summaries. Default 12. */
-    histogramBins?: number;
-    /** Decision threshold — the smallest composite lift the caller cares
-     *  about. Used by the recommendations engine to call ship vs hold.
-     *  Default 0.02. */
-    decisionThreshold?: number;
-    /** Optional prior-period runs. When set, the report includes
-     *  `priorPeriodComparison` with per-metric Welch-CI deltas and
-     *  recommendations fire on statistically significant regressions.
-     *  The two windows do NOT have to share scenarios — the comparison
-     *  is two-sample unpaired (the substrate's `lift` field uses paired
-     *  bootstrap on shared (experimentId, seed) tuples; this is the
-     *  shape for "this week vs last week" rather than "candidate vs
-     *  baseline within a campaign"). */
-    baselineRuns?: RunRecord[];
-    /** Human-readable label for the baseline window, e.g. "vs prior 7
-     *  days", "vs v3.1 release". Surfaces in recommendations + UI. */
-    baselineLabel?: string;
-}
-declare function analyzeRuns(opts: AnalyzeRunsOptions): Promise<InsightReport>;
-
 /**
  * # `@tangle-network/agent-eval/contract` — eval-run diff primitive.
  *
@@ -710,4 +639,4 @@ interface FromOtelSpansOptions {
 }
 declare function fromOtelSpans(opts: FromOtelSpansOptions): RunRecord[];
 
-export { type AgentTraceContributor, type AgentTraceContributorType, type AgentTraceConversation, type AgentTraceFile, type AgentTraceIndex, type AgentTraceRange, type AgentTraceRecord, type AnalyzeRunsOptions, type AuthoringProvenance, CampaignStorage, DispatchContext, type EvalCellScoreDelta, type EvalDimensionDelta, type EvalGenerationDiff, type EvalRunDiff, type FeedbackTableMeta, type FeedbackTableRow, type FromFeedbackTableOptions, type FromFeedbackTableResult, type FromOtelSpansOptions, Gate, GateDecision, ImprovementDriver, InsightReport, JudgeConfig, MutableSurface, type PartitionByAuthoringModelResult, RunImprovementLoopResult, Scenario, type SelfImproveBudget, type SelfImproveLlm, type SelfImproveOptions, type SelfImproveProgressEvent, type SelfImproveResult, analyzeRuns, diffGenerations, diffRunBaselineToWinner, diffRuns, fromFeedbackTable, fromOtelSpans, parseAgentTrace, partitionRunsByAuthoringModel, selfImprove };
+export { type AgentTraceContributor, type AgentTraceContributorType, type AgentTraceConversation, type AgentTraceFile, type AgentTraceIndex, type AgentTraceRange, type AgentTraceRecord, type AuthoringProvenance, CampaignStorage, DispatchContext, type EvalCellScoreDelta, type EvalDimensionDelta, type EvalGenerationDiff, type EvalRunDiff, type FeedbackTableMeta, type FeedbackTableRow, type FromFeedbackTableOptions, type FromFeedbackTableResult, type FromOtelSpansOptions, Gate, GateDecision, ImprovementDriver, InsightReport, JudgeConfig, MutableSurface, type PartitionByAuthoringModelResult, RunImprovementLoopResult, Scenario, type SelfImproveBudget, type SelfImproveLlm, type SelfImproveOptions, type SelfImproveProgressEvent, type SelfImproveResult, diffGenerations, diffRunBaselineToWinner, diffRuns, fromFeedbackTable, fromOtelSpans, parseAgentTrace, partitionRunsByAuthoringModel, selfImprove };