@tangle-network/agent-eval 0.122.0 → 0.122.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +6 -0
- package/dist/analyst/index.js +4 -4
- package/dist/benchmarks/index.js +6 -6
- package/dist/campaign/index.d.ts +19 -3
- package/dist/campaign/index.js +5 -5
- package/dist/{chunk-QWMPPZ3X.js → chunk-5BYTIDZ7.js} +2 -2
- package/dist/{chunk-S5TT5R3L.js → chunk-5CVUPHJ4.js} +2 -2
- package/dist/{chunk-3A246TSA.js → chunk-BD2OK6ZW.js} +2 -2
- package/dist/{chunk-TDHX3BVA.js → chunk-EXMKNOP5.js} +37 -29
- package/dist/{chunk-TDHX3BVA.js.map → chunk-EXMKNOP5.js.map} +1 -1
- package/dist/{chunk-XDIRG3TO.js → chunk-N3QPYIVG.js} +201 -103
- package/dist/chunk-N3QPYIVG.js.map +1 -0
- package/dist/{chunk-4I2E3LLO.js → chunk-NMN4WGSJ.js} +2 -2
- package/dist/{chunk-PICTDURQ.js → chunk-OZFTLMGS.js} +2 -2
- package/dist/{chunk-VZ6VKOJT.js → chunk-U4R2AGY5.js} +52 -10
- package/dist/{chunk-VZ6VKOJT.js.map → chunk-U4R2AGY5.js.map} +1 -1
- package/dist/{chunk-P62I523A.js → chunk-XKSDSGSC.js} +3 -3
- package/dist/{chunk-JHOJHHU7.js → chunk-ZMXDQ4K7.js} +8 -5
- package/dist/{chunk-JHOJHHU7.js.map → chunk-ZMXDQ4K7.js.map} +1 -1
- package/dist/cli.js +2 -2
- package/dist/contract/index.d.ts +15 -0
- package/dist/contract/index.js +6 -6
- package/dist/fuzz.js +1 -1
- package/dist/index.js +9 -9
- package/dist/openapi.json +1 -1
- package/dist/{run-campaign-HNFPJET4.js → run-campaign-2YJ3Z6JS.js} +3 -3
- package/dist/wire/index.js +2 -2
- package/docs/eval-surface-map.md +1 -1
- package/package.json +2 -2
- package/dist/chunk-XDIRG3TO.js.map +0 -1
- /package/dist/{chunk-QWMPPZ3X.js.map → chunk-5BYTIDZ7.js.map} +0 -0
- /package/dist/{chunk-S5TT5R3L.js.map → chunk-5CVUPHJ4.js.map} +0 -0
- /package/dist/{chunk-3A246TSA.js.map → chunk-BD2OK6ZW.js.map} +0 -0
- /package/dist/{chunk-4I2E3LLO.js.map → chunk-NMN4WGSJ.js.map} +0 -0
- /package/dist/{chunk-PICTDURQ.js.map → chunk-OZFTLMGS.js.map} +0 -0
- /package/dist/{chunk-P62I523A.js.map → chunk-XKSDSGSC.js.map} +0 -0
- /package/dist/{run-campaign-HNFPJET4.js.map → run-campaign-2YJ3Z6JS.js.map} +0 -0
package/CHANGELOG.md
CHANGED
|
@@ -4,6 +4,12 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
|
|
|
4
4
|
|
|
5
5
|
---
|
|
6
6
|
|
|
7
|
+
## [0.122.2] — 2026-07-17 — premeasured optimization continuation
|
|
8
|
+
|
|
9
|
+
### Added
|
|
10
|
+
|
|
11
|
+
- `runOptimization()` accepts a surface-bound, split-validated complete campaign as `premeasuredBaseline`, preserves its artifacts for analysis, and skips duplicate baseline dispatch.
|
|
12
|
+
|
|
7
13
|
## [0.121.0] — 2026-07-15 — one measured-comparison contract
|
|
8
14
|
|
|
9
15
|
### Changed
|
package/dist/analyst/index.js
CHANGED
|
@@ -9,13 +9,13 @@ import {
|
|
|
9
9
|
diffFindings,
|
|
10
10
|
emitSkillUsageFindings,
|
|
11
11
|
runSemanticConceptJudge
|
|
12
|
-
} from "../chunk-
|
|
12
|
+
} from "../chunk-XKSDSGSC.js";
|
|
13
13
|
import {
|
|
14
14
|
behavioralAnalyst,
|
|
15
15
|
buildDefaultAnalystRegistry,
|
|
16
16
|
createChatClient,
|
|
17
17
|
deriveEfficiencyFindings
|
|
18
|
-
} from "../chunk-
|
|
18
|
+
} from "../chunk-5BYTIDZ7.js";
|
|
19
19
|
import "../chunk-HHWE3POT.js";
|
|
20
20
|
import "../chunk-3YYRZDON.js";
|
|
21
21
|
import {
|
|
@@ -72,11 +72,11 @@ import {
|
|
|
72
72
|
validatePolicyEdit,
|
|
73
73
|
validatePolicyEditCandidateRecord,
|
|
74
74
|
validateUsageSettlementTimeout
|
|
75
|
-
} from "../chunk-
|
|
75
|
+
} from "../chunk-5CVUPHJ4.js";
|
|
76
76
|
import "../chunk-NJC7U437.js";
|
|
77
77
|
import {
|
|
78
78
|
CostLedger
|
|
79
|
-
} from "../chunk-
|
|
79
|
+
} from "../chunk-ZMXDQ4K7.js";
|
|
80
80
|
import "../chunk-VI2UW6B6.js";
|
|
81
81
|
import "../chunk-PXD6ZFNY.js";
|
|
82
82
|
import "../chunk-PC4UYEBM.js";
|
package/dist/benchmarks/index.js
CHANGED
|
@@ -16,16 +16,16 @@ import {
|
|
|
16
16
|
routing_exports,
|
|
17
17
|
runBenchmarkAdapter,
|
|
18
18
|
summarizeBenchmarkCampaign
|
|
19
|
-
} from "../chunk-
|
|
20
|
-
import "../chunk-
|
|
21
|
-
import "../chunk-
|
|
22
|
-
import "../chunk-
|
|
19
|
+
} from "../chunk-OZFTLMGS.js";
|
|
20
|
+
import "../chunk-EXMKNOP5.js";
|
|
21
|
+
import "../chunk-U4R2AGY5.js";
|
|
22
|
+
import "../chunk-N3QPYIVG.js";
|
|
23
23
|
import "../chunk-3YYRZDON.js";
|
|
24
|
-
import "../chunk-
|
|
24
|
+
import "../chunk-5CVUPHJ4.js";
|
|
25
25
|
import "../chunk-ARU2PZFM.js";
|
|
26
26
|
import "../chunk-NJC7U437.js";
|
|
27
27
|
import "../chunk-PJQFMIOX.js";
|
|
28
|
-
import "../chunk-
|
|
28
|
+
import "../chunk-ZMXDQ4K7.js";
|
|
29
29
|
import "../chunk-VI2UW6B6.js";
|
|
30
30
|
import "../chunk-WW2A73HW.js";
|
|
31
31
|
import "../chunk-GGE4NNQT.js";
|
package/dist/campaign/index.d.ts
CHANGED
|
@@ -4131,9 +4131,24 @@ declare function parameterSweepProposer(opts: ParameterSweepProposerOptions): Su
|
|
|
4131
4131
|
* re-score + release gate + optional PR.
|
|
4132
4132
|
*/
|
|
4133
4133
|
|
|
4134
|
+
interface PremeasuredOptimizationBaseline<TArtifact, TScenario extends Scenario> {
|
|
4135
|
+
/** Hash of the exact surface that produced `campaign`. */
|
|
4136
|
+
surfaceHash: string;
|
|
4137
|
+
/** Complete prior measurement reused by identity, including artifactsByPath. */
|
|
4138
|
+
campaign: CampaignResult<TArtifact, TScenario>;
|
|
4139
|
+
}
|
|
4134
4140
|
interface RunOptimizationBaseOptions<TScenario extends Scenario, TArtifact> extends Omit<RunCampaignOptions<TScenario, TArtifact>, 'dispatch'> {
|
|
4135
4141
|
/** Initial mutable surface (typically system prompt or addendum). */
|
|
4136
4142
|
baselineSurface: MutableSurface;
|
|
4143
|
+
/**
|
|
4144
|
+
* Complete prior measurement of `baselineSurface`. When present,
|
|
4145
|
+
* `runOptimization` validates its surface, scenario split, seed, reps, and
|
|
4146
|
+
* normal campaign coverage, then skips the baseline campaign entirely — no
|
|
4147
|
+
* dispatch or resumability-cache lookup. Candidate campaigns still run
|
|
4148
|
+
* normally. Prior spend remains in the imported campaign aggregates and is
|
|
4149
|
+
* not added again to this continuation's CostLedger.
|
|
4150
|
+
*/
|
|
4151
|
+
premeasuredBaseline?: PremeasuredOptimizationBaseline<TArtifact, TScenario>;
|
|
4137
4152
|
/** Dispatcher that takes the CURRENT surface + scenario → artifact. */
|
|
4138
4153
|
dispatchWithSurface: (surface: MutableSurface, scenario: TScenario, ctx: Parameters<RunCampaignOptions<TScenario, TArtifact>['dispatch']>[1]) => Promise<TArtifact>;
|
|
4139
4154
|
/** The candidate-generation strategy. Wrap a population `Mutator` via
|
|
@@ -7324,7 +7339,7 @@ interface SingleRunLockOptions {
|
|
|
7324
7339
|
readonly alsoCheck?: readonly string[];
|
|
7325
7340
|
/** Install a process 'exit' hook that releases the lock. Default true. */
|
|
7326
7341
|
readonly releaseOnExit?: boolean;
|
|
7327
|
-
/** Owner pid recorded in the lockfile. Default process.pid. */
|
|
7342
|
+
/** Owner pid recorded in the lockfile metadata. Default process.pid. */
|
|
7328
7343
|
readonly pid?: number;
|
|
7329
7344
|
}
|
|
7330
7345
|
interface SingleRunLock {
|
|
@@ -7333,7 +7348,8 @@ interface SingleRunLock {
|
|
|
7333
7348
|
}
|
|
7334
7349
|
/**
|
|
7335
7350
|
* Acquire the lock or throw naming the live holder. A stale lock (holder pid
|
|
7336
|
-
* no longer running) is reclaimed
|
|
7351
|
+
* no longer running) is reclaimed by one contender. An interrupted reclaim
|
|
7352
|
+
* leaves a marker that fails closed instead of admitting overlapping runs.
|
|
7337
7353
|
*/
|
|
7338
7354
|
declare function acquireSingleRunLock(opts: SingleRunLockOptions): SingleRunLock;
|
|
7339
7355
|
|
|
@@ -7464,4 +7480,4 @@ declare function verifyCodeSurface(surface: CodeSurface, worktreeDir?: string):
|
|
|
7464
7480
|
* identity against the checkout at `worktreeRef`. */
|
|
7465
7481
|
declare function resolveWorktreePath(surface: CodeSurface, worktreeDir?: string): string;
|
|
7466
7482
|
|
|
7467
|
-
export { type AcceptedEdit, type AceProposerOptions, type AnalystArtifact, type AnalystScenario, type AnalyzeCrossSurfaceInteractionsInput, type ApplySkillPatchResult, type AxisEvidence, type AxisVerdict, type BuildAnalystSurfaceDispatchOptions, type BuildEvidenceVectorOptions, type BuildLoopProvenanceArgs, type CampaignAggregates, type CampaignArtifactWriter, type CampaignBreakdown, type CampaignCellResult, type CampaignCostMeter, type CampaignResult, type CampaignRunPlan, type CampaignRunPlanCell, type CampaignScenarioIdentity, type CampaignStorage, type CampaignTokenUsage, type CampaignTraceWriter, type CodeSurface, type CodeSurfaceVerification, type CompareProposersOptions, type CompositeProposerOptions, type CostLedgerHandle, type CrossSurfaceAdditionDecision, type CrossSurfaceAdditionRejectionReason, type CrossSurfaceAttemptCompleteness, type CrossSurfaceBestSingleSelection, type CrossSurfaceBootstrapPolicy, type CrossSurfaceCandidate, type CrossSurfaceCandidateComparison, type CrossSurfaceCandidateEvidence, type CrossSurfaceCandidateOutcome, type CrossSurfaceCandidateSummary, type CrossSurfaceComponent, type CrossSurfaceComponentEvidence, type CrossSurfaceCompositionStep, type CrossSurfaceDistribution, type CrossSurfaceEligibility, type CrossSurfaceEvidenceBreakdown, type CrossSurfaceIneligibilityReason, type CrossSurfaceInteractionAwareSelection, type CrossSurfaceInteractionEffect, type CrossSurfaceInteractionPath, type CrossSurfaceInteractionReport, type CrossSurfaceInteractionTask, type CrossSurfaceNaiveStackSelection, type CrossSurfacePairCompatibility, type CrossSurfacePairEvidence, type CrossSurfacePairIncompatibilityReason, type CrossSurfacePairwiseEntry, type CrossSurfaceRankedSingle, type CrossSurfaceRelativeCost, type CrossSurfaceSelectionPolicy, type CrossSurfaceSelections, type CrossSurfaceTaskRow, DEFAULT_POLICY_EDIT_HISTORY_LIMITS, type DefaultProductionGateOptions, type DimensionRegression, type DiscriminationScore, type DispatchContext, type DispatchFn, type EmitLoopProvenanceArgs, type EmitLoopProvenanceResult, type EvalFixture, type EvalFixtureFile, type EvalFixtureLoadOptions, type EvalFixtureRunPlan, type EvalFixtureScenario, type EvalFixtureValidationMode, type EvidenceVector, type EvolutionaryProposerOptions, type FailureModeRecallJudgeOptions, type FapoAttributionSignals, type FapoEntryConfig, type FapoFailureCluster, type FapoOptimizationLevel, type FapoProposerOptions, type FapoReviewInput, type FapoReviewIssue, type FapoReviewResult, type FapoScopeContract, FileSearchLedger, FsLabeledScenarioStore, type FsLabeledScenarioStoreOptions, type Gate, type GateContext, type GateDecision, type GateResult, type GenerationCandidate, type GenerationRecord, type GepaProposerConstraints, type GepaProposerOptions, type GitWorktreeAdapterOptions, type Governor, type GovernorContext, type GovernorOp, type HaloProposerOptions, type HeldOutGateOptions, type HeldoutSignificance, type HeldoutSignificanceOptions, type HeuristicGovernorOptions, type JsonPolicyEditTargetSurface, type JsonPrimitive, type JsonValue, type JudgeAggregate, type JudgeConfig, type JudgeDimension, type JudgeScore, type LabelTrust, type LabeledScenarioRecord, type LabeledScenarioSampleArgs, type LabeledScenarioSource, type LabeledScenarioStore, LabeledScenarioStoreError, type LabeledScenarioWrite, Lineage, type LineageEdge, type LineageGraph, type LineageNode, type LineageNodeInput, type LineageStore, type LlmJudgeDimension, type LlmJudgeOptions, type LlmPolicyEditProposerOptions, type LoadEvalFixtureScenariosOptions, type LoopProvenanceArgsFromResult, type LoopProvenanceBackend, type LoopProvenanceCandidate, type LoopProvenanceEvidence, type LoopProvenanceRecord, type MemoryCurationProposerOptions, type MutableSurface, type Mutator, type NeutralizationGateOptions, type ObjectiveSource, type OpenAutoPrOptions, type OpenAutoPrResult, type OpenSearchLedgerOptions, type OptimizationProposer, type OptimizerConfig, type OptimizerEntryConfig, POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, type PairedHoldout, type ParameterCandidate, type ParameterChange, type ParameterSweepProposerOptions, type ParetoParent, type ParetoSignificanceGateOptions, type PlanCampaignRunOptions, type PlanEvalFixtureRunOptions, type PlaybackContext, type PlaybackDriver, type PlaybackStep, type PolicyEditAuthorScenarioRow, type PolicyEditCandidateRecord, type PolicyEditCandidateSummary, type PolicyEditFindingInput, type PolicyEditFindingSource, type PolicyEditHistoryCandidateContext, type PolicyEditHistoryGenerationContext, type PolicyEditHistoryProjectionOptions, type PolicyEditObjective, type PolicyEditOutcomeContext, type PolicyEditProposerOptions, type PowerPreflight, type PowerPreflightOptions, type ProfileDispatchFn, ProfileMatrixError, type ProfileSummary, type PromotionObjective, type PromotionPolicy, type ProposeContext, type ProposePatchesArgs, type ProposedCandidate, type ProposerComparison, type ProposerEntry, type ProposerPairwise, type ProposerScore, type RedactionStatus, type ReferenceEquivalenceJudgeOptions, type ReferenceEquivalenceScenario, type RejectedEdit, type RolloutArgumentDiff, type RolloutArgumentDiffOptions, type RolloutCall, type RunCampaignOptions, type RunEvalOptions, type RunImprovementLoopOptions, type RunImprovementLoopResult, type RunLineageLoopOptions, type RunLineageLoopResult, type RunLineageLoopSeed, type RunLineageOptions, type RunLineageResult, type RunLineageSeed, type RunLineageStepResult, type RunOptimizationOptions, type RunOptimizationResult, type RunProfileMatrixOptions, type RunProfileMatrixResult, type RunSkillOptOptions, type RunSkillOptResult, SEARCH_LEDGER_SCHEMA, type Scenario, type ScenarioAggregate, type ScenarioRollup, type ScenarioSignal, type ScoreboardRenderOptions, type ScoreboardRow, type ScoreboardSummary, type ScoredRollout, type ScoredSurfaceOutcome, type SearchAccountingAudit, type SearchArtifactRef, type SearchAttemptAccounting, type SearchCandidateDecidedEvent, type SearchCandidateLineage, type SearchCandidateRegisteredEvent, type SearchCandidateSlot, type SearchCandidateSlotClosedEvent, type SearchCandidateSurface, type SearchCompletedEvent, type SearchCostAccounting, type SearchFailureReason, type SearchLedger, type SearchLedgerAppendResult, SearchLedgerConflictError, type SearchLedgerEntry, SearchLedgerError, type SearchLedgerEvent, type SearchLedgerHash, SearchLedgerIntegrityError, type SearchLedgerReplay, type SearchModelIdentity, type SearchOperationKind, type SearchOperationRecordedEvent, type SearchPlan, type SearchPlannedEvent, type SearchPlannedOperation, type SearchPlannedTask, type SearchSourceRef, type SearchSurfaceEffect, type SearchSurfaceEvidence, type SearchSurfaceKind, type SearchTaskAttemptedEvent, type SearchTaskOutcome, type SearchTokenAccounting, type SelectPolicyEditAuthorRowsOptions, type SequentialDecideFn, type SequentialDecideOptions, type SequentialDecision, type SequentialObservation, type SequentialPairedGate, type SequentialPairedGateOptions, type SerializedJsonBudget, type SessionScript, type SingleRunLock, type SingleRunLockOptions, type SkillOptEpochRecord, type SkillOptEvidence, type SkillOptProposer, type SkillOptProposerOptions, type SkillPatch, type SkillPatchOp, SkillPatchParseError, type SkillPatchRejection, type SurfaceProposer, type SurfaceScore, type TraceAnalystProposerOptions, type TraceSpan, type TransientFailureOptions, type UngroundedLiteralReport, type UserStory, type UserStoryVerdict, type Worktree, type WorktreeAdapter, WorktreeAdapterError, aceProposer, acquireSingleRunLock, analyzeCrossSurfaceInteractions, applySkillPatch, assertCampaignDesign, assertCampaignSplitIdentity, assertCodeSurfaceIdentity, assertPolicyEditAuthorContextBudget, buildAnalystSurfaceDispatch, buildEvidenceVector, buildLoopProvenanceRecord, callbackGovernor, campaignBreakdown, campaignMeanComposite, campaignMeasurementDigest, campaignScenarioIdentity, campaignSplitDigest, campaignSplitDigestFromIdentities, canonicalDigest, classifyUngroundedLiterals, codeSurfaceIdentityMaterial, compareProposers, composeGate, compositeProposer, countSentenceEdits, createReferenceEquivalenceJudge, createRunCostLedger, defaultProductionGate, detectScale, dimensionRegressions, discoverEvalFixtures, emitLoopProvenance, evolutionaryProposer, extractFapoAttributionSignals, extractH2Sections, failureModeRecallJudge, fapoEscalationEntry, fapoProposer, fsCampaignStorage, fsLineageStore, gepaParetoEntry, gepaProposer, gepaReflectionEntry, gitWorktreeAdapter, haloProposer, heldOutGate, heldoutSignificance, heuristicGovernor, inMemoryCampaignStorage, isProposedCandidate, isTransientTransportFailure, labelTrustRank, lineageNodeId, llmJudge, llmPolicyEditProposer, loadEvalFixture, loadEvalFixtureScenarios, loopProvenanceArgsFromResult, loopProvenanceSpans, makePlaybackDispatch, memLineageStore, memoryCurationProposer, neutralizationGate, neutralizeText, openAutoPr, openSearchLedger, pairHoldout, parameterSweepProposer, paretoPolicy, paretoSignificanceGate, parseSkillPatchResponse, patchEditCount, planCampaignRun, planEvalFixtureRun, policyEditProposer, powerPreflight, projectPolicyEditHistory, provenanceRecordPath, provenanceSpansPath, renderScoreboardMarkdown, renderSurfaceDiff, resolveRunDir, resolveWorktreePath, rolloutArgumentDiff, runCampaign, runEval, runImprovementLoop, runLineage, runLineageLoop, runOptimization, runProfileMatrix, runSkillOpt, scoreDiscrimination, scoreUserStory, scoreboardSummary, selectDiscriminative, selectPolicyEditAuthorRows, sequentialDecide, sequentialPairedGate, skillOptEntry, skillOptProposer, surfaceContentHash, surfaceHash, tangleTracesRoot, traceAnalystProposer, userStoryScoreboard, validatePolicyEditCandidateRecord, validateSearchLedgerEvent, verifyCodeSurface, verifyLoopProvenanceRecord };
|
|
7483
|
+
export { type AcceptedEdit, type AceProposerOptions, type AnalystArtifact, type AnalystScenario, type AnalyzeCrossSurfaceInteractionsInput, type ApplySkillPatchResult, type AxisEvidence, type AxisVerdict, type BuildAnalystSurfaceDispatchOptions, type BuildEvidenceVectorOptions, type BuildLoopProvenanceArgs, type CampaignAggregates, type CampaignArtifactWriter, type CampaignBreakdown, type CampaignCellResult, type CampaignCostMeter, type CampaignResult, type CampaignRunPlan, type CampaignRunPlanCell, type CampaignScenarioIdentity, type CampaignStorage, type CampaignTokenUsage, type CampaignTraceWriter, type CodeSurface, type CodeSurfaceVerification, type CompareProposersOptions, type CompositeProposerOptions, type CostLedgerHandle, type CrossSurfaceAdditionDecision, type CrossSurfaceAdditionRejectionReason, type CrossSurfaceAttemptCompleteness, type CrossSurfaceBestSingleSelection, type CrossSurfaceBootstrapPolicy, type CrossSurfaceCandidate, type CrossSurfaceCandidateComparison, type CrossSurfaceCandidateEvidence, type CrossSurfaceCandidateOutcome, type CrossSurfaceCandidateSummary, type CrossSurfaceComponent, type CrossSurfaceComponentEvidence, type CrossSurfaceCompositionStep, type CrossSurfaceDistribution, type CrossSurfaceEligibility, type CrossSurfaceEvidenceBreakdown, type CrossSurfaceIneligibilityReason, type CrossSurfaceInteractionAwareSelection, type CrossSurfaceInteractionEffect, type CrossSurfaceInteractionPath, type CrossSurfaceInteractionReport, type CrossSurfaceInteractionTask, type CrossSurfaceNaiveStackSelection, type CrossSurfacePairCompatibility, type CrossSurfacePairEvidence, type CrossSurfacePairIncompatibilityReason, type CrossSurfacePairwiseEntry, type CrossSurfaceRankedSingle, type CrossSurfaceRelativeCost, type CrossSurfaceSelectionPolicy, type CrossSurfaceSelections, type CrossSurfaceTaskRow, DEFAULT_POLICY_EDIT_HISTORY_LIMITS, type DefaultProductionGateOptions, type DimensionRegression, type DiscriminationScore, type DispatchContext, type DispatchFn, type EmitLoopProvenanceArgs, type EmitLoopProvenanceResult, type EvalFixture, type EvalFixtureFile, type EvalFixtureLoadOptions, type EvalFixtureRunPlan, type EvalFixtureScenario, type EvalFixtureValidationMode, type EvidenceVector, type EvolutionaryProposerOptions, type FailureModeRecallJudgeOptions, type FapoAttributionSignals, type FapoEntryConfig, type FapoFailureCluster, type FapoOptimizationLevel, type FapoProposerOptions, type FapoReviewInput, type FapoReviewIssue, type FapoReviewResult, type FapoScopeContract, FileSearchLedger, FsLabeledScenarioStore, type FsLabeledScenarioStoreOptions, type Gate, type GateContext, type GateDecision, type GateResult, type GenerationCandidate, type GenerationRecord, type GepaProposerConstraints, type GepaProposerOptions, type GitWorktreeAdapterOptions, type Governor, type GovernorContext, type GovernorOp, type HaloProposerOptions, type HeldOutGateOptions, type HeldoutSignificance, type HeldoutSignificanceOptions, type HeuristicGovernorOptions, type JsonPolicyEditTargetSurface, type JsonPrimitive, type JsonValue, type JudgeAggregate, type JudgeConfig, type JudgeDimension, type JudgeScore, type LabelTrust, type LabeledScenarioRecord, type LabeledScenarioSampleArgs, type LabeledScenarioSource, type LabeledScenarioStore, LabeledScenarioStoreError, type LabeledScenarioWrite, Lineage, type LineageEdge, type LineageGraph, type LineageNode, type LineageNodeInput, type LineageStore, type LlmJudgeDimension, type LlmJudgeOptions, type LlmPolicyEditProposerOptions, type LoadEvalFixtureScenariosOptions, type LoopProvenanceArgsFromResult, type LoopProvenanceBackend, type LoopProvenanceCandidate, type LoopProvenanceEvidence, type LoopProvenanceRecord, type MemoryCurationProposerOptions, type MutableSurface, type Mutator, type NeutralizationGateOptions, type ObjectiveSource, type OpenAutoPrOptions, type OpenAutoPrResult, type OpenSearchLedgerOptions, type OptimizationProposer, type OptimizerConfig, type OptimizerEntryConfig, POLICY_EDIT_CANDIDATE_RECORD_SCHEMA, type PairedHoldout, type ParameterCandidate, type ParameterChange, type ParameterSweepProposerOptions, type ParetoParent, type ParetoSignificanceGateOptions, type PlanCampaignRunOptions, type PlanEvalFixtureRunOptions, type PlaybackContext, type PlaybackDriver, type PlaybackStep, type PolicyEditAuthorScenarioRow, type PolicyEditCandidateRecord, type PolicyEditCandidateSummary, type PolicyEditFindingInput, type PolicyEditFindingSource, type PolicyEditHistoryCandidateContext, type PolicyEditHistoryGenerationContext, type PolicyEditHistoryProjectionOptions, type PolicyEditObjective, type PolicyEditOutcomeContext, type PolicyEditProposerOptions, type PowerPreflight, type PowerPreflightOptions, type PremeasuredOptimizationBaseline, type ProfileDispatchFn, ProfileMatrixError, type ProfileSummary, type PromotionObjective, type PromotionPolicy, type ProposeContext, type ProposePatchesArgs, type ProposedCandidate, type ProposerComparison, type ProposerEntry, type ProposerPairwise, type ProposerScore, type RedactionStatus, type ReferenceEquivalenceJudgeOptions, type ReferenceEquivalenceScenario, type RejectedEdit, type RolloutArgumentDiff, type RolloutArgumentDiffOptions, type RolloutCall, type RunCampaignOptions, type RunEvalOptions, type RunImprovementLoopOptions, type RunImprovementLoopResult, type RunLineageLoopOptions, type RunLineageLoopResult, type RunLineageLoopSeed, type RunLineageOptions, type RunLineageResult, type RunLineageSeed, type RunLineageStepResult, type RunOptimizationOptions, type RunOptimizationResult, type RunProfileMatrixOptions, type RunProfileMatrixResult, type RunSkillOptOptions, type RunSkillOptResult, SEARCH_LEDGER_SCHEMA, type Scenario, type ScenarioAggregate, type ScenarioRollup, type ScenarioSignal, type ScoreboardRenderOptions, type ScoreboardRow, type ScoreboardSummary, type ScoredRollout, type ScoredSurfaceOutcome, type SearchAccountingAudit, type SearchArtifactRef, type SearchAttemptAccounting, type SearchCandidateDecidedEvent, type SearchCandidateLineage, type SearchCandidateRegisteredEvent, type SearchCandidateSlot, type SearchCandidateSlotClosedEvent, type SearchCandidateSurface, type SearchCompletedEvent, type SearchCostAccounting, type SearchFailureReason, type SearchLedger, type SearchLedgerAppendResult, SearchLedgerConflictError, type SearchLedgerEntry, SearchLedgerError, type SearchLedgerEvent, type SearchLedgerHash, SearchLedgerIntegrityError, type SearchLedgerReplay, type SearchModelIdentity, type SearchOperationKind, type SearchOperationRecordedEvent, type SearchPlan, type SearchPlannedEvent, type SearchPlannedOperation, type SearchPlannedTask, type SearchSourceRef, type SearchSurfaceEffect, type SearchSurfaceEvidence, type SearchSurfaceKind, type SearchTaskAttemptedEvent, type SearchTaskOutcome, type SearchTokenAccounting, type SelectPolicyEditAuthorRowsOptions, type SequentialDecideFn, type SequentialDecideOptions, type SequentialDecision, type SequentialObservation, type SequentialPairedGate, type SequentialPairedGateOptions, type SerializedJsonBudget, type SessionScript, type SingleRunLock, type SingleRunLockOptions, type SkillOptEpochRecord, type SkillOptEvidence, type SkillOptProposer, type SkillOptProposerOptions, type SkillPatch, type SkillPatchOp, SkillPatchParseError, type SkillPatchRejection, type SurfaceProposer, type SurfaceScore, type TraceAnalystProposerOptions, type TraceSpan, type TransientFailureOptions, type UngroundedLiteralReport, type UserStory, type UserStoryVerdict, type Worktree, type WorktreeAdapter, WorktreeAdapterError, aceProposer, acquireSingleRunLock, analyzeCrossSurfaceInteractions, applySkillPatch, assertCampaignDesign, assertCampaignSplitIdentity, assertCodeSurfaceIdentity, assertPolicyEditAuthorContextBudget, buildAnalystSurfaceDispatch, buildEvidenceVector, buildLoopProvenanceRecord, callbackGovernor, campaignBreakdown, campaignMeanComposite, campaignMeasurementDigest, campaignScenarioIdentity, campaignSplitDigest, campaignSplitDigestFromIdentities, canonicalDigest, classifyUngroundedLiterals, codeSurfaceIdentityMaterial, compareProposers, composeGate, compositeProposer, countSentenceEdits, createReferenceEquivalenceJudge, createRunCostLedger, defaultProductionGate, detectScale, dimensionRegressions, discoverEvalFixtures, emitLoopProvenance, evolutionaryProposer, extractFapoAttributionSignals, extractH2Sections, failureModeRecallJudge, fapoEscalationEntry, fapoProposer, fsCampaignStorage, fsLineageStore, gepaParetoEntry, gepaProposer, gepaReflectionEntry, gitWorktreeAdapter, haloProposer, heldOutGate, heldoutSignificance, heuristicGovernor, inMemoryCampaignStorage, isProposedCandidate, isTransientTransportFailure, labelTrustRank, lineageNodeId, llmJudge, llmPolicyEditProposer, loadEvalFixture, loadEvalFixtureScenarios, loopProvenanceArgsFromResult, loopProvenanceSpans, makePlaybackDispatch, memLineageStore, memoryCurationProposer, neutralizationGate, neutralizeText, openAutoPr, openSearchLedger, pairHoldout, parameterSweepProposer, paretoPolicy, paretoSignificanceGate, parseSkillPatchResponse, patchEditCount, planCampaignRun, planEvalFixtureRun, policyEditProposer, powerPreflight, projectPolicyEditHistory, provenanceRecordPath, provenanceSpansPath, renderScoreboardMarkdown, renderSurfaceDiff, resolveRunDir, resolveWorktreePath, rolloutArgumentDiff, runCampaign, runEval, runImprovementLoop, runLineage, runLineageLoop, runOptimization, runProfileMatrix, runSkillOpt, scoreDiscrimination, scoreUserStory, scoreboardSummary, selectDiscriminative, selectPolicyEditAuthorRows, sequentialDecide, sequentialPairedGate, skillOptEntry, skillOptProposer, surfaceContentHash, surfaceHash, tangleTracesRoot, traceAnalystProposer, userStoryScoreboard, validatePolicyEditCandidateRecord, validateSearchLedgerEvent, verifyCodeSurface, verifyLoopProvenanceRecord };
|
package/dist/campaign/index.js
CHANGED
|
@@ -66,7 +66,7 @@ import {
|
|
|
66
66
|
userStoryScoreboard,
|
|
67
67
|
validateSearchLedgerEvent,
|
|
68
68
|
verifyCodeSurface
|
|
69
|
-
} from "../chunk-
|
|
69
|
+
} from "../chunk-EXMKNOP5.js";
|
|
70
70
|
import {
|
|
71
71
|
assertCodeSurfaceIdentity,
|
|
72
72
|
buildEvidenceVector,
|
|
@@ -107,7 +107,7 @@ import {
|
|
|
107
107
|
surfaceContentHash,
|
|
108
108
|
surfaceHash,
|
|
109
109
|
verifyLoopProvenanceRecord
|
|
110
|
-
} from "../chunk-
|
|
110
|
+
} from "../chunk-U4R2AGY5.js";
|
|
111
111
|
import {
|
|
112
112
|
SearchLedgerConflictError,
|
|
113
113
|
SearchLedgerError,
|
|
@@ -124,16 +124,16 @@ import {
|
|
|
124
124
|
resolveRunDir,
|
|
125
125
|
runCampaign,
|
|
126
126
|
tangleTracesRoot
|
|
127
|
-
} from "../chunk-
|
|
127
|
+
} from "../chunk-N3QPYIVG.js";
|
|
128
128
|
import "../chunk-3YYRZDON.js";
|
|
129
129
|
import {
|
|
130
130
|
POLICY_EDIT_CANDIDATE_RECORD_SCHEMA,
|
|
131
131
|
validatePolicyEditCandidateRecord
|
|
132
|
-
} from "../chunk-
|
|
132
|
+
} from "../chunk-5CVUPHJ4.js";
|
|
133
133
|
import "../chunk-ARU2PZFM.js";
|
|
134
134
|
import "../chunk-NJC7U437.js";
|
|
135
135
|
import "../chunk-PJQFMIOX.js";
|
|
136
|
-
import "../chunk-
|
|
136
|
+
import "../chunk-ZMXDQ4K7.js";
|
|
137
137
|
import "../chunk-VI2UW6B6.js";
|
|
138
138
|
import "../chunk-WW2A73HW.js";
|
|
139
139
|
import "../chunk-GGE4NNQT.js";
|
|
@@ -6,7 +6,7 @@ import {
|
|
|
6
6
|
DEFAULT_TRACE_ANALYST_KINDS,
|
|
7
7
|
createTraceAnalystKind,
|
|
8
8
|
makeFinding
|
|
9
|
-
} from "./chunk-
|
|
9
|
+
} from "./chunk-5CVUPHJ4.js";
|
|
10
10
|
import {
|
|
11
11
|
LlmClient
|
|
12
12
|
} from "./chunk-NJC7U437.js";
|
|
@@ -547,4 +547,4 @@ export {
|
|
|
547
547
|
behavioralAnalyst,
|
|
548
548
|
buildDefaultAnalystRegistry
|
|
549
549
|
};
|
|
550
|
-
//# sourceMappingURL=chunk-
|
|
550
|
+
//# sourceMappingURL=chunk-5BYTIDZ7.js.map
|
|
@@ -6,7 +6,7 @@ import {
|
|
|
6
6
|
} from "./chunk-NJC7U437.js";
|
|
7
7
|
import {
|
|
8
8
|
CostLedger
|
|
9
|
-
} from "./chunk-
|
|
9
|
+
} from "./chunk-ZMXDQ4K7.js";
|
|
10
10
|
import {
|
|
11
11
|
buildTraceAnalystTools,
|
|
12
12
|
runTraceAnalysisLoop
|
|
@@ -2665,4 +2665,4 @@ export {
|
|
|
2665
2665
|
aggregateRunScore,
|
|
2666
2666
|
clamp012 as clamp01
|
|
2667
2667
|
};
|
|
2668
|
-
//# sourceMappingURL=chunk-
|
|
2668
|
+
//# sourceMappingURL=chunk-5CVUPHJ4.js.map
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import {
|
|
2
2
|
summarizeBackendIntegrity
|
|
3
|
-
} from "./chunk-
|
|
3
|
+
} from "./chunk-N3QPYIVG.js";
|
|
4
4
|
import {
|
|
5
5
|
paretoChart
|
|
6
6
|
} from "./chunk-DPZAEKA6.js";
|
|
@@ -995,4 +995,4 @@ export {
|
|
|
995
995
|
summarizeExecution,
|
|
996
996
|
analyzeRuns
|
|
997
997
|
};
|
|
998
|
-
//# sourceMappingURL=chunk-
|
|
998
|
+
//# sourceMappingURL=chunk-BD2OK6ZW.js.map
|
|
@@ -15,7 +15,7 @@ import {
|
|
|
15
15
|
runImprovementLoop,
|
|
16
16
|
surfaceContentHash,
|
|
17
17
|
surfaceHash
|
|
18
|
-
} from "./chunk-
|
|
18
|
+
} from "./chunk-U4R2AGY5.js";
|
|
19
19
|
import {
|
|
20
20
|
SearchLedgerConflictError,
|
|
21
21
|
SearchLedgerError,
|
|
@@ -26,11 +26,13 @@ import {
|
|
|
26
26
|
createRunCostLedger,
|
|
27
27
|
fsCampaignStorage,
|
|
28
28
|
planCampaignRun,
|
|
29
|
+
probeAtomicFileLock,
|
|
29
30
|
resolveRunDir,
|
|
30
31
|
runCampaign,
|
|
31
32
|
summarizeBackendIntegrity,
|
|
33
|
+
tryAcquireAtomicFileLock,
|
|
32
34
|
withSearchLedgerFileLock
|
|
33
|
-
} from "./chunk-
|
|
35
|
+
} from "./chunk-N3QPYIVG.js";
|
|
34
36
|
import {
|
|
35
37
|
Mutex
|
|
36
38
|
} from "./chunk-3YYRZDON.js";
|
|
@@ -49,7 +51,7 @@ import {
|
|
|
49
51
|
makePolicyEditCandidateRecord,
|
|
50
52
|
policyEditsFromFindings,
|
|
51
53
|
validatePolicyEditCandidateRecord
|
|
52
|
-
} from "./chunk-
|
|
54
|
+
} from "./chunk-5CVUPHJ4.js";
|
|
53
55
|
import {
|
|
54
56
|
callLlm,
|
|
55
57
|
callLlmJson,
|
|
@@ -68,7 +70,7 @@ import {
|
|
|
68
70
|
import {
|
|
69
71
|
CostAccountingIncompleteError,
|
|
70
72
|
CostLedger
|
|
71
|
-
} from "./chunk-
|
|
73
|
+
} from "./chunk-ZMXDQ4K7.js";
|
|
72
74
|
import {
|
|
73
75
|
analyzeTraces
|
|
74
76
|
} from "./chunk-WW2A73HW.js";
|
|
@@ -7377,37 +7379,43 @@ function formatZodError(error) {
|
|
|
7377
7379
|
}
|
|
7378
7380
|
|
|
7379
7381
|
// src/campaign/single-run-lock.ts
|
|
7380
|
-
|
|
7381
|
-
|
|
7382
|
-
if (
|
|
7383
|
-
|
|
7384
|
-
|
|
7385
|
-
|
|
7386
|
-
|
|
7387
|
-
|
|
7388
|
-
|
|
7389
|
-
return null;
|
|
7382
|
+
function assertAvailable(path) {
|
|
7383
|
+
const unavailable = probeAtomicFileLock({ lockPath: path, acceptLegacyPid: true });
|
|
7384
|
+
if (unavailable) throw unavailableError(path, unavailable);
|
|
7385
|
+
}
|
|
7386
|
+
function unavailableError(path, unavailable) {
|
|
7387
|
+
if (unavailable.reason === "recovery") {
|
|
7388
|
+
return new Error(
|
|
7389
|
+
`single-run lock recovery is already in progress (${path}); refusing a concurrent run`
|
|
7390
|
+
);
|
|
7390
7391
|
}
|
|
7392
|
+
return new Error(
|
|
7393
|
+
`single-run lock held by live pid ${unavailable.holder.pid} (${path}); refusing a concurrent run on the shared resource`
|
|
7394
|
+
);
|
|
7391
7395
|
}
|
|
7392
7396
|
function acquireSingleRunLock(opts) {
|
|
7393
7397
|
const pid = opts.pid ?? process.pid;
|
|
7394
|
-
|
|
7395
|
-
|
|
7396
|
-
|
|
7397
|
-
|
|
7398
|
-
|
|
7399
|
-
|
|
7400
|
-
|
|
7401
|
-
}
|
|
7402
|
-
|
|
7398
|
+
if (!Number.isSafeInteger(pid) || pid <= 0)
|
|
7399
|
+
throw new Error("single-run lock pid must be a positive integer");
|
|
7400
|
+
for (const path of opts.alsoCheck ?? []) assertAvailable(path);
|
|
7401
|
+
const acquisition = tryAcquireAtomicFileLock({
|
|
7402
|
+
lockPath: opts.lockPath,
|
|
7403
|
+
pid,
|
|
7404
|
+
acceptLegacyPid: true
|
|
7405
|
+
});
|
|
7406
|
+
if (!acquisition.acquired) throw unavailableError(opts.lockPath, acquisition);
|
|
7403
7407
|
const release = () => {
|
|
7404
7408
|
try {
|
|
7405
|
-
|
|
7406
|
-
unlinkSync(opts.lockPath);
|
|
7407
|
-
}
|
|
7409
|
+
acquisition.lock.release();
|
|
7408
7410
|
} catch {
|
|
7409
7411
|
}
|
|
7410
7412
|
};
|
|
7413
|
+
try {
|
|
7414
|
+
for (const path of opts.alsoCheck ?? []) assertAvailable(path);
|
|
7415
|
+
} catch (error) {
|
|
7416
|
+
release();
|
|
7417
|
+
throw error;
|
|
7418
|
+
}
|
|
7411
7419
|
if (opts.releaseOnExit ?? true) process.on("exit", release);
|
|
7412
7420
|
return { release };
|
|
7413
7421
|
}
|
|
@@ -7428,7 +7436,7 @@ import { execFileSync } from "child_process";
|
|
|
7428
7436
|
import { createHash as createHash8 } from "crypto";
|
|
7429
7437
|
import {
|
|
7430
7438
|
closeSync,
|
|
7431
|
-
existsSync as
|
|
7439
|
+
existsSync as existsSync4,
|
|
7432
7440
|
constants as fsConstants,
|
|
7433
7441
|
fstatSync,
|
|
7434
7442
|
lstatSync,
|
|
@@ -7829,7 +7837,7 @@ function assertRawTreeMatchesWorktree(git, root, candidateCommit) {
|
|
|
7829
7837
|
}
|
|
7830
7838
|
function verifyCodeSurfaceWithGit(surface, path, git) {
|
|
7831
7839
|
assertCodeSurfaceIdentity(surface);
|
|
7832
|
-
if (!
|
|
7840
|
+
if (!existsSync4(path)) {
|
|
7833
7841
|
throw new WorktreeAdapterError(`CodeSurface worktree does not exist: ${path}`);
|
|
7834
7842
|
}
|
|
7835
7843
|
const lexicalRoot = resolve3(path);
|
|
@@ -8108,4 +8116,4 @@ export {
|
|
|
8108
8116
|
verifyCodeSurface,
|
|
8109
8117
|
resolveWorktreePath
|
|
8110
8118
|
};
|
|
8111
|
-
//# sourceMappingURL=chunk-
|
|
8119
|
+
//# sourceMappingURL=chunk-EXMKNOP5.js.map
|