@tangle-network/agent-eval 0.145.12 → 0.145.14
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 +17 -0
- package/dist/analyst/index.d.ts +1 -1
- package/dist/analyst/index.js +1 -1
- package/dist/{benchmark-command-JSjVoAxT.js → benchmark-command-DSlSLfYO.js} +2 -2
- package/dist/{benchmark-command-JSjVoAxT.js.map → benchmark-command-DSlSLfYO.js.map} +1 -1
- package/dist/cli.js +1 -1
- package/dist/contract/index.d.ts +26 -1
- package/dist/contract/index.d.ts.map +1 -1
- package/dist/contract/index.js +70 -7
- package/dist/contract/index.js.map +1 -1
- package/dist/openapi.json +1 -1
- package/package.json +3 -3
package/dist/cli.js
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
#!/usr/bin/env node
|
|
2
2
|
import { o as runRolloutReleaseCli } from "./hf-dataset-XggBupCr.js";
|
|
3
|
-
import { n as runAnalystBenchmarkCommand } from "./benchmark-command-
|
|
3
|
+
import { n as runAnalystBenchmarkCommand } from "./benchmark-command-DSlSLfYO.js";
|
|
4
4
|
import { a as runRpcBatch, o as runRpcOnce, p as handleVersion, r as startServerAsync, s as buildOpenApi } from "./server-ulsOdrTI.js";
|
|
5
5
|
import { writeFileSync } from "node:fs";
|
|
6
6
|
//#region src/cli-config.ts
|
package/dist/contract/index.d.ts
CHANGED
|
@@ -595,7 +595,32 @@ interface CodeAgentSessionIntakeOptions {
|
|
|
595
595
|
* JSON stream has no terminal event, as with `opencode run --format json`. */
|
|
596
596
|
execution?: CodeAgentSessionExecutionReceipt;
|
|
597
597
|
}
|
|
598
|
+
/** One transcript line after the intake rule ran on it. A blank line produces
|
|
599
|
+
* nothing, so every value here is either a parsed entry or a counted defect. */
|
|
600
|
+
type CodeAgentJsonlLine = {
|
|
601
|
+
kind: 'entry';
|
|
602
|
+
lineNumber: number;
|
|
603
|
+
entry: unknown;
|
|
604
|
+
} | {
|
|
605
|
+
kind: 'malformed';
|
|
606
|
+
lineNumber: number;
|
|
607
|
+
};
|
|
598
608
|
declare function parseCodeAgentJsonl(jsonl: string): ParsedCodeAgentJsonl;
|
|
609
|
+
/** Reads a transcript one line at a time and never holds the file as a single
|
|
610
|
+
* string. `parseCodeAgentJsonl` needs the whole file in one string, so a
|
|
611
|
+
* session above V8's ~512MB string ceiling throws `ERR_STRING_TOO_LONG` and
|
|
612
|
+
* cannot be ingested at all; the largest real Codex rollout on record is 695MB.
|
|
613
|
+
*
|
|
614
|
+
* Lines break on `\n` only, which is what the string path's `split('\n')` does.
|
|
615
|
+
* `node:readline` also breaks on a bare `\r`, so it is deliberately not used
|
|
616
|
+
* here: a lone carriage return inside a line must stay inside that line for the
|
|
617
|
+
* two paths to report the same malformed count. */
|
|
618
|
+
declare function streamCodeAgentJsonlFile(path: string): AsyncGenerator<CodeAgentJsonlLine>;
|
|
619
|
+
/** Streaming counterpart to `parseCodeAgentJsonl` for a transcript on disk.
|
|
620
|
+
* It returns the same shape, so a caller that holds every entry keeps working
|
|
621
|
+
* above the string ceiling. The entry array still grows with the transcript;
|
|
622
|
+
* consume `streamCodeAgentJsonlFile` directly when memory must stay flat. */
|
|
623
|
+
declare function parseCodeAgentJsonlFile(path: string): Promise<ParsedCodeAgentJsonl>;
|
|
599
624
|
declare function fromCodexSession(options: CodeAgentSessionIntakeOptions): CodeAgentSessionIntakeResult;
|
|
600
625
|
declare function fromClaudeCodeSession(options: CodeAgentSessionIntakeOptions): CodeAgentSessionIntakeResult;
|
|
601
626
|
declare function fromOpenCodeSession(options: CodeAgentSessionIntakeOptions): CodeAgentSessionIntakeResult;
|
|
@@ -691,5 +716,5 @@ interface FromOtelSpansOptions {
|
|
|
691
716
|
}
|
|
692
717
|
declare function fromOtelSpans(opts: FromOtelSpansOptions): RunRecord[];
|
|
693
718
|
//#endregion
|
|
694
|
-
export { type AgentEvalAgent, type AgentEvalEvaluateOptions, type AgentEvalImproveOptions, type AgentProfileImprovementExperimentExecutionInput, type AgentProfileImprovementExperimentRun, type AgentTraceContributor, type AgentTraceContributorType, type AgentTraceConversation, type AgentTraceFile, type AgentTraceIndex, type AgentTraceRange, type AgentTraceRecord, type AnalystFinding, type AnalyzeRunsOptions, type AuthoringProvenance, type AxisEvidence, type AxisVerdict, type BuildEvidenceVectorOptions, type CampaignAggregates, type CampaignArtifactWriter, type CampaignCellFailureReceipt, type CampaignCellResult, type CampaignCostMeter, type CampaignResult, type CampaignStorage, type CampaignTraceWriter, type CandidateExperimentExecutionInput, type CandidateExperimentRun, type ChatClient, type CodeAgentSessionAction, type CodeAgentSessionActionKind, type CodeAgentSessionActionStatus, type CodeAgentSessionActionSurface, type CodeAgentSessionDiagnostic, type CodeAgentSessionExecutionReceipt, type CodeAgentSessionIntakeOptions, type CodeAgentSessionIntakeResult, type CodeAgentSessionMetrics, type CodeAgentSessionObservation, type CodeAgentSessionSource, type CodeAgentSessionTerminalStatus, type CodeSurface, type CompareAgentProfileImprovementExperimentOptions, type CompareCandidateExperimentOptions, type CompareOptimizationMethodsOptions, type ComparisonCost, type CostLedgerHandle, type CostProvenanceSummary, type CreateChatClientOpts, type DefaultAnalystRegistryOptions, type DefaultProductionGateCheck, type DefaultProductionGateOptions, type DefaultProductionRewardHackingOptions, type DefineAgentEvalOptions, type DefinedAgentEval, type DeploymentOutcome, type DispatchFn as Dispatch, type DispatchContext, type EvalCellScoreDelta, type EvalDimensionDelta, type EvalGenerationDiff, type EvalReportingSuiteInput, type EvalReportingSuiteOptions, type EvalReportingSuiteResult, type EvalRunDiff, type EvaluatePairedMeasurementsOptions, type EvidenceVector, type ExecutionErrorOutcomeCell, type ExecutionInsight, type ExecutionReport, type ExternalOptimizationExample, type ExternalTextEvaluationResponse, type ExternalTextOptimizationMethodConfig, type ExternalTextOptimizerContext, type ExternalTextOptimizerResult, type FailureClassTally, type FailureClusterInsight, type FeedbackTableMeta, type FeedbackTableRow, FileSystemOutcomeStore, type FileSystemOutcomeStoreOptions, type FromFeedbackTableOptions, type FromFeedbackTableResult, type FromOtelSpansOptions, type FromRunRecordDirOptions, type FromRunRecordDirResult, type Gate, type GateCheckStatus, type GateContext, type GateContribution, type GateDecision, type GateResult, type GenerationCandidate, type GenerationRecord, type GepaAdaptiveEngineRun, type GepaEngineOptions, type GepaEngineRun, type GepaOptimizationMethodConfig, type GepaOptimizationRecipe, type GepaRunnerCommand, type HeldOutGateOptions, type HostedTenant, InMemoryOutcomeStore, type InsightReport, type InterRaterInsight, type JudgeConfig, type JudgeDimension, type JudgeInsight, type JudgeScore, type LiftInsight, type LlmJudgeDimension, type LlmJudgeOptions, type MutableSurface, type ObjectiveSource, type OpenAICompatibleOptimizerModel, type OptimizationMethod, type OptimizationMethodComparison, type OptimizationMethodInput, type OptimizationMethodProvenance, type OptimizationMethodResult, type OptimizationPackageSource, type OptimizationTokenUsage, type OptimizerConfig, type OptimizerModelBudget, type OutcomeCorrelationInsight, type OutcomeStore, type PairedMeasurement, type PairedMeasurementAdapter, type PairedMeasurementEvaluation, type ParetoSignificanceGateOptions, type ParsedCodeAgentJsonl, type PartitionByAuthoringModelResult, type PromotionObjective, type PromotionPolicy, type ProposalFinding, type ProposalFindingOrigin, REFERENCE_EQUIVALENCE_INPUT_LIMITS, REFERENCE_EQUIVALENCE_JUDGE_VERSION, type Recommendation, type ReferenceEquivalenceJudgeInput, type ReferenceEquivalenceJudgeOptions, type ReferenceEquivalenceJudgeResult, type ReferenceEquivalenceScenario, type ReleaseSummary, type RunAgentProfileImprovementExperimentOptions, type RunCampaignOptions, type RunCandidateExperimentOptions, type RunEvalOptions, type RunImprovementLoopOptions, type RunImprovementLoopResult, type RunRecordRejection, type ScalarDistribution, type Scenario, type SealAgentProfileImprovementSuiteOptions, type SealCandidateBenchmarkSuiteOptions, type SelfImproveBudget, type SelfImproveOptions, type SelfImproveProgressEvent, type SelfImproveResult, SelfImproveRunError, type SessionScript, type SkillOptOptimizationMethodConfig, type SkillOptRunnerCommand, type SkillOptTrainerConfig, type SummarizeExecutionOptions, type SurfaceProposer, type TokenUsageInsight, analyzeRuns, buildDefaultAnalystRegistry, buildEvidenceVector, campaignSplitDigest, compareOptimizationMethods, composeGate, createChatClient, createReferenceEquivalenceJudge, defaultProductionGate, defineAgentEval, diffGenerations, diffRunBaselineToWinner, diffRuns, evalReportingSuite, evaluatePairedMeasurements, externalTextOptimizationMethod, fromClaudeCodeSession, fromCodexSession, fromFeedbackTable, fromKimiCodeSession, fromOpenCodeSession, fromOtelSpans, fromPiSession, fromPigraphSession, fromRunRecordDir, fsCampaignStorage, gepaOptimizationMethod, heldOutGate, inMemoryCampaignStorage, llmJudge, makeProposalFinding, measuredComparisonFromAgentProfileImprovementExperiment, measuredComparisonFromCandidateExperiment, observeCodeAgentSession, paretoPolicy, paretoSignificanceGate, parseAgentTrace, parseCodeAgentJsonl, partitionRunsByAuthoringModel, runAgentProfileImprovementExperiment, runCampaign, runCandidateExperiment, runEval, runImprovementLoop, runReferenceEquivalenceJudge, sealAgentProfileImprovementExperiment, sealAgentProfileImprovementSuite, sealAgentProfileImprovementTask, sealCandidateBenchmarkSuite, sealCandidateBenchmarkTask, sealCandidateExperiment, selfImprove, skillOptOptimizationMethod, summarizeExecution, verifyAgentProfileImprovementExperiment, verifyAgentProfileImprovementExperimentComparison, verifyAgentProfileImprovementSuiteInputs, verifyAgentProfileImprovementTask, verifyCandidateBenchmarkSuite, verifyCandidateBenchmarkSuiteInputs, verifyCandidateBenchmarkTask, verifyCandidateExperiment, verifyCandidateExperimentComparison };
|
|
719
|
+
export { type AgentEvalAgent, type AgentEvalEvaluateOptions, type AgentEvalImproveOptions, type AgentProfileImprovementExperimentExecutionInput, type AgentProfileImprovementExperimentRun, type AgentTraceContributor, type AgentTraceContributorType, type AgentTraceConversation, type AgentTraceFile, type AgentTraceIndex, type AgentTraceRange, type AgentTraceRecord, type AnalystFinding, type AnalyzeRunsOptions, type AuthoringProvenance, type AxisEvidence, type AxisVerdict, type BuildEvidenceVectorOptions, type CampaignAggregates, type CampaignArtifactWriter, type CampaignCellFailureReceipt, type CampaignCellResult, type CampaignCostMeter, type CampaignResult, type CampaignStorage, type CampaignTraceWriter, type CandidateExperimentExecutionInput, type CandidateExperimentRun, type ChatClient, type CodeAgentJsonlLine, type CodeAgentSessionAction, type CodeAgentSessionActionKind, type CodeAgentSessionActionStatus, type CodeAgentSessionActionSurface, type CodeAgentSessionDiagnostic, type CodeAgentSessionExecutionReceipt, type CodeAgentSessionIntakeOptions, type CodeAgentSessionIntakeResult, type CodeAgentSessionMetrics, type CodeAgentSessionObservation, type CodeAgentSessionSource, type CodeAgentSessionTerminalStatus, type CodeSurface, type CompareAgentProfileImprovementExperimentOptions, type CompareCandidateExperimentOptions, type CompareOptimizationMethodsOptions, type ComparisonCost, type CostLedgerHandle, type CostProvenanceSummary, type CreateChatClientOpts, type DefaultAnalystRegistryOptions, type DefaultProductionGateCheck, type DefaultProductionGateOptions, type DefaultProductionRewardHackingOptions, type DefineAgentEvalOptions, type DefinedAgentEval, type DeploymentOutcome, type DispatchFn as Dispatch, type DispatchContext, type EvalCellScoreDelta, type EvalDimensionDelta, type EvalGenerationDiff, type EvalReportingSuiteInput, type EvalReportingSuiteOptions, type EvalReportingSuiteResult, type EvalRunDiff, type EvaluatePairedMeasurementsOptions, type EvidenceVector, type ExecutionErrorOutcomeCell, type ExecutionInsight, type ExecutionReport, type ExternalOptimizationExample, type ExternalTextEvaluationResponse, type ExternalTextOptimizationMethodConfig, type ExternalTextOptimizerContext, type ExternalTextOptimizerResult, type FailureClassTally, type FailureClusterInsight, type FeedbackTableMeta, type FeedbackTableRow, FileSystemOutcomeStore, type FileSystemOutcomeStoreOptions, type FromFeedbackTableOptions, type FromFeedbackTableResult, type FromOtelSpansOptions, type FromRunRecordDirOptions, type FromRunRecordDirResult, type Gate, type GateCheckStatus, type GateContext, type GateContribution, type GateDecision, type GateResult, type GenerationCandidate, type GenerationRecord, type GepaAdaptiveEngineRun, type GepaEngineOptions, type GepaEngineRun, type GepaOptimizationMethodConfig, type GepaOptimizationRecipe, type GepaRunnerCommand, type HeldOutGateOptions, type HostedTenant, InMemoryOutcomeStore, type InsightReport, type InterRaterInsight, type JudgeConfig, type JudgeDimension, type JudgeInsight, type JudgeScore, type LiftInsight, type LlmJudgeDimension, type LlmJudgeOptions, type MutableSurface, type ObjectiveSource, type OpenAICompatibleOptimizerModel, type OptimizationMethod, type OptimizationMethodComparison, type OptimizationMethodInput, type OptimizationMethodProvenance, type OptimizationMethodResult, type OptimizationPackageSource, type OptimizationTokenUsage, type OptimizerConfig, type OptimizerModelBudget, type OutcomeCorrelationInsight, type OutcomeStore, type PairedMeasurement, type PairedMeasurementAdapter, type PairedMeasurementEvaluation, type ParetoSignificanceGateOptions, type ParsedCodeAgentJsonl, type PartitionByAuthoringModelResult, type PromotionObjective, type PromotionPolicy, type ProposalFinding, type ProposalFindingOrigin, REFERENCE_EQUIVALENCE_INPUT_LIMITS, REFERENCE_EQUIVALENCE_JUDGE_VERSION, type Recommendation, type ReferenceEquivalenceJudgeInput, type ReferenceEquivalenceJudgeOptions, type ReferenceEquivalenceJudgeResult, type ReferenceEquivalenceScenario, type ReleaseSummary, type RunAgentProfileImprovementExperimentOptions, type RunCampaignOptions, type RunCandidateExperimentOptions, type RunEvalOptions, type RunImprovementLoopOptions, type RunImprovementLoopResult, type RunRecordRejection, type ScalarDistribution, type Scenario, type SealAgentProfileImprovementSuiteOptions, type SealCandidateBenchmarkSuiteOptions, type SelfImproveBudget, type SelfImproveOptions, type SelfImproveProgressEvent, type SelfImproveResult, SelfImproveRunError, type SessionScript, type SkillOptOptimizationMethodConfig, type SkillOptRunnerCommand, type SkillOptTrainerConfig, type SummarizeExecutionOptions, type SurfaceProposer, type TokenUsageInsight, analyzeRuns, buildDefaultAnalystRegistry, buildEvidenceVector, campaignSplitDigest, compareOptimizationMethods, composeGate, createChatClient, createReferenceEquivalenceJudge, defaultProductionGate, defineAgentEval, diffGenerations, diffRunBaselineToWinner, diffRuns, evalReportingSuite, evaluatePairedMeasurements, externalTextOptimizationMethod, fromClaudeCodeSession, fromCodexSession, fromFeedbackTable, fromKimiCodeSession, fromOpenCodeSession, fromOtelSpans, fromPiSession, fromPigraphSession, fromRunRecordDir, fsCampaignStorage, gepaOptimizationMethod, heldOutGate, inMemoryCampaignStorage, llmJudge, makeProposalFinding, measuredComparisonFromAgentProfileImprovementExperiment, measuredComparisonFromCandidateExperiment, observeCodeAgentSession, paretoPolicy, paretoSignificanceGate, parseAgentTrace, parseCodeAgentJsonl, parseCodeAgentJsonlFile, partitionRunsByAuthoringModel, runAgentProfileImprovementExperiment, runCampaign, runCandidateExperiment, runEval, runImprovementLoop, runReferenceEquivalenceJudge, sealAgentProfileImprovementExperiment, sealAgentProfileImprovementSuite, sealAgentProfileImprovementTask, sealCandidateBenchmarkSuite, sealCandidateBenchmarkTask, sealCandidateExperiment, selfImprove, skillOptOptimizationMethod, streamCodeAgentJsonlFile, summarizeExecution, verifyAgentProfileImprovementExperiment, verifyAgentProfileImprovementExperimentComparison, verifyAgentProfileImprovementSuiteInputs, verifyAgentProfileImprovementTask, verifyCandidateBenchmarkSuite, verifyCandidateBenchmarkSuiteInputs, verifyCandidateBenchmarkTask, verifyCandidateExperiment, verifyCandidateExperimentComparison };
|
|
695
720
|
//# sourceMappingURL=index.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/contract/measured-comparison.ts","../../src/contract/profile-measured-comparison.ts","../../src/contract/intake/run-record-dir.ts","../../src/contract/eval-reporting-suite.ts","../../src/contract/diff.ts","../../src/contract/intake/agent-trace.ts","../../src/contract/intake/code-agent-observation.ts","../../src/contract/intake/code-agent-session.ts","../../src/contract/intake/feedback-table.ts","../../src/contract/intake/otel-spans.ts"],"mappings":";;;;;;;;;;;;;;;;UAiCiB;EACf,QAAQ,gCAAgC;EACxC;EACA;;UAGe;EACf,YAAY;EACZ;EACA,QAAQ;EACR,MAAM;EACN,eAAe;EACf;EACA,SAAS;;UAGM;EACf,YAAY;EACZ,QAAQ,OAAO,oCAAoC,QAAQ;;EAE3D;;EAEA,aAAa;EACb,SAAS;;UAGM;EACf,cAAc;EACd;IACE;IACA,MAAM;;;UAIO;EACf,YAAY;EACZ,cAAc;EACd;IACE;IACA,MAAM;;EAER,aAAa;EACb;EACA,YAAY;EACZ;EACA,WAAW;;;UAII,kBAAkB;EACjC;EACA,UAAU;EACV,WAAW;;;UAII,yBAAyB;EACxC,MAAM,KAAK;EACX,WAAW,KAAK;IAAkB;IAAc;;EAChD,QAAQ,KAAK;EACb,eAAe,KAAK,OAAO;EAC3B,UAAU,KAAK;EACf,UAAU,KAAK;EACf,OAAO,KAAK;;UAGG,kCAAkC;EACjD,uBAAuB,kBAAkB;EACzC,QAAQ;EACR,SAAS,yBAAyB;;EAElC;;EAEA,kBAAkB;;EAElB,kBAAkB;;;KAIR,8BAA8B,KACxC;EAGA,iBAAiB;EACjB,WAAW;EACX;;;iBAIc,2BACd,UAAU,sCACT;;iBAQa,4BACd,SAAS,qCACR;;iBAiBa,wBACd,UAAU,mCACT;iBAQa,0BAA0B,iBAAiB;;iBAarC,uBACpB,SAAS,gCACR,QAAQ;;;;;;;;iBAkEK,2BAA2B,MACzC,SAAS,kCAAkC,QAC1C;;iBA8Ya,0CACd,SAAS,oCACR;;iBAuEa,oCACd,iBACC;iBAkHa,6BAA6B,iBAAiB;iBAM9C,oCACd,iBACC;iBAkBa,8BAA8B;;;;;;;;;;UC7zB7B;EACf,aAAa;EACb,QAAQ,gCAAgC;EACxC;EACA;;UAGe;EACf,YAAY;EACZ;EACA,aAAa;EACb,MAAM;EACN,SAAS;EACT;EACA,SAAS;;UAGM;EACf,YAAY;EACZ,QACE,OAAO,kDACN,QAAQ;;EAEX;;EAEA,aAAa;EACb,SAAS;;UAGM;EACf,cAAc;EACd;IACE;IACA,MAAM;;;UAIO;EACf,YAAY;EACZ,cAAc;EACd;IACE;IACA,MAAM;;EAER,aAAa;EACb;EACA,YAAY;EACZ;EACA,WAAW;;;iBAIG,gCACd,UAAU,sCACT;;iBAQa,iCACd,SAAS,0CACR;;iBAqBa,sCACd,UAAU,4CACT;iBAOa,kCAAkC,iBAAiB;iBAInD,yCACd,iBACC;iBAIa,wCACd,iBACC;;;;;;iBASmB,qCACpB,SAAS,8CACR,QAAQ;;iBA0CK,wDACd,SAAS,kDACR;;iBAuEa,kDACd,iBACC;;;;UCnPc;;EAEf;;EAEA;;EAEA;;UAGe;;;;;;EAMf;;;;;;;EAOA,WAAW;;;;;;EAMX;;UAGe;;EAEf,MAAM;;EAEN,UAAU;;EAEV;;;;;;;;;;iBAkBoB,iBACpB,cACA,UAAS,0BACR,QAAQ;;;;;KC7CC,0BAA0B;UAErB;;;;;EAKf,UAAU,KAAK;;EAEf,OAAO;;;;;;;;;;EAUP;;;;UAKe;;;EAGf,QAAQ;;EAER;;IAEE;;IAEA;;;IAGA;;IAEA;;;IAGA,UAAU;;;EAGZ;;;;;;;iBAUoB,mBACpB,OAAO,yBACP,UAAS,4BACR,QAAQ;;;;;;UC5DM;EACf;EACA;EACA;;;UAIe;EACf;EACA;EACA;EACA;EACA;;;EAGA,YAAY,eAAe,eAAe;;;;UAK3B;EACf;EACA;EACA;EACA;EACA;;EAEA,SAAS;;EAET,SAAS;;EAET,OAAO;;EAEP;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;;;;UAMe;EACf;EACA;EACA;EACA;EACA,oBAAoB;EACpB,mBAAmB;EACnB;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA,cAAc;;;;EAId,aAAa;;;;;;;;;iBAoDC,gBACd,QAAQ,2BACR,OAAO,4BACN;;;;;;iBAqEa,SAAS,QAAQ,cAAc,OAAO,eAAe;;;;;;;iBAsCrD,wBAAwB,KAAK,eAAe;;;KC1OhD;UAEK;EACf,MAAM;;EAEN;;UAGe;EACf;EACA;EACA;;;EAGA,cAAc;;UAGC;EACf;EACA,cAAc;EACd,QAAQ;;UAGO;EACf;EACA,eAAe;;UAGA;EACf;EACA;EACA;EACA;IAAQ;IAAc;;EACtB;IAAS;IAAe;;EACxB,OAAO;;;;UAOQ;EACf;;EAEA;;EAEA;EACA;EACA;;EAEA;;EAEA;;KAGU,kBAAkB,YAAY;;;;;;iBAW1B,gBAAgB,SAAS,qBAAqB;UAmE7C;;;;EAIf,SAAS,YAAY;;;EAGrB,cAAc;;;;;;;;iBASA,8BACd,MAAM,aACN,OAAO,kBACN;;;KCjLS;KAEA;KAEA;KAEA;KASA;UAEK;EACf;EACA;EACA;;UAGe;EACf;EACA;EACA,MAAM;EACN,SAAS;EACT;EACA,QAAQ;EACR;EACA;EACA,UAAU;;UAGK;EACf,QAAQ;EACR;EACA;EACA;IACE,QAAQ;IACR;;EAEF,SAAS;;UAGM;EACf,QAAQ;EACR;EACA;EACA,YAAY;;;;;;;iBAeE,wBACd,SAAS,iCACR;;;
|
|
1
|
+
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/contract/measured-comparison.ts","../../src/contract/profile-measured-comparison.ts","../../src/contract/intake/run-record-dir.ts","../../src/contract/eval-reporting-suite.ts","../../src/contract/diff.ts","../../src/contract/intake/agent-trace.ts","../../src/contract/intake/code-agent-observation.ts","../../src/contract/intake/code-agent-session.ts","../../src/contract/intake/feedback-table.ts","../../src/contract/intake/otel-spans.ts"],"mappings":";;;;;;;;;;;;;;;;UAiCiB;EACf,QAAQ,gCAAgC;EACxC;EACA;;UAGe;EACf,YAAY;EACZ;EACA,QAAQ;EACR,MAAM;EACN,eAAe;EACf;EACA,SAAS;;UAGM;EACf,YAAY;EACZ,QAAQ,OAAO,oCAAoC,QAAQ;;EAE3D;;EAEA,aAAa;EACb,SAAS;;UAGM;EACf,cAAc;EACd;IACE;IACA,MAAM;;;UAIO;EACf,YAAY;EACZ,cAAc;EACd;IACE;IACA,MAAM;;EAER,aAAa;EACb;EACA,YAAY;EACZ;EACA,WAAW;;;UAII,kBAAkB;EACjC;EACA,UAAU;EACV,WAAW;;;UAII,yBAAyB;EACxC,MAAM,KAAK;EACX,WAAW,KAAK;IAAkB;IAAc;;EAChD,QAAQ,KAAK;EACb,eAAe,KAAK,OAAO;EAC3B,UAAU,KAAK;EACf,UAAU,KAAK;EACf,OAAO,KAAK;;UAGG,kCAAkC;EACjD,uBAAuB,kBAAkB;EACzC,QAAQ;EACR,SAAS,yBAAyB;;EAElC;;EAEA,kBAAkB;;EAElB,kBAAkB;;;KAIR,8BAA8B,KACxC;EAGA,iBAAiB;EACjB,WAAW;EACX;;;iBAIc,2BACd,UAAU,sCACT;;iBAQa,4BACd,SAAS,qCACR;;iBAiBa,wBACd,UAAU,mCACT;iBAQa,0BAA0B,iBAAiB;;iBAarC,uBACpB,SAAS,gCACR,QAAQ;;;;;;;;iBAkEK,2BAA2B,MACzC,SAAS,kCAAkC,QAC1C;;iBA8Ya,0CACd,SAAS,oCACR;;iBAuEa,oCACd,iBACC;iBAkHa,6BAA6B,iBAAiB;iBAM9C,oCACd,iBACC;iBAkBa,8BAA8B;;;;;;;;;;UC7zB7B;EACf,aAAa;EACb,QAAQ,gCAAgC;EACxC;EACA;;UAGe;EACf,YAAY;EACZ;EACA,aAAa;EACb,MAAM;EACN,SAAS;EACT;EACA,SAAS;;UAGM;EACf,YAAY;EACZ,QACE,OAAO,kDACN,QAAQ;;EAEX;;EAEA,aAAa;EACb,SAAS;;UAGM;EACf,cAAc;EACd;IACE;IACA,MAAM;;;UAIO;EACf,YAAY;EACZ,cAAc;EACd;IACE;IACA,MAAM;;EAER,aAAa;EACb;EACA,YAAY;EACZ;EACA,WAAW;;;iBAIG,gCACd,UAAU,sCACT;;iBAQa,iCACd,SAAS,0CACR;;iBAqBa,sCACd,UAAU,4CACT;iBAOa,kCAAkC,iBAAiB;iBAInD,yCACd,iBACC;iBAIa,wCACd,iBACC;;;;;;iBASmB,qCACpB,SAAS,8CACR,QAAQ;;iBA0CK,wDACd,SAAS,kDACR;;iBAuEa,kDACd,iBACC;;;;UCnPc;;EAEf;;EAEA;;EAEA;;UAGe;;;;;;EAMf;;;;;;;EAOA,WAAW;;;;;;EAMX;;UAGe;;EAEf,MAAM;;EAEN,UAAU;;EAEV;;;;;;;;;;iBAkBoB,iBACpB,cACA,UAAS,0BACR,QAAQ;;;;;KC7CC,0BAA0B;UAErB;;;;;EAKf,UAAU,KAAK;;EAEf,OAAO;;;;;;;;;;EAUP;;;;UAKe;;;EAGf,QAAQ;;EAER;;IAEE;;IAEA;;;IAGA;;IAEA;;;IAGA,UAAU;;;EAGZ;;;;;;;iBAUoB,mBACpB,OAAO,yBACP,UAAS,4BACR,QAAQ;;;;;;UC5DM;EACf;EACA;EACA;;;UAIe;EACf;EACA;EACA;EACA;EACA;;;EAGA,YAAY,eAAe,eAAe;;;;UAK3B;EACf;EACA;EACA;EACA;EACA;;EAEA,SAAS;;EAET,SAAS;;EAET,OAAO;;EAEP;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;;;;UAMe;EACf;EACA;EACA;EACA;EACA,oBAAoB;EACpB,mBAAmB;EACnB;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA,cAAc;;;;EAId,aAAa;;;;;;;;;iBAoDC,gBACd,QAAQ,2BACR,OAAO,4BACN;;;;;;iBAqEa,SAAS,QAAQ,cAAc,OAAO,eAAe;;;;;;;iBAsCrD,wBAAwB,KAAK,eAAe;;;KC1OhD;UAEK;EACf,MAAM;;EAEN;;UAGe;EACf;EACA;EACA;;;EAGA,cAAc;;UAGC;EACf;EACA,cAAc;EACd,QAAQ;;UAGO;EACf;EACA,eAAe;;UAGA;EACf;EACA;EACA;EACA;IAAQ;IAAc;;EACtB;IAAS;IAAe;;EACxB,OAAO;;;;UAOQ;EACf;;EAEA;;EAEA;EACA;EACA;;EAEA;;EAEA;;KAGU,kBAAkB,YAAY;;;;;;iBAW1B,gBAAgB,SAAS,qBAAqB;UAmE7C;;;;EAIf,SAAS,YAAY;;;EAGrB,cAAc;;;;;;;;iBASA,8BACd,MAAM,aACN,OAAO,kBACN;;;KCjLS;KAEA;KAEA;KAEA;KASA;UAEK;EACf;EACA;EACA;;UAGe;EACf;EACA;EACA,MAAM;EACN,SAAS;EACT;EACA,QAAQ;EACR;EACA;EACA,UAAU;;UAGK;EACf,QAAQ;EACR;EACA;EACA;IACE,QAAQ;IACR;;EAEF,SAAS;;UAGM;EACf,QAAQ;EACR;EACA;EACA,YAAY;;;;;;;iBAeE,wBACd,SAAS,iCACR;;;UClCc;EACf;EACA;;UAGe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf,QAAQ;EACR;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA,WAAW;EACX;;UAGe;EACf,MAAM;EACN,aAAa;EACb,SAAS;EACT,cAAc;;UAGC;EACf;EACA;EACA;EACA;EACA;EACA;EACA,WAAW;EACX;EACA;EACA;EACA;EACA;EACA;;;EAGA,iBAAiB;;;EAGjB,YAAY;;;;KAKF;EACN;EAAe;EAAoB;;EACnC;EAAmB;;iBAcT,oBAAoB,gBAAgB;;;;;;;;;;iBAuB7B,yBAAyB,eAAe,eAAe;;;;;iBA6BxD,wBAAwB,eAAe,QAAQ;iBAUrD,iBACd,SAAS,gCACR;iBAIa,sBACd,SAAS,gCACR;iBAIa,oBACd,SAAS,gCACR;iBAIa,oBACd,SAAS,gCACR;iBAIa,cACd,SAAS,gCACR;cAIU,2BAAkB;;;UCxNd;;;EAGf;;EAEA;;;EAGA;;;EAGA,WAAW;;UAGI;EACf;;;EAGA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA,WAAW;;;EAGX,SAAS;;UAGM;;EAEf,SAAS;;;EAGT,OAAO;;;;EAIP;IAAU;IAAa;;;;;EAIvB;;UAGe;EACf,MAAM;;;EAGN,aAAa;IAAQ;IAAe;IAAe;;;iBAGrC,kBAAkB,MAAM,2BAA2B;;;UCvBlD;EACf,OAAO;;EAEP,eAAe;;EAEf;;;;;;EAMA,eAAe,eAAe,gBAAgB;;iBAGhC,cAAc,MAAM,uBAAuB"}
|
package/dist/contract/index.js
CHANGED
|
@@ -19,6 +19,7 @@ import { a as externalTextOptimizationMethod, c as REFERENCE_EQUIVALENCE_INPUT_L
|
|
|
19
19
|
import { n as paretoPolicy, r as paretoSignificanceGate, t as buildEvidenceVector } from "../promotion-policy-xzA40Evo.js";
|
|
20
20
|
import { createHash } from "node:crypto";
|
|
21
21
|
import { agentCandidateBenchmarkSuiteSchema, agentCandidateBenchmarkTaskSchema, agentCandidateBundleSchema, agentCandidateEvaluationPolicySchema, agentCandidateExperimentSchema, agentImprovementMeasuredComparisonSchema, agentProfileImprovementExperimentSchema, agentProfileImprovementMeasuredComparisonSchema, agentProfileImprovementRunCellSchema, agentProfileImprovementRunReceiptSchema, agentProfileImprovementSuiteInputsSchema, agentProfileImprovementSuiteSchema, agentProfileImprovementTaskSchema, candidateExecutionEvidenceSchema, canonicalCandidateDigest, canonicalCandidateJson, numbersApproximatelyEqual, omitTopLevelDigest } from "@tangle-network/agent-interface";
|
|
22
|
+
import { createReadStream } from "node:fs";
|
|
22
23
|
import { dirname, join } from "node:path";
|
|
23
24
|
import { mkdir, readFile, readdir, stat, writeFile } from "node:fs/promises";
|
|
24
25
|
//#region src/contract/fixed-spend.ts
|
|
@@ -2214,18 +2215,80 @@ function timestamp(value) {
|
|
|
2214
2215
|
}
|
|
2215
2216
|
//#endregion
|
|
2216
2217
|
//#region src/contract/intake/code-agent-session.ts
|
|
2218
|
+
/** The single per-line rule. Both the string path and the file path call this,
|
|
2219
|
+
* so malformed-line handling and entry validation cannot drift apart. */
|
|
2220
|
+
function readCodeAgentJsonlLine(line, lineNumber) {
|
|
2221
|
+
const trimmed = line.trim();
|
|
2222
|
+
if (!trimmed) return void 0;
|
|
2223
|
+
try {
|
|
2224
|
+
return {
|
|
2225
|
+
kind: "entry",
|
|
2226
|
+
lineNumber,
|
|
2227
|
+
entry: JSON.parse(trimmed)
|
|
2228
|
+
};
|
|
2229
|
+
} catch {
|
|
2230
|
+
return {
|
|
2231
|
+
kind: "malformed",
|
|
2232
|
+
lineNumber
|
|
2233
|
+
};
|
|
2234
|
+
}
|
|
2235
|
+
}
|
|
2217
2236
|
function parseCodeAgentJsonl(jsonl) {
|
|
2218
2237
|
const entries = [];
|
|
2219
2238
|
let malformedLines = 0;
|
|
2239
|
+
let lineNumber = 0;
|
|
2220
2240
|
for (const line of jsonl.split("\n")) {
|
|
2221
|
-
|
|
2222
|
-
|
|
2223
|
-
|
|
2224
|
-
|
|
2225
|
-
|
|
2226
|
-
|
|
2241
|
+
lineNumber += 1;
|
|
2242
|
+
const read = readCodeAgentJsonlLine(line, lineNumber);
|
|
2243
|
+
if (!read) continue;
|
|
2244
|
+
if (read.kind === "malformed") malformedLines += 1;
|
|
2245
|
+
else entries.push(read.entry);
|
|
2246
|
+
}
|
|
2247
|
+
return {
|
|
2248
|
+
entries,
|
|
2249
|
+
malformedLines
|
|
2250
|
+
};
|
|
2251
|
+
}
|
|
2252
|
+
/** Reads a transcript one line at a time and never holds the file as a single
|
|
2253
|
+
* string. `parseCodeAgentJsonl` needs the whole file in one string, so a
|
|
2254
|
+
* session above V8's ~512MB string ceiling throws `ERR_STRING_TOO_LONG` and
|
|
2255
|
+
* cannot be ingested at all; the largest real Codex rollout on record is 695MB.
|
|
2256
|
+
*
|
|
2257
|
+
* Lines break on `\n` only, which is what the string path's `split('\n')` does.
|
|
2258
|
+
* `node:readline` also breaks on a bare `\r`, so it is deliberately not used
|
|
2259
|
+
* here: a lone carriage return inside a line must stay inside that line for the
|
|
2260
|
+
* two paths to report the same malformed count. */
|
|
2261
|
+
async function* streamCodeAgentJsonlFile(path) {
|
|
2262
|
+
const stream = createReadStream(path, { encoding: "utf8" });
|
|
2263
|
+
let pending = "";
|
|
2264
|
+
let lineNumber = 0;
|
|
2265
|
+
try {
|
|
2266
|
+
for await (const chunk of stream) {
|
|
2267
|
+
pending += chunk;
|
|
2268
|
+
let start = 0;
|
|
2269
|
+
for (let at = pending.indexOf("\n"); at !== -1; at = pending.indexOf("\n", start)) {
|
|
2270
|
+
lineNumber += 1;
|
|
2271
|
+
const read = readCodeAgentJsonlLine(pending.slice(start, at), lineNumber);
|
|
2272
|
+
if (read) yield read;
|
|
2273
|
+
start = at + 1;
|
|
2274
|
+
}
|
|
2275
|
+
pending = pending.slice(start);
|
|
2227
2276
|
}
|
|
2277
|
+
} finally {
|
|
2278
|
+
stream.destroy();
|
|
2228
2279
|
}
|
|
2280
|
+
const last = readCodeAgentJsonlLine(pending, lineNumber + 1);
|
|
2281
|
+
if (last) yield last;
|
|
2282
|
+
}
|
|
2283
|
+
/** Streaming counterpart to `parseCodeAgentJsonl` for a transcript on disk.
|
|
2284
|
+
* It returns the same shape, so a caller that holds every entry keeps working
|
|
2285
|
+
* above the string ceiling. The entry array still grows with the transcript;
|
|
2286
|
+
* consume `streamCodeAgentJsonlFile` directly when memory must stay flat. */
|
|
2287
|
+
async function parseCodeAgentJsonlFile(path) {
|
|
2288
|
+
const entries = [];
|
|
2289
|
+
let malformedLines = 0;
|
|
2290
|
+
for await (const read of streamCodeAgentJsonlFile(path)) if (read.kind === "malformed") malformedLines += 1;
|
|
2291
|
+
else entries.push(read.entry);
|
|
2229
2292
|
return {
|
|
2230
2293
|
entries,
|
|
2231
2294
|
malformedLines
|
|
@@ -3358,6 +3421,6 @@ function collectNumericAttrs(spans) {
|
|
|
3358
3421
|
return raw;
|
|
3359
3422
|
}
|
|
3360
3423
|
//#endregion
|
|
3361
|
-
export { FileSystemOutcomeStore, InMemoryOutcomeStore, REFERENCE_EQUIVALENCE_INPUT_LIMITS, REFERENCE_EQUIVALENCE_JUDGE_VERSION, SelfImproveRunError, analyzeRuns, buildDefaultAnalystRegistry, buildEvidenceVector, campaignSplitDigest, compareOptimizationMethods, composeGate, createChatClient, createReferenceEquivalenceJudge, defaultProductionGate, defineAgentEval, diffGenerations, diffRunBaselineToWinner, diffRuns, evalReportingSuite, evaluatePairedMeasurements, externalTextOptimizationMethod, fromClaudeCodeSession, fromCodexSession, fromFeedbackTable, fromKimiCodeSession, fromOpenCodeSession, fromOtelSpans, fromPiSession, fromPigraphSession, fromRunRecordDir, fsCampaignStorage, gepaOptimizationMethod, heldOutGate, inMemoryCampaignStorage, llmJudge, makeProposalFinding, measuredComparisonFromAgentProfileImprovementExperiment, measuredComparisonFromCandidateExperiment, observeCodeAgentSession, paretoPolicy, paretoSignificanceGate, parseAgentTrace, parseCodeAgentJsonl, partitionRunsByAuthoringModel, runAgentProfileImprovementExperiment, runCampaign, runCandidateExperiment, runEval, runImprovementLoop, runReferenceEquivalenceJudge, sealAgentProfileImprovementExperiment, sealAgentProfileImprovementSuite, sealAgentProfileImprovementTask, sealCandidateBenchmarkSuite, sealCandidateBenchmarkTask, sealCandidateExperiment, selfImprove, skillOptOptimizationMethod, summarizeExecution, verifyAgentProfileImprovementExperiment, verifyAgentProfileImprovementExperimentComparison, verifyAgentProfileImprovementSuiteInputs, verifyAgentProfileImprovementTask, verifyCandidateBenchmarkSuite, verifyCandidateBenchmarkSuiteInputs, verifyCandidateBenchmarkTask, verifyCandidateExperiment, verifyCandidateExperimentComparison };
|
|
3424
|
+
export { FileSystemOutcomeStore, InMemoryOutcomeStore, REFERENCE_EQUIVALENCE_INPUT_LIMITS, REFERENCE_EQUIVALENCE_JUDGE_VERSION, SelfImproveRunError, analyzeRuns, buildDefaultAnalystRegistry, buildEvidenceVector, campaignSplitDigest, compareOptimizationMethods, composeGate, createChatClient, createReferenceEquivalenceJudge, defaultProductionGate, defineAgentEval, diffGenerations, diffRunBaselineToWinner, diffRuns, evalReportingSuite, evaluatePairedMeasurements, externalTextOptimizationMethod, fromClaudeCodeSession, fromCodexSession, fromFeedbackTable, fromKimiCodeSession, fromOpenCodeSession, fromOtelSpans, fromPiSession, fromPigraphSession, fromRunRecordDir, fsCampaignStorage, gepaOptimizationMethod, heldOutGate, inMemoryCampaignStorage, llmJudge, makeProposalFinding, measuredComparisonFromAgentProfileImprovementExperiment, measuredComparisonFromCandidateExperiment, observeCodeAgentSession, paretoPolicy, paretoSignificanceGate, parseAgentTrace, parseCodeAgentJsonl, parseCodeAgentJsonlFile, partitionRunsByAuthoringModel, runAgentProfileImprovementExperiment, runCampaign, runCandidateExperiment, runEval, runImprovementLoop, runReferenceEquivalenceJudge, sealAgentProfileImprovementExperiment, sealAgentProfileImprovementSuite, sealAgentProfileImprovementTask, sealCandidateBenchmarkSuite, sealCandidateBenchmarkTask, sealCandidateExperiment, selfImprove, skillOptOptimizationMethod, streamCodeAgentJsonlFile, summarizeExecution, verifyAgentProfileImprovementExperiment, verifyAgentProfileImprovementExperimentComparison, verifyAgentProfileImprovementSuiteInputs, verifyAgentProfileImprovementTask, verifyCandidateBenchmarkSuite, verifyCandidateBenchmarkSuiteInputs, verifyCandidateBenchmarkTask, verifyCandidateExperiment, verifyCandidateExperimentComparison };
|
|
3362
3425
|
|
|
3363
3426
|
//# sourceMappingURL=index.js.map
|