@tangle-network/traces 0.15.0 → 0.15.3

package/dist/index.js

@@ -51,10 +51,8 @@ import {
   SPAN_SYNTHESIZED_ATTR,
   SPAWN_JOURNAL_FILE,
   STATUS_MARK,
-  SUBMIT_ACTION_SIGNATURE,
   SYNTHESIZED_SOURCE_ATTR,
   SYNTHESIZED_SPAN_ATTR,
-  SandboxCounterfactualRunner,
   SessionWorkflowError,
   TRACES_REDACTION_RULES,
   TRACE_QUESTION_PREVIEW_HEAD_CHARS,
@@ -73,9 +71,7 @@ import {
   assembleSessionBundle,
   assertAnswerSchema,
   assertSandboxReachable,
-  buildFixPrompt,
   buildPolicyEvidenceRecord,
-  buildRetryFixPrompt,
   buildSessionFactsReport,
   buildSessionIndex,
   buildSessionIndexFromRows,
@@ -87,12 +83,7 @@ import {
   classifyFailureFollowUps,
   classifyLiveActions,
   classifyReaction,
-  classifyVerdict,
   claudeActor,
-  clipText,
-  cmdReplayVerify,
-  cmdReplayVerifyBatch,
-  cmdVerifyFindings,
   codexActor,
   collectPolicyEvidence,
   collectSessionIndex,
@@ -104,19 +95,14 @@ import {
   computeSessionFacts,
   conformanceOfSpans,
   connectors,
-  countScriptCommands,
   countSkillRunsJsonl,
   createBundleSourceReader,
   createFindingRejectionTally,
   cwdMatchesSelection,
   cwdMatchesSelectionResolved,
   defaultTraceLiveAnalysts,
-  deriveFailureSignature,
-  derivedImageTag,
   describeSessionRelationship,
-  dockerImagePreparer,
   engineCallReservationFloorUsd,
-  enumerateReplayableCases,
   environmentAttributes,
   environmentFromSpanAttributes,
   equivalentGitCwds,
@@ -125,7 +111,6 @@ import {
   exportTraceEvidenceRows,
   exportTraceEvidenceText,
   externalFailureMessage,
-  extractFixCommand,
   fileRunContextSupervisorRunReader,
   fileSha256,
   findContextRoot,
@@ -133,12 +118,8 @@ import {
   findSpanSourceFiles,
   findingRejection,
   findingRejectionDetail,
-  findingReplayStep,
-  findingTrajectoryId,
   forEachTreeNode,
   formatFindingRejections,
-  generateFixCommand,
-  goldIncorrectSteps,
   haloAnalyzer,
   hashSessionBundleFiles,
   hodoscopeAnalyzer,
@@ -148,7 +129,6 @@ import {
   indexSessionIdsByTrace,
   inferRole,
   inferSpanKind,
-  ingestCodeTraceBenchSteps,
   inspectSessionIndex,
   int,
   isBridgeMismatchError,
@@ -156,7 +136,6 @@ import {
   isChatTrajectoryMessage,
   isFileRunContextDir,
   isInheritedSpan,
-  isSubmitAction,
   isSynthesizedSpan,
   knownHarnesses,
   listAdapters,
@@ -174,27 +153,18 @@ import {
   normalizeTraceQuestions,
   otlpRowToSpan,
   outboundHash,
-  parseCorpusFlag,
-  parseIncorrectStepsSubject,
   parseIsoToEpochMs,
   parseJsonAnswer,
-  parseObservationOutput,
-  parseRecordedReturncode,
-  parseReplayBatchArgs,
-  parseReplayVerifyArgs,
   parseSession,
   parseSince,
-  parseVerifyFindingsArgs,
   planTraceAgenticRoute,
   planUpload,
   primeAnalyzer,
   projectSessionBundle,
   readCoordinationLog,
   readFileRunContext,
-  readFindingsFile,
   readJsonFile,
   readJsonl,
-  readLabelEntries,
   readOtlpInput,
   readPullRequests,
   readRunContextJournal,
@@ -205,7 +175,6 @@ import {
   readSpanRunSnapshot,
   recordSessionCorruption,
   redactSpans,
-  renderBatchReport,
   renderConformance,
   renderInspectionReport,
   renderRunContextSnapshot,
@@ -217,15 +186,7 @@ import {
   renderTraceQuestionPrompt,
   renderTraceQuestionsReport,
   renderValidation,
-  renderVerifiedFindingsSection,
-  replayBatchUsage,
-  replayVerify,
-  replayVerifyFinding,
-  replayVerifyUsage,
   resolveAdapter,
-  resolveCaseResources,
-  resolveFindingInvocation,
-  resolveFindingReplayability,
   resolveOtlpInputFiles,
   resolveRepoAttrs,
   resolveRunContextDir,
@@ -233,9 +194,7 @@ import {
   resolveSessionRepoAttrs,
   runCommand,
   runExternalAnalyzers,
-  runFixLoop,
   runPipelines,
-  runReplayBatch,
   runTotals,
   runTraceImprovement,
   runTraceInvestigation,
@@ -245,7 +204,6 @@ import {
   saveReport,
   saveState,
   scanSessions,
-  seededSample,
   selectAdapters,
   serializeInspectionReport,
   serializePolicyEvidence,
@@ -287,13 +245,12 @@ import {
   userPromptSpan,
   validateOtlpSpans,
   validationExitCode,
-  verifyFindings,
-  verifyFindingsUsage,
+  verifyFindingsInSandbox,
+  verifySessionBundle,
   watchRunTarget,
   watchSessions,
   withSessionFactsContext,
   wrap,
-  wrapActionForExec,
   writeHodoscopeInput,
   writeInspectionReportFile,
   writeOtlpFile,
@@ -305,7 +262,7 @@ import {
   writeTraceQuestionsArtifacts,
   zaiChatCaller,
   zeroPricedCost
-} from "./chunk-7DOBU7BJ.js";
+} from "./chunk-U4EB4HK7.js";
 
 // src/collect.ts
 async function collectSessions(opts = {}) {
@@ -322,6 +279,7 @@ async function collectSessions(opts = {}) {
 }
 
 // src/index.ts
+export * from "@tangle-network/agent-eval/trajectory-replay";
 import {
   AnalystRegistry,
   buildDefaultAnalystRegistry,
@@ -384,10 +342,8 @@ export {
   SPAN_SYNTHESIZED_ATTR,
   SPAWN_JOURNAL_FILE,
   STATUS_MARK,
-  SUBMIT_ACTION_SIGNATURE,
   SYNTHESIZED_SOURCE_ATTR,
   SYNTHESIZED_SPAN_ATTR,
-  SandboxCounterfactualRunner,
   SessionWorkflowError,
   TRACES_REDACTION_RULES,
   TRACE_QUESTION_PREVIEW_HEAD_CHARS,
@@ -407,9 +363,7 @@ export {
   assertAnswerSchema,
   assertSandboxReachable,
   buildDefaultAnalystRegistry,
-  buildFixPrompt,
   buildPolicyEvidenceRecord,
-  buildRetryFixPrompt,
   buildSessionFactsReport,
   buildSessionIndex,
   buildSessionIndexFromRows,
@@ -421,12 +375,7 @@ export {
   classifyFailureFollowUps,
   classifyLiveActions,
   classifyReaction,
-  classifyVerdict,
   claudeActor,
-  clipText,
-  cmdReplayVerify,
-  cmdReplayVerifyBatch,
-  cmdVerifyFindings,
   codexActor,
   collectPolicyEvidence,
   collectSessionIndex,
@@ -439,7 +388,6 @@ export {
   computeSessionFacts,
   conformanceOfSpans,
   connectors,
-  countScriptCommands,
   countSkillRunsJsonl,
   createBundleSourceReader,
   createFindingRejectionTally,
@@ -447,12 +395,8 @@ export {
   cwdMatchesSelection,
   cwdMatchesSelectionResolved,
   defaultTraceLiveAnalysts,
-  deriveFailureSignature,
-  derivedImageTag,
   describeSessionRelationship,
-  dockerImagePreparer,
   engineCallReservationFloorUsd,
-  enumerateReplayableCases,
   environmentAttributes,
   environmentFromSpanAttributes,
   equivalentGitCwds,
@@ -461,7 +405,6 @@ export {
   exportTraceEvidenceRows,
   exportTraceEvidenceText,
   externalFailureMessage,
-  extractFixCommand,
   fileRunContextSupervisorRunReader,
   fileSha256,
   findContextRoot,
@@ -469,12 +412,8 @@ export {
   findSpanSourceFiles,
   findingRejection,
   findingRejectionDetail,
-  findingReplayStep,
-  findingTrajectoryId,
   forEachTreeNode,
   formatFindingRejections,
-  generateFixCommand,
-  goldIncorrectSteps,
   haloAnalyzer,
   hashSessionBundleFiles,
   hodoscopeAnalyzer,
@@ -485,7 +424,6 @@ export {
   indexSessionIdsByTrace,
   inferRole,
   inferSpanKind,
-  ingestCodeTraceBenchSteps,
   inspectSessionIndex,
   int,
   isBridgeMismatchError,
@@ -493,7 +431,6 @@ export {
   isChatTrajectoryMessage,
   isFileRunContextDir,
   isInheritedSpan,
-  isSubmitAction,
   isSynthesizedSpan,
   knownHarnesses,
   listAdapters,
@@ -512,27 +449,18 @@ export {
   normalizeTraceQuestions,
   otlpRowToSpan,
   outboundHash,
-  parseCorpusFlag,
-  parseIncorrectStepsSubject,
   parseIsoToEpochMs,
   parseJsonAnswer,
-  parseObservationOutput,
-  parseRecordedReturncode,
-  parseReplayBatchArgs,
-  parseReplayVerifyArgs,
   parseSession,
   parseSince,
-  parseVerifyFindingsArgs,
   planTraceAgenticRoute,
   planUpload,
   primeAnalyzer,
   projectSessionBundle,
   readCoordinationLog,
   readFileRunContext,
-  readFindingsFile,
   readJsonFile,
   readJsonl,
-  readLabelEntries,
   readOtlpInput,
   readPullRequests,
   readRunContextJournal,
@@ -545,7 +473,6 @@ export {
   redactSpans,
   redactString,
   redactValue,
-  renderBatchReport,
   renderConformance,
   renderInspectionReport,
   renderRunContextSnapshot,
@@ -557,15 +484,7 @@ export {
   renderTraceQuestionPrompt,
   renderTraceQuestionsReport,
   renderValidation,
-  renderVerifiedFindingsSection,
-  replayBatchUsage,
-  replayVerify,
-  replayVerifyFinding,
-  replayVerifyUsage,
   resolveAdapter,
-  resolveCaseResources,
-  resolveFindingInvocation,
-  resolveFindingReplayability,
   resolveOtlpInputFiles,
   resolveRepoAttrs,
   resolveRunContextDir,
@@ -573,9 +492,7 @@ export {
   resolveSessionRepoAttrs,
   runCommand,
   runExternalAnalyzers,
-  runFixLoop,
   runPipelines,
-  runReplayBatch,
   runTotals,
   runTraceImprovement,
   runTraceInvestigation,
@@ -585,7 +502,6 @@ export {
   saveReport,
   saveState,
   scanSessions,
-  seededSample,
   selectAdapters,
   serializeInspectionReport,
   serializePolicyEvidence,
@@ -627,13 +543,12 @@ export {
   userPromptSpan,
   validateOtlpSpans,
   validationExitCode,
-  verifyFindings,
-  verifyFindingsUsage,
+  verifyFindingsInSandbox,
+  verifySessionBundle,
   watchRunTarget,
   watchSessions,
   withSessionFactsContext,
   wrap,
-  wrapActionForExec,
   writeHodoscopeInput,
   writeInspectionReportFile,
   writeOtlpFile,

package/docs/replay-verify.md

@@ -7,7 +7,8 @@ An analyst finding today is a cited claim: "step 37 is where the run went wrong.
 - **Arm B** executes a corrected step k and checks the failure vanishes (exit 0 and the signature absent).
 
 Each arm gets a fresh sandbox with its own prefix replay, so arm B is never contaminated by arm A's side effects.
-Under the hood each arm is an agent-eval `runCounterfactual` meta-run (`layer='meta'`, `parentRunId` = the ingested trajectory run), with the sandbox-backed `CounterfactualRunner` from `src/replay-verify.ts` supplying the `executeFrom` callback that scaffold leaves to consumers.
+Under the hood each arm is an agent-eval `runCounterfactual` meta-run (`layer='meta'`, `parentRunId` = the ingested trajectory run), with the `SandboxCounterfactualRunner` from `@tangle-network/agent-eval/trajectory-replay` supplying the `executeFrom` callback that scaffold leaves to consumers.
+The replay engine is agent-eval's; traces supplies only the Tangle Sandbox execution backend and the CLI (`src/replay.ts`).
 
 ## Usage
 
@@ -79,7 +80,7 @@ Fix generation (`--fix generate`) runs ONE chat completion per arm-A-reproduced
 
 ### Iterative fix loop (`--fix loop`)
 
-`--fix loop` (in `src/replay-fix-loop.ts`) replaces the single shot with an execution-feedback loop of up to `--fix-attempts` attempts (default 3) per case:
+`--fix loop` (agent-eval `runFixLoop`) replaces the single shot with an execution-feedback loop of up to `--fix-attempts` attempts (default 3) per case:
 
 - Attempt 1 is byte-identical to the one-shot prompt, so the report's `fixFlipAttempt1` stays directly comparable to `--fix generate`.
 - When an arm fails (nonzero exit or the failure signature persists) or the model call itself fails, the next prompt carries every prior command with its REAL executed stdout/stderr (clipped tails), never a paraphrase.
@@ -97,14 +98,14 @@ Outputs: `batch-report.json`, `batch-report.md` (headline + exclusion + pull-fai
 
 ## Analyst wire — finding in, executed proof out
 
-`replayVerifyFinding` (exported from the package root, `src/replay-wire.ts`) is the entry point the analyst product calls:
+`replayVerifyFinding` (from `@tangle-network/agent-eval/trajectory-replay`, re-exported by this package) is the entry point the analyst product calls:
 
 ```ts
-import { replayVerifyFinding } from '@tangle-network/traces'
+import { replayVerifyFinding, sandboxReplayBackend } from '@tangle-network/traces'
 
 const { invocation, fixCommand, verdict } = await replayVerifyFinding(
   { trajId: 'miniswe-…-zstd-1733-786102c1', subject: 'incorrect-steps-37-37-unescaped-consequence-39' },
-  { corpora, out: './proof', fixCaller, baseUrl, apiKey },
+  { corpora, out: './proof', fixCaller, backendFactory: (image) => sandboxReplayBackend({ image, apiKey, baseUrl }) },
 )
 ```
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/traces",
-  "version": "0.15.0",
+  "version": "0.15.3",
   "description": "Point it at your coding-agent session traces (Claude Code, Codex, OpenCode, Gemini, Pi, …) and get failure-mode + efficiency findings. CLI + SDK over the @tangle-network/agent-eval analyst suite — observe live sessions, run your own analysts, redact, and upload to the Tangle Intelligence Platform.",
   "type": "module",
   "license": "MIT",
@@ -60,19 +60,19 @@
     "prepublishOnly": "pnpm check:source && pnpm build && pnpm check:package"
   },
   "dependencies": {
-    "@tangle-network/agent-eval": "0.182.0",
-    "@tangle-network/agent-interface": "2.6.1",
-    "@tangle-network/agent-runtime": "0.231.1",
-    "@tangle-network/agent-trace-contract": "1.0.2",
-    "@tangle-network/sandbox": "0.40.2"
+    "@tangle-network/agent-eval": "0.186.2",
+    "@tangle-network/agent-interface": "2.12.0",
+    "@tangle-network/agent-runtime": "0.257.0",
+    "@tangle-network/agent-trace-contract": "1.2.0",
+    "@tangle-network/sandbox": "0.45.0"
   },
   "devDependencies": {
-    "@types/node": "^22.0.0",
-    "postcss": "^8.5.26",
-    "tsup": "^8.0.0",
-    "tsx": "^4.23.13",
-    "typescript": "^5.7.0",
-    "vite": "^8.2.2",
-    "vitest": "^4.1.11"
+    "@types/node": "^22.20.4",
+    "postcss": "^8.5.28",
+    "tsup": "^8.5.1",
+    "tsx": "^4.23.15",
+    "typescript": "~5.9.3",
+    "vite": "^8.3.0",
+    "vitest": "^5.0.1"
   }
 }