@tangle-network/traces 0.11.6 → 0.12.0

package/dist/{chunk-MEAEXKUV.js → chunk-CV22VGCK.js}

@@ -1816,6 +1816,76 @@ function summarizeSpanExecution(spans, opts = {}) {
   });
 }
 
+// src/failure-followup.ts
+import { OPENINFERENCE_SPAN_KIND as OPENINFERENCE_SPAN_KIND2, TOOL_NAME as TOOL_NAME3 } from "@tangle-network/agent-eval/trace-attributes";
+function argKeyOf(span2) {
+  const sha = span2.attributes["traces.input.sha256"];
+  if (typeof sha === "string" && sha.length > 0) return `sha:${sha}`;
+  const value = span2.attributes["input.value"];
+  if (typeof value === "string" && value.length > 0) return `raw:${value}`;
+  return null;
+}
+function schemeOf(key) {
+  return key.slice(0, key.indexOf(":"));
+}
+function classifyFailureFollowUps(spans) {
+  const calls = [];
+  for (const s of spans) {
+    if (s.attributes[OPENINFERENCE_SPAN_KIND2] !== "TOOL") continue;
+    const toolName4 = s.attributes[TOOL_NAME3];
+    if (typeof toolName4 !== "string" || toolName4.length === 0) continue;
+    calls.push({ span: s, toolName: toolName4, argKey: argKeyOf(s) });
+  }
+  calls.sort(
+    (a, b) => a.span.start_time.localeCompare(b.span.start_time) || (Number(a.span.attributes.step) || 0) - (Number(b.span.attributes.step) || 0)
+  );
+  const sequences = /* @__PURE__ */ new Map();
+  for (const call of calls) {
+    const key = `${call.span.trace_id}|${call.toolName}`;
+    const seq = sequences.get(key);
+    if (seq) seq.push(call);
+    else sequences.set(key, [call]);
+  }
+  const items = [];
+  const blindByTool = {};
+  for (const seq of sequences.values()) {
+    for (let i = 0; i < seq.length; i++) {
+      const call = seq[i];
+      if (call.span.status.code !== "ERROR") continue;
+      if (call.span.attributes["traces.expected_blocking"] === true) continue;
+      const followUp = seq[i + 1] ?? null;
+      let kind;
+      if (!followUp) {
+        kind = "none";
+      } else if (call.argKey !== null && followUp.argKey !== null && schemeOf(call.argKey) === schemeOf(followUp.argKey)) {
+        kind = call.argKey === followUp.argKey ? "blind" : "adapted";
+      } else {
+        kind = "args-unknown";
+      }
+      if (kind === "blind") blindByTool[call.toolName] = (blindByTool[call.toolName] ?? 0) + 1;
+      items.push({
+        traceId: call.span.trace_id,
+        toolName: call.toolName,
+        failedSpanId: call.span.span_id,
+        followUpSpanId: followUp?.span.span_id ?? null,
+        kind,
+        followUpSucceeded: followUp ? followUp.span.status.code !== "ERROR" : null
+      });
+    }
+  }
+  const followedItems = items.filter((item) => item.kind !== "none");
+  return {
+    failures: items.length,
+    followed: followedItems.length,
+    blind: items.filter((item) => item.kind === "blind").length,
+    adapted: items.filter((item) => item.kind === "adapted").length,
+    argsUnknown: items.filter((item) => item.kind === "args-unknown").length,
+    followUpSucceeded: followedItems.filter((item) => item.followUpSucceeded === true).length,
+    blindByTool,
+    items
+  };
+}
+
 // src/runtime-store.ts
 import { InMemoryTraceStore } from "@tangle-network/agent-eval";
 var KIND = {
@@ -1904,7 +1974,7 @@ async function runPipelines(spans, opts = {}) {
     failureClusterView(store),
     Promise.all(runIds.map((runId) => computeToolUseMetrics(store, runId)))
   ]);
-  return { stuckLoops, failureClusters, toolUse };
+  return { stuckLoops, failureClusters, toolUse, failureFollowUps: classifyFailureFollowUps(spans) };
 }
 
 // src/integrity.ts
@@ -6056,7 +6126,7 @@ function selectAdapters(opts) {
 }
 
 // src/repo.ts
-import { stat as stat13 } from "fs/promises";
+import { realpath as realpath3, stat as stat13 } from "fs/promises";
 import { dirname as dirname6, isAbsolute as isAbsolute4, join as join17, normalize, parse as parsePath } from "path";
 function normalizeRemote(url) {
   const raw = url.trim();
@@ -6135,6 +6205,13 @@ function uniqNormalized(paths) {
 function parseWorktreePorcelain(output) {
   return output.split("\n").filter((line) => line.startsWith("worktree ")).map((line) => line.slice("worktree ".length).trim()).filter(Boolean);
 }
+async function realpathOrNull(path) {
+  try {
+    return await realpath3(path);
+  } catch {
+    return null;
+  }
+}
 async function equivalentGitCwds(cwd) {
   if (!cwd) return [];
   const selected = normalize(cwd);
@@ -6142,7 +6219,20 @@ async function equivalentGitCwds(cwd) {
   const top = await gitOutput(selected, ["rev-parse", "--show-toplevel"]);
   if (!top) return [selected];
   const worktrees = await gitOutput(top, ["worktree", "list", "--porcelain"]);
-  return uniqNormalized([selected, top, ...worktrees ? parseWorktreePorcelain(worktrees) : []]);
+  const candidates = [selected, top, ...worktrees ? parseWorktreePorcelain(worktrees) : []];
+  const expanded = [];
+  for (const candidate of candidates) {
+    expanded.push(candidate);
+    const resolved = await realpathOrNull(candidate);
+    if (resolved) expanded.push(resolved);
+  }
+  return uniqNormalized(expanded);
+}
+async function cwdMatchesSelectionResolved(cwd, selections) {
+  if (cwdMatchesSelection(cwd, selections)) return true;
+  if (!cwd || !isAbsolute4(cwd)) return false;
+  const resolved = await realpathOrNull(normalize(cwd));
+  return resolved !== null && resolved !== normalize(cwd) && cwdMatchesSelection(resolved, selections);
 }
 function cwdMatchesSelection(cwd, selections) {
   if (!cwd) return false;
@@ -6469,8 +6559,15 @@ async function locateSessions(adapter, opts = {}) {
   const cwdSelections = await equivalentGitCwds(opts.cwd);
   const refs = await adapter.locate({ sinceMs: opts.sinceMs });
   const byKey = /* @__PURE__ */ new Map();
+  const resolvedMatch = /* @__PURE__ */ new Map();
   for (const ref of refs) {
-    if (!cwdMatchesSelection(ref.cwd, cwdSelections) && !(ref.cwd === null && ref.integrity)) continue;
+    let matched = cwdMatchesSelection(ref.cwd, cwdSelections);
+    if (!matched && ref.cwd !== null) {
+      const cached = resolvedMatch.get(ref.cwd);
+      matched = cached ?? await cwdMatchesSelectionResolved(ref.cwd, cwdSelections);
+      if (cached === void 0) resolvedMatch.set(ref.cwd, matched);
+    }
+    if (!matched && !(ref.cwd === null && ref.integrity)) continue;
     byKey.set(refKey(ref), ref);
   }
   return [...byKey.values()].sort((a, b) => b.mtimeMs - a.mtimeMs);
@@ -6605,8 +6702,8 @@ import { mkdtemp as mkdtemp2, writeFile as writeFile3 } from "fs/promises";
 import { tmpdir as tmpdir2 } from "os";
 import { join as join18 } from "path";
 import {
-  OPENINFERENCE_SPAN_KIND as OPENINFERENCE_SPAN_KIND2,
-  TOOL_NAME as TOOL_NAME3
+  OPENINFERENCE_SPAN_KIND as OPENINFERENCE_SPAN_KIND3,
+  TOOL_NAME as TOOL_NAME4
 } from "@tangle-network/agent-eval/trace-attributes";
 var POLICY_EVIDENCE_NOTE = "This is normalized coding-agent session evidence for downstream policy mining; it is not an eval campaign cell.";
 function stringAttr(span2, key) {
@@ -6614,7 +6711,7 @@ function stringAttr(span2, key) {
   return typeof value === "string" && value.length > 0 ? value : void 0;
 }
 function spanKind(span2) {
-  return stringAttr(span2, OPENINFERENCE_SPAN_KIND2);
+  return stringAttr(span2, OPENINFERENCE_SPAN_KIND3);
 }
 function repoFromSpans(spans) {
   const attrs = {};
@@ -6640,7 +6737,7 @@ function summarizeTools(spans) {
   const byTool = /* @__PURE__ */ new Map();
   for (const span2 of spans) {
     if (spanKind(span2) !== "TOOL") continue;
-    const name = stringAttr(span2, TOOL_NAME3) ?? span2.name.replace(/^tool\./, "");
+    const name = stringAttr(span2, TOOL_NAME4) ?? span2.name.replace(/^tool\./, "");
     const current = byTool.get(name) ?? { calls: 0, errors: 0 };
     current.calls += 1;
     if (span2.status.code === "ERROR") current.errors += 1;
@@ -13478,11 +13575,27 @@ function renderPipelines(pr, unavailable) {
         byTool.set(name, aggregate);
       }
     }
-    const failureFollowUp = errorCalls > 0 ? `; ${followedFailures}/${errorCalls} failed calls followed by another same-tool call (${(followedFailures / errorCalls * 100).toFixed(0)}%)` : "";
+    const followUps = pr.failureFollowUps;
+    let failureFollowUp = "";
+    if (followUps && followUps.failures > 0) {
+      const split = [
+        `${followUps.blind} blind (identical args)`,
+        `${followUps.adapted} adapted (changed args)`
+      ];
+      if (followUps.argsUnknown > 0) split.push(`${followUps.argsUnknown} not comparable`);
+      const succeeded = followUps.followed > 0 ? `; ${followUps.followUpSucceeded}/${followUps.followed} follow-ups succeeded` : "";
+      failureFollowUp = `; ${followUps.followed}/${followUps.failures} failed calls followed by another same-tool call (${(followUps.followed / followUps.failures * 100).toFixed(0)}%) \u2014 ${split.join(", ")}${succeeded}`;
+    } else if (errorCalls > 0) {
+      failureFollowUp = `; ${followedFailures}/${errorCalls} failed calls followed by another same-tool call (${(followedFailures / errorCalls * 100).toFixed(0)}%)`;
+    }
     const duplicateSummary = callsWithCapturedArgs > 0 ? `${duplicateCalls}/${callsWithCapturedArgs} captured calls repeated exactly` : "exact repeats unavailable";
     lines.push(
       `- **Tool use:** ${totalCalls} calls across ${runsWithTools.length}/${pr.toolUse.length} traces; ${callsWithCapturedArgs}/${totalCalls} arguments captured; ${duplicateSummary}; ${errorCalls}/${totalCalls} failed${failureFollowUp}`
     );
+    if (followUps && followUps.blind > 0) {
+      const byTool2 = Object.entries(followUps.blindByTool).sort((left, right) => right[1] - left[1] || left[0].localeCompare(right[0])).map(([name, count2]) => `\`${tableCell2(name)}\` \xD7${count2}`);
+      lines.push(`  - \u{1F501} Blind retries (same args re-sent after a failure): ${byTool2.join(", ")}`);
+    }
     if (byTool.size > 0) {
       lines.push("");
       lines.push("| Tool | Calls | Args captured | Exact repeats | Errors | Mean latency |");
@@ -17339,6 +17452,7 @@ export {
   importCodeTraceBench,
   summarizeOtlpExecution,
   summarizeSpanExecution,
+  classifyFailureFollowUps,
   toRuntimeStore,
   runPipelines,
   recordSessionCorruption,
@@ -17374,6 +17488,7 @@ export {
   selectAdapters,
   normalizeRemote,
   equivalentGitCwds,
+  cwdMatchesSelectionResolved,
   cwdMatchesSelection,
   resolveRepoAttrs,
   resolveSessionRepoAttrs,

package/dist/cli.js

@@ -58,7 +58,7 @@ import {
   writePolicyEvidenceFile,
   writeSessionIndexFile,
   writeTraceEvidenceExportFile
-} from "./chunk-MEAEXKUV.js";
+} from "./chunk-CV22VGCK.js";
 
 // src/cli.ts
 import { readFileSync } from "fs";

package/dist/index.d.ts

@@ -1121,8 +1121,19 @@ interface RepoResolution {
  * `git@host:owner/repo.git` SCP-style forms. Returns null when unparseable.
  */
 declare function normalizeRemote(url: string): string | null;
-/** Expand a selected cwd to every checked-out worktree for the same git repo. */
+/**
+ * Expand a selected cwd to every checked-out worktree for the same git repo.
+ * Each candidate is emitted in both its selected and its resolved form: git
+ * prints symlink-resolved paths (macOS `/var` → `/private/var`), while a
+ * recorded session cwd usually keeps the unresolved spelling, and a match
+ * must not depend on which spelling either side used.
+ */
 declare function equivalentGitCwds(cwd: string | null | undefined): Promise<string[]>;
+/**
+ * `cwdMatchesSelection`, then a second attempt with the cwd's resolved
+ * (symlink-free) form. A deleted cwd cannot resolve and keeps the raw verdict.
+ */
+declare function cwdMatchesSelectionResolved(cwd: string | null | undefined, selections: readonly string[]): Promise<boolean>;
 /** Boundary-aware cwd matching; `/repo` matches `/repo/src`, not `/repo-old`. */
 declare function cwdMatchesSelection(cwd: string | null | undefined, selections: readonly string[]): boolean;
 /**
@@ -1516,6 +1527,55 @@ declare function classifyReaction(text: string): Reaction[];
  *  corrective-to-positive ratio). `praise` is the positive denominator. */
 declare const CORRECTIVE_REACTIONS: readonly Reaction[];
 
+/**
+ * Failure follow-up classification — the retry-with-state marker.
+ *
+ * The tool-use line has long counted "failed calls followed by another
+ * same-tool call", but that count cannot separate a blind retry (the same
+ * arguments re-sent after a failure, so the failure produced no new state)
+ * from an adapted one (the arguments changed, so the failure taught the agent
+ * something). Without the split the number reads as pure waste even in a run
+ * where every retry adapted — or as healthy adaptation in a run that re-sent
+ * the same command 300 times.
+ *
+ * Arguments are compared by the full-input hash (`traces.input.sha256`) when
+ * both sides carry it, because it survives `input.value` truncation, and by
+ * the verbatim `input.value` otherwise. A pair whose arguments cannot be
+ * compared is counted as unknown, never guessed. Calls marked
+ * `traces.expected_blocking` are excluded: re-sending an identical poll after
+ * a not-ready failure is the protocol working, not thrash.
+ */
+
+type FollowUpKind = 'blind' | 'adapted' | 'args-unknown' | 'none';
+interface FailureFollowUp {
+    traceId: string;
+    toolName: string;
+    failedSpanId: string;
+    /** Null when the tool was never called again after the failure. */
+    followUpSpanId: string | null;
+    kind: FollowUpKind;
+    /** Whether the follow-up call ended without an error; null when kind is 'none'. */
+    followUpSucceeded: boolean | null;
+}
+interface FailureFollowUpReport {
+    /** Failed TOOL calls classified; expected-blocking failures are excluded. */
+    failures: number;
+    /** Failures with a later call of the same tool in the same trace. */
+    followed: number;
+    /** Follow-up re-sent identical arguments. */
+    blind: number;
+    /** Follow-up changed the arguments. */
+    adapted: number;
+    /** Arguments were not captured or not comparable on one side of the pair. */
+    argsUnknown: number;
+    /** Follow-ups that ended without an error, out of `followed`. */
+    followUpSucceeded: number;
+    /** Blind-retry count per tool, largest offender first when rendered. */
+    blindByTool: Record<string, number>;
+    items: FailureFollowUp[];
+}
+declare function classifyFailureFollowUps(spans: readonly OtlpSpan[]): FailureFollowUpReport;
+
 /**
  * Run agent-eval's shipped trace pipelines over normalized harness spans.
  *
@@ -1524,6 +1584,10 @@ declare const CORRECTIVE_REACTIONS: readonly Reaction[];
  *   - stuckLoopView         — same tool + same args ≥ N times in a short interval
  *   - computeToolUseMetrics — duplicate-call / retry / error rates per run
  *
+ * One local view runs beside them: `classifyFailureFollowUps` splits each
+ * failed call's same-tool follow-up into blind (identical args) vs adapted
+ * (changed args), which the aggregate retry rate cannot express.
+ *
  * `toolWasteView` is intentionally NOT used: its default heuristic needs
  * verbatim tool *results* + per-turn LLM `messages`, which traces
  * doesn't capture from harness logs (we store tool args + error status). Its
@@ -1540,6 +1604,8 @@ interface PipelineReport {
     stuckLoops: StuckLoopReport;
     failureClusters: FailureClusterReport;
     toolUse: ToolUseMetrics[];
+    /** Absent only on hand-built reports; runPipelines always computes it. */
+    failureFollowUps?: FailureFollowUpReport;
 }
 interface PipelineOptions {
     /** Minimum repeated identical calls to flag a loop (default 3). */
@@ -4589,4 +4655,4 @@ interface ExecuteOptions {
 /** Send the final, new outbound items (or write their dry-run OTLP). */
 declare function executeUpload(plan: UploadPlan, opts?: ExecuteOptions): Promise<UploadResult>;
 
-export { ACTOR_ATTR, ATTR, type Actor, type AdoptionOptions, type AdoptionReport, AmpAdapter, type AnalystReplayFinding, type AnalyzeOptions, type AnalyzeResult, type ArmExecutionResult, type AssembleSessionBundleOptions, BLOBS_DIR, type BuildPolicyEvidenceOptions, type BuildSessionIndexOptions, type BuildTraceFindingPacketOptions, CODING_REDACTION_RULES, CONTENT_CAP, COORDINATION_LOG_FILE, CORRECTIVE_REACTIONS, type CaseResources, type ChatCompletionCaller, type ChatOutcome, type ChatTrajectory, type ChatTrajectoryMessage, type ChatTrajectoryOptions, type ChatUsage, ClaudeAdapter, type CodeTraceBenchFileRef, type CodeTraceBenchImportOptions, type CodeTraceBenchImportReceipt, type CodeTraceBenchImportResult, type CodeTraceBenchImportedTrace, type CodeTraceBenchStep, CodexAdapter, CodexExecAdapter, CodexExecStreamError, CodexTaskScopeError, type CollectOptions, type CollectPolicyEvidenceOptions, type CollectSessionIndexOptions, type CollectSessionSelectionOptions, type CollectSessionWorkflowOptions, type ConformanceRenderOptions, type CoordinationLog, type CoordinationQuestion, type CoordinationSteer, CopilotAdapter, type CorpusSpec, type CorruptionMode, type CwdSource, DEFAULT_HARNESS, DEFAULT_SANDBOX_BASE_URL, type DockerImagePreparerOptions, EMPTY_SPAN_RUN_SNAPSHOT, EXPORT_RULE, EmptySessionError, type EnumerationResult, type EvidenceOnlyVerdict, type ExcludedCase, type ExecuteOptions, type ExportRule, type ExternalAnalysisKind, type ExternalAnalysisPayload, type ExternalAnalysisResult, type ExternalAnalyzer, type ExternalAnalyzerOptions, type ExternalDiscoveryCandidate, FactoryAdapter, type FailedFixAttempt, type FileRunContextReaderOptions, type FindSessionTextLeaksOptions, type FindingReplaySource, type FindingReplayability, type FindingVerification, type FindingVerificationStatus, type FixArmExecution, type FixArmExecutor, type FixGenerationOutcome, type FixLoopAttemptRecord, type FixLoopOptions, type FixLoopResult, type FixPromptInput, ForgeAdapter, GeminiAdapter, type HaloAnalyzerOptions, type HaloReasoningEffort, type HarnessTraceAdapter, type HodoscopeAnalyzerOptions, type HodoscopeInput, type HodoscopeMessage, type HodoscopeSourceMapEntry, type HodoscopeTrajectory, INGEST_SOURCE_CLI, type ImagePreparation, type ImagePreparer, type IncorrectStepsSubject, type IngestedTrajectory, type InspectSessionIndexOptions, JsonSourceError, type JsonlCorruptionReceipt, JsonlParseError, type JsonlReadOptions, type LiveBatchOptions, type LocateOptions, type LoopConvergence, type LoopConvergenceReport, type LoopIteration, type LoopTrend, type LoopTrendBasis, type ObservedLoop, type ObserverOptions, OpencodeAdapter, type OtlpIngestIssue, type OtlpIngestIssueKind, type OtlpInput, type OtlpInputFile, type OtlpRowConversion, type OtlpSpan, type OtlpSpanKind, type OtlpSpanLink, type OtlpStatusCode, POLICY_EVIDENCE_NOTE, type ParentTaskResolution, type ParseOptions, PartialUploadError, PiAdapter, type PiAdapterOptions, type PipelineOptions, type PipelineReport, type PlanOptions, type PolicyEvidenceLoopSummary, type PolicyEvidenceRecord, type PolicyEvidenceToolSummary, type PrefixDivergence, type PrefixReplayResult, type PrimeAnalyzerOptions, type PrimeTransport, type PrimeTransportRequest, type PrimeTransportResponse, type ProjectSessionBundleOptions, type ProjectSessionBundleResult, QwenAdapter, REACTIONS, RESULT_FILE, RULE, type Reaction, type ReactionReport, type ReactionSessionReport, type Redactor, type ReplayArmVerdict, type ReplayBatchCaseRow, type ReplayBatchCliArgs, type ReplayBatchFixResult, type ReplayBatchOptions, type ReplayBatchReport, type ReplayExclusionReason, type ReplayExecBackend, type ReplayExecResult, type ReplayExecSession, type ReplayFindingOptions, type ReplayFindingResult, type ReplayVerdict, type ReplayVerifyCliArgs, type ReplayVerifyOptions, type ReplayableCase, type RepoAttrs, type RepoResolution, type RepoResolutionSource, type ResolvedFindingReplay, type ResolvedOtlpInput, type ResolvedReplayInvocation, type ResourceResolution, type RunContextJournal, type RunContextNode, type RunContextResult, type RunContextSnapshot, type RunResult, type RunWatchSnapshot, type RunWatchTarget, type RuntimeTrace, SPAWN_JOURNAL_FILE, STATUS_MARK, SUBMIT_ACTION_SIGNATURE, SandboxCounterfactualRunner, type SandboxCounterfactualRunnerOptions, type SandboxReplayBackendOptions, type ScanOptions, type ScannedSession, type SessionBatch, type SessionBundleAbsence, type SessionBundleCarriedSignatures, type SessionBundleExclusion, type SessionBundleExclusionRule, type SessionBundleFile, type SessionBundleLeakCheck, type SessionBundleLedgerSlice, type SessionBundleManifest, type SessionBundleProjection, type SessionBundleResult, type SessionBundleSliceRule, type SessionBundleView, type SessionCorruptionReceipt, type SessionEnvironment, type SessionIdentityConflict, type SessionIdentityIndex, type SessionIntegrity, type SessionRef, type SessionRelationship, type SessionRole, type SessionSeedGroup, type SessionSelection, type SessionSelectionRow, type SessionSourceFile, type SessionTextLeak, type SessionTextLeakReport, type SessionWorkflow, SessionWorkflowError, type SessionWorkflowErrorCode, type SessionWorkflowIssue, type SessionWorkflowSession, type SessionWorkflowSummary, type SkippedAnalysis, type SkippedInputFile, type SpanInput, type SpanNodeStatus, type SpanRedaction, type SpanRunNode, type SpanRunSnapshot, type SpanStructure, type SpendRoll, type SteeringChainReport, type SteeringEdge, TRACES_REDACTION_RULES, type TraceAgenticRoute, type TraceAgenticRouteReason, type TraceAgenticRouteReasonCode, type TraceContextIndex, type TraceContextRoot, type TraceEvidenceExportOptions, type TraceEvidenceExportResult, type TraceEvidenceFormatOption, type TraceEvidenceInputFormat, type TraceEvidenceRow, type TraceExecutionOptions, type TraceFindingPacket, type TraceImprovementArtifacts, type TraceImprovementOptions, type TraceImprovementResult, type TraceIndexedFile, type TraceInspectionArea, type TraceInspectionFinding, type TraceInspectionReport, type TraceInspectionSeverity, type TraceInvestigationOptions, type TraceInvestigationResult, type TraceLiveAction, type TraceLiveActionKind, type TraceLiveAnalysisContext, type TraceLiveAnalyst, type TraceLiveBatch, type TraceLiveEvidence, type TraceLiveFinding, type TraceLiveLoop, type TraceLiveSeverity, type TraceSessionIndex, type TraceSessionIndexRow, type TraceStoreInvestigationOptions, type TraceStoreInvestigationResult, type TraceStreamEvent, type TraceStreamOptions, type TraceStreamReplayOptions, type TraceStreamSession, type TracesConfig, type TriggerPair, type UnavailableCapabilities, type UploadBackend, type UploadIdentity, type UploadItem, type UploadPlan, type UploadRecord, type UploadResult, type UploadState, type UserPromptInput, type VerifiableFinding, type VerifyFindingsCliArgs, type VerifyFindingsOptions, type VerifyFindingsRun, type WatchRunTargetOptions, ZERO_ROLL, type ZaiChatCallerOptions, addSpend, alreadyUploaded, analyzeAdoption, analyzeLiveBatch, analyzeLoopConvergence, analyzeReactions, analyzeSpans, analyzeSteeringChain, applyRedactor, assembleSessionBundle, assertSandboxReachable, buildFixPrompt, buildPolicyEvidenceRecord, buildRetryFixPrompt, buildSessionIndex, buildSessionIndexFromRows, buildSpanRunTree, buildTraceFindingPacket, capText, chatTrajectoryToSpans, classifyEvidenceOnlyPath, classifyLiveActions, classifyReaction, classifyVerdict, claudeActor, clipText, cmdReplayVerify, cmdReplayVerifyBatch, cmdVerifyFindings, codexActor, collectPolicyEvidence, collectSessionIndex, collectSessionSelection, collectSessionWorkflow, collectSessions, commandAnalyzer, commandRedactor, conformanceOfSpans, connectors, countScriptCommands, countSkillRunsJsonl, cwdMatchesSelection, defaultTraceLiveAnalysts, deriveFailureSignature, derivedImageTag, describeSessionRelationship, dockerImagePreparer, enumerateReplayableCases, environmentAttributes, environmentFromSpanAttributes, equivalentGitCwds, executeUpload, exportTraceEvidenceFile, exportTraceEvidenceRows, exportTraceEvidenceText, extractFixCommand, fileRunContextSupervisorRunReader, findContextRoot, findSessionTextLeaks, findSpanSourceFiles, findingReplayStep, findingTrajectoryId, forEachTreeNode, generateFixCommand, goldIncorrectSteps, haloAnalyzer, hashSessionBundleFiles, hodoscopeAnalyzer, httpJsonTransport, importCodeTraceBench, incompleteInputsNote, indexSessionIdsByTrace, inferRole, inferSpanKind, ingestCodeTraceBenchSteps, inspectSessionIndex, int, isChatTrajectory, isChatTrajectoryMessage, isFileRunContextDir, isSubmitAction, knownHarnesses, listAdapters, listSessionBundleFiles, loadState, loadTracesConfig, locateSessions, looksLikeAgentPrompt, mergeRolls, mergeTracesConfig, ms, normalizeRemote, otlpRowToSpan, outboundHash, parseCorpusFlag, parseIncorrectStepsSubject, parseIsoToEpochMs, parseObservationOutput, parseRecordedReturncode, parseReplayBatchArgs, parseReplayVerifyArgs, parseSession, parseSince, parseVerifyFindingsArgs, planTraceAgenticRoute, planUpload, primeAnalyzer, projectSessionBundle, readCoordinationLog, readFileRunContext, readFindingsFile, readJsonFile, readJsonl, readLabelEntries, readOtlpInput, readRunContextJournal, readRunContextSnapshot, readRunWatchSnapshot, readSessionBundleManifest, readSessionIndexFile, readSpanRunSnapshot, recordSessionCorruption, redactSpans, renderBatchReport, renderConformance, renderInspectionReport, renderRunContextSnapshot, renderRunWatchSnapshot, renderSpanRunSnapshot, renderValidation, renderVerifiedFindingsSection, replayBatchUsage, replayVerify, replayVerifyFinding, replayVerifyUsage, resolveAdapter, resolveCaseResources, resolveFindingInvocation, resolveFindingReplayability, resolveOtlpInputFiles, resolveRepoAttrs, resolveRunContextDir, resolveRunWatchTarget, resolveSessionRepoAttrs, runCommand, runExternalAnalyzers, runFixLoop, runPipelines, runReplayBatch, runTotals, runTraceImprovement, runTraceInvestigation, runTraceStoreInvestigation, sandboxReplayBackend, saveReport, saveState, scanSessions, seededSample, selectAdapters, serializeInspectionReport, serializePolicyEvidence, serializeSessionIndex, serializeSpans, serializeTraceStreamEvent, sessionHash, sessionIdFromAttributes, sessionIntegrityAttributes, sessionTextSignatures, skippedAnalyses, span, spanRunTotals, stampEnvironmentAttrs, stampRepoAttrs, stampSessionIdentity, stampSessionIntegrity, stampSpanWorkdirRepoAttrs, streamSessions, summarizeOtlpExecution, summarizeSpanExecution, summarizeSpanStructure, takeJsonl, textIsSynthetic, toOpenInferenceSpan, toRuntimeStore, toTraceSpanEvents, tokens, totalAgenticFailureMessage, traceAgenticKinds, traceStreamEventsFromSpans, unavailableCapabilities, uploadKey, usd, userPromptSpan, validateOtlpSpans, validationExitCode, verifyFindings, verifyFindingsUsage, watchRunTarget, watchSessions, wrap, wrapActionForExec, writeHodoscopeInput, writeInspectionReportFile, writeOtlpFile, writePolicyEvidenceFile, writeSessionBundleManifest, writeSessionIndexFile, writeTraceEvidenceExportFile, writeTraceImprovementArtifacts, zaiChatCaller, zeroPricedCost };
+export { ACTOR_ATTR, ATTR, type Actor, type AdoptionOptions, type AdoptionReport, AmpAdapter, type AnalystReplayFinding, type AnalyzeOptions, type AnalyzeResult, type ArmExecutionResult, type AssembleSessionBundleOptions, BLOBS_DIR, type BuildPolicyEvidenceOptions, type BuildSessionIndexOptions, type BuildTraceFindingPacketOptions, CODING_REDACTION_RULES, CONTENT_CAP, COORDINATION_LOG_FILE, CORRECTIVE_REACTIONS, type CaseResources, type ChatCompletionCaller, type ChatOutcome, type ChatTrajectory, type ChatTrajectoryMessage, type ChatTrajectoryOptions, type ChatUsage, ClaudeAdapter, type CodeTraceBenchFileRef, type CodeTraceBenchImportOptions, type CodeTraceBenchImportReceipt, type CodeTraceBenchImportResult, type CodeTraceBenchImportedTrace, type CodeTraceBenchStep, CodexAdapter, CodexExecAdapter, CodexExecStreamError, CodexTaskScopeError, type CollectOptions, type CollectPolicyEvidenceOptions, type CollectSessionIndexOptions, type CollectSessionSelectionOptions, type CollectSessionWorkflowOptions, type ConformanceRenderOptions, type CoordinationLog, type CoordinationQuestion, type CoordinationSteer, CopilotAdapter, type CorpusSpec, type CorruptionMode, type CwdSource, DEFAULT_HARNESS, DEFAULT_SANDBOX_BASE_URL, type DockerImagePreparerOptions, EMPTY_SPAN_RUN_SNAPSHOT, EXPORT_RULE, EmptySessionError, type EnumerationResult, type EvidenceOnlyVerdict, type ExcludedCase, type ExecuteOptions, type ExportRule, type ExternalAnalysisKind, type ExternalAnalysisPayload, type ExternalAnalysisResult, type ExternalAnalyzer, type ExternalAnalyzerOptions, type ExternalDiscoveryCandidate, FactoryAdapter, type FailedFixAttempt, type FailureFollowUp, type FailureFollowUpReport, type FileRunContextReaderOptions, type FindSessionTextLeaksOptions, type FindingReplaySource, type FindingReplayability, type FindingVerification, type FindingVerificationStatus, type FixArmExecution, type FixArmExecutor, type FixGenerationOutcome, type FixLoopAttemptRecord, type FixLoopOptions, type FixLoopResult, type FixPromptInput, type FollowUpKind, ForgeAdapter, GeminiAdapter, type HaloAnalyzerOptions, type HaloReasoningEffort, type HarnessTraceAdapter, type HodoscopeAnalyzerOptions, type HodoscopeInput, type HodoscopeMessage, type HodoscopeSourceMapEntry, type HodoscopeTrajectory, INGEST_SOURCE_CLI, type ImagePreparation, type ImagePreparer, type IncorrectStepsSubject, type IngestedTrajectory, type InspectSessionIndexOptions, JsonSourceError, type JsonlCorruptionReceipt, JsonlParseError, type JsonlReadOptions, type LiveBatchOptions, type LocateOptions, type LoopConvergence, type LoopConvergenceReport, type LoopIteration, type LoopTrend, type LoopTrendBasis, type ObservedLoop, type ObserverOptions, OpencodeAdapter, type OtlpIngestIssue, type OtlpIngestIssueKind, type OtlpInput, type OtlpInputFile, type OtlpRowConversion, type OtlpSpan, type OtlpSpanKind, type OtlpSpanLink, type OtlpStatusCode, POLICY_EVIDENCE_NOTE, type ParentTaskResolution, type ParseOptions, PartialUploadError, PiAdapter, type PiAdapterOptions, type PipelineOptions, type PipelineReport, type PlanOptions, type PolicyEvidenceLoopSummary, type PolicyEvidenceRecord, type PolicyEvidenceToolSummary, type PrefixDivergence, type PrefixReplayResult, type PrimeAnalyzerOptions, type PrimeTransport, type PrimeTransportRequest, type PrimeTransportResponse, type ProjectSessionBundleOptions, type ProjectSessionBundleResult, QwenAdapter, REACTIONS, RESULT_FILE, RULE, type Reaction, type ReactionReport, type ReactionSessionReport, type Redactor, type ReplayArmVerdict, type ReplayBatchCaseRow, type ReplayBatchCliArgs, type ReplayBatchFixResult, type ReplayBatchOptions, type ReplayBatchReport, type ReplayExclusionReason, type ReplayExecBackend, type ReplayExecResult, type ReplayExecSession, type ReplayFindingOptions, type ReplayFindingResult, type ReplayVerdict, type ReplayVerifyCliArgs, type ReplayVerifyOptions, type ReplayableCase, type RepoAttrs, type RepoResolution, type RepoResolutionSource, type ResolvedFindingReplay, type ResolvedOtlpInput, type ResolvedReplayInvocation, type ResourceResolution, type RunContextJournal, type RunContextNode, type RunContextResult, type RunContextSnapshot, type RunResult, type RunWatchSnapshot, type RunWatchTarget, type RuntimeTrace, SPAWN_JOURNAL_FILE, STATUS_MARK, SUBMIT_ACTION_SIGNATURE, SandboxCounterfactualRunner, type SandboxCounterfactualRunnerOptions, type SandboxReplayBackendOptions, type ScanOptions, type ScannedSession, type SessionBatch, type SessionBundleAbsence, type SessionBundleCarriedSignatures, type SessionBundleExclusion, type SessionBundleExclusionRule, type SessionBundleFile, type SessionBundleLeakCheck, type SessionBundleLedgerSlice, type SessionBundleManifest, type SessionBundleProjection, type SessionBundleResult, type SessionBundleSliceRule, type SessionBundleView, type SessionCorruptionReceipt, type SessionEnvironment, type SessionIdentityConflict, type SessionIdentityIndex, type SessionIntegrity, type SessionRef, type SessionRelationship, type SessionRole, type SessionSeedGroup, type SessionSelection, type SessionSelectionRow, type SessionSourceFile, type SessionTextLeak, type SessionTextLeakReport, type SessionWorkflow, SessionWorkflowError, type SessionWorkflowErrorCode, type SessionWorkflowIssue, type SessionWorkflowSession, type SessionWorkflowSummary, type SkippedAnalysis, type SkippedInputFile, type SpanInput, type SpanNodeStatus, type SpanRedaction, type SpanRunNode, type SpanRunSnapshot, type SpanStructure, type SpendRoll, type SteeringChainReport, type SteeringEdge, TRACES_REDACTION_RULES, type TraceAgenticRoute, type TraceAgenticRouteReason, type TraceAgenticRouteReasonCode, type TraceContextIndex, type TraceContextRoot, type TraceEvidenceExportOptions, type TraceEvidenceExportResult, type TraceEvidenceFormatOption, type TraceEvidenceInputFormat, type TraceEvidenceRow, type TraceExecutionOptions, type TraceFindingPacket, type TraceImprovementArtifacts, type TraceImprovementOptions, type TraceImprovementResult, type TraceIndexedFile, type TraceInspectionArea, type TraceInspectionFinding, type TraceInspectionReport, type TraceInspectionSeverity, type TraceInvestigationOptions, type TraceInvestigationResult, type TraceLiveAction, type TraceLiveActionKind, type TraceLiveAnalysisContext, type TraceLiveAnalyst, type TraceLiveBatch, type TraceLiveEvidence, type TraceLiveFinding, type TraceLiveLoop, type TraceLiveSeverity, type TraceSessionIndex, type TraceSessionIndexRow, type TraceStoreInvestigationOptions, type TraceStoreInvestigationResult, type TraceStreamEvent, type TraceStreamOptions, type TraceStreamReplayOptions, type TraceStreamSession, type TracesConfig, type TriggerPair, type UnavailableCapabilities, type UploadBackend, type UploadIdentity, type UploadItem, type UploadPlan, type UploadRecord, type UploadResult, type UploadState, type UserPromptInput, type VerifiableFinding, type VerifyFindingsCliArgs, type VerifyFindingsOptions, type VerifyFindingsRun, type WatchRunTargetOptions, ZERO_ROLL, type ZaiChatCallerOptions, addSpend, alreadyUploaded, analyzeAdoption, analyzeLiveBatch, analyzeLoopConvergence, analyzeReactions, analyzeSpans, analyzeSteeringChain, applyRedactor, assembleSessionBundle, assertSandboxReachable, buildFixPrompt, buildPolicyEvidenceRecord, buildRetryFixPrompt, buildSessionIndex, buildSessionIndexFromRows, buildSpanRunTree, buildTraceFindingPacket, capText, chatTrajectoryToSpans, classifyEvidenceOnlyPath, classifyFailureFollowUps, classifyLiveActions, classifyReaction, classifyVerdict, claudeActor, clipText, cmdReplayVerify, cmdReplayVerifyBatch, cmdVerifyFindings, codexActor, collectPolicyEvidence, collectSessionIndex, collectSessionSelection, collectSessionWorkflow, collectSessions, commandAnalyzer, commandRedactor, conformanceOfSpans, connectors, countScriptCommands, countSkillRunsJsonl, cwdMatchesSelection, cwdMatchesSelectionResolved, defaultTraceLiveAnalysts, deriveFailureSignature, derivedImageTag, describeSessionRelationship, dockerImagePreparer, enumerateReplayableCases, environmentAttributes, environmentFromSpanAttributes, equivalentGitCwds, executeUpload, exportTraceEvidenceFile, exportTraceEvidenceRows, exportTraceEvidenceText, extractFixCommand, fileRunContextSupervisorRunReader, findContextRoot, findSessionTextLeaks, findSpanSourceFiles, findingReplayStep, findingTrajectoryId, forEachTreeNode, generateFixCommand, goldIncorrectSteps, haloAnalyzer, hashSessionBundleFiles, hodoscopeAnalyzer, httpJsonTransport, importCodeTraceBench, incompleteInputsNote, indexSessionIdsByTrace, inferRole, inferSpanKind, ingestCodeTraceBenchSteps, inspectSessionIndex, int, isChatTrajectory, isChatTrajectoryMessage, isFileRunContextDir, isSubmitAction, knownHarnesses, listAdapters, listSessionBundleFiles, loadState, loadTracesConfig, locateSessions, looksLikeAgentPrompt, mergeRolls, mergeTracesConfig, ms, normalizeRemote, otlpRowToSpan, outboundHash, parseCorpusFlag, parseIncorrectStepsSubject, parseIsoToEpochMs, parseObservationOutput, parseRecordedReturncode, parseReplayBatchArgs, parseReplayVerifyArgs, parseSession, parseSince, parseVerifyFindingsArgs, planTraceAgenticRoute, planUpload, primeAnalyzer, projectSessionBundle, readCoordinationLog, readFileRunContext, readFindingsFile, readJsonFile, readJsonl, readLabelEntries, readOtlpInput, readRunContextJournal, readRunContextSnapshot, readRunWatchSnapshot, readSessionBundleManifest, readSessionIndexFile, readSpanRunSnapshot, recordSessionCorruption, redactSpans, renderBatchReport, renderConformance, renderInspectionReport, renderRunContextSnapshot, renderRunWatchSnapshot, renderSpanRunSnapshot, renderValidation, renderVerifiedFindingsSection, replayBatchUsage, replayVerify, replayVerifyFinding, replayVerifyUsage, resolveAdapter, resolveCaseResources, resolveFindingInvocation, resolveFindingReplayability, resolveOtlpInputFiles, resolveRepoAttrs, resolveRunContextDir, resolveRunWatchTarget, resolveSessionRepoAttrs, runCommand, runExternalAnalyzers, runFixLoop, runPipelines, runReplayBatch, runTotals, runTraceImprovement, runTraceInvestigation, runTraceStoreInvestigation, sandboxReplayBackend, saveReport, saveState, scanSessions, seededSample, selectAdapters, serializeInspectionReport, serializePolicyEvidence, serializeSessionIndex, serializeSpans, serializeTraceStreamEvent, sessionHash, sessionIdFromAttributes, sessionIntegrityAttributes, sessionTextSignatures, skippedAnalyses, span, spanRunTotals, stampEnvironmentAttrs, stampRepoAttrs, stampSessionIdentity, stampSessionIntegrity, stampSpanWorkdirRepoAttrs, streamSessions, summarizeOtlpExecution, summarizeSpanExecution, summarizeSpanStructure, takeJsonl, textIsSynthetic, toOpenInferenceSpan, toRuntimeStore, toTraceSpanEvents, tokens, totalAgenticFailureMessage, traceAgenticKinds, traceStreamEventsFromSpans, unavailableCapabilities, uploadKey, usd, userPromptSpan, validateOtlpSpans, validationExitCode, verifyFindings, verifyFindingsUsage, watchRunTarget, watchSessions, wrap, wrapActionForExec, writeHodoscopeInput, writeInspectionReportFile, writeOtlpFile, writePolicyEvidenceFile, writeSessionBundleManifest, writeSessionIndexFile, writeTraceEvidenceExportFile, writeTraceImprovementArtifacts, zaiChatCaller, zeroPricedCost };

package/dist/index.js

@@ -60,6 +60,7 @@ import {
   capText,
   chatTrajectoryToSpans,
   classifyEvidenceOnlyPath,
+  classifyFailureFollowUps,
   classifyLiveActions,
   classifyReaction,
   classifyVerdict,
@@ -80,6 +81,7 @@ import {
   countScriptCommands,
   countSkillRunsJsonl,
   cwdMatchesSelection,
+  cwdMatchesSelectionResolved,
   defaultTraceLiveAnalysts,
   deriveFailureSignature,
   derivedImageTag,
@@ -249,7 +251,7 @@ import {
   writeTraceImprovementArtifacts,
   zaiChatCaller,
   zeroPricedCost
-} from "./chunk-MEAEXKUV.js";
+} from "./chunk-CV22VGCK.js";
 
 // src/collect.ts
 async function collectSessions(opts = {}) {
@@ -338,6 +340,7 @@ export {
   capText,
   chatTrajectoryToSpans,
   classifyEvidenceOnlyPath,
+  classifyFailureFollowUps,
   classifyLiveActions,
   classifyReaction,
   classifyVerdict,
@@ -360,6 +363,7 @@ export {
   countSkillRunsJsonl,
   createHostedClient,
   cwdMatchesSelection,
+  cwdMatchesSelectionResolved,
   defaultTraceLiveAnalysts,
   deriveFailureSignature,
   derivedImageTag,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/traces",
-  "version": "0.11.6",
+  "version": "0.12.0",
   "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",