@tangle-network/traces 0.9.0 → 0.9.2

package/dist/cli.js

@@ -3,9 +3,7 @@ import {
   ACTOR_ATTR,
   ATTR,
   CORRUPTION_RECEIPT_DISPLAY_LIMIT,
-  analyzeAdoption,
-  analyzeReactions,
-  analyzeSpans,
+  appendAll,
   buildPolicyEvidenceRecord,
   buildSessionIndexFromRows,
   commandAnalyzer,
@@ -23,14 +21,8 @@ import {
   parseSince,
   planUpload,
   readSessionIndexFile,
-  renderAdoption,
   renderInspectionReport,
-  renderPipelines,
-  renderReactions,
-  renderReport,
   resolveAdapter,
-  runExternalAnalyzers,
-  runPipelines,
   runTraceImprovement,
   runTraceInvestigation,
   saveReport,
@@ -41,7 +33,6 @@ import {
   serializeTraceStreamEvent,
   sessionIdFromAttributes,
   streamSessions,
-  summarizeDeterministicSignals,
   traceStreamEventsFromSpans,
   watchSessions,
   writeInspectionReportFile,
@@ -49,11 +40,11 @@ import {
   writePolicyEvidenceFile,
   writeSessionIndexFile,
   writeTraceEvidenceExportFile
-} from "./chunk-5SBXL62V.js";
+} from "./chunk-RUG26PIS.js";
 
 // src/cli.ts
 import { readFileSync } from "fs";
-import { readFile, stat, writeFile } from "fs/promises";
+import { readFile, stat } from "fs/promises";
 import { basename, resolve } from "path";
 function packageVersion() {
   const pkg = JSON.parse(readFileSync(new URL("../package.json", import.meta.url), "utf8"));
@@ -217,6 +208,50 @@ async function discover(args) {
   }
   return out;
 }
+function missingPath(error) {
+  return typeof error === "object" && error !== null && error.code === "ENOENT";
+}
+async function resolveSelectedSession(args) {
+  if (!args.session) throw new Error("resolveSelectedSession requires --session");
+  const adapter = resolveAdapter(args.harness);
+  if (!adapter && !args.all) throw new Error(`unknown harness "${args.harness}"`);
+  try {
+    const st = await stat(args.session);
+    if (!adapter) {
+      throw new Error("--session <path> requires --harness when --all is set");
+    }
+    return {
+      adapter,
+      ref: {
+        harness: adapter.harness,
+        sessionId: args.session,
+        path: args.session,
+        // --session is an explicit file; honor --cwd so adoption can find the
+        // project's .evolve/skill-runs.jsonl (locate() infers cwd in the scan path).
+        cwd: args.cwd ?? null,
+        mtimeMs: st.mtimeMs
+      }
+    };
+  } catch (error) {
+    if (!missingPath(error)) throw error;
+  }
+  const matches = [];
+  for (const candidate of args.all ? adaptersFor(args) : [adapter]) {
+    const refs = await locateSessions(candidate, { cwd: args.cwd });
+    for (const ref of refs) {
+      if (ref.sessionId === args.session) matches.push({ adapter: candidate, ref });
+    }
+  }
+  if (matches.length === 1) return matches[0];
+  if (matches.length === 0) {
+    throw new Error(
+      `no ${args.all ? "selected" : adapter.harness} session with ID "${args.session}"; run \`traces list\` to select an ID, or pass an explicit session path`
+    );
+  }
+  throw new Error(
+    `session ID "${args.session}" is ambiguous across ${matches.length} files; select one harness or pass its explicit session path`
+  );
+}
 async function buildAxService() {
   const apiKey = process.env.OPENAI_API_KEY;
   if (!apiKey) {
@@ -226,7 +261,7 @@ async function buildAxService() {
   }
   const apiURL = process.env.OPENAI_BASE_URL || void 0;
   const { AxAI } = await import("@ax-llm/ax");
-  return new AxAI({ name: "openai", apiKey, ...apiURL ? { apiURL } : {} });
+  return AxAI.create({ name: "openai", apiKey, ...apiURL ? { apiURL } : {} });
 }
 async function cmdList(args) {
   const groups = await discover({ ...args, last: args.last || 20 });
@@ -274,18 +309,7 @@ function selectedSessionSource(ref, spans, sessionIdOverride) {
 async function collectSpans(args) {
   if (args.input) return collectImportedSpans(args);
   if (args.session) {
-    const adapter = resolveAdapter(args.harness);
-    if (!adapter) throw new Error(`unknown harness "${args.harness}"`);
-    const st = await stat(args.session);
-    const ref = {
-      harness: adapter.harness,
-      sessionId: args.session,
-      path: args.session,
-      // --session is an explicit file; honor --cwd so adoption can find the
-      // project's .evolve/skill-runs.jsonl (locate() infers cwd in the scan path).
-      cwd: args.cwd ?? null,
-      mtimeMs: st.mtimeMs
-    };
+    const { adapter, ref } = await resolveSelectedSession(args);
     const spans2 = await parseSession(adapter, ref);
     return {
       spans: spans2,
@@ -303,7 +327,7 @@ async function collectSpans(args) {
     if (refs.length > 0) harnesses.push(adapter.harness);
     for (const ref of refs) {
       const parsed = await parseSession(adapter, ref);
-      spans.push(...parsed);
+      appendAll(spans, parsed);
       sources.push(selectedSessionSource(ref, parsed));
       if (ref.cwd) cwds.push(ref.cwd);
     }
@@ -318,16 +342,7 @@ async function cmdConvert(args) {
 }
 async function collectSessionRows(args) {
   if (args.session) {
-    const adapter = resolveAdapter(args.harness);
-    if (!adapter) throw new Error(`unknown harness "${args.harness}"`);
-    const st = await stat(args.session);
-    const ref = {
-      harness: adapter.harness,
-      sessionId: args.session,
-      path: args.session,
-      cwd: args.cwd ?? null,
-      mtimeMs: st.mtimeMs
-    };
+    const { adapter, ref } = await resolveSelectedSession(args);
     return [{ ref, spans: await parseSession(adapter, ref) }];
   }
   const groups = await discover({ ...args, last: args.last || 20 });
@@ -438,52 +453,14 @@ async function loadExportAttributes(args) {
   return attributes;
 }
 async function cmdAnalyze(args) {
-  const { spans, harness, cwds, sources } = await collectSpans(args);
-  if (spans.length === 0) throw new Error("no spans found for the given selection");
-  const ai = args.llm ? await buildAxService() : void 0;
-  const { otlpPath, execution, result } = await analyzeSpans(spans, {
-    ai,
-    model: args.model,
-    budgetUsd: args.budget,
-    otlpOutPath: args.otlp,
-    log: (msg) => process.stderr.write(`${msg}
-`)
-  });
-  const pipelines = await runPipelines(spans, { minLoopOccurrences: args.minLoop });
-  const reactions = analyzeReactions(spans);
-  const adoption = await analyzeAdoption(spans, { cwds });
-  const deterministic = summarizeDeterministicSignals(pipelines, reactions);
-  let report = `${renderReport(result, {
-    harness,
-    sessionCount: adoption.identifiedSessionCount,
-    unassignedTraceCount: adoption.unassignedTraceCount,
-    spanCount: spans.length,
-    otlpPath,
-    execution,
-    deterministic,
-    sources
-  })}
-${renderPipelines(pipelines)}
-${renderReactions(reactions)}
-${renderAdoption(adoption)}`;
-  if (args.analyzers.length > 0 && otlpPath) {
-    const engines = externalAnalyzersFromArgs(args);
-    const results = await runExternalAnalyzers(otlpPath, engines, { prompt: args.analyzerPrompt });
-    for (const r of results) {
-      report += `
-
-## ${r.analyzer} (external analyzer)
-
-${r.ok ? r.output || "(no output)" : `failed: ${r.error}`}`;
-    }
-  }
+  const result = await investigate(args, { loadDefaultConfig: false });
   if (args.out) {
-    await writeFile(args.out, report, "utf8");
+    await saveReport(args.out, result.report);
     console.log(
-      `report \u2192 ${args.out}  (${result.findings.length} findings, ${pipelines.stuckLoops.findings.length} loops, OTLP: ${otlpPath})`
+      `report \u2192 ${args.out}  (${result.findings.length} findings, ${result.pipelines.stuckLoops.findings.length} loops, OTLP: ${result.otlpPath})`
     );
   } else {
-    console.log(report);
+    console.log(result.report);
   }
 }
 async function collectImportedSpans(args) {
@@ -521,11 +498,20 @@ function externalAnalyzersFromArgs(args) {
   );
 }
 async function cmdInvestigate(args) {
+  const result = await investigate(args);
+  if (args.out) {
+    await saveReport(args.out, result.report);
+    console.log(`investigation report \u2192 ${args.out}  (${result.findings.length} findings with actions, OTLP: ${result.otlpPath})`);
+  } else {
+    console.log(result.report);
+  }
+}
+async function investigate(args, options = {}) {
   const { spans, harness, cwds, sources } = await collectSpans(args);
   if (spans.length === 0) throw new Error("no spans found for the given selection");
-  const config = await loadTracesConfig(args.config);
+  const config = args.config !== void 0 || options.loadDefaultConfig !== false ? await loadTracesConfig(args.config) : void 0;
   const ai = args.llm ? await buildAxService() : void 0;
-  const result = await runTraceInvestigation(mergeTracesConfig({
+  return runTraceInvestigation(mergeTracesConfig({
     spans,
     harness,
     sources,
@@ -540,12 +526,6 @@ async function cmdInvestigate(args) {
     log: (msg) => process.stderr.write(`${msg}
 `)
   }, config));
-  if (args.out) {
-    await saveReport(args.out, result.report);
-    console.log(`investigation report \u2192 ${args.out}  (${result.findings.length} findings with actions, OTLP: ${result.otlpPath})`);
-  } else {
-    console.log(result.report);
-  }
 }
 async function cmdImprove(args) {
   const { spans, harness, cwds, sources } = await collectSpans(args);
@@ -822,7 +802,7 @@ Options:
   --harness <id>   Harness or alias (default: claude-code). Known: ${knownHarnesses().join(", ")}
   --all            Sweep every known harness
   --last <n>       Most-recent N sessions
-  --session <path> Analyze/stream one explicit harness session file
+  --session <id|path> Analyze one listed session ID or one explicit harness session file
   --cwd <dir>      Filter sessions by working directory
   --since <t>      upload: window, 30m / 2h / 7d or an ISO date (default 24h); analyze: ISO cutoff
   --out <path>     Write report to a file

package/dist/index.d.ts

@@ -815,13 +815,9 @@ declare function analyzeReactions(spans: readonly OtlpSpan[]): ReactionReport;
  *   2. per-skill invocation frequency
  *   3. subagent (Task/Agent) spawn frequency
  *
- * CRITICAL undercount fix: the trace-level `Skill` tool-use count only sees
- * skills a session invoked *interactively*. Skills dispatched inside an
- * automation loop (e.g. `/evolve`, `/converge`) are logged to
- * `.evolve/skill-runs.jsonl` under the session's cwd and never surface as a
- * `Skill` tool span — historically ~370× more runs than the trace count. So we
- * read those files too and report "explicit invocations" and "loop-dispatched
- * runs" SEPARATELY, so neither silently reads as zero.
+ * A repository's `.evolve/skill-runs.jsonl` is historical context, not session
+ * evidence, unless a row carries the selected session ID. This keeps old runs
+ * from being attributed to the current trace.
  */
 
 interface AdoptionReport {
@@ -843,6 +839,11 @@ interface AdoptionReport {
     sessionsWithMaterializedSkills: number;
     /** Sessions with tool inputs that reference a SKILL.md path. */
     sessionsWithSkillFileReference: number;
+    /** Skill documents opened by a successful captured shell-read command.
+     *  Inspection is not evidence that the skill changed the outcome. */
+    skillDocumentReads: Record<string, number>;
+    /** Total observed skill-document reads. */
+    totalSkillDocumentReads: number;
     /** Explicit `Skill` tool invocations per skill name, corpus-wide. */
     skillInvocations: Record<string, number>;
     /** Total explicit `Skill` tool invocations. */
@@ -853,19 +854,18 @@ interface AdoptionReport {
     totalSubagentSpawns: number;
     /** Sessions that spawned ≥1 subagent. */
     sessionsWithSubagent: number;
-    /** Loop-dispatched skill runs per skill, read from `.evolve/skill-runs.jsonl`
-     *  (NOT visible at trace level — counted separately to avoid silent zeros). */
+    /** Loop-dispatched skill runs whose row names a selected session ID. */
     loopDispatchedRuns: Record<string, number>;
-    /** Total loop-dispatched skill runs. */
+    /** Total session-linked loop-dispatched skill runs. */
     totalLoopDispatchedRuns: number;
     /** `.evolve/skill-runs.jsonl` files that were read. */
     skillRunFilesRead: number;
+    /** Same-repository loop history without a selected session ID. Never treat
+     *  this as evidence that a selected session used a skill. */
+    unlinkedLoopDispatchedRuns: Record<string, number>;
+    /** Total unlinked same-repository loop history rows. */
+    totalUnlinkedLoopDispatchedRuns: number;
 }
-/**
- * Count loop-dispatched skill runs from a `.evolve/skill-runs.jsonl` file's raw
- * contents. Each line records one or more skills under `skills` (array) or
- * `skill` (string) — the schema varies across producers, so accept both.
- */
 declare function countSkillRunsJsonl(raw: string): Record<string, number>;
 interface AdoptionOptions {
     /** Session cwds to probe for `.evolve/skill-runs.jsonl`. The CLI passes the

package/dist/index.js

@@ -115,7 +115,7 @@ import {
   writeSessionIndexFile,
   writeTraceEvidenceExportFile,
   writeTraceImprovementArtifacts
-} from "./chunk-5SBXL62V.js";
+} from "./chunk-RUG26PIS.js";
 
 // src/collect.ts
 async function collectSessions(opts = {}) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/traces",
-  "version": "0.9.0",
+  "version": "0.9.2",
   "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",
@@ -44,8 +44,8 @@
     "node": ">=22"
   },
   "dependencies": {
-    "@ax-llm/ax": "^19.0.45",
-    "@tangle-network/agent-eval": "^0.118.3"
+    "@ax-llm/ax": "^23.0.1",
+    "@tangle-network/agent-eval": "^0.122.8"
   },
   "devDependencies": {
     "@types/node": "^22.0.0",