@tangle-network/traces 0.20.0 → 0.21.1

package/README.md

@@ -364,6 +364,7 @@ See [Replay verification](./docs/replay-verify.md) for setup, semantics, and hon
 | `--question <text>` | `ask`: one question, repeatable. Kept short so the engine sees it whole |
 | `--questions <file>` | `ask`: JSON array of questions — strings, or `{ id?, question, instructions?, answerSchema? }` |
 | `--question-budget <usd>` | `ask`: provider ceiling for ONE question; `--budget` is the ceiling shared by all of them |
+| `--question-timeout <seconds>` | `ask`: wall-clock deadline for ONE question's investigation (default 600); raise it for questions over many large sessions |
 | `--concurrency <n>` | `ask`: questions running at once (default 4); `import-codetracebench`: trajectories imported at once |
 | `--config <path>` | `analyze` / `investigate` / `improve` / `stream`: load BYO analysts, live analysts, and external analyzers |
 | `--interval <s>` / `--window <m>` | `watch` / live `stream`: poll seconds (sessions 5, run tree 2) / active-session window minutes (default 30) |

package/dist/{chunk-2NINDZCW.js → chunk-LB6AD6SX.js}

@@ -17497,12 +17497,14 @@ function asRecord(v) {
 function str(v) {
   return typeof v === "string" && v.length > 0 ? v : null;
 }
+var RETIRED_WAIT_KINDS = /* @__PURE__ */ new Set(["waiting", "woken"]);
 function readRunContextJournal(text2) {
   const parsed = readJsonlLines(text2);
   let root = null;
   let begunAt = null;
   const events = [];
   let invalidRows = parsed.invalidRows;
+  let retiredWaitRows = 0;
   for (const record2 of parsed.rows) {
     const recordRoot = str(record2.root);
     if (root === null && recordRoot !== null) root = recordRoot;
@@ -17519,11 +17521,15 @@ function readRunContextJournal(text2) {
       invalidRows += 1;
       continue;
     }
+    if (RETIRED_WAIT_KINDS.has(event.kind)) {
+      retiredWaitRows += 1;
+      continue;
+    }
     events.push(event);
   }
   const supervisorIds = /* @__PURE__ */ new Set();
   for (const event of events) {
-    if (event.kind !== "spawned" && event.kind !== "waiting") continue;
+    if (event.kind !== "spawned") continue;
     const parent = event.parent;
     if (typeof parent === "string" && parent.length > 0) supervisorIds.add(parent);
   }
@@ -17533,6 +17539,7 @@ function readRunContextJournal(text2) {
     events,
     supervisorIds,
     invalidRows,
+    retiredWaitRows,
     partialTail: parsed.partialTail,
     present: parsed.present
   };
@@ -17722,6 +17729,7 @@ async function readRunContextSnapshot(runDir) {
     result,
     terminal: resultText !== null,
     invalidJournalRows: journal.invalidRows,
+    retiredWaitRows: journal.retiredWaitRows,
     invalidCoordinationRows: coordination.invalidRows,
     partialTail: journal.partialTail || coordination.partialTail,
     treeError: view === null && journal.present ? "spawn journal could not be folded into a tree (a settle or cancel names a node that was never spawned)" : null
@@ -17745,7 +17753,7 @@ function orderNodes(view, journal) {
   const parentOf = new Map(view.nodes.map((n) => [n.id, n.parent]));
   for (const event of journal.events) {
     const at = parseInstant(event.at);
-    if (event.kind === "spawned" || event.kind === "waiting") {
+    if (event.kind === "spawned") {
       spawnedAt.set(event.id, at);
     } else if (event.kind === "metered") {
       driver.set(event.id, addSpend(driver.get(event.id) ?? ZERO_ROLL, event.spend));
@@ -17759,8 +17767,6 @@ function orderNodes(view, journal) {
     } else if (event.kind === "cancelled") {
       settledAt.set(event.id, at);
       cancelReason.set(event.id, event.reason);
-    } else if (event.kind === "woken") {
-      settledAt.set(event.id, at);
     }
   }
   const byParent = /* @__PURE__ */ new Map();
@@ -17844,7 +17850,6 @@ var STATUS_MARK = {
   pending: "\xB7",
   acquiring: "\u25D0",
   running: "\u25B6",
-  waiting: "\u23F8",
   done: "\u2713",
   failed: "\u2717",
   cancelled: "\u2298"
@@ -18265,6 +18270,9 @@ function renderRunContextSnapshot(snapshot) {
   }
   const problems = [];
   if (snapshot.invalidJournalRows > 0) problems.push(`${snapshot.invalidJournalRows} unreadable journal row(s)`);
+  if (snapshot.retiredWaitRows > 0) {
+    problems.push(`${snapshot.retiredWaitRows} wait-state journal row(s) from an older Runtime left out of the tree`);
+  }
   if (snapshot.invalidCoordinationRows > 0) {
     problems.push(`${snapshot.invalidCoordinationRows} unreadable coordination row(s)`);
   }
@@ -18447,6 +18455,7 @@ import { profileOptimizerModelCall } from "@tangle-network/agent-runtime/kernel"
 import { createDspyRlmTraceEngine } from "@tangle-network/agent-eval/analyst";
 var DEFAULT_ANALYST_MODEL = "gpt-5.6-luna";
 var DEFAULT_QUESTION_MAX_COST_USD = 1;
+var DEFAULT_QUESTION_TIMEOUT_MS = 6e5;
 var TANGLE_ROUTER_BASE_URL = "https://router.tangle.tools/v1";
 var ANALYST_MAX_OUTPUT_TOKENS = 16384;
 var GPT_5_6_ANALYST_MAX_OUTPUT_TOKENS = 8192;
@@ -18462,7 +18471,11 @@ function createAnalystModelOwner(opts) {
       provider: opts.provider,
       default: opts.model,
       reasoningEffort: "none",
-      maxVisibleOutputTokens: maxOutputTokens
+      maxVisibleOutputTokens: maxOutputTokens,
+      // Stream every analyst call. The Tangle Router bounds a buffered request by a deadline that
+      // ends when response headers are sent, and a thinking model's long recursive step outlives
+      // it (`request_deadline_exceeded`). A streamed response sends headers at the first byte.
+      metadata: { stream: true }
     }
   };
   return {
@@ -18521,6 +18534,7 @@ function analysisEngineFromEnv(opts) {
     // reservation grows ~4x, so that default can bind before the caller's own
     // allocation and kill investigations mid-run.
     ...opts.maxCostUsd !== void 0 && Number.isFinite(opts.maxCostUsd) && opts.maxCostUsd > 0 ? { maxCostUsd: opts.maxCostUsd } : {},
+    timeoutMs: opts.timeoutMs ?? DEFAULT_QUESTION_TIMEOUT_MS,
     ...python ? { runner: { command: python } } : {}
   });
 }
@@ -19108,6 +19122,7 @@ async function runTraceQuestions(opts) {
     effectiveConcurrency,
     budgetUsd: budgetUsd ?? null,
     questionBudgetUsd: typeof opts.engine.executionConfig.max_cost_usd === "number" ? opts.engine.executionConfig.max_cost_usd : null,
+    questionTimeoutMs: typeof opts.engine.executionConfig.timeout_ms === "number" ? opts.engine.executionConfig.timeout_ms : null,
     spanCount: opts.spans.length,
     otlpPath: traceFile.otlpPath,
     traces,
@@ -20874,6 +20889,7 @@ export {
   watchRunTarget,
   DEFAULT_ANALYST_MODEL,
   DEFAULT_QUESTION_MAX_COST_USD,
+  DEFAULT_QUESTION_TIMEOUT_MS,
   analysisEngineFromEnv,
   assertAnswerSchema,
   answerSchemaErrors,

package/dist/cli.js

@@ -2,6 +2,7 @@
 import {
   DEFAULT_ANALYST_MODEL,
   DEFAULT_QUESTION_MAX_COST_USD,
+  DEFAULT_QUESTION_TIMEOUT_MS,
   DEFAULT_SANDBOX_BASE_URL,
   DEFAULT_SWEEP_THRESHOLD,
   MAX_TRACE_QUESTION_CHARS,
@@ -89,7 +90,7 @@ import {
   writeSessionIndexFile,
   writeTraceEvidenceExportFile,
   writeTraceQuestionsArtifacts
-} from "./chunk-2NINDZCW.js";
+} from "./chunk-LB6AD6SX.js";
 
 // src/cli.ts
 import { readFileSync } from "fs";
@@ -590,6 +591,9 @@ function parseArgs(argv) {
       case "--question-budget":
         a.questionBudget = Number(next());
         break;
+      case "--question-timeout":
+        a.questionTimeout = Number(next());
+        break;
       case "--redactor":
         a.redactorCmd = next();
         break;
@@ -1523,9 +1527,14 @@ async function cmdAsk(args) {
   if (args.budget !== void 0 && (!Number.isFinite(args.budget) || args.budget <= 0)) {
     throw new Error("--budget must be a positive number of USD");
   }
+  const questionTimeoutMs = args.questionTimeout === void 0 ? void 0 : Math.round(args.questionTimeout * 1e3);
+  if (questionTimeoutMs !== void 0 && (!Number.isSafeInteger(questionTimeoutMs) || questionTimeoutMs < 1e3)) {
+    throw new Error("--question-timeout must be a number of seconds, at least 1");
+  }
   const engine = analysisEngineFromEnv({
     model: analystModelFor(args),
     maxCostUsd: args.questionBudget ?? Math.min(args.budget ?? Infinity, DEFAULT_QUESTION_MAX_COST_USD),
+    ...questionTimeoutMs !== void 0 ? { timeoutMs: questionTimeoutMs } : {},
     log: analystLog
   });
   const collected = await collectSpans(args);
@@ -2161,6 +2170,9 @@ Options:
   --question-budget <usd>
                    ask: provider ceiling for one question (default: the smaller
                    of --budget and $${DEFAULT_QUESTION_MAX_COST_USD})
+  --question-timeout <seconds>
+                   ask: wall-clock deadline for one question's investigation
+                   (default: ${DEFAULT_QUESTION_TIMEOUT_MS / 1e3})
   --concurrency <n> ask: questions running at once (default 4);
                    import-codetracebench: trajectories imported at once
   --analyzer <id>  analyze: also run halo, hodoscope, prime, or an installed command (repeatable);

package/dist/index.d.ts

@@ -3593,6 +3593,8 @@ interface TraceQuestionsResult {
     readonly budgetUsd: number | null;
     /** The engine's own ceiling for one question, when it declares one. */
     readonly questionBudgetUsd: number | null;
+    /** The engine's wall-clock deadline for one question, when it declares one. */
+    readonly questionTimeoutMs: number | null;
     readonly spanCount: number;
     readonly otlpPath: string;
     readonly traces: readonly TraceQuestionTrace[];
@@ -4154,6 +4156,8 @@ interface AnalysisEngineFromEnvOptions {
      * cap here and the total to the ledger.
      */
     maxCostUsd?: number;
+    /** Wall-clock deadline for ONE investigation. Default: {@link DEFAULT_QUESTION_TIMEOUT_MS}. */
+    timeoutMs?: number;
     /** Receives one line per model call, in the CLI's analyst log format. */
     log?: (msg: string, fields?: Record<string, unknown>) => void;
     /** Environment to read credentials and the Python interpreter from. Default: process.env. */
@@ -4389,6 +4393,8 @@ interface RunContextJournal {
     /** Node ids that appear as some spawn's `parent` — the exact supervisor set. */
     readonly supervisorIds: ReadonlySet<string>;
     readonly invalidRows: number;
+    /** Readable `waiting`/`woken` rows from a Runtime that still had wait-state nodes. */
+    readonly retiredWaitRows: number;
     readonly partialTail: boolean;
     readonly present: boolean;
 }
@@ -4526,6 +4532,8 @@ interface RunContextSnapshot {
     readonly terminal: boolean;
     /** Rows that were present and unreadable, per artifact. */
     readonly invalidJournalRows: number;
+    /** Wait-state rows from a pre-0.305 Runtime, left out of the tree. */
+    readonly retiredWaitRows: number;
     readonly invalidCoordinationRows: number;
     /** True when an artifact's last line was still being written. */
     readonly partialTail: boolean;

package/dist/index.js

@@ -284,7 +284,7 @@ import {
   writeTraceQuestionsArtifacts,
   zaiChatCaller,
   zeroPricedCost
-} from "./chunk-2NINDZCW.js";
+} from "./chunk-LB6AD6SX.js";
 
 // src/adapters/factory.ts
 import { readdir, stat } from "fs/promises";

package/docs/trace-analysts.md

@@ -124,6 +124,7 @@ This package carries no JSON Schema library, and a constraint that is quietly ig
 - Ctrl-C keeps what the run already bought. The signal reaches the engine, not the checks that follow it: an answer that came back is kept with its citations resolved, and each question the run never reached is recorded as `aborted`.
 - The artifacts are written before the exit code is decided. `ask` exits 1 when any question failed, returned no answer, broke its schema, or cited a missing span.
 - `totals.wallTimeMs` covers the whole run, including writing and indexing the trace file; `totals.setupTimeMs` names that part. `result.effectiveConcurrency` is the number of workers the run created, `min(--concurrency, questions)`, and `totals.peakConcurrency` is how many actually overlapped.
+- `--question-timeout <seconds>` is the wall-clock deadline for one question's investigation (default 600). A question over many large sessions can need longer; the timeout fails that question alone.
 
 ### Budget under concurrency
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/traces",
-  "version": "0.20.0",
+  "version": "0.21.1",
   "description": "Point it at your coding-agent session traces (Claude Code, Codex, OpenCode, Gemini, Pi, \u2026) and get failure-mode + efficiency findings. CLI + SDK over the @tangle-network/agent-eval analyst suite \u2014 observe live sessions, run your own analysts, redact, and upload to the Tangle Intelligence Platform.",
   "type": "module",
   "license": "MIT",
@@ -61,9 +61,9 @@
     "prepublishOnly": "pnpm check:source && pnpm build && pnpm check:package"
   },
   "dependencies": {
-    "@tangle-network/agent-eval": "0.205.1",
-    "@tangle-network/agent-interface": "2.17.0",
-    "@tangle-network/agent-runtime": "0.293.0",
+    "@tangle-network/agent-eval": "0.209.1",
+    "@tangle-network/agent-interface": "3.1.2",
+    "@tangle-network/agent-runtime": "0.313.3",
     "@tangle-network/agent-trace-contract": "2.0.0",
     "@tangle-network/harness-sessions": "0.1.0",
     "@tangle-network/sandbox": "0.60.10"