fix: let replay recover from dead letter jobs

src/autonomy/llm.js

@@ -11,8 +11,14 @@ function extractJson(text) {
 async function callCoordinator(config, prompt) {
   const apiKey = String(config?.openRouter?.apiKey || '').trim();
   if (!apiKey) throw new Error('OpenRouter API key is not configured');
-  const response = await fetch('https://openrouter.ai/api/v1/chat/completions', {
+  const timeoutMs = Math.max(1000, Number(config?.openRouter?.timeoutMs || process.env.OPEN_ROUTER_TIMEOUT_MS) || 60000);
+  const controller = new AbortController();
+  const timeout = setTimeout(() => controller.abort(), timeoutMs);
+  let response;
+  try {
+    response = await fetch('https://openrouter.ai/api/v1/chat/completions', {
       method: 'POST',
+      signal: controller.signal,
       headers: { Authorization: `Bearer ${apiKey}`, 'Content-Type': 'application/json' },
       body: JSON.stringify({
         model: config.openRouter.llmModel,
@@ -24,7 +30,16 @@ async function callCoordinator(config, prompt) {
         ],
       }),
     });
-  if (!response.ok) throw new Error(`coordinator request failed with ${response.status}`);
+  } catch (error) {
+    const cause = error?.cause?.code || error?.code || error?.name || 'network_error';
+    throw new Error(`coordinator request failed before response (${cause})`);
+  } finally {
+    clearTimeout(timeout);
+  }
+  if (!response.ok) {
+    const body = await response.text().catch(() => '');
+    throw new Error(`coordinator request failed with ${response.status}: ${body.slice(0, 300)}`);
+  }
   const body = await response.json();
   return extractJson(body?.choices?.[0]?.message?.content);
 }

test/autonomy.test.js

@@ -11,6 +11,7 @@ const { validatePaperIntent, createSimulator } = require('../src/autonomy/execut
 const { calculateOutcome } = require('../src/autonomy/outcomes');
 const { createOrderIntent } = require('../src/autonomy/orderIntents');
 const { reconcileArchiveBatch, reconcileLiveBatch } = require('../workers/autonomyWorker');
+const { scheduleNext } = require('../workers/replayWorker');
 
 test('autonomy schema and leased jobs are restart-safe', () => {
   const db = new Database(':memory:');
@@ -73,6 +74,40 @@ test('replay evidence cannot look beyond its information cutoff and remains non-
   assert.equal(executable.length, 0);
 });
 
+test('replay scheduler skips terminal replay jobs instead of pinning the cursor', () => {
+  const archive = new Database(':memory:');
+  archive.exec(`
+    CREATE TABLE articles (
+      id INTEGER PRIMARY KEY,
+      title TEXT,
+      description TEXT,
+      content TEXT,
+      pub_date_effective TEXT,
+      is_index_page INTEGER
+    )
+  `);
+  archive.prepare("INSERT INTO articles VALUES (1, 'bad', '', 'content', '2020-01-01T00:00:00Z', 0)").run();
+  archive.prepare("INSERT INTO articles VALUES (2, 'next', '', 'content', '2020-01-02T00:00:00Z', 0)").run();
+
+  const intelligence = new Database(':memory:');
+  initAutonomySchema(intelligence);
+  const runId = intelligence.prepare(`
+    INSERT INTO autonomy_replay_runs (watermark_at, strategy_version, prompt_version, coordinator_model)
+    VALUES ('2020-01-03T00:00:00Z', 'test', 'test', 'test')
+  `).run().lastInsertRowid;
+  enqueueJob(intelligence, {
+    jobType: 'replay_article', lane: 'historical', priority: 1, entityType: 'article', entityId: 1,
+    idempotencyKey: `replay:${runId}:article:1`,
+  });
+  intelligence.prepare("UPDATE autonomy_jobs SET status='dead_letter', attempts=5, last_error='fetch failed'").run();
+
+  const run = intelligence.prepare('SELECT * FROM autonomy_replay_runs WHERE id=?').get(runId);
+  const next = scheduleNext(intelligence, archive, run);
+  assert.equal(next.id, 2);
+  assert.equal(intelligence.prepare('SELECT cursor_article_id FROM autonomy_replay_runs WHERE id=?').get(runId).cursor_article_id, 1);
+  assert.equal(intelligence.prepare("SELECT COUNT(*) count FROM autonomy_jobs WHERE status='pending' AND entity_id='2'").get().count, 1);
+});
+
 test('calibration and policy abstain on insufficient evidence', () => {
   const calibration = calibrateOutcomes([
     { excess_return: 0.02, direction_correct: 1 },

workers/replayWorker.js

@@ -23,7 +23,7 @@ function articleTimeColumns(archiveDb) {
   const columns = new Set(archiveDb.prepare('PRAGMA table_info(articles)').all().map((row) => row.name));
   const candidates = ['pub_date_effective', 'pub_date', 'ingested_at'].filter((name) => columns.has(name));
   if (!candidates.length) throw new Error('archive articles need publication or ingestion timestamps for replay');
-  return { columns, effective: `COALESCE(${candidates.join(', ')})` };
+  return { columns, effective: candidates.length === 1 ? candidates[0] : `COALESCE(${candidates.join(', ')})` };
 }
 
 function replayPrompt(article) {
@@ -52,11 +52,15 @@ function scheduleNext(db, archiveDb, run) {
   const { columns, effective } = articleTimeColumns(archiveDb);
   const content = columns.has('content') ? "content IS NOT NULL AND content != ''" : '1=1';
   const indexFilter = columns.has('is_index_page') ? 'AND (is_index_page = 0 OR is_index_page IS NULL)' : '';
-  const cursorFilter = run.cursor_effective_at
+  let cursorEffectiveAt = run.cursor_effective_at;
+  let cursorArticleId = run.cursor_article_id;
+
+  for (let skipped = 0; skipped < 100; skipped += 1) {
+    const cursorFilter = cursorEffectiveAt
       ? `AND (datetime(${effective}) > datetime(?) OR (datetime(${effective}) = datetime(?) AND id > ?))`
       : '';
     const params = [run.watermark_at];
-  if (run.cursor_effective_at) params.push(run.cursor_effective_at, run.cursor_effective_at, run.cursor_article_id);
+    if (cursorEffectiveAt) params.push(cursorEffectiveAt, cursorEffectiveAt, cursorArticleId);
     const article = archiveDb.prepare(`
       SELECT id, title, description, content, ${effective} AS effective_at
       FROM articles
@@ -64,11 +68,29 @@ function scheduleNext(db, archiveDb, run) {
       ORDER BY datetime(${effective}) ASC, id ASC LIMIT 1
     `).get(...params);
     if (!article) return null;
+
+    const idempotencyKey = `replay:${run.id}:article:${article.id}`;
+    const existing = db.prepare('SELECT status, last_error FROM autonomy_jobs WHERE idempotency_key = ?').get(idempotencyKey);
+    if (existing && ['complete', 'dead_letter'].includes(existing.status)) {
+      db.prepare(`
+        UPDATE autonomy_replay_runs
+        SET cursor_article_id=?, cursor_effective_at=?, last_error=?, updated_at=datetime('now')
+        WHERE id=?
+      `).run(article.id, article.effective_at, existing.status === 'dead_letter'
+        ? `Skipped dead-letter replay job for article ${article.id}: ${existing.last_error || 'unknown error'}`
+        : run.last_error, run.id);
+      cursorEffectiveAt = article.effective_at;
+      cursorArticleId = article.id;
+      continue;
+    }
+
     enqueueJob(db, {
       jobType: 'replay_article', lane: 'historical', priority: 1, entityType: 'article', entityId: article.id,
-    idempotencyKey: `replay:${run.id}:article:${article.id}`,
+      idempotencyKey,
     });
     return article;
+  }
+  throw new Error('replay scheduler skipped too many terminal jobs in one pass');
 }
 
 async function runReplayWorker({ archivePath, intelligencePath, workerId = `replay-${os.hostname()}-${process.pid}`, pollMs = 15000 } = {}) {