@exaudeus/workrail 3.59.6 → 3.59.8

package/dist/manifest.json

@@ -238,8 +238,8 @@
       "bytes": 31
     },
     "cli-worktrain.js": {
-      "sha256": "34c2a2d4596a1f1b4c385366c601391cc329fd85abccdebce5d74a9bd1631aac",
-      "bytes": 60787
+      "sha256": "ffe47ef9595c11968952fe573e7ec11bf87813bd0197d24c173d7740479b8f5b",
+      "bytes": 58480
     },
     "cli.d.ts": {
       "sha256": "43e818adf60173644896298637f47b01d5819b17eda46eaa32d0c7d64724d012",
@@ -258,12 +258,12 @@
       "bytes": 745
     },
     "cli/commands/index.d.ts": {
-      "sha256": "3c4cad3ea64ac1df28d57f1bebcd211a0ff059898c5481821358ae259aae5233",
-      "bytes": 2667
+      "sha256": "088a55b17f60d1b3f6eb289d51b4af75b44a7ef72b772a95b867321ebd3cb3ed",
+      "bytes": 2522
     },
     "cli/commands/index.js": {
-      "sha256": "f4596c05ea9c587700bf53d7bf8f21d157a23deac966fd17bea039dd024c6896",
-      "bytes": 6122
+      "sha256": "60eed410bbf8faae535cb5f27971159a943bbc480450108683623d79d1fe1ad9",
+      "bytes": 5818
     },
     "cli/commands/init.d.ts": {
       "sha256": "b5f8b88a072c68509dab3938ba1d6b4a949ad32f8fc55e91c5039b8c77301c1b",
@@ -393,14 +393,6 @@
       "sha256": "b0286fef461835a0b73070fd278e43af5f3a1fbebbe1c6de1fc39ace4075df8f",
       "bytes": 1395
     },
-    "cli/commands/worktrain-trigger-poll.d.ts": {
-      "sha256": "73cffd1f6a44d8b9ef35bd15492dabc32d3d6e907c52617c7f482fd5229aed93",
-      "bytes": 934
-    },
-    "cli/commands/worktrain-trigger-poll.js": {
-      "sha256": "aedf752565bbf6053bfdd79a7b1452b67e4c786e01a160aede9390b2313ccf91",
-      "bytes": 3920
-    },
     "cli/commands/worktrain-trigger-test.d.ts": {
       "sha256": "3b85edacabf0657b208892f13b8fb540f794f47f18b5a1263562d3518f7fce43",
       "bytes": 1357
@@ -481,16 +473,16 @@
       "sha256": "5fe866e54f796975dec5d8ba9983aefd86074db212d3fccd64eed04bc9f0b3da",
       "bytes": 8011
     },
+    "console-ui/assets/index-CtUfpXCx.js": {
+      "sha256": "a0106452fd0eb629474a299b2eb757cae92578a4c5ed792bd6a2daefb203645f",
+      "bytes": 760528
+    },
     "console-ui/assets/index-DGj8EsFR.css": {
       "sha256": "3bdb55ec0957928e0ebbb86a7d6b36d28f7ba7d5c0f3e236fd8f2e2aacee2fa4",
       "bytes": 60631
     },
-    "console-ui/assets/index-xMwhHmR2.js": {
-      "sha256": "9fbff5f59a5e014930778fd79a1de39f532b20862a4fa10d7320d12602b445fc",
-      "bytes": 760528
-    },
     "console-ui/index.html": {
-      "sha256": "1aa1ca8a09fba68cb675918108702a338ec426ec0a6778736dbd1842e8a1a8bc",
+      "sha256": "5cb6ae1aa3a61a3c9eb6138d92f686d9880cb4822cc96c2399c9eb0c223ce67c",
       "bytes": 417
     },
     "console/standalone-console.d.ts": {
@@ -498,8 +490,8 @@
       "bytes": 788
     },
     "console/standalone-console.js": {
-      "sha256": "759d8be0b9804aa2e6301f14afb2dc97d45231434058f64d1cfbec6e025a1991",
-      "bytes": 6325
+      "sha256": "bfe72ea5161d6d3af917180fad905abae05b9ae5cb329dadbe3d8acaed6aed9a",
+      "bytes": 6314
     },
     "context-assembly/deps.d.ts": {
       "sha256": "d699ae8f8f081d92708eba5969e35cf24a45d3f86de72d308ffc4a542b954bc3",
@@ -650,8 +642,8 @@
       "bytes": 7307
     },
     "daemon/workflow-runner.js": {
-      "sha256": "2184d5202eadecff166aeb5371d84c1660a9b6cf3fdf7598d311172fdfdb910a",
-      "bytes": 95312
+      "sha256": "0d4991c3589e75679d4035d506d84ebe595df8328c6e632d352597c9e23ad741",
+      "bytes": 95348
     },
     "di/container.d.ts": {
       "sha256": "003bb7fb7478d627524b9b1e76bd0a963a243794a687ff233b96dc0e33a06d9f",
@@ -1661,14 +1653,6 @@
       "sha256": "6728a2169f4007b9ea0414fade6b21500500d9c79d0b09296d92ef8bcabb9c79",
       "bytes": 2763
     },
-    "trigger/daemon-console.d.ts": {
-      "sha256": "a3b9a9f58482c6ea379c0e02c30f55a5820c7c37fa3fae55fc336cd518f35462",
-      "bytes": 1162
-    },
-    "trigger/daemon-console.js": {
-      "sha256": "f2f09c05e48b42ebf1c7be137fc6eced46673048471b7114434710b5691fe6f2",
-      "bytes": 5497
-    },
     "trigger/delivery-action.d.ts": {
       "sha256": "559e2b2645aa60528f73de351cd35ebf45c5b82f47797aa15ddd681319315d39",
       "bytes": 1759
@@ -1722,8 +1706,8 @@
       "bytes": 1126
     },
     "trigger/polling-scheduler.js": {
-      "sha256": "0588534bacec382e00eb766d33438569f31563e5bbad8d3498ef64c00a0f87db",
-      "bytes": 23912
+      "sha256": "61b94e35aae2e9578a9e9cc32548791166b9ec98abb8f2cff58135fc6b3e5593",
+      "bytes": 23945
     },
     "trigger/trigger-listener.d.ts": {
       "sha256": "1eebb3d4829030b264c3798b0b0d55d7357d313ab83e3f344ad455eaafcedb44",
@@ -3070,12 +3054,12 @@
       "bytes": 4795
     },
     "v2/usecases/console-routes.d.ts": {
-      "sha256": "f97d6db5323c14a310064657dc3ea160baa77f387a28514efcdbd2b83f53b79f",
-      "bytes": 880
+      "sha256": "bb8abe9b6510ab09ab9aec1af24c253bf16b538c18c5ac583caffe48ef347536",
+      "bytes": 596
     },
     "v2/usecases/console-routes.js": {
-      "sha256": "76e34345d329bfc8a998ebc50ef8bd6ead1b7914873f6350ba479909c9c097be",
-      "bytes": 31648
+      "sha256": "80d7629410da6b991471497b209d85969192d1fe3601af9e643e030b93b9aa54",
+      "bytes": 28556
     },
     "v2/usecases/console-service.d.ts": {
       "sha256": "fc8fe65427fa9f4f3535344b385b36f66ca06b7e3bfaea708931817a3edcad2b",

package/dist/trigger/polling-scheduler.js

@@ -475,7 +475,7 @@ function extractDotPath(obj, rawPath) {
 async function countActiveSessions(sessionsDir) {
     try {
         const files = await fs.readdir(sessionsDir);
-        return files.filter((f) => f.endsWith('.json')).length;
+        return files.filter((f) => f.endsWith('.json') && !f.startsWith('queue-issue-')).length;
     }
     catch {
         return 0;

package/dist/v2/usecases/console-routes.d.ts

@@ -2,7 +2,5 @@ import type { Application } from 'express';
 import type { ConsoleService } from './console-service.js';
 import type { WorkflowService } from '../../application/services/workflow-service.js';
 import type { ToolCallTimingRingBuffer } from '../../mcp/tool-call-timing.js';
-import type { TriggerRouter } from '../../trigger/trigger-router.js';
 import type { V2ToolContext } from '../../mcp/types.js';
-import type { SteerRegistry } from '../../daemon/workflow-runner.js';
-export declare function mountConsoleRoutes(app: Application, consoleService: ConsoleService, workflowService?: WorkflowService, timingRingBuffer?: ToolCallTimingRingBuffer, toolCallsPerfFile?: string, serverVersion?: string, v2ToolContext?: V2ToolContext, triggerRouter?: TriggerRouter, steerRegistry?: SteerRegistry, pollingScheduler?: import('../../trigger/polling-scheduler.js').PollingScheduler): () => void;
+export declare function mountConsoleRoutes(app: Application, consoleService: ConsoleService, workflowService?: WorkflowService, timingRingBuffer?: ToolCallTimingRingBuffer, toolCallsPerfFile?: string, serverVersion?: string, v2ToolContext?: V2ToolContext): () => void;

package/dist/v2/usecases/console-routes.js

@@ -91,7 +91,7 @@ function loadWorkflowTags() {
         return { version: 0, tags: [], workflows: {} };
     }
 }
-function mountConsoleRoutes(app, consoleService, workflowService, timingRingBuffer, toolCallsPerfFile, serverVersion, v2ToolContext, triggerRouter, steerRegistry, pollingScheduler) {
+function mountConsoleRoutes(app, consoleService, workflowService, timingRingBuffer, toolCallsPerfFile, serverVersion, v2ToolContext) {
     const sseClients = new Set();
     let sseDebounceTimer = null;
     function broadcastChange() {
@@ -429,13 +429,11 @@ function mountConsoleRoutes(app, consoleService, workflowService, timingRingBuff
                 repoRootsExpiresAt = Date.now() + REPO_ROOTS_TTL_MS;
             }
             const repoRoots = cachedRepoRoots;
-            const data = await Promise.race([
-                (0, worktree_service_js_1.getWorktreeList)(repoRoots, activeSessions).finally(() => {
-                    if (timeoutId !== null)
-                        clearTimeout(timeoutId);
-                }),
-                timeoutPromise,
-            ]);
+            const worktreeWork = (0, worktree_service_js_1.getWorktreeList)(repoRoots, activeSessions)
+                .finally(() => { if (timeoutId !== null)
+                clearTimeout(timeoutId); })
+                .catch(() => ({ repos: [] }));
+            const data = await Promise.race([worktreeWork, timeoutPromise]);
             if (timeoutId !== null)
                 clearTimeout(timeoutId);
             res.json({ success: true, data });
@@ -532,7 +530,7 @@ function mountConsoleRoutes(app, consoleService, workflowService, timingRingBuff
     }
     app.post('/api/v2/auto/dispatch', express_1.default.json(), async (req, res) => {
         if (!v2ToolContext) {
-            res.status(503).json({ success: false, error: 'Autonomous dispatch requires v2 tools enabled.' });
+            res.status(503).json({ success: false, error: 'Autonomous dispatch requires the WorkTrain daemon. Run worktrain console alongside worktrain daemon to enable browser dispatch.' });
             return;
         }
         const body = req.body;
@@ -590,107 +588,30 @@ function mountConsoleRoutes(app, consoleService, workflowService, timingRingBuff
             sessionHandle = workflowId;
         }
         const trigger = { workflowId, goal, workspacePath, context, _preAllocatedStartResponse: startResponse };
-        if (triggerRouter) {
-            triggerRouter.dispatch(trigger);
-        }
-        else {
-            void (0, workflow_runner_js_1.runWorkflow)(trigger, v2ToolContext, apiKey ?? '', undefined, undefined, steerRegistry).then((result) => {
-                if (result._tag === 'success') {
-                    console.log(`[ConsoleRoutes] Auto dispatch completed: workflowId=${workflowId} stopReason=${result.stopReason}`);
-                }
-                else if (result._tag === 'delivery_failed') {
-                    console.log(`[ConsoleRoutes] Auto dispatch delivery failed: workflowId=${workflowId}`);
-                }
-                else if (result._tag === 'timeout') {
-                    console.log(`[ConsoleRoutes] Auto dispatch timed out: workflowId=${workflowId}`);
-                }
-                else if (result._tag === 'error') {
-                    console.log(`[ConsoleRoutes] Auto dispatch failed: workflowId=${workflowId} error=${result.message}`);
-                }
-                else if (result._tag === 'stuck') {
-                    console.log(`[ConsoleRoutes] Auto dispatch stuck: workflowId=${workflowId} reason=${result.reason} message=${result.message}`);
-                }
-                else {
-                    (0, assert_never_js_1.assertNever)(result);
-                }
-            });
-        }
+        void (0, workflow_runner_js_1.runWorkflow)(trigger, v2ToolContext, apiKey ?? '', undefined, undefined, undefined).then((result) => {
+            if (result._tag === 'success') {
+                console.log(`[ConsoleRoutes] Auto dispatch completed: workflowId=${workflowId} stopReason=${result.stopReason}`);
+            }
+            else if (result._tag === 'delivery_failed') {
+                console.log(`[ConsoleRoutes] Auto dispatch delivery failed: workflowId=${workflowId}`);
+            }
+            else if (result._tag === 'timeout') {
+                console.log(`[ConsoleRoutes] Auto dispatch timed out: workflowId=${workflowId}`);
+            }
+            else if (result._tag === 'error') {
+                console.log(`[ConsoleRoutes] Auto dispatch failed: workflowId=${workflowId} error=${result.message}`);
+            }
+            else if (result._tag === 'stuck') {
+                console.log(`[ConsoleRoutes] Auto dispatch stuck: workflowId=${workflowId} reason=${result.reason} message=${result.message}`);
+            }
+            else {
+                (0, assert_never_js_1.assertNever)(result);
+            }
+        });
         res.json({ success: true, data: { status: 'dispatched', workflowId, sessionHandle } });
     });
     app.get('/api/v2/triggers', (_req, res) => {
-        if (!triggerRouter) {
-            res.json({ success: true, data: { triggers: [] } });
-            return;
-        }
-        const triggers = triggerRouter.listTriggers().map((t) => ({
-            id: t.id,
-            provider: t.provider,
-            workflowId: t.workflowId,
-            workspacePath: t.workspacePath,
-            goal: t.goal,
-        }));
-        res.json({ success: true, data: { triggers } });
-    });
-    app.post('/api/v2/triggers/:triggerId/poll', async (req, res) => {
-        if (!pollingScheduler) {
-            res.status(503).json({ success: false, error: 'Force poll not available (not a daemon context).' });
-            return;
-        }
-        const triggerId = req.params['triggerId'] ?? '';
-        if (!triggerId) {
-            res.status(400).json({ success: false, error: 'Missing triggerId' });
-            return;
-        }
-        const result = await pollingScheduler.forcePoll(triggerId);
-        switch (result.kind) {
-            case 'ok':
-                res.json({
-                    success: true,
-                    data: {
-                        triggerId,
-                        cycleRan: result.cycleRan,
-                        message: result.cycleRan
-                            ? `Poll cycle started for trigger '${triggerId}'.`
-                            : `Poll cycle skipped for trigger '${triggerId}' -- a previous cycle is still running.`,
-                    },
-                });
-                return;
-            case 'not_found':
-                res.status(400).json({ success: false, error: `Trigger '${triggerId}' not found` });
-                return;
-            case 'wrong_provider':
-                res.status(400).json({
-                    success: false,
-                    error: `Trigger '${triggerId}' is not a queue poll trigger (provider: ${result.provider})`,
-                });
-                return;
-            default: {
-                const _exhaustive = result;
-                res.status(500).json({ success: false, error: 'Unexpected forcePoll result' });
-                void _exhaustive;
-                return;
-            }
-        }
-    });
-    app.post('/api/v2/sessions/:sessionId/steer', express_1.default.json(), (req, res) => {
-        if (!steerRegistry) {
-            res.status(503).json({ success: false, error: 'Steer not available (not a daemon context).' });
-            return;
-        }
-        const { sessionId } = req.params;
-        const body = req.body;
-        const text = typeof body.text === 'string' ? body.text.trim() : '';
-        if (!text) {
-            res.status(400).json({ success: false, error: 'text is required and must be a non-empty string.' });
-            return;
-        }
-        const callback = steerRegistry.get(sessionId);
-        if (!callback) {
-            res.status(404).json({ success: false, error: 'Session not found or not a daemon session.' });
-            return;
-        }
-        callback(text);
-        res.json({ success: true });
+        res.json({ success: true, data: { triggers: [] } });
     });
     const consoleDist = resolveConsoleDist();
     if (consoleDist) {

package/docs/design/dispatch-dedup-prealloc-bypass-implementation-plan.md

@@ -1,3 +1,5 @@
+> Note: worktrain trigger poll was removed in PR #755.
+
 # Implementation Plan: Bypass Dispatch Dedup for Pre-Allocated Sessions
 
 **Date:** 2026-04-19

package/docs/ideas/backlog.md

@@ -7486,3 +7486,94 @@ Medium for the cleanup command (quality of life, stops log noise). High for star
 **Files:** `src/coordinators/modes/implement-shared.ts`, `src/coordinators/pr-review.ts`.
 
 **Priority:** Medium. Correctness issues that won't crash in production but make future refactors unsafe.
+
+---
+
+## Current state update (Apr 21, 2026)
+
+**npm version: v3.59.6** | Daemon PID: 54113 | Status: Running, pipeline active
+
+### What shipped in this session (Apr 19-21, 2026)
+
+**All five autonomous pipeline items (previously recorded) plus:**
+
+- ✅ **Discovery loop fix** (#748) -- three coupled fixes: thread `maxSessionMinutes` through `spawnSession` (sessions now get 55/35/65 min instead of 30 min default), inspect `PipelineOutcome` in polling-scheduler and apply `worktrain:in-progress` label on escalation, write issue-ownership sidecar for cross-restart idempotency
+- ✅ **In-process `awaitSessions` and `getAgentResult`** (#741) -- replaced HTTP calls to the daemon's own console with direct `ConsoleService` access
+- ✅ **Try/catch on all coordinator I/O** (#740) -- `getAgentResult`, `pollForPR`, `postToOutbox` all wrapped; coordinator no longer crashes on I/O failure
+- ✅ **Dispatch dedup prealloc bypass** (#744) -- `dispatch()` now bypasses dedup for pre-allocated sessions, fixing the zombie session bug that prevented discovery from starting
+- ✅ **Promise.race crash fix** (#733) -- worktrees scan timeout no longer crashes the daemon via unhandled rejection
+- ✅ **Trigger validator** (#690) -- `worktrain trigger validate` command, `validateTriggerStrict()` pure function
+- ✅ **`worktrain trigger poll`** (#697) -- force immediate poll cycle on any queue trigger
+- ✅ **`worktrain trigger test`** (#656) -- dry-run showing what would dispatch
+- ✅ **Auto-load ~/.workrail/.env** (#673) -- daemon reads secrets from .env automatically
+- ✅ **Daemon lifecycle events** (#674) -- `session_aborted` on SIGTERM, `daemon_heartbeat` every 30s
+- ✅ **Attribution signals** (#658) -- `[WT]` PR title prefix, `Co-authored-by: WorkTrain` commit trailers, `worktrain:generated` label
+- ✅ **Secret scan before push** (#660) -- pattern-based scan blocks commits with leaked credentials
+- ✅ **Unified logs stream** (#680) -- `worktrain logs` now merges daemon events, queue-poll.jsonl, and filtered stderr
+- ✅ **Stale lock file handling** (#705) -- validates lock file PID before trusting port discovery
+- ✅ **5 architectural audits** (docs/design/) -- coordinator access, error handling, testability, type bloat, memory management
+- ✅ **Stale user workflow cleanup** -- removed old copies from `~/.workrail/workflows/` that were causing ValidationError noise
+
+### Current pipeline state (live)
+
+Discovery session `ecf359d7` running: 77 turns, 11 step advances (active, making real progress on issue #393). Session `b7df0c8b` also running (just started). First clean run after all pipeline fixes landed.
+
+### Accurate limitations (v3.59.6)
+
+1. **Ghost sessions in event log** -- sessions killed by daemon crashes don't get `session_aborted` events from old daemon instances. New daemons emit it on shutdown, but historical sessions show as RUNNING.
+2. **Worktree orphan leak** -- if `maybeRunDelivery()` worktree removal fails after sidecar deletion, orphan is invisible to `runStartupRecovery`. See backlog.
+3. **`queue-poll.jsonl` never rotated** -- disk exhaustion risk on long-running daemons. See backlog.
+4. **`ReviewSeverity` missing `assertNever`** -- future variants silently fall through. See backlog.
+5. **`process.stderr.write` in `readVerdictArtifact`** -- bypasses injected dep, invisible to test fakes. See backlog.
+6. **WorkRail MCP stale state** -- `workrail cleanup` command doesn't exist yet. Manual cleanup needed for dead managed sources, old session accumulation.
+7. **Trigger validation static/runtime gap** -- some runtime checks not in static validator. See trigger-validation-gap-audit.md.
+8. **WorkflowTrigger type bloat** -- mixes trigger config, session runtime state, delivery config. See workflow-trigger-lifecycle-audit.md.
+9. **Conversation history not persisted** -- LLM conversation history is in-memory only. On crash, context is lost. See backlog.
+
+### Next priorities (groomed Apr 21)
+
+1. **Watch the current pipeline run** -- discovery `ecf359d7` is active at 77 turns/11 steps. If it completes, shaping and coding should fire automatically. First end-to-end validation.
+2. **Execution time tracking** -- add session timing to `execution-stats.jsonl` for timeout calibration. Small change in `runWorkflow()` finally block.
+3. **Three audit findings from above** -- worktree orphan leak, queue-poll rotation, assertNever fixes. All small, targeted.
+4. **`workrail cleanup` command** -- removes dead managed sources, rotates old session files, clears stale git caches. Stops ValidationError noise in MCP server logs.
+5. **Conversation history persistence** -- `conversation.jsonl` per session, append-only. Prerequisite for true crash recovery.
+6. **Autonomous crash recovery and interrupted-session resume** -- see full entry below (Apr 21).
+
+---
+
+## Autonomous crash recovery and interrupted-session resume (Apr 21, 2026)
+
+**The problem we hit today:** A daemon crash loop (console `worktrees scan` unhandled rejection) killed all in-flight sessions. The queue correctly detected the sidecar and skipped re-dispatch for 56 min (TTL), but when the sidecar expired the session was re-dispatched from scratch with zero context from the previous attempt. The agent had already spent ~10 min in Phase 0, read codebase files, and formed a plan -- all of that work was lost.
+
+**What we want:** WorkTrain should be able to detect orphaned sessions on startup and make an autonomous decision: resume if the session had meaningful progress, discard and re-dispatch from scratch if it was too early to be worth resuming.
+
+**Resumability decision criteria (heuristics):**
+- Session had >= 1 `continue_workflow` call (at least one step advance): worth resuming -- the agent made real progress.
+- Session is at step 0 with 0 advances but > 5 LLM turns: borderline -- context was accumulated but no checkpoint written. Resume is risky (stale context), discard is safer. Could surface to console for human decision.
+- Session is at step 0, < 5 turns, < 2 min: discard -- nothing was lost.
+- Session's worktree is missing or corrupted: discard -- can't resume cleanly.
+- Session is on a coding workflow and has uncommitted changes in the worktree: pause for human review before discarding (could have partial work).
+
+**Implementation sketch:**
+
+1. **On daemon startup**, `runStartupRecovery()` already scans `daemon-sessions/` for orphaned token files. Extend it to also inspect the session event log for each orphan:
+   - Count `continue_workflow` calls and LLM turns from `~/.workrail/events/<sessionId>.jsonl`
+   - Apply decision criteria above
+   - For resume candidates: call `continue_workflow` with the checkpoint token and a synthesized re-entry prompt: "You are resuming a session that was interrupted by a daemon crash. Your last known step was [stepLabel]. Continue from where you left off."
+   - For discard candidates: emit `session_aborted` event, delete the sidecar, re-add the issue to the queue (or just let the TTL expire and the queue re-select naturally)
+
+2. **Conversation history prerequisite**: Resume is only useful if the agent can reconstruct its context. Today, conversation history is in-memory only -- it is lost on crash. The `conversation.jsonl` per-session persistence (backlog item #5 above) is a prerequisite for high-quality resume. Without it, resume starts from the workflow system prompt plus the current step recap only. This is enough for mid-pipeline phases (shaping, coding) since they read artifacts from disk. It may be insufficient for early discovery phases.
+
+3. **`worktrain session resume <sessionId>` CLI** -- manual override for human-initiated resume. Useful when the daemon's automatic heuristic chose to discard but the user sees partial work worth keeping.
+
+4. **Queue sidecar TTL for resume vs. discard**: Today the sidecar TTL prevents re-dispatch during the entire pipeline window (56 min). With autonomous resume, the TTL for a discarded session should be much shorter (5 min) so the queue can quickly re-select. For a resumed session, keep the full TTL and extend it by the time already spent.
+
+**Files to change:**
+- `src/daemon/workflow-runner.ts` -- `runStartupRecovery()`: add event log inspection and conditional resume
+- `src/trigger/polling-scheduler.ts` -- `doPollGitHubQueue()`: accept a `ttlOverride` param so discard path uses short TTL
+- `src/trigger/adapters/github-queue-poller.ts` -- `checkIdempotency()`: handle expired sidecars with `ttlOverride`
+- New: `src/daemon/session-recovery-policy.ts` -- pure function `evaluateRecovery(orphan, eventLog) -> 'resume' | 'discard' | 'human_review'`
+
+**Priority:** High. Every daemon crash currently wastes all in-flight work and waits up to 56 min before retrying. With even basic resume (step > 0 → resume, step = 0 → discard + fast re-dispatch), we'd recover most of the lost work and reduce retry latency from 56 min to < 5 min.
+
+**Depends on:** Conversation history persistence (for high-quality resume context).

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@exaudeus/workrail",
-  "version": "3.59.6",
+  "version": "3.59.8",
   "description": "Step-by-step workflow enforcement for AI agents via MCP",
   "license": "MIT",
   "repository": {

package/dist/cli/commands/worktrain-trigger-poll.d.ts

@@ -1,23 +0,0 @@
-import type { CliResult } from '../types/cli-result.js';
-export interface WorktrainTriggerPollDeps {
-    readonly fetch: (url: string, opts: {
-        method: string;
-        signal?: AbortSignal;
-    }) => Promise<{
-        readonly ok: boolean;
-        readonly status: number;
-        readonly json: () => Promise<unknown>;
-    }>;
-    readonly readFile: (path: string) => Promise<string>;
-    readonly deleteFile: (path: string) => Promise<void>;
-    readonly isPidAlive: (pid: number) => boolean;
-    readonly print: (line: string) => void;
-    readonly stderr: (line: string) => void;
-    readonly homedir: () => string;
-    readonly joinPath: (...paths: string[]) => string;
-}
-export interface WorktrainTriggerPollOpts {
-    readonly triggerId: string;
-    readonly port?: number;
-}
-export declare function executeWorktrainTriggerPollCommand(deps: WorktrainTriggerPollDeps, opts: WorktrainTriggerPollOpts): Promise<CliResult>;

package/dist/cli/commands/worktrain-trigger-poll.js

@@ -1,94 +0,0 @@
-"use strict";
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.executeWorktrainTriggerPollCommand = executeWorktrainTriggerPollCommand;
-const cli_result_js_1 = require("../types/cli-result.js");
-const DEFAULT_POLL_PORT = 3200;
-const LOCK_FILE_NAMES = ['daemon-console.lock', 'dashboard.lock'];
-async function discoverConsolePort(deps, portOverride) {
-    if (portOverride !== undefined && portOverride > 0) {
-        return portOverride;
-    }
-    let staleLockPath;
-    for (const lockFileName of LOCK_FILE_NAMES) {
-        const lockPath = deps.joinPath(deps.homedir(), '.workrail', lockFileName);
-        try {
-            const raw = await deps.readFile(lockPath);
-            const parsed = JSON.parse(raw);
-            if (typeof parsed.pid === 'number' && parsed.pid > 0) {
-                if (!deps.isPidAlive(parsed.pid)) {
-                    deps.stderr(`[Poll] ${lockFileName} points to dead PID ${parsed.pid} -- skipping stale lock, falling back to port ${DEFAULT_POLL_PORT}`);
-                    staleLockPath = lockPath;
-                    continue;
-                }
-            }
-            if (typeof parsed.port === 'number' && parsed.port > 0) {
-                return parsed.port;
-            }
-        }
-        catch {
-        }
-    }
-    if (staleLockPath !== undefined) {
-        try {
-            await deps.deleteFile(staleLockPath);
-        }
-        catch {
-        }
-    }
-    return DEFAULT_POLL_PORT;
-}
-async function executeWorktrainTriggerPollCommand(deps, opts) {
-    const triggerId = opts.triggerId.trim();
-    if (!triggerId) {
-        deps.stderr('[Poll] Error: triggerId must not be empty.');
-        return (0, cli_result_js_1.failure)('triggerId must not be empty.');
-    }
-    const port = await discoverConsolePort(deps, opts.port);
-    const url = `http://127.0.0.1:${port}/api/v2/triggers/${encodeURIComponent(triggerId)}/poll`;
-    deps.print(`[Poll] Forcing immediate poll cycle for trigger: ${triggerId}`);
-    let responseBody;
-    try {
-        const response = await deps.fetch(url, {
-            method: 'POST',
-            signal: AbortSignal.timeout(30000),
-        });
-        responseBody = await response.json();
-        if (!response.ok) {
-            const errMsg = typeof responseBody['error'] === 'string'
-                ? responseBody['error']
-                : `HTTP ${response.status}`;
-            deps.stderr(`[Poll] Error: ${errMsg}`);
-            return (0, cli_result_js_1.failure)(errMsg);
-        }
-    }
-    catch (e) {
-        const msg = e instanceof Error ? e.message : String(e);
-        const isConnRefused = msg.includes('ECONNREFUSED') || msg.includes('fetch failed');
-        const isTimeout = e instanceof Error && e.name === 'TimeoutError';
-        if (isConnRefused) {
-            deps.stderr(`[Poll] Error: Could not connect to WorkTrain daemon on port ${port}. ` +
-                `Ensure the daemon is running with: worktrain daemon`);
-            return (0, cli_result_js_1.failure)(`Could not connect to daemon on port ${port}`);
-        }
-        if (isTimeout) {
-            deps.stderr(`[Poll] Error: Request timed out after 30s. The poll cycle may still be running.`);
-            return (0, cli_result_js_1.failure)('Request timed out after 30s');
-        }
-        deps.stderr(`[Poll] Error: ${msg}`);
-        return (0, cli_result_js_1.failure)(msg);
-    }
-    const body = responseBody;
-    const data = body['data'];
-    if (data !== undefined) {
-        const cycleRan = data['cycleRan'];
-        const message = typeof data['message'] === 'string' ? data['message'] : '';
-        if (cycleRan === true) {
-            deps.print(`[Poll] ${message || 'Poll cycle started.'}`);
-        }
-        else {
-            deps.print(`[Poll] ${message || 'Poll cycle skipped (previous cycle still running).'}`);
-        }
-    }
-    deps.print('[Poll] Done.');
-    return (0, cli_result_js_1.success)();
-}

package/dist/trigger/daemon-console.d.ts

@@ -1,28 +0,0 @@
-import 'reflect-metadata';
-import type { V2ToolContext } from '../mcp/types.js';
-import type { TriggerRouter } from './trigger-router.js';
-import type { PollingScheduler } from './polling-scheduler.js';
-import type { WorkflowService } from '../application/services/workflow-service.js';
-import type { SteerRegistry } from '../daemon/workflow-runner.js';
-import type { Result } from '../runtime/result.js';
-export interface DaemonConsoleHandle {
-    readonly port: number;
-    stop(): Promise<void>;
-}
-export type DaemonConsoleError = {
-    readonly kind: 'port_conflict';
-    readonly port: number;
-} | {
-    readonly kind: 'io_error';
-    readonly message: string;
-};
-export interface StartDaemonConsoleOptions {
-    readonly port?: number;
-    readonly triggerRouter?: TriggerRouter;
-    readonly serverVersion?: string;
-    readonly workflowService?: WorkflowService;
-    readonly lockFilePath?: string;
-    readonly steerRegistry?: SteerRegistry;
-    readonly pollingScheduler?: PollingScheduler;
-}
-export declare function startDaemonConsole(ctx: V2ToolContext, options?: StartDaemonConsoleOptions): Promise<Result<DaemonConsoleHandle, DaemonConsoleError>>;