Fix stuck processing state: auto-fail stale jobs, improve active job detection

- Frontend: only treat parsing_documents/distilling as actively running;
  pending jobs older than 2 minutes are ignored as stale
- Backend: add fail_stale_jobs() that marks pending/active jobs older than
  5 minutes as failed before checking for active jobs in trigger_processing
- Prevents UI from getting stuck on old jobs that never completed

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/app/api/knowledge_base_routes.py

@@ -238,6 +238,9 @@ async def trigger_processing(
     if not kb:
         raise HTTPException(status_code=404, detail="Knowledge base not found")
 
+    # Auto-fail stale pending jobs (older than 5 minutes) before checking
+    await repo.fail_stale_jobs(kb_id)
+
     # Check for active jobs
     has_active = await repo.has_active_job(kb_id)
     if has_active:

backend/app/repositories/knowledge_base_repository.py

@@ -1,5 +1,5 @@
 import uuid
-from datetime import datetime, timezone
+from datetime import datetime, timedelta, timezone
 from typing import Optional
 
 from sqlalchemy import select, func
@@ -286,6 +286,23 @@ class KnowledgeBaseRepository:
         result = await self.session.execute(query)
         return result.scalar_one_or_none()
 
+    async def fail_stale_jobs(self, kb_id: uuid.UUID, stale_minutes: int = 5) -> None:
+        """Mark stale pending/active jobs as failed (older than stale_minutes)."""
+        cutoff = datetime.now(timezone.utc) - timedelta(minutes=stale_minutes)
+        active_statuses = ["pending", "parsing_documents", "distilling"]
+        query = (
+            select(ProcessingJob)
+            .where(ProcessingJob.knowledge_base_id == kb_id)
+            .where(ProcessingJob.status.in_(active_statuses))
+            .where(ProcessingJob.created_at < cutoff)
+        )
+        result = await self.session.execute(query)
+        for job in result.scalars().all():
+            job.status = "failed"
+            job.error_message = "Job timed out (stale)"
+            job.completed_at = datetime.now(timezone.utc)
+        await self.session.flush()
+
     async def has_active_job(self, kb_id: uuid.UUID) -> bool:
         """Check if there's an active (non-terminal) processing job for this KB."""
         active_statuses = ["pending", "parsing_documents", "distilling"]

frontend/components/KnowledgeBase.tsx

@@ -93,9 +93,17 @@ export const KnowledgeBase: React.FC = () => {
             setSelectedKb(detail);
             setVersions(vers);
 
-            // Check for active job
-            if (detail.latest_job && ['pending', 'parsing_documents', 'distilling'].includes(detail.latest_job.status)) {
+            // Check for active job - only treat as running if actually in progress
+            // A "pending" job older than 2 minutes is considered stale/stuck
+            if (detail.latest_job && ['parsing_documents', 'distilling'].includes(detail.latest_job.status)) {
                 setActiveJob(detail.latest_job);
+            } else if (detail.latest_job && detail.latest_job.status === 'pending' && detail.latest_job.started_at) {
+                const ageMs = Date.now() - new Date(detail.latest_job.started_at).getTime();
+                if (ageMs < 2 * 60 * 1000) {
+                    setActiveJob(detail.latest_job);
+                } else {
+                    setActiveJob(null);
+                }
             } else {
                 setActiveJob(null);
             }