Queue Monitoring

Enrichment and extraction run asynchronously and can accumulate while models or DB load degrade.

What this page covers

How to monitor queue depth, failures, and stale work.

Core workflow

1. Check queue from CLI:

pgm queue

2. Check queue from REST:

curl http://127.0.0.1:3100/api/queue?include_failures=true \
  -H "Authorization: Bearer $PGM_API_KEY"

3. Include failure context:

• pending/completed/failed counts for extraction and embedding
• recent failure messages and paths when include_failures=true

4. For deeper operational view, query direct SQL:

pgm-admin sql "SELECT extraction_status, COUNT(*) FROM entities GROUP BY 1"

Notes

• failure_limit defaults to 20 with a maximum of 100 when requesting include_failures=true.
• If pending stays high, check worker startup and extraction provider availability before scaling the app.