Previously /api/prescreen/batch blocked for 4-10 minutes waiting for Replicate/
DeepSeek, causing browser connection timeout and zero results saved.
- Phase 1 (HTTP check) runs synchronously and saves results immediately
- Phase 2 (DeepSeek classify) fires as asyncio.create_task and runs in background
- Response is returned to client as soon as phase 1 completes (~30-90s)
- Frontend toast shows "classifying N in background" so user knows niche/type
will appear shortly without waiting
- Each DeepSeek sub-batch saves independently so partial results are preserved
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Malin
daccb99a0c