Flask+Gunicorn+Apache service on backend2 handling /fellow requests with Firestore-backed chat logic.
Details
- POST /fellow accepts {participant_id, start_date, block, chat, build_block, missionNumber?}
- chat.py implemented with real Firestore queries and hierarchical filtering (~125-200ms warm)
- reconfigure.py forwards completed trials to adapt and builds the next block from its response — no longer a stub
- Apache only proxies /health and /fellow
- Service name is gf-dataplane-api (systemd unit + /var/log/gf-dataplane-api/); worker_tmp_dir now survives reboot via RuntimeDirectory in the unit
Health check
https://backend2.bgbehavior.com/health