Deploying AITracer for Production Workloads
A practical deployment checklist for self-hosted AITracer with Agent Lab and optional API ingest.
Self-hosted AITracer needs Postgres, auth, and (for Agent Lab) network access to Ollama.
Recommended launch checks
DATABASE_URLset and migrations applied (prisma migrate deploy)- Auth0 or
DEV_AUTH_BYPASSfor local dev only OLLAMA_BASE_URLreachable from the app container if using Agent Lab- Smoke test: sign in → Run agent → confirm a trace appears under Traces
- Optional: API key created →
POST /api/tracesingest from staging
Optional production ingest
If you also send cloud model traffic:
- provider credentials in Settings (for coach or direct execution)
- ingest rate limits and retention policies reviewed
- governance / verification features enabled only if your compliance program requires them
These checks keep the core loop (run → trace → train) predictable before you add advanced controls.