Deploy a production MLflow tracking server backed by PostgreSQL for run metadata and S3 for artifacts. Includes docker-compose setup, basic auth proxy and Python client configuration.
A production-ready FastAPI application using the app factory pattern, async SQLAlchemy 2.0 with PostgreSQL, Alembic migrations, structured logging, and health-check endpoint.
Tuned postgresql.conf settings for a dedicated 8-core / 32 GB RAM database server. Covers shared buffers, WAL, autovacuum, parallel query and connection settings — with explanations for each knob.
A docker-compose.yml for local development that spins up the app, PostgreSQL 16, Redis 7 and an optional Adminer instance — volumes ensure data survives container restarts.