Dockerize FastAPI: Async-First Containers for Modern Python APIs

FastAPI's async capabilities depend on the correct Uvicorn configuration — wrong worker types, missing uvloop, or running with --reload in production can tank performance. Docker ensures the exact same runtime, the exact same Uvicorn config, and the exact same Python version runs everywhere.

Containerization also makes scaling trivial — run multiple replicas of the same FastAPI image behind a load balancer, each with its own Uvicorn worker pool.

Two-stage build optimized for async Python:

• Build stage: FROM python:3.12-slim AS builder — creates virtualenv, installs dependencies with pip install --no-cache-dir -r requirements.txt including uvicorn[standard] for uvloop and httptools.
• Runtime stage: FROM python:3.12-slim — copies virtualenv, sets PYTHONUNBUFFERED=1 and PYTHONDONTWRITEBYTECODE=1. CMD: uvicorn app.main:app --host 0.0.0.0 --port 8000 --workers 4 --loop uvloop --http httptools.

For CPU-bound workloads, we configure Gunicorn with Uvicorn workers: gunicorn app.main:app -k uvicorn.workers.UvicornWorker --workers 4 --bind 0.0.0.0:8000. Health check uses FastAPI's built-in /health endpoint with proper async response.

• Multi-stage Dockerfile with uvloop and httptools — 120-160 MB
• docker-compose.yml with health checks and resource limits
• Uvicorn production config with correct worker count formula
• Development Compose override with --reload and volume mounts
• Optional Gunicorn wrapper for CPU-bound apps
• CI pipeline with dependency caching