docker-compose.yml

# docker-compose.yml (racine oasis/models/crime_predictor/)
version: '3.8'

services:
  # MLflow Tracking Server + Backend Store
  mlflow:
    image: python:3.11-slim
    ports:
      - "5000:5000"
    command: >
      bash -c "pip install mlflow psycopg2-binary &&
               mlflow server
               --backend-store-uri postgresql://mlflow:mlflow@postgres:5432/mlflow_db
               --default-artifact-root ./mlruns
               --host 0.0.0.0 --port 5000"
    volumes:
      - mlflow_data:/mlruns
      - ./models:/models
    depends_on:
      - postgres
    networks:
      - ml_network
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:5000/health"]
      interval: 30s
      timeout: 10s
      retries: 3

  # Postgres Backend pour MLflow (persistance expérimentations)
  postgres:
    image: postgres:15-alpine
    environment:
      POSTGRES_DB: mlflow_db
      POSTGRES_USER: mlflow
      POSTGRES_PASSWORD: mlflow
    volumes:
      - postgres_data:/var/lib/postgresql/data
    ports:
      - "5432:5432"
    networks:
      - ml_network
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U mlflow -d mlflow_db"]
      interval: 10s
      timeout: 5s
      retries: 5

  # API FastAPI Crime Predictor
  api:
    build: .
    ports:
      - "8000:8000"
    environment:
      - MLFLOW_TRACKING_URI=http://mlflow:5000
      - MODEL_PATH=/app/models/crime_predictor.pkl
    volumes:
      - ./models:/app/models
    depends_on:
      mlflow:
        condition: service_healthy
      postgres:
        condition: service_healthy
    networks:
      - ml_network
    command: >
      bash -c "
      until curl -f http://mlflow:5000/health; do
        echo '⏳ Attente MLflow...';
        sleep 5;
      done;
      uvicorn predict:app --host 0.0.0.0 --port 8000
      "

  # Training Worker (optionnel - CI/CD)
  trainer:
    build: .
    environment:
      - MLFLOW_TRACKING_URI=http://mlflow:5000
    volumes:
      - ./models:/app/models
    depends_on:
      - mlflow
      - postgres
    networks:
      - ml_network
    command: >
      bash -c "
      until curl -f http://mlflow:5000/health; do
        echo '⏳ Attente MLflow...';
        sleep 5;
      done;
      python train.py --data-url 'https://static.data.gouv.fr/.../delinquance.csv'
      "

volumes:
  mlflow_data:
  postgres_data:

networks:
  ml_network:
    driver: bridge