services:

  ollama:
    image: ollama/ollama
    ports:
      - "11434:11434"
    volumes:
      - ollama_data:/root/.ollama          # model files persist across restarts
      - ./ollama-entrypoint.sh:/entrypoint.sh
    entrypoint: ["/bin/sh", "/entrypoint.sh"]
    environment:
      - MODEL=${MODEL:-qwen3:4b}
    healthcheck:
      test: ["CMD-SHELL", "ollama list | grep -q ${MODEL:-qwen3:4b}"]
      interval: 15s
      timeout: 5s
      retries: 120                         # first-time model pull (~5 GB) can take 20+ minutes
      start_period: 30s

  backend:
    build: ./backend
    ports:
      - "8000:8000"
    depends_on:
      ollama:
        condition: service_healthy         # waits until model is ready, not just server up
    env_file:
      - .env
    volumes:
      - ./backend:/app                     # live reload during dev

  frontend:
    build: ./frontend
    ports:
      - "8501:8501"
    depends_on:
      - backend
    env_file:
      - .env
    volumes:
      - ./frontend:/app                    # live reload during dev

volumes:
  ollama_data:                             # named volume so model survives docker compose down