rag/docker-compose.yml

x-app: &app
  build:
    context: .
    dockerfile: .docker/python/Dockerfile
  environment:
    - PYTHONPATH=/app
    - SENTENCE_TRANSFORMERS_HOME=/app/.cache
  volumes:
    - ./:/app:rw,cached
  depends_on:
    - chroma
  networks:
    default:
    rag:
      aliases:
        - inference

services:
  inference:
    <<: *app
    # command: uvicorn app:app --app-dir src --reload --host 0.0.0.0 --port 8000
    command: uvicorn src.app:app --reload --reload-dir src --host 0.0.0.0 --port 8000
    ports:
      - "8000:8000"

  # ollama:
  #   build:
  #     context: .docker/ollama
  #     dockerfile: Dockerfile
  #   volumes:
  #     - ollama-data:/root/.ollama
  #   environment:
  #     OLLAMA_MODELS: ${OLLAMA_MODELS}
  #     OLLAMA_MODEL: ${OLLAMA_MODEL}
  #   healthcheck:
  #     test: ["CMD", "ollama", "ps"]
  #     interval: 15s
  #     retries: 5
  #     start_period: 5s
  #     timeout: 3s

  chroma:
    image: chromadb/chroma
    volumes:
      - chroma-data:/data
    ports:
      - "8008:8000"

  cli:
    <<: *app

volumes:
  ollama-data: {}
  chroma-data: {}

networks:
  default: {}
  rag:
    name: rag
    external: true