services:
  code-llm:
    build: .
    container_name: code-llm
    restart: unless-stopped
    ports:
      - "8000:8000"
    environment:
      # Variables Ollama / FastAPI
      OLLAMA_HOST: "http://host.docker.internal:11434"
      GENERATE_MODEL: "deepseek-coder:6.7b-instruct"
      CHAT_MODEL: "gpt-oss:20b"
      CHROMA_PERSIST_DIR: "/chroma_db"
      RESPONSE_CACHE_DIR: "/response_cache"
      PYTHONUNBUFFERED: 1
      OLLAMA_NUM_PARALLEL: 6
    volumes:
      - "/Users/coissac/.ollama/models:/app/ollama:/models"       # modèles persistants
      - "/Users/coissac/Sync/maison/Petite_maisons/src/pmomusic:/code"       # code Go à indexer
      - "./app.py:/app/app.py"
#      - ./chroma_db:/chroma_db
#      - ./response_cache:/response_cache
    stdin_open: true
    tty: true