services:
  whisper:
    image: fedirz/faster-whisper-server:latest-cpu
    container_name: andytranscribe-whisper
    ports:
      # Host 8090 avoids clashing with `php artisan serve` on 8000
      - "${WHISPER_HOST_PORT:-8090}:8000"
    volumes:
      - whisper-huggingface-cache:/root/.cache/huggingface
    environment:
      # OpenAI-compatible API; model is also passed per request from Laravel
      WHISPER__MODEL: ${LOCAL_WHISPER_MODEL:-Systran/faster-whisper-base}
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health')"]
      interval: 15s
      timeout: 5s
      retries: 10
      start_period: 30s

  # GPU variant (NVIDIA). Start with:
  #   docker compose --profile gpu up -d whisper-gpu
  whisper-gpu:
    profiles: ["gpu"]
    image: fedirz/faster-whisper-server:latest-cuda
    container_name: andytranscribe-whisper-gpu
    ports:
      - "${WHISPER_HOST_PORT:-8090}:8000"
    volumes:
      - whisper-huggingface-cache:/root/.cache/huggingface
    environment:
      WHISPER__MODEL: ${LOCAL_WHISPER_MODEL:-Systran/faster-whisper-base}
    restart: unless-stopped
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

volumes:
  whisper-huggingface-cache: