# Docker Compose configuration for DeepAnalyze vLLM service
# Usage: docker-compose up -d

version: '3.8'

services:
  vllm:
    # Use pre-built image from Docker Hub or locally built image
    image: deepanalyze-env:latest
    # Or pull from Docker Hub:
    # image: yourusername/deepanalyze-env:latest
    
    container_name: deepanalyze-vllm
    
    # GPU support
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    
    # Port mapping
    ports:
      - "8000:8000"
    
    # Volume mounts
    volumes:
      - ./workspace:/workspace
      - ./models:/models
    
    # Environment variables (optional)
    environment:
      - CUDA_VISIBLE_DEVICES=0
      # - VLLM_USE_MODELSCOPE=False
    
    # Startup command
    # Uncomment and modify to start vLLM server automatically
    # command: python3 -m vllm.entrypoints.openai.api_server --model /models/your-model-name --host 0.0.0.0 --port 8000
    
    # Default: interactive bash shell
    command: /bin/bash
    stdin_open: true
    tty: true
    
    # Restart policy
    restart: unless-stopped