# LLM Proxy Production Deployment with GPU Support and Monitoring Stack
# Usage: docker-compose up -d to start all services

version: '3.8'

services:
  # ==========================================
  # LLM Proxy Service (Main API Gateway)
  # ==========================================
  proxy:
    build:
      context: .
      dockerfile: Dockerfile
    image: llm-proxy:${PROXY_TAG:-latest}
    
    # Expose proxy API port to host
    ports:
      - "9999:9999"
    
    # Environment variables from .env file (loaded automatically)
    env_file:
      - .env
    
    # Mount config files for hot-reload capability
    volumes:
      - ./config:/app/config:ro
      - ./logs:/app/logs:rw
    
    # Health check configuration
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:9999/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 10s
    
    # Restart policy for production resilience
    restart: unless-stopped
    
    # Resource limits (adjust based on hardware)
    deploy:
      resources:
        limits:
          memory: 2G
        reservations:
          memory: 512M
    
    # Logging configuration
    logging:
      driver: "json-file"
      options:
        max-size: "10m"
        max-file: "3"
    
    # Network configuration
    networks:
      - llm-proxy-network

  # ==========================================
  # LM Studio Service (Optional - Local LLM Inference)
  # ==========================================
  # Uncomment below and set DISCOVERY_ENABLED=true in .env to enable auto-discovery
  lm-studio:
    image: lmstudio/lmstudio:latest
    # or use specific version for stability: lmstudio/lmstudio:0.2.64
    
    # GPU Support - requires NVIDIA runtime on Docker host
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    
    ports:
      - "1234:1234"
    
    # Volume for model downloads and persistence
    volumes:
      - ./lmstudio-data:/data
    
    # Environment (optional LM Studio settings)
    environment:
      LM_STUDIO_MODEL: ""  # Set default model to load automatically (optional)
    
    restart: unless-stopped
    networks:
      - llm-proxy-network

  # ==========================================
  # Prometheus for Metrics Collection
  # ==========================================
  prometheus:
    image: prom/prometheus:v2.50.1
    container_name: llm-proxy-prometheus
    volumes:
      - ./monitoring/prometheus.yml:/etc/prometheus/prometheus.yml:ro
      - prometheus_data:/prometheus
    command:
      - '--config.file=/etc/prometheus/prometheus.yml'
      - '--storage.tsdb.path=/prometheus'
      - '--web.enable-lifecycle'
      - '--web.console.libraries=/usr/share/prometheus/console_libraries'
      - '--web.console.templates=/usr/share/prometheus/consoles'
    ports:
      - "9090:9090"  # Prometheus UI
    depends_on:
      - proxy
    networks:
      - llm-proxy-network
    restart: unless-stopped

  # ==========================================
  # Grafana for Visualization & Dashboards
  # ==========================================
  grafana:
    image: grafana/grafana:10.2.3
    container_name: llm-proxy-grafana
    environment:
      - GF_SECURITY_ADMIN_USER=admin
      - GF_SECURITY_ADMIN_PASSWORD=admin123
      - GF_SERVER_ROOT_URL=http://localhost:3000
      - GF_SECURITY_ADMIN_EMAIL=admin@llmproxy.local
    volumes:
      - grafana_data:/var/lib/grafana
      - ./monitoring/provisioning:/etc/grafana/provisioning:ro
      - ./dashboards:/var/lib/grafana/dashboards:ro
    ports:
      - "3000:3000"  # Grafana UI (note: if LM Studio is enabled, use 127.0.0.1:3000)
    depends_on:
      - prometheus
    networks:
      - llm-proxy-network
    restart: unless-stopped

# Named volumes for persistent data
volumes:
  prometheus_data:
    driver: local
  grafana_data:
    driver: local

# Network configuration for service discovery
networks:
  llm-proxy-network:
    driver: bridge