docker-compose.yaml

networks:
  shared_net:
  backend_llm_net: # network for backend and llm_server communication

services:
  nginx:
    image: nginx:stable-alpine
    ports:
      - "80:80"
    networks:
      - shared_net
    volumes:
      - "./nginx.conf:/etc/nginx/nginx.conf"
    depends_on:
      - backend
      - frontend
      - llm_server
  frontend:
    build:
      context: frontend
    ports:
      - "3000:3000"
    networks:
      - shared_net
    depends_on:
      - llm_server
      - backend
  backend:
    build:
      context: backend
    ports:
      - "5192:5192"
    networks:
      - shared_net
      - backend_llm_net  # Network for backend-vllm communication
    depends_on:
      - llm_server
  llm_server:
    image: vllm/vllm-openai:latest
    command: --model /mnt/model --dtype=half --max_model_len 8192
    ports:
      - "8000:8000"
    volumes:
      - "./models/Meta-Llama-3.1-8B-Instruct:/mnt/model/"
    networks:
      - backend_llm_net  # Only accessible by backend
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]