services:
  nirmaan:
    build: .
    volumes:
      - .:/workspace
    environment:
      - GROQ_API_KEY=${GROQ_API_KEY:-}
      - VLLM_BASE_URL=http://vllm:8000/v1
    stdin_open: true
    tty: true
    command: ["--provider", "groq"]

  vllm:
    image: vllm/vllm-openai:latest
    profiles: ["vllm"]
    ports:
      - "8000:8000"
    command:
      - --model
      - Qwen/Qwen2.5-Coder-7B-Instruct
      - --enable-auto-tool-choice
      - --tool-call-parser
      - hermes
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]