compose.yaml

# env variables needed
# NGC_API_KEY

services:
  nv-embedqa-e5-v5:
    image: "nvcr.io/nim/nvidia/nv-embedqa-e5-v5:1.0.1"
    profiles: ["Local LLM + Embedding", "Local LLM + Embedding + Reranking"]
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              capabilities: ["gpu"]
              count: 1
    ipc: host
    environment:
      - NGC_API_KEY=${NGC_API_KEY}
    volumes:
      - nim-cache:/opt/nim/.cache
    healthcheck:
      test: ["CMD", "python3", "-c", "import requests; resp = requests.get('http://localhost:8000/v1/health/ready'); resp.raise_for_status()"]
      interval: 30s
      start_period: 600s
      timeout: 20s
      retries: 3
    networks:
      - default

  nv-rerankqa-mistral-4b-v3:
    image: "nvcr.io/nim/nvidia/nv-rerankqa-mistral-4b-v3:1.0.2"
    profiles: ["Local LLM + Embedding + Reranking"]
    runtime: "nvidia"
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              capabilities: ["gpu"]
              count: 1
    ipc: host
    environment:
      - NGC_API_KEY=${NGC_API_KEY}
    volumes:
      - nim-cache:/opt/nim/.cache
    healthcheck:
      test: ["CMD", "python3", "-c", "import requests; resp = requests.get('http://localhost:8000/v1/health/ready'); resp.raise_for_status()"]
      interval: 30s
      start_period: 600s
      timeout: 20s
      retries: 3
    networks:
      - default

  llm-nim:
    image: "nvcr.io/nim/meta/llama3-8b-instruct:1"
    profiles: ["Local LLM", "Local LLM + Embedding", "Local LLM + Embedding + Reranking"]
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              capabilities: ["gpu"]
              count: 1
    ipc: host
    environment:
      - NGC_API_KEY=${NGC_API_KEY}
    volumes:
      - nim-cache:/opt/nim/.cache
    healthcheck:
      test: ["CMD", "python", "-c", "import requests; resp = requests.get('http://localhost:8000/v1/health/ready'); resp.raise_for_status()"]
      interval: 30s
      start_period: 600s
      timeout: 20s
      retries: 3
    networks:
      - default

  milvus:
    image: "milvusdb/milvus:v2.4.6"
    security_opt:
      - seccomp:unconfined
    environment:
      - ETCD_USE_EMBED=true
      - ETCD_DATA_DIR=/var/lib/milvus/etcd
      - COMMON_STORAGETYPE=local
    volumes:
      - milvus:/var/lib/milvus
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:9091/healthz"]
      interval: 30s
      start_period: 90s
      timeout: 20s
      retries: 3
    command: "milvus run standalone"
    networks:
      - default

  redis:
    image: "redis:7"
    volumes:
      - redis:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 30s
      start_period: 30s
      timeout: 5s
      retries: 5
    command: "redis-server --save 20 1 --loglevel warning"
    networks:
      - default

networks:
  default:

volumes:
  milvus:
  redis:
  nim-cache: