llm_vector_server/docker-compose.yml

services:
  # vLLM - Qwen3 0.6B
  vllm-0.6b:
    image: swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/vllm/vllm-openai:latest
    container_name: vllm-qwen3-0.6b
    ports:
      - "8100:8000"
    command: >
      sh -c "
      pip install --upgrade vllm &&
      pip install --upgrade transformers &&
      python -m vllm.entrypoints.openai.api_server --model /app/models/Qwen3-Embedding-0.6B --dtype float16 --port 8000
      "
    volumes:
      # 挂载预下载的模型目录
      - ./models/Qwen3-Embedding-0.6B:/app/models/Qwen3-Embedding-0.6B
      # 可选：挂载缓存目录
      - ./cache:/root/.cache/huggingface
    environment:
      # 设置 Hugging Face 国内镜像
      - HF_ENDPOINT=https://hf-mirror.com
      # 可选：设置 Python 包索引镜像
      - PIP_INDEX_URL=https://pypi.tuna.tsinghua.edu.cn/simple
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [ gpu ]
    # 添加健康检查
    healthcheck:
      test: [ "CMD", "curl", "-f", "http://localhost:8000/health" ]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s
#  # vLLM - Qwen3 4B
#  vllm-4b:
#    image: vllm/vllm-openai:latest
#    container_name: vllm-qwen3-4b
#    ports:
#      - "8200:8000"
#    command: >
#      --model Qwen/Qwen3-Embedding-4B
#      --dtype float16
#      --api-port 8000
#    volumes:
#      - ./models:/root/.cache/huggingface
#
#  # vLLM - Qwen3 8B
#  vllm-8b:
#    image: vllm/vllm-openai:latest
#    container_name: vllm-qwen3-8b
#    ports:
#      - "8300:8000"
#    command: >
#      --model Qwen/Qwen3-Embedding-8B
#      --dtype float16
#      --api-port 8000
#    volumes:
#      - ./models:/root/.cache/huggingface

  # Milvus 向量数据库
#  milvus:
#    image: swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/milvusdb/milvus:v2.4.5
##    image: milvusdb/milvus:v2.4.4-standalone
#    container_name: milvus
#    ports:
#      - "19530:19530"
#      - "9091:9091"
#    environment:
#      - ETCD_USE_EMBED=true
#      - MINIO_USE_EMBED=true
#      - PULSAR_USE_EMBED=true
#    volumes:
#      - ./milvus_data:/var/lib/milvus
#    command: ["milvus", "run", "standalone"]