llm_vector_server/docker-compose.yml

version: "3.8"
services:
  # vLLM - Qwen3 0.6B
  vllm-0.6b:
    image: swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/vllm/vllm-openai:v0.9.2
    container_name: vllm-qwen3-0.6b
    ports:
      - "8100:8000"
    command: >
      --model Qwen/Qwen3-Embedding-0.6B
      --dtype float16
      --api-port 8000
    volumes:
      - ./models:/root/.cache/huggingface

#  # 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:v2.4.4-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
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [ gpu ]