MinerU v3.4.0 容器化部署：模型下载与多服务配置

# daocloud 镜像加速（国内环境）
docker pull docker.m.daocloud.io/vllm/vllm-openai:v0.21.0

FROM docker.m.daocloud.io/vllm/vllm-openai:v0.21.0

# 安装中文字体与 OpenGL 依赖（OCR 需要）
RUN apt-get update && \
    apt-get install -y --no-install-recommends \
        fonts-noto-core \
        fonts-noto-cjk \
        fontconfig \
        libgl1 && \
    fc-cache -fv && \
    apt-get clean && \
    rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*

# 安装 MinerU（latest），用阿里云 PyPI 镜像加速
RUN python3 -m pip install -U 'mineru[core]>=3.4.0' \
        -i https://mirrors.aliyun.com/pypi/simple \
        --break-system-packages && \
    python3 -m pip cache purge

# 默认使用本地模型（不自动下载）
ENTRYPOINT ["/bin/bash", "-c", "export MINERU_MODEL_SOURCE=local && exec \"$@\"", "--"]

# 进入容器
docker exec -it mineru-api bash

# 下载 pipeline 模型（OCR + 版面分析）
mineru-models-download -s modelscope -m pipeline

# 下载 VLM 模型
mineru-models-download -s modelscope -m vlm

services:
  mineru-api:
    image: mineru:latest
    container_name: mineru-api
    ports:
      - "8000:8000"
    volumes:
      - mineru-cache:/root/.cache
      - mineru-config:/root/mineru.json:ro
    environment:
      - MINERU_MODEL_SOURCE=local
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    restart: unless-stopped

  mineru-openai-server:
    image: mineru:latest
    container_name: mineru-openai-server
    ports:
      - "30000:30000"
    volumes:
      - mineru-cache:/root/.cache
      - mineru-config:/root/mineru.json:ro
    environment:
      - MINERU_MODEL_SOURCE=local
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    restart: unless-stopped

  mineru-gradio:
    image: mineru:latest
    container_name: mineru-gradio
    ports:
      - "7860:7860"
    volumes:
      - mineru-cache:/root/.cache
      - mineru-config:/root/mineru.json:ro
    environment:
      - MINERU_MODEL_SOURCE=local
    restart: unless-stopped

volumes:
  mineru-cache:
  mineru-config:

curl -X POST http://localhost:8000/file_parse \
  -F "files=@document.jpg" \
  -F "backend=pipeline"

curl -X POST http://localhost:8000/file_parse \
  -F "files=@document.jpg" \
  -F "backend=hybrid-auto-engine"

docker stop mineru-openai-server
# 运行解析
curl -X POST http://localhost:8000/file_parse -F "backend=hybrid-auto-engine" ...
# 解析完恢复
docker start mineru-openai-server

docker volume inspect mineru-config
# 确认 Mountpoint 存在且包含 mineru.json

docker exec mineru-api bash -c \
  'jq ".vlm = \"/root/.cache/modelscope/hub/models/.../MinerU2___5-Pro-2605-1___2B\"" \
   /root/mineru.json > /tmp/mineru.json && mv /tmp/mineru.json /root/mineru.json'

# 清除 lock
rm -rf ~/.cache/modelscope/hub/.lock/
# 清除临时目录
rm -rf ~/.cache/modelscope/hub/._____temp/
# 重新下载
mineru-models-download -s modelscope -m pipeline

ln -s MinerU2___5-Pro-2605-1___2B MinerU2.5-Pro-2605-1.2B

# mineru-api
curl http://localhost:8000/health
# {"status":"healthy","version":"3.4.0",...}

# mineru-openai-server
curl http://localhost:30000/health
# {"status":"healthy",...}

# mineru-gradio
curl -o /dev/null -w "%{http_code}" http://localhost:7860/
# 200

curl -X POST http://localhost:8000/file_parse \
  -F "files=@test.jpg" \
  -F "backend=pipeline" \
  -F "return_md=true" \
  -F "return_content_list=true"

curl http://localhost:30000/v1/models