mineru
Skill0 starsChanged 3 months agoMITView on GitHub
MinerU PDF→Markdown: install, CLI usage, backend selection, Windows no-GPU notes.
- Installs packages
6 other projects use a version of this.
Adopted by
Most recently taken up by
- LeoLin990405/grimoire-skilltheir own version, first seen Nov 8, 2026
- opendatalab/MinerUtheir own version, first seen Oct 8, 2026
- Nebutra/MinerU-Skilltheir own version, first seen Sep 24, 2026
- xjsongphy/skillstheir own version, first seen Aug 17, 2026
- walkingjohnny/ai-workbenchtheir own version, first seen May 5, 2026
- liangdabiao/fetch-everythingtheir own version, first seen Mar 30, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- opendatalab/MinerU★ 83ksame name
- Nebutra/MinerU-Skill★ 119same name
- liangdabiao/fetch-everything★ 47same name
- LeoLin990405/grimoire-skill★ 8same name
- xjsongphy/skills★ 7same name
- walkingjohnny/ai-workbench★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository219 other files it gives its agents.
- agent-doctrineSkill
- agent-evaluation-frameworkSkill
- agent-visualization-flintSkill
- hermes-agentSkill
- bailian-cliSkill
- edge-browser-automationSkill
- computer-useSkill
- jupyter-live-kernelSkill
- backup-strategiesSkill
- browser-automationSkill
- chinese-ai-free-apiSkill
- cicd-pipeline-templatesSkill
- copybaraSkill
- devops-toolkitSkill
- email-deliverabilitySkill
- monitoring-observabilitySkill
- obsidian-maintenanceSkill
- obsidian-troubleshootingSkill
- openclaw-windowsSkill
- openclaw-workerSkill
- webhook-subscriptionsSkill
- windows-dev-environmentSkill
- windows-masterSkill
- zero-downtime-deploySkill
- codebase-inspectionSkill
- github-workflowSkill
- adaptive-communicationSkill
- agent-cost-dashboardSkill
- agent-evolutionSkill
- agent-governanceSkill
- agent-osSkill
- context-efficient-toolsSkill
- context-loss-awarenessSkill
- creative-synthesisSkill
- hermes-desktop-pluginsSkill
- emotional-intelligenceSkill
- graph-engineering-workflowSkill
- hermes-maintenanceSkill
- hermes-voiceSkill
- hermes-hygieneSkill
- identity-coreSkill
- income-loopSkill
- master-synthesisSkill
- meta-cognitionSkill
- reasoning-engineSkill
- self-preservationSkill
- teaching-abilitySkill
- temporal-awarenessSkill
- hermes-themesSkill
- thinking-visualizationSkill
- user-modelingSkill
- voice-full-restoreSkill
- knowledge-graph-traversalSkill
- mcp-browser-canvasSkill
- free-api-gatewaySkill
- native-mcpSkill
- web-reverse-apiSkill
- ai-edaSkill
- huggingface-hubSkill
- industrial-ai-siemens-mcpSkill
- The first 60 shown.
The file
Files
MinerU — PDF → Markdown
Opendatalab 出品,magic-pdf 的后继者。v3.4.0 起是 namespace package,安装 mineru[core]。
约 2GB 依赖(torch + transformers + onnxruntime),无需 GPU 可用 pipeline 后端。
快速安装 (Windows, no GPU)
# 标准安装
uv pip install "mineru[core]" --python "D:/Hermes/hermes-agent/.venv/Scripts/python.exe"
# 国内网络用腾讯云镜像(2026-06 实测最稳)
uv pip install "mineru[core]" --python "D:/Hermes/hermes-agent/.venv/Scripts/python.exe" \
-i https://mirrors.cloud.tencent.com/pypi/simple/安装后验证:
uv pip list --python "<venv-python>" | grep mineru
# mineru 3.4.0, mineru-vl-utils 1.0.5
python -c "from mineru import version; print(version.__version__)"
# 3.4.0CLI 核心用法
# 基础转换(pipeline 后端,中文 OCR)
python -m mineru.cli.client -p input.pdf -o ./output -b pipeline -l ch
# 高精度(hybrid engine,含图表分析)
python -m mineru.cli.client -p input.pdf -o ./output -b hybrid-engine --effort high -l ch
# 纯文本提取(跳过 OCR,最快)
python -m mineru.cli.client -p input.pdf -o ./output -m txt -b pipeline -l ch
# 指定页码范围(0-indexed)
python -m mineru.cli.client -p input.pdf -o ./output -s 0 -e 5 -b pipeline -l ch
# 远程 API(无需本地 GPU)
python -m mineru.cli.client -p input.pdf -o ./output -b vlm-http-client -u http://server:30000 -l chCLI 参数速查
| Flag | 默认值 | 说明 |
-p / --path | (必填) | 输入文件或目录,支持 pdf/image/docx/pptx/xlsx |
-o / --output | (必填) | 输出目录 |
-b / --backend | hybrid-engine | pipeline / vlm-engine / hybrid-engine / vlm-http-client / hybrid-http-client |
-m / --method | auto | auto / txt / ocr(pipeline 和 hybrid 后端适用) |
-l / --lang | ch | ch, ch_server, korean, arabic, east_slavic, cyrillic, devanagari 等 |
--effort | medium | hybrid 后端可用 high 开启图表/图片分析 |
-s / --start | 0 | 起始页(0-indexed) |
-e / --end | 末页 | 结束页(0-indexed) |
后端选型(无 GPU Windows)
| 后端 | 需要 | 适用场景 |
pipeline | CPU only | 通用文档、中文 PDF首选 |
vlm-http-client | 远程 VLM 服务 | 无本地 GPU,高精度 |
hybrid-http-client | 远程服务 + 本地 CPU | 无本地 GPU,平衡精度和速度 |
vlm-engine | 本地 GPU | 最高精度、复杂版面 |
hybrid-engine | 本地 GPU + CPU | 默认,速度与精度平衡 |
无 GPU 推荐:先用 pipeline(稳定),需要更高精度时切 vlm-http-client 指向远程 GPU 服务。
包结构(非显而易见)
mineru是 namespace package:__init__.py仅 50 字节,dir(mineru)为空- 实际代码在子目录:
mineru.cli.client(CLI 入口)、mineru.backend、mineru.model、mineru.utils - 没有
magic_pdf包——v3.x 已将旧 magic-pdf 合并进 mineru - 正确的 import:
from mineru.cli.client import ...(或直接用 CLI)
常见坑
uv 锁冲突
同时跑多个 uv pip install 会抢 D:\sofeware\cache\uv\ 的锁。错误信息:
Timeout (300s) when waiting for lock on `D:\sofeware\cache\uv\wheels-v6\pypi\...`修法:kill 掉多余的进程,留一个就行。或设 UV_LOCK_TIMEOUT=600。
pip 不在 venv 里
Hermes 的 venv 精简过,没有 pip。用 uv pip install,不要用 python -m pip。
python -m mineru.cli.client --help 卡住
首次运行可能在做模型预热,用 --version 快速验证安装即可。
中文镜像不稳定
清华镜像 (tuna) 和阿里云镜像在 2026-06 实测超时。
腾讯云镜像 mirrors.cloud.tencent.com 可用。
PyPI 直连 files.pythonhosted.org 也可用(慢但不断)。
与其他工具对比
| pymupdf (~25MB) | marker-pdf (~3-5GB) | MinerU (~2GB) | |
| 文本 PDF | ✅ | ✅ | ✅ |
| 复杂版面 | ❌ | ✅ | ✅ |
| 表格 | 基础 | ✅ | ✅ |
| OCR 扫描件 | ❌ | ✅ | ✅ |
| 远程 API | ❌ | ❌ | ✅ |
| 多后端 | 无 | 单一 | 5 种 |
| 安装速度 | 秒级 | 分钟级 | 5-15 分钟 |
MinerU 是 marker-pdf 的强力替代品——更多后端选择,原生支持远程推理。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

