MinerU PDF-to-Markdown Document Parser

Transforms complex PDFs into LLM-ready markdown and JSON using MinerU, a high-accuracy document intelligence pipeline. Extracts text, tables, formulas, and images from scientific papers, reports, and scanned documents with layout-aware parsing.

Installation

Use the upstream install or setup path that matches your environment:

• pip install --upgrade pip
• pip install uv
• uv pip install -U "mineru[all]"
• git clone https://github.com/opendatalab/MinerU.git

Requirements and caveats from upstream:

• 
• | Development | Python / Go / TypeScript SDK · CLI · REST API · Docker |
• The official online version has the same functionality as the client, with a beautiful interface and rich features, requires login to use

Basic usage or getting-started notes:

• While maintaining high accuracy, it keeps resource usage extremely low and continues to support inference in pure CPU environments.

• Optimized the parsing pipeline with a sliding-window mechanism, significantly reducing peak memory usage in long-document scenarios, so documents with tens of thousands of pages no longer need to be split manually.

• This update is not just a set of feature enhancements, but a key leap forward in MinerU's overall system capabilities. We specifically addressed the peak memory usage issue in long-document parsing. Through optimizati...

• Source: https://github.com/opendatalab/MinerU

• Extracted from upstream docs: https://raw.githubusercontent.com/opendatalab/MinerU/HEAD/README.md

Source

• Agent Skill Exchange