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