pdf-parser

What This Skill Does

The PDF Parser skill integrates the powerful MinerU API directly into your OpenClaw agent workflow, enabling high-fidelity conversion of complex PDF documents into structured Markdown. Unlike standard text extractors, this skill is specifically engineered to handle the nuances of technical and academic documents. It features advanced OCR for image-based PDFs, precise LaTeX formula extraction, and automatic structural conversion of tables into Markdown format. By leveraging the MinerU engine, the agent can transform static, unreadable PDF layouts into machine-actionable knowledge assets that can be further processed by LLMs or stored in your local knowledge base.

Installation

To begin, ensure you have the necessary dependencies installed by running pip install requests in your environment. You must obtain a valid API token from https://mineru.net/. Once you have your token, you must set it as an environment variable named MINERU_TOKEN. On Linux/macOS, use export MINERU_TOKEN=your-token-here, and on Windows PowerShell, use $env:MINERU_TOKEN = "your-token-here". Finally, install the skill using the OpenClaw command: clawhub install openclaw/skills/skills/alex-zxyz/mineru-pdf-parser.

Use Cases

• Academic Research: Rapidly convert long research papers and journal articles into Markdown to summarize, query, or extract data points using AI.
• Technical Documentation: Parse manuals and blueprints that include complex tables and mathematical notations for integration into documentation systems.
• Knowledge Management: Convert local archives of reports and financial statements into a standardized format for building private RAG (Retrieval-Augmented Generation) knowledge bases.
• Document Archiving: Streamline the process of digitizing legacy scanned files that are otherwise locked in static image formats.

Example Prompts

1. "解析 PDF ~/Documents/research_paper.pdf，并将内容转换成 Markdown 格式。"
2. "将 PDF 转为 Markdown，请使用 pipeline 引擎进行精细化解析。"
3. "解析当前的 PDF 文件，并提取其中的公式与表格数据到知识库。"

Tips & Limitations

• Engine Selection: While the 'vlm' engine is default and highly versatile, switch to 'pipeline' if you require stricter adherence to document flow for complex layouts.
• Privacy: Be aware that this skill sends your document content to the MinerU external API for processing. Avoid processing highly sensitive or classified documents.
• File Management: Note that processed outputs are automatically stored in ~/.openclaw/MinerU_Results/. Ensure you have sufficient disk space when processing large batches of high-resolution PDFs.
• Latency: The parsing process involves network requests and heavy computation on the server side; always use the --wait flag if you want the agent to confirm completion before proceeding to downstream tasks.