docask

Ask questions about any PDF using Docling + Claude.

Most AI tools only read PDF text. docask preserves the full layout — tables, sections, columns — so Claude actually understands your document.

Install

pip install docask

Usage

export ANTHROPIC_API_KEY=your_key_here

docask report.pdf "summarize this document"
docask invoice.pdf "what is the total amount?"
docask research.pdf "what are the key findings?"

How it works

PDF → Docling (layout-aware extraction) → structured markdown → Claude API → answer

• Docling converts PDF preserving tables, sections, reading order
• Claude reasons over the structured content
• Fully local extraction, only the structured text hits the API

Requirements

• Python 3.9+
• ANTHROPIC_API_KEY environment variable

License

MIT