released · 2025
LayoutScribe
A Python package that converts PDF, PowerPoint, and Word files into Markdown and layout-aware JSON using multimodal models.
- Python
- multimodal
- documents
- PyPI
Document parsing is usually treated as text extraction. Many useful documents communicate through layout: callouts, diagrams, reading order, tables, and relationships between visual blocks.
LayoutScribe uses multimodal models to produce clean Markdown alongside layout-aware JSON. It includes validation, retries, and fallbacks for blank or difficult pages.
The package is designed as a reusable preprocessing layer for retrieval, extraction, and document understanding pipelines where plain OCR loses too much structure.