Chandra OCR
OCR model converting scanned pages and PDFs into layout-aware Markdown
What it is
Chandra is an OCR model with a Python package and CLI that turns images and PDFs into Markdown, HTML, or JSON while keeping page layout. It handles tables, math, forms with checkboxes, handwriting, and more than 90 languages, and can extract embedded images with captions. It runs locally through Hugging Face or against a vLLM server, with an optional Streamlit demo app. Developers building document pipelines use it to get structured text out of messy scans before indexing or analysis.