What problem does it solve?
Extracting accurate text from PDFs and scanned documents is slow and error-prone, especially when documents are images, have complex layouts, or require OCR.
Core Features & Use Cases
- URL-first extraction: When a document URL is available, it can convert PDFs to Markdown using a web extraction workflow (e.g., via Firecrawl) to avoid local setup.
- Lightweight local text extraction (pymupdf): Instantly extract text, tables, metadata, selected pages, and embedded images for text-based PDFs.
- High-quality OCR extraction (marker-pdf): Perform OCR with better layout handling for scanned PDFs, equations/LaTeX, forms, tables, and image-to-text use cases, including optional LLM-boosted accuracy.
- Research/document pipelines: Fetch and extract Arxiv content and support split/merge/search workflows across pages using pymupdf.
Quick Start
Use the ocr-and-documents skill to extract Markdown text from a local scanned PDF by running python scripts/extract_marker.py scanned.pdf.