ocr-and-documents

Use pymupdf to open the PDF and call page.get_text() for each page, or pymupdf4llm for Markdown output. It installs in seconds at ~25MB and handles text-based PDFs, tables, images, and metadata without model downloads.

pymupdf is lightweight (~25MB) and instant, handling text-based PDFs, basic tables, and embedded images. marker-pdf (~3-5GB with PyTorch) adds OCR for scanned documents in 90+ languages, equations, forms, and reading-order detection.

Scanned PDFs require OCR, which pymupdf cannot do. Use marker-pdf, which performs OCR across 90+ languages and outputs Markdown; note it downloads ~2.5GB of models to the HuggingFace cache on first use.

Yes, try web_extract on the URL first, which converts PDFs to Markdown via Firecrawl with no local dependencies. Fall back to local pymupdf or marker-pdf extraction only if web_extract fails or batch processing is needed.

pymupdf handles split and merge natively with no extra dependencies. Use insert_pdf to copy page ranges into a new document for splitting, or to combine multiple source PDFs into one merged file.

marker-pdf requires ~5GB of free disk space for PyTorch and its models, plus ~2.5GB of model downloads on first run. Check disk space first; if space is limited, use web_extract for URLs or pymupdf for text-based PDFs.