PyMuPDF

Built by Artifex on top of its MuPDF C engine, this Python package gives scripts fine-grained control over PDFs and related formats such as XPS, EPUB, MOBI, CBZ, SVG and images. Text comes out as plain strings or as nested blocks carrying font, size, colour and position for every span; find_tables returns tables as Markdown or pandas frames; pages rasterize at any DPI; and scans can be OCRed through a separately installed Tesseract. It writes too: annotations, permanent redactions, form filling, page reordering, merging, encryption, and PDFs generated from Markdown. The companion PyMuPDF4LLM package emits reading-order Markdown or JSON chunks for RAG ingestion, and a paid Pro add-on opens Office and HWP files. Prebuilt wheels exist for Windows, Linux and macOS, covering Python 3.10 through 3.14, nothing else is required, and no network calls happen at runtime, so air-gapped use is fine. The open build is AGPL-3.0, whose copyleft also reaches software offered over a network, so closed-source products generally need Artifex's commercial license. Data engineers get far more speed here than pure-Python parsers offer.