pymupdf (@pymupdf) header
pymupdf (@pymupdf) profile picture

PDF → LLM-ready markdown. Built for RAG pipelines that actually work. Open source.

San Francisco
Joined January 2026
pymupdf retweeted
my top 5 python libraries this month: - requests - sqlite3 - tqdm - concurrent.futures - PyMuPDF
5
2
31
2,707
Needed to share bank statements but only the lines for one specific vendor. Asked Claude Code to black out everything else across 5 PDFs. It pulled in PyMuPDF, found the right rows, drew redaction boxes, and saved clean copies. Marker-on-paper, automated.
1
1
29
So how did I make it work? I extracted the PDF text with PyMuPDF, split it into chunks, turned them into embeddings with Sentence Transformers, and stored them in ChromaDB. PDF → chunks → embeddings → ChromaDB That was the basic setup
1
2
5
68
Great article explaining routing shortcuts with PDF Extraction: devrev.ai/blog/pdf-extractio…
1
27
proposal for ending the document processing benchmark wars: 1. assume every benchmark is biased (alas) 2. test vendors on your own data 3. choose the one that works 4. everyone stops arguing and we go back to building thank you for coming to my ted talk
12
4
2
42
3,434
pymupdf retweeted
With PyMuPDF4LLM 1.28.2, we move to a fully AGPL v3 licensing with the PyMuPDF Layout package dependency and all the source goes public. The fastest CPU-only layout engine for PDF is now genuinely open source. Read more: pymupdf.io/blog/open-source-…
1
2
116
pymupdf retweeted
Sometimes an open source contribution starts as a problem in your own workflow. The author of this piece wanted PyMuPDF4LLM as a LangChain document loader, found it didn't exist, and built it. This is a great writeup of what followed, from first PR to standalone package: medium.com/@lakinduboteju/bu…
1
3
121
pymupdf retweeted
Replying to @bbddpp @GrablyR
J'utilise pymupdf4llm qui est adaptée de la librairie PyMuPDF. Pour des pdf avec du texte et des tableaux, ça marche très bien. Et on peut même demander au llm de fournir un script basé sur cette librairie pour faire le job de conversion
3
4
7
154
pymupdf retweeted
Tables can be one of the hardest problems in document intelligence. The latest PyMuPDF4LLM release majorly upgrades table extraction, and rebuilds table structure detection from the ground up: rows, columns, and full cell grids. Read more: pymupdf.io/blog/enhanced-tab…
1
3
93
pymupdf retweeted
PyMuPDF 1.28 is here! It brings four major upgrades, all built for the AI era: 📊 Better table extraction 🔍 Smarter adaptive OCR 📄 Native Markdown → PDF 📝 Native Office → Markdown You can now generate PDFs straight from .md with full control over fonts, colors, tables & layout via CSS. Learn more: pymupdf.io/blog/markdown-in-…
2
4
80
pymupdf retweeted
PyMuPDF makes it into the top 3 PDF parsers out there dude. This thing is magnificent, It also has PyMuPDF4LLM as an extension for converting PDFs into structured data for a RAG pipeline. Also has integrations with LangChain and LlamaIndex btw. Goated stuff.
2
3
10
311
pymupdf retweeted
Before you send a PDF page to OCR or an LLM, ask whether it needs to go at all. PyMuPDF reads char counts, image coverage, tables, and form widgets in milliseconds per page — enough to route each page and skip the spend. Code + walkthrough: artifex.com/blog/using-pymup…
2
1
3
117
pymupdf retweeted
If you're building anything that extracts data from PDFs with an LLM, you eventually hit the question: where on the page did this actually come from? PyMuPDF answers it with exact coordinates pulled from the PDF's own geometry. More about why this matters in document AI: artifex.com/blog/grounding-i…
3
3
85
Building this project made me realize that AI apps are really multiple tools working together behind the scenes. FastAPI, FAISS, Groq, SentenceTransformers, PyMuPDF each one solves a different problem, and together they power my RAG system. 🚀 #RAG #BuildInPublic #AI
🤖 Made with AI
2
2
6
149