PixelRAG
https://arxiv.org/abs/2606.28344. The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/
What is PixelRAG?
https://arxiv.org/abs/2606.28344. The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/
Overview of PixelRAG
What it is
PixelRAG renders documents — web pages, PDFs, images — as screenshots and retrieves over the images directly. Visual structure that HTML parsing throws away — tables, charts, layout, infographics — stays intact, so the reader model can actually answer questions about it. Wikipedia's 8.28M articles ship as a pre-built index; the pipeline itself is general-purpose.
Key features
- Renders documents to screenshot tiles and retrieves over images directly (pixel-native search).
- Pre-built index of 8.28M Wikipedia pages available via a hosted API — no setup, no API key required.
- Includes a Claude Code plugin (pixelbrowse skill) for AI agents to visually browse pages.
- Supports multiple vector search backends: FAISS (default) and Qdrant.
- Modular pipeline: capture (pixelshot), chunk, embed, build-index, serve.
- Works on Linux (CUDA) and macOS (Apple Silicon / MPS).
- Open source under Apache-2.0 license.
Use cases
Search any document by how it looks, not just the text it contains. PixelRAG is suited for answering questions about tables, charts, diagrams, and infographics that text-based parsing loses. It also enables AI agents (Claude, opencode) to visually browse web pages and extract information from rendered screenshots.
Getting started
pip install pixelrag
Render a page to screenshot tiles:
pixelshot https://en.wikipedia.org/wiki/Python --output ./tiles
Search the hosted Wikipedia index (no setup):
curl -X POST https://api.pixelrag.ai/search \
-H "Content-Type: application/json" \
-d '{"queries": [{"text": "What is the capital of France?"}], "n_docs": 5}'
For building your own index, see the repository README.
License and pricing
Licensed under Apache-2.0. The project is open source with no stated pricing.
Based on the project's public README and repository metadata.