PixelRAG skips HTML parsing, uses screenshots for web retrieval, and beats text RAG by 18.1%
Roger_M_Taylor · x · 2026-07-25
PixelRAG is an open-source retrieval system that skips HTML parsing entirely.
Instead of converting web pages into text, it screenshots the page and retrieves over the rendered image. A vision-language model then reads the answer directly from pixels, which the authors argue preserves layout, tables, and charts that text parsers often destroy.
Key claims from the post:
- HTML-to-text parsing can drop 40%+ of a page.
- Changing parsers alone can shift accuracy by about 10 points on the same documents.
- PixelRAG indexes pages as users actually see them.
- The team reportedly built a visual index of all Wikipedia and 30M+ screenshots.
- It beats the strongest text-RAG baseline by 18.1% on text-only QA.
The repo also includes a Claude Code plugin that lets Claude screenshot any URL and read the rendered page.
More from coding & agent
- Spring AI 2.1.0-RC1 adds tool-call confirmation, pausing agent loops before real-data access — therealdanvega · 2026-10-26
- Morgan Stanley and US banks block third-party AI agents over credentials — thione · 2026-10-26
- METR red-teams OpenAI's agent monitoring, finds bypasses and gaps — thione · 2026-10-26
- OpenAI: coding agents weakened safety monitoring, one lied about it — thione · 2026-10-26
- ConvFill: A 0.6B Talker Buys Time While a Large Reasoner Whispers the Real Answer — andimarafioti · 2026-10-26
- 250 Builders Hack With Codex at Berlin Community Hackathon Ahead of OpenAI DevDay — paw_lean · 2026-10-26