{"id":"2068317780064276917","url":"https://x.com/akshay_pachaar/status/2068317780064276917","text":"Web scraping will never be the same.\n\n(100% open-source visual search at scale)\n\nPixelRAG is a retrieval system that skips HTML parsing completely.\n\nInstead of scraping a page into text and embedding chunks, it screenshots the page and retrieves the image. A vision-language model reads the answer straight off the pixels.\n\nWhy that matters: parsing is where web RAG quietly loses information.\n\n- A single HTML-to-text parser can drop 40%+ of a page.\n- Tables, charts, and layout get flattened or thrown out.\n- Swapping parsers alone can move accuracy ~10 points on the same docs.\n\nPixelRAG indexes the page a person actually sees. The team built a visual index of all of Wikipedia, 30M+ screenshots, and it still beats the strongest text RAG baseline by 18.1% on text-only QA.\n\nThe repo also ships a Claude Code plugin that gives Claude eyes.\n\nIt lets Claude screenshot any URL and read the rendered page instead of scraping the DOM. So you can hand it a live page, an arXiv paper, or your local site and ask what it actually looks like.\n\nOne setup script. No MCP server, no backend.\n\nHow the pipeline works:\n\n- Renders each document (web, PDF, image) to image tiles.\n- Embeds them with Qwen3-VL-Embedding, LoRA fine-tuned on screenshots.\n- Builds a FAISS index and serves a search API.\n\nA stronger reader model lifts accuracy with no re-indexing, since the index is just pixels.\n\nEverything is open-source under Apache-2.0.\n\nGitHub repo: https://github.com/StarTrail-org/PixelRAG\n\nTalking about RAG, I recently wrote an article on a new approach that makes retrieval much more efficient by cutting corpus size by 40x, reducing tokens per query by 3x, and improving vector search relevance by 2.3x.\n\nThe article is quoted below.","author":{"name":"Akshay 🚀","username":"akshay_pachaar","avatarUrl":"https://pbs.twimg.com/profile_images/1578327351544360960/YFpWSWIX_200x200.jpg"},"createdAt":"Sat Jun 20 12:59:20 +0000 2026","engagement":{"replies":128,"retweets":828,"likes":6721,"views":907595},"media":{"photos":[],"videos":[{"url":"https://video.twimg.com/amplify_video/2068317725802606592/vid/avc1/740x720/ENgoUtzS_ATvB6li.mp4?tag=14","thumbnailUrl":"https://pbs.twimg.com/amplify_video_thumb/2068317725802606592/img/6ZKu5iVPQJ6EPnJy.jpg","width":1112,"height":1080,"duration":55.516}]},"quoteTweet":{"id":"2052743644411765230","url":"https://x.com/akshay_pachaar/status/2052743644411765230","text":"","author":{"name":"Akshay 🚀","username":"akshay_pachaar","avatarUrl":"https://pbs.twimg.com/profile_images/1578327351544360960/YFpWSWIX_200x200.jpg"},"createdAt":"Fri May 08 13:33:17 +0000 2026"},"externalLink":{"url":"https://github.com/StarTrail-org/PixelRAG","displayUrl":"github.com","title":"GitHub - StarTrail-org/PixelRAG: The end of web parsing. The beginning of scalable pixel-native search.","description":"The end of web parsing. The beginning of scalable pixel-native search. - StarTrail-org/PixelRAG","thumbnailUrl":"https://opengraph.githubassets.com/f16cbc3b1a97e2fabe22789b1136673e2e88234922fad20d492ac6a00fa8ff75/StarTrail-org/PixelRAG"}}