92 websites, 129,931 chunks, each site its own AI assistant — crawled, embedded, and published by a Rust daemon running on one machine. Here's how WWW-RAG works and where it's going.
WWW-RAGWeb CrawlingRAGRustEmbeddingsTursolibSQLLocal-First AI
An e-commerce assistant linked products correctly but recommended the wrong ones. The fix was discovering that decoration, ranking, and recommendation are three different problems — and that prominence is only ever a tiebreaker on relevance.
RAGProduct RecommendationsE-commerceBM25ProminenceStorefrontLLM GenerationBuilding in Public
We swapped OpenParse for LiteParse in our RAG ingestion pipeline. The headline is 20–50× faster parsing. The useful part is the four ways we got it wrong first.
Open-source Headroom compresses RAG context with an ONNX model. We wired it into our pipeline and raced it against a 50-line in-process extractor. Neither won outright.
Divinci AI se une al Workers Launchpad Cohorte #6 de Cloudflare. Edge-RAG sub-100ms, el pitch del Demo Day y un deep-dive de nuestro stack de producción.
Divinci AI rejoint la cohorte #6 du Workers Launchpad de Cloudflare. Edge-RAG sous 100 ms, le pitch du Demo Day, et un deep-dive sur notre stack de prod.