Vector search isn’t the only answer to RAG anymore. VectifyAI’s PageIndex just crossed 29K GitHub stars (+953 today) with a different pitch: skip embeddings, give the LLM a tree index of your document, let it reason its way to the right page like a human would.
How it works
Feed in a PDF. PageIndex builds a hierarchical table of contents — sections, subsections, page numbers. At query time the LLM walks the tree, picking which branch to open. No chunking, no top-k similarity, no vector DB to maintain. Mafin 2.5, Vectify’s RAG model running on PageIndex, hits 98.7% on FinanceBench — the open-book QA benchmark for financial filings.
API and SDK access
PageIndex ships as an open-source Python library and a hosted cloud API. An MCP server plugs into Claude Agent SDK, OpenAI Agents SDK, Vercel AI SDK, LangChain. Sweet spot: long structured documents where citations matter more than millisecond latency. 10-K filings, legal contracts, technical manuals.
Why it’s trending
Vector databases became a $1B+ category on the bet that semantic similarity is the right retrieval primitive. PageIndex says: for long structured docs, navigation beats similarity. Built by Mingtian Zhang (UCL) and Yu Tang, launched September 2025. 29K stars in a few months is the market voting on whether RAG is solved.
You Might Also Like
- Pageindex Just hit Github Trending and it Might Make you Rethink rag Entirely
- 700 Github Stars in a Week Apfel Exposes the Free llm Apple Locked Behind Siri
- Hume ai Open Sources Tada an llm Based tts With Zero Hallucinations and 0 09 rtf
- Google Releases Gemini Embedding 2 one Vector Space for Text Images Video and Audio
- Astrbot Crosses 22k Github Stars as Developers Flock to its 18 Platform ai Chatbot Framework
Top AI Product 
Leave a comment