PageIndex Hits 31,709 Stars — Vectorless Reasoning-Based RAG Achieves 98.7% Accuracy on FinanceBench

GitHub·high signal

VectifyAI's PageIndex eliminates vector databases entirely from RAG by building a hierarchical Table of Contents tree from documents and using LLM reasoning to navigate to the most relevant section. The approach, inspired by AlphaGo's tree search, achieved 98.7% accuracy on the FinanceBench benchmark via the Mafin 2.5 financial analysis system — significantly outperforming traditional vector-based RAG. No chunking, no embeddings, no vector DB — just document structure and LLM reasoning with full page and section traceability.