[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"readme:pageindex":3},"\u003Cdiv align=\"center\">\u003Ca href=\"https:\u002F\u002Fvectify.ai\u002Fpageindex\" rel=\"nofollow ugc noopener\">\n  \u003Cimg src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F46201e72-675b-43bc-bfbd-081cc6b65a1d\" alt=\"PageIndex Banner\" \u002F>\n\u003C\u002Fa>\u003Cbr \u002F>\n\u003Cbr \u002F>\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F14736\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Ftrendshift.io\u002Fapi\u002Fbadge\u002Frepositories\u002F14736\" alt=\"VectifyAI%2FPageIndex | Trendshift\" width=\"250\" height=\"55\" \u002F>\u003C\u002Fa>\n\u003C\u002Fp>\u003Ch1>PageIndex: Vectorless, Reasoning-based RAG\u003C\u002Fh1>\n\u003Cp align=\"center\">\u003Cb>Reasoning-based RAG  ◦  No Vector DB, No Chunking  ◦  Context-Aware Retrieval  ◦  Reads Like a Human\u003C\u002Fb>\u003C\u002Fp>\u003Ch4>\n  \u003Ca href=\"https:\u002F\u002Fvectify.ai\" rel=\"nofollow ugc noopener\">🌐 Website\u003C\u002Fa>  •  \n  \u003Ca href=\"https:\u002F\u002Fchat.pageindex.ai\" rel=\"nofollow ugc noopener\">🖥️ Chat Platform\u003C\u002Fa>  •  \n  \u003Ca href=\"https:\u002F\u002Fpageindex.ai\u002Fdeveloper\" rel=\"nofollow ugc noopener\">🔌 MCP & API\u003C\u002Fa>  •  \n  \u003Ca href=\"https:\u002F\u002Fdocs.pageindex.ai\" rel=\"nofollow ugc noopener\">📖 Docs\u003C\u002Fa>  •  \n  \u003Ca href=\"https:\u002F\u002Fdiscord.com\u002Finvite\u002FVuXuf29EUj\" rel=\"nofollow ugc noopener\">💬 Discord\u003C\u002Fa>  •  \n  \u003Ca href=\"https:\u002F\u002Fii2abc2jejf.typeform.com\u002Fto\u002FtK3AXl8T\" rel=\"nofollow ugc noopener\">✉️ Contact\u003C\u002Fa> \n\u003C\u002Fh4>\u003C\u002Fdiv>\u003Cdetails open>\n\u003Csummary>\u003Ch2>📢 Updates\u003C\u002Fh2>\u003C\u002Fsummary>\u003Cul>\n\u003Cli>🔥 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FVectifyAI\u002FPageIndex\u002Fblob\u002Fmain\u002Fexamples\u002Fagentic_vectorless_rag_demo.py\" rel=\"nofollow ugc noopener\">\u003Cstrong>Agentic Vectorless RAG\u003C\u002Fstrong>\u003C\u002Fa> — A simple agentic, vectorless RAG \u003Ca href=\"#-agentic-vectorless-rag-an-example\" rel=\"nofollow ugc noopener\">example\u003C\u002Fa> with \u003Cem>self-hosted PageIndex\u003C\u002Fem>, using OpenAI Agents SDK.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fpageindex.ai\u002Fblog\u002Fpageindex-filesystem\" rel=\"nofollow ugc noopener\">\u003Cstrong>Scale PageIndex to Millions of Documents\u003C\u002Fstrong>\u003C\u002Fa> — \u003Cem>PageIndex File System\u003C\u002Fem> is a file-level tree indexing layer that lets PageIndex reason over an entire corpus, not just a single document, enabling massive-scale document search.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fchat.pageindex.ai\" rel=\"nofollow ugc noopener\">PageIndex Chat\u003C\u002Fa> — Human-like document analysis agent \u003Ca href=\"https:\u002F\u002Fchat.pageindex.ai\" rel=\"nofollow ugc noopener\">platform\u003C\u002Fa> for professional long documents. Also available via \u003Ca href=\"https:\u002F\u002Fpageindex.ai\u002Fdeveloper\" rel=\"nofollow ugc noopener\">MCP\u003C\u002Fa> or \u003Ca href=\"https:\u002F\u002Fpageindex.ai\u002Fdeveloper\" rel=\"nofollow ugc noopener\">API\u003C\u002Fa>.\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fpageindex.ai\u002Fblog\u002Fpageindex-intro\" rel=\"nofollow ugc noopener\">PageIndex Framework\u003C\u002Fa> — Deep dive into PageIndex: an \u003Cem>agentic, in-context tree index\u003C\u002Fem> that enables LLMs to perform \u003Cem>reasoning-based, context-aware retrieval\u003C\u002Fem> over long documents.\u003C\u002Fli>\n\u003C\u002Ful>\n \u003C\u002Fdetails>\u003Chr \u002F>\n\u003Ch1>📑 Introduction to PageIndex\u003C\u002Fh1>\n\u003Cp>Are you frustrated with vector database retrieval accuracy for long professional documents? Traditional vector-based RAG relies on semantic \u003Cem>similarity\u003C\u002Fem> rather than true \u003Cem>relevance\u003C\u002Fem>. But \u003Cstrong>similarity ≠ relevance\u003C\u002Fstrong> — what we truly need in retrieval is \u003Cstrong>relevance\u003C\u002Fstrong>, and that requires \u003Cstrong>reasoning\u003C\u002Fstrong>. When working with professional documents that demand \u003Cem>contextual understanding\u003C\u002Fem>, domain expertise, and multi-step reasoning, similarity search often falls short — missing what's relevant but not similar, and returning what's similar yet not relevant.\u003C\u002Fp>\n\u003Cp>Inspired by AlphaGo, we propose \u003Cstrong>\u003Ca href=\"https:\u002F\u002Fvectify.ai\u002Fpageindex\" rel=\"nofollow ugc noopener\">PageIndex\u003C\u002Fa>\u003C\u002Fstrong> — a \u003Cstrong>vectorless\u003C\u002Fstrong>, \u003Cstrong>reasoning-based RAG\u003C\u002Fstrong> system that builds a \u003Cstrong>hierarchical tree index\u003C\u002Fstrong> from long documents, and uses LLMs to \u003Cstrong>reason\u003C\u002Fstrong> \u003Cem>over that index\u003C\u002Fem> for \u003Cstrong>agentic, context-aware retrieval\u003C\u002Fstrong>. The retrieval is \u003Cem>traceable\u003C\u002Fem> and \u003Cem>explainable\u003C\u002Fem>, with no vector DBs or chunking.\nPageIndex simulates how \u003Cem>human experts\u003C\u002Fem> navigate and extract knowledge from complex documents through \u003Cem>tree search\u003C\u002Fem>, enabling LLMs to \u003Cem>think\u003C\u002Fem> and \u003Cem>reason\u003C\u002Fem> their way to the most relevant document sections. It performs retrieval in two steps:\u003C\u002Fp>\n\u003Col>\n\u003Cli>Generate a “Table-of-Contents” \u003Cstrong>tree structure index\u003C\u002Fstrong> of documents\u003C\u002Fli>\n\u003Cli>Perform (agentic) reasoning-based retrieval through\u003C\u002Fli>\n\u003C\u002Fol>\n",1785106159475]