PageIndex is a vectorless RAG system that builds a hierarchical tree index from lengthy PDF or Markdown documents and uses LLM reasoning for retrieval. It requires an external LLM API to operate and the open-source package handles simple PDFs using standard parsing.
Project overview
The system replaces traditional vector similarity search with LLM reasoning over a semantic tree index, providing traceable retrieval grounded in explicit page and section references.
Project type
AI Agent · RAG · Vector Database
Use cases
Knowledge Q&A
Deployment
Refer to project documentation
License
MIT
Best for
Developers, AI engineers, and enterprise teams needing reasoning-based, context-aware retrieval over lengthy professional documents.
Key capabilities
Transforms lengthy PDF or Markdown documents into a semantic tree structure similar to a table of contents, optimized for use with LLMs and AI agents.
Performs reasoning-based, context-aware retrieval over long documents using LLMs without requiring a vector database or chunking.
Provides a simple, end-to-end agentic vectorless RAG example using self-hosted PageIndex with the OpenAI Agents SDK.
Limitations and risks
The open-source repo is designed for generating the tree structure for simple PDFs using standard PDF parsing.
Complex PDFs are hard to parse by classic Python tools and may lose document context and hierarchy.
Getting started
Installation involves running pip3 install --upgrade -r requirements.txt, creating a .env file with an LLM API key, and executing python3 run_pageindex.py with a specified PDF path.
Alternatives and comparisons
A streamlined RAG workflow with quality document chunking, reduced hallucinations, and configurable LLMs adaptable to enterprises of any scale.
Provides a memory layer for AI agents by ingesting unstructured documents and outputting persistent, navigable memory with full semantic context.