VectifyAI/PageIndex
Retrieves from long documents by reasoning over a tree index instead of vector search.
PageIndex builds a hierarchical tree index of a document's structure instead of chunking it into a vector database, then has an LLM reason through that tree to retrieve the relevant section, the way a person would flip to the right page of a report. It targets long professional documents — financial reports, legal filings, technical manuals — where similarity search returns text that's similar but not actually relevant. It's open source (pip install pageindex), works with your own OpenAI key or PageIndex Cloud, and reports 98.7% accuracy on FinanceBench at about $0.001 per page indexed.
What holds up
- +Open source and installable via pip — no vector database or chunking pipeline to set up
- +Retrieval is traceable to explicit document references, not an opaque similarity score
- +98.7% accuracy on FinanceBench at about $0.001 per page, both published and checkable
Mind the limits
- −The retrieval step needs a strong, expensive LLM — cheap models are only recommended for indexing
- −Built for long, structured documents — likely overkill for short text or simple lookups
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/vectifyai-pageindex/"><img src="https://stillworks.dev/badge/vectifyai-pageindex.svg" alt="Picked by StillWorks" width="230" height="54"></a>