This repo is a Streamlit application that uses Recursive Abstractive Processing for Tree Organized Retrieval or RAPTOR to fetch data from a RAG. The rag in this case is a document that the users uploads. You can discuss with the document using the model chatGPT 3.5 turbo.
I will be looking into local model support for RAPTOR at some point, but this is a POC to check to see what kind of applications can be spiked quickly.
This work is based on a tweet from Franck SN https://twitter.com/ndzfs/status/1765367593299288453