ppipada

I am trying to understand and build RAG in few of the LLM Agents I am working on.

I have noted a small collection of things that are currently being done in the field at: https://pankajpipada.com/posts/2024-12-17-rag-resources/

From all the reading and experimenting it looks like Graph + vectors + optional pagerank for certain usecases gives best results.

Question:

For enterprise unstructured data (docs/logs/scripts/etc), what is the current state of art technology wrt establishing a knowledge base to be used in either search or context population for some code generation usecases?

Thanks in advance