How to build RAG cum Recommender system?

Below is the description of the issue.
I have large dataset where I see a particular user question can be solved in more than one ways. For example: If someone asks “how to login into Facebook?” their could be more than one possible way to login to Facebook like using app or from the web portal etc etc.

I want may to generate one way as an response with it’s source information which is present in Documents metadata. And recommend question for other approaches. For example:
Human: “How to login into facebook?”
AI: "You can login to Facebook using mobile app following below step:



You might also be interested in:
recommended question: “How to login to facebook from Desktop?”


Source:
Section: “…”
Title: “…”
Source_uri: “…”

I will appreciate any suggestion and guidance.