What's the best mechanism to store my embeddings documents isolated and retrieve a dynamic number of them to interact ?

[AhmedHefnawy]

HI All,

I'm working on a Document ChatBot Project, we want our user upload any number of documents to his/her database, then select a specific document/s to chat with, that mean we want to store each document in an isolated container, but we need to a dynamic retriever that allow user to choose a specific document/s to chat with.

not: user may ask a question that need an answer from many documents.

use-case:

" our user called Ahmed, he is one of the HR team, he want to upload company policies and candidates CVs, after that he want to select some of company policies and 3 CVs to interact with.
Ahmed will ask some multi-hop questions like: - based on the CVs, which one of them meet location policy condition of our country?
... etc "

can you recommend me the best way to embed, store, and retrieve those document to cover our user needs ?!
i can use another vector DB or any recommended frame work, feel free to recommend.

your support is appreciated :)