I made a chatbot agent for my website, with a RAG sytem built on pinecone with data from my company.
My agent has a tool, which is directly connted to pinecone to access the data (left part of the screenshot).
Everything seems to work fine using this. But, I saw people using the “Answer questions with a vector store” module in between (right part of the screenshot).
I wonder if this is really necessary and if it improves the performances, given the fact that the solution without this module is working fine.
Does anyone have any opinion on this and maybe want to share their tests ?