How do you choose the right open-source LLM for a RAG (Retrieval-Augmented Generation) system?

[Suhebmultani]

If you’re building an AI application that searches documents and answers questions (RAG), how do you decide which open-source AI model to use?