RAG Methodology

Hey RAG experts, I have a question. Please provide your thoughts on the below.
Lets say, we have multiple documents which are almost similar to each other with only a few contextual difference but in almost similar topics , and the structure is almost the same with common headings like summary, background etc., I would like to pick only the document with highest relevance related to a particular topic. (Sometimes, humans can struggle to pick the correct document) . How do you design the RAG? How do you ensure that the context is set correctly ? Also, if we have multiple documents that can answer the similar questions, and I want all the documents to be referred, what is the best approach ?