What is RAG (search before answering)?
RAG is retrieval-augmented generation: first the relevant passages are found, then the model answers from them.
Why it matters
The answer stays tied to what the clinic wrote, not to what the model learned somewhere else.