What RAG is

Retrieve relevant text, then prompt with it.

Retrieval-Augmented Generation finds documents relevant to the question and puts them in the prompt, so the model answers from real, current sources.