Retrieval-augmented generation supplies a generative model with relevant external evidence at inference time so answers can reflect current or private knowledge. This guide explains the mechanism, trade-offs, evaluation, and controls that matter in practice.