A technique where the system first searches your own data for information relevant to a question, then the model answers based on it.
RAG stands for retrieval-augmented generation. The idea is simple: before the model answers, the system searches your documents or database for the passages most relevant to the question and places them in the request. The model then writes its answer based on that material instead of its general memory.
RAG matters for any app that must answer questions about your own information: product catalogs, policies, user guides. It is usually simpler than fine-tuning, lets you update knowledge just by editing documents, and noticeably reduces hallucinations. Practical tip: RAG is only as good as your data. Organize documents, remove outdated versions that contradict each other, and split content into clear sections. Ask the assistant to cite the source it used so you can check answers easily.