Retrieval-augmented generation (RAG)
RAG retrieves relevant passages from your corpus and feeds them into a model so answers cite current sources instead of memory alone.
How it works
Documents are chunked and embedded into a vector index. At query time, similar chunks are retrieved and passed to the model with instructions to use them. Quality depends on chunking, metadata filters and eval - not only model size.
RAG is not a substitute for a clear job. It helps when documents are the evidence path for answers or agent steps.
Example
A policy assistant retrieves the latest leave policy sections before answering an employee question, with links back to source paragraphs.