RAG (retrieval-augmented generation)

RAG (retrieval-augmented generation) is a technique that combines a language model with a search over an external data source. Before answering, the system finds the relevant snippets in documents, a knowledge base or the web, and hands them to the model as context so it writes the answer from them. This achieves two important things: the model can use up-to-date or private information that was not in its training, and hallucinations ✕ A hallucination is when an AI presents made-up or incorrect information as true, in a confident tone. It is why you should verify its answers. More in the glossary → are reduced, because it answers grounded in real texts it can also cite. It is the basis of many assistants that answer about a company’s documentation or about a specific set of files. It does not remove errors entirely, but it makes answers far more reliable and verifiable.