METANOWBring knowledge to the question
Retrieval-augmented generation retrieves relevant information and supplies it as context to a generative model. It can help an application answer using approved documents that were not part of the model’s original training.
The basic pipeline
Prepare and index source material, retrieve relevant passages for a question, then generate an answer from the selected context. Retrieval may use keyword, vector or combined search. Updating the external index is different from retraining the model.
Evidence still needs checking
RAG can retrieve the wrong passage or miss an important exception. A model can also misinterpret a correct source. Preserve document identity, version and citations, and test whether claims are supported. Retrieval does not eliminate hallucinations.
Apply the same access rules
Filter retrieval to documents the requesting user is allowed to read. Keep confidential content out of shared caches and logs. If the approved sources do not answer the question, return an uncertainty message or route it to a person.
Sources & further reading
Reviewed 19 October 2026. This is a technology learning session; examples describe possible designs, not a claim that a particular integration is active.