Retrieval-augmented generation (RAG)
Grounding a model's answer in documents retrieved at query time.
AI
Definition
RAG retrieves relevant passages from a knowledge source and supplies them to a language model as context before generation.
It separates what the model knows from what the organisation knows, and makes answers citable.
Why it matters
It is the standard pattern for enterprise assistants because it keeps answers current and auditable without retraining.
How to apply it
- Invest in chunking and metadata before model choice.
- Evaluate retrieval quality separately from generation quality.
- Log the retrieved context with every answer for review.