Retrieval-augmented generation
Retrieving relevant documents or records first and giving them to the model, so its answer rests on current material.
Retrieval-augmented generation is a design in which an application first retrieves relevant documents or records, such as a product's terms or a customer's recent transactions, and passes them to the model together with the question. The answer can then rest on current, citable material rather than only on what the model learned in training. Retrieved content is also a path for indirect prompt injection, so it should be handled as data, never as instructions.
Agent Minute explains this term on 25 January 2027.
Related terms
Large language modelA model trained on very large amounts of text to predict likely text, used by most agents to interpret instructions and choose steps.Indirect prompt injectionPrompt injection hidden in content an AI system retrieves, such as a web page, document or email.ConfabulationConfidently stated but false or erroneous output from a generative AI system, often called hallucination.System promptThe standing instructions an operator gives a model before any user input, describing its role, rules and tools.