Retrieval-Augmented Generation
An architectural approach that optimizes the output of a Large Language Model by querying verified, external knowledge sources before generating responses.
RAG combines dense vector search databases with generative models. Instead of relying solely on static training parameters, a RAG system runs a query against a vector index, retrieves relevant documents, and passes them to the LLM to supply factual, context-grounded context. This minimizes LLM hallucinations on business metrics.