Definition
RAG, or retrieval-augmented generation, is a method for building an AI solution in which the system searches specified sources for the required passages before preparing an answer. These sources include instructions, procedures, contracts, a knowledge base, and other organizational documents. The retrieved context is passed to the language model together with the user’s question. RAG reduces answers that are disconnected from company materials and allows sources to be cited, but it does not guarantee correctness. Quality depends on current documents, content segmentation, retrieval, and access control. The result still requires human review. Explore Astrafox AI solutions.