Methodology (RAG) for overcoming LLM knowledge cut-offs and integrating proprietary corporate knowledge. Pipeline: embed the user's question into a vector, run a similarity search against a vector database of proprietary documents, retrieve relevant chunks, construct a prompt merging instructions + retrieved context + user question, and have the LLM generate an answer. Common use case: QA systems over corporate/HR documents (e.g. 'how long is paternity leave at my company'). Naive RAG uses a single index; advanced strategies use multiple indices (e.g. a summary index pointing to a detail index).