Retrieval-augmented generation (RAG) is a way to give LLMs relevant context at inference time. RAG stack:
