What is RAG (Retrieval-Augmented Generation)

Question: What is RAG (Retrieval-Augmented Generation)? Key points: RAG combines an LLM with an external retrieval step - relevant documents/passages are fetched from a knowledge base (typically via vector similarity search over embeddings) and injected into the prompt as context before generation. This grounds the model's output in retrieved facts, reducing hallucination and allowing the model to answer questions about knowledge outside its training data or that changes over time, without retraining the model itself.