RAG Retrieval-Augmented Generation
The technique where an AI first searches a knowledge base, then formulates the answer based on what it found.
The fix for two problems with raw LLMs: they can't know things that happened after their training, and they tend to hallucinate. With RAG, you search your own data (site content, docs, knowledge base) and let the model answer only based on that.
Technically: split data into chunks → create embeddings → on query, embed the question, find nearest chunks, send them as context to the LLM.