Build the RAG pipeline
RAG (Retrieval-Augmented Generation): the AI fetches context before answering.