Implement Retrieval-Augmented Generation end-to-end — chunking, embedding, vector storage, retrieval, and prompt composition. Use when building a chatbot or agent that needs to answer from a corpus the LLM wasn't trained on.