What is RAG (Retrieval-Augmented Generation)?

RAG answers a user question by first searching relevant information, retrieving documents, sending that context to the LLM and then generating the answer.

Knowledge sources can be HR policies, product manuals, technical docs or internal knowledge. RAG grounds answers in retrieved information instead of relying only on model memory.

1. 1User question
2. 2Search relevant information
3. 3Retrieve documents
4. 4Send context to the LLM
5. 5Generate the answer