SynaBot

What is RAG (retrieval-augmented generation)?

RAG is a technique where the model looks up relevant documents before answering, so its answers are grounded in your data instead of guesses.

Mark Barclay
Answer curated and reviewed byMark Barclay

Key takeaways

  • RAG is a technique where the model looks up relevant documents before answering, so its answers are grounded in your data instead of guesses.

Large language models only "know" what was in their training data. RAG (retrieval-augmented generation) lets an assistant search a knowledge base at question-time and stuff the top-matching chunks into the prompt.

The upshot: the assistant can cite your policies, product docs, or past emails without being retrained. It is the standard way to make an AI assistant expert in your business.