Skip to content
Glossary

RAG (retrieval-augmented generation), explained

RAG is the technique of retrieving relevant passages from a knowledge base and giving them to the language model before it answers, so the response relies on the company's own information.

Without RAG, the model answers from what it learned in training, which may be outdated or not know your products. With RAG, it receives passages from the company's manuals, policies and FAQs at question time.

The process has three steps: split documents into passages, find the passages most similar to the question and build the answer from them.

RAG reduces hallucinations but does not eliminate them. Quality depends on clear, up-to-date documents without contradictions.

Conversations aren't support.They're infrastructure.

Step into the new era of the relationship between companies and customers.

Create agent now