RAG flow
- User asks a question
- The system uses vector search to find relevant documents
- Documents are added to the prompt as context
- The LLM generates an answer based on the context
Benefits
- Up-to-date data without retraining the model
- Source control — answers are grounded in specific documents
- Reduced hallucinations for factual questions
When needed
Q&A over corporate documentation, search in a large knowledge base, a chatbot answering product questions.