RAG

RAG reduces LLM hallucinations by dynamically injecting fresh, verifiable context into the model prompt. It typically relies on vector databases to compute semantic search similarities between user queries and private document embeddings.