Agentic RAG

An advanced retrieval pattern where the AI agent actively decides what to search for, evaluates retrieved results, and iterates if the initial retrieval is insufficient. Unlike basic RAG (single query → single retrieval), agentic RAG involves multi-step research: the agent decomposes complex questions, queries multiple sources, cross-references results, and synthesizes a comprehensive answer.