Why does an AI agent need RAG? Long-term memory makes a difference in performance

The more context information relevant to the task an AI agent obtains, the better its performance. However, the agent's memory has limitations. This is because the large-scale language model (LLM) on which the agent is based is a stateless structure.