We need to separate the agent's memory (stateful) from the LLM (stateless). Has anyone tried bi-temporal architectures for agent state?
· Sign up or log in to comment
