Nope - just use memory layer with model routing system.
replyhttps://github.com/rush86999/atom/blob/main/docs/EPISODIC_ME...
Memory is usually slow and haven't seen many voice agents atleast leverage it. Are you building in text modality or audio as well?
reply