Hacker News Reader

Nope - just use memory layer with model routing system.

Memory is usually slow and haven't seen many voice agents atleast leverage it. Are you building in text modality or audio as well?