Comment by mjbonanno

This is awesome! Exactly the kind of low-latency agent tooling I've been looking for. How are you handling long-term memory/context between calls?