Best Practices: Chatterbox Agent

Use this skill to keep Chatterbox voice agents controllable, interruptible, and evidence-grounded. Chatterbox is the speech renderer; the agent stack still owns turn state, memory recall, long-running skill execution, cancellation, receipts, and what is worth saying out loud.

Core Pattern

Design the system as four separate loops:

user audio / text turn
  -> turn manager
  -> skill orchestrator running asyncio task batches
  -> structured result queue
  -> voice coordinator
  -> cancellable Chatterbox TTS queue