The error

cannot schedule new speech, the speech scheduling is draining/pausing

means that the system is currently in a draining or pausing state and is not accepting new speech tasks. This typically happens during: • Agent lifecycle transitions, such as shutdown or switching between agents • Graceful shutdown, where existing speech tasks are allowed to finish but new ones are blocked • Agent session management, when the system is cleaning up resources During this state, new speech can only be scheduled if the

force

parameter is set to

True

, which is usually reserved for critical tool responses that must be generated before the agent fully shuts down. This mechanism ensures that all active speech tasks complete properly and resources are cleaned up safely. Sources: livekit-agents/livekit/agents/tts/tts.py | Distributed Setup | LiveKit Docs