We need the ability to configure the assistant so that it does not listen or react to user input while it's reading a prompt aloud. Right now, the assistant starts listening as it speaks, which causes interruptions, false detections, or premature transitions when the user makes a noise or speaks accidentally.

This is particularly problematic in regulated workflows where full legal disclaimers or consent texts must be delivered uninterrupted before collecting a response.

Or just an option to delay listening until the assistant finishes its TTS output.

This would help ensure more reliable and compliant conversation flows.

Feature Area:

Conversation nodes / Voice configuration