[ Feature Request ] In some noisy scenarios, we need the ability to finish a single round of conversation before moving on to the next round.

[weynechen]

In simple terms, we need to interact between the front and back ends, with flags and callbacks for starting and ending a conversation。

A simple way is to add flag bits to the current audio data to indicate whether it is the last audio. After the front-end playback is completed, information can be returned to the endpoint of fastrtc or FastAPI

[weynechen]

By the way , I think
handler= ReplyOnPause(response, can_interrupt=False,
can_interrupt flag should used for this.

[freddyaboulton]

Yes, does can_interrupt=false not work for you?

[weynechen]

Yes, does can_interrupt=false not work for you?

not work. The current situation is that the playback will continue when new sounds come. But ASR has already begun. It doesn't start only after the playback is completed.

A typical scenario is that if I need to press the keys to control the start of recording the sound and transmit it to the back end. In fact, the front end has no way of knowing whether it has completed the playback of the last conversation