Issue: Unable to Clear Audio Buffer When Caller Interrupts AI Response

[Whisperer]

Version: 3CX Version 20.0 Update 5 (Build 551)
Note: The application I am discussing below is built in python

Hello,
I am working on an integration where we use the 3CX Call Control API to manage calls and bridge them with an AI agent (OpenAI Realtime API). The flow is as follows:

• We listen to call events via WebSocket.
• On an inbound call, we attach to both the input stream (caller’s audio) and the output stream (AI-generated audio).
• Caller speech is streamed into the AI, and the AI’s generated audio is sent back into the 3CX call leg.

This works well in normal back-and-forth conversation.
However, when the caller speaks while the AI is already speaking, OpenAI’s voice activity detection (VAD) cancels the AI’s response.

At this point:

1. We clear our local output buffer, so no additional AI audio is sent.
2. But the audio already sent to 3CX remains in its buffer and is still played out to the caller.
3. This results in the caller hearing the “tail end” of the canceled AI response, even though the AI has already stopped.

---

Question​

Is there any way, via Call Control API or another method, to flush or clear the audio buffer inside 3CX once a cancel event occurs?

---

Request​

• Does 3CX provide any mechanism to flush the audio buffer of a call leg immediately?
• If not, what is the recommended way to handle this scenario so that the caller does not hear canceled AI speech after they start talking?