so I have a AI assistant set up for my business that's been running fall asleep for quite some time and all the sudden it starts having a lot of issues with the phone calls like interrupting the caller and not understanding what the caller saying, even though when I listen to the recording, it's very clear what was said. I was looking at the backend and I noticed that there was some kind of a update made to the stop speaking plan that took away the settings I had and forced me into a live kit with an aggressive setting, which I'm assuming is what the glitches here. How can this be prevented so somebody is not literally taking their time to perfect their voice agent and then it gets messed up by updates I think that's terrible.

To restore your assistant's preferred interruption behavior, explicitly set your

stopSpeakingPlan

and

smartEndpointingPlan

via the API. For example:

json
{
  "stopSpeakingPlan": {
    "numWords": 2,
    "voiceSeconds": 0.3,
    "backoffSeconds": 1
  },
  "startSpeakingPlan": {
    "smartEndpointingPlan": {
      "provider": "vapi" // or "off" if you don't want LiveKit
    }
  }
}

This ensures your custom settings are applied, even after platform updates. For persistent issues or to request a feature that locks your configuration, contact [Vapi support](https://docs.vapi.ai/support). Sources: - [Speech configuration](https://docs.vapi.ai/customization/speech-configuration) - [How Vapi works](https://docs.vapi.ai/how-vapi-works) - [Vapi Support](https://docs.vapi.ai/support)

This honestly makes no sense. I understand that changes are to make things better but to take someone's configurations and change them without permission or any consideration on how it will impact things is a terrible practice.

PATCH [https://api.vapi.ai/assistant/](https://api.vapi.ai/assistant/id)<ASSISTANT_ID> You can read more about it here in the API documentation: