Has anyone experienced the Assistant ending calls too quickly? For example, if the AI is reading of a list of featured and the caller says "okay thats enough, thank you" the AI then goes to end the call. Any combination of endCall tool details, prompt details, endCallPhrases, etc that you have found to work well would be helpful!

It sounds like you're encountering a common issue with the Assistant ending calls prematurely! Here are some recommendations to help configure the

endCall

tool and manage endCallPhrases effectively in Vapi: ### Recommended Configurations for the

endCall

Tool 1. **Condition-Based Termination**: Set specific conditions for when the call should end. For example, you might want to avoid ending the call if the user says something like "that's enough" unless it's clear they want to end the conversation. You can configure it like this:

json
   "content": "If the user is being rude or uses offensive language, terminate the call."

2. **Testing and Feedback**: Regularly test the Assistant's performance in real-world scenarios to ensure it only ends calls when appropriate. This helps avoid false positives. 3. **Speech Configuration**: Adjust the assistant’s speech settings to better handle background noise and interruptions. This can help prevent misunderstandings that lead to premature call endings. 4. **Monitoring Call End Reasons**: Use the "Ended Reason" reports to analyze why calls are terminated. This can help you refine your configurations over time. ### Effective EndCallPhrases Using the

say

method can help provide a courteous exit before ending the call. For example:

javascript
vapi.say("Our time's up, goodbye!", true);

This ensures the Assistant communicates a closing message before the call ends. ### Example of Implementation You can set up your model to include the

endCall

function based on specific phrases:

json
{
  "model": {
    "provider": "openai",
    "model": "gpt-3.5-turbo",
    "messages": [
      {
        "role": "system",
        "content": "If the user says 'that's enough', do not end the call immediately."
      }
    ],
    "tools": [
      {
        "type": "endCall"
      }
    ]
  }
}

### Caveats

I've tried these solutions and they have no impact on solving the issue or making it any better :/

Hey! To help track down this issue, could you share: - The call ID This would really help us figure out what went wrong!

Sample call IDs: e9f832b7-cfb6-4206-b254-4b7ffa60766f 09ac1643-fd9a-4126-bb45-170ec7f46b2a e9bbd0a2-355c-4acc-b5cd-a49d1be2d7a3

It's a common issue when AI assistants interpret phrases like "that's enough, thank you" as a signal to end the entire call rather than just stop the current explanation. Based on your call examples, the AI is ending calls when users are actually trying to interrupt the current information flow but want to continue the conversation.

okay I will give this a try, previous attempts at updating the prompt have not been successful, I will try your exact langauge and see if that makes a difference

Let us know how that goes, a definite solution would be to specify exact phrases if adjusting your prompt does not work as shown above.

Can you provide an example of what you mean please?

Sure absolutely, you can specify custom end call phrases to anything you want, here's an example: endCallPhrases: \[ "goodbye", "bye", "end the call", "hang up", "that will be all", "I'm done" \]

Ah ok, yeah we have those and it isn't going the trick on improving things

Hey LLMuser995, following up on this ticket if it is resolved for you.