I have this call: https://dashboard.vapi.ai/calls/019a7a23-607b-7bb9-88c7-58d12e14f287 I don't understand why it hung up while the person was talking, without warning. How do we prevent this from happening again?

To understand why the assistant hung up, check the call's

endedReason

in your Vapi dashboard or via the API. For example, you can use the following API call to retrieve the call object:

bash
curl -X GET "https://api.vapi.ai/call/019a7a23-607b-7bb9-88c7-58d12e14f287" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json"

Review the

endedReason

field in the response to identify the cause and consult the [Call ended reasons documentation](https://docs.vapi.ai/calls/call-ended-reason) for next steps. Sources: - [Debug call forwarding drops](https://docs.vapi.ai/phone-calling/transfer-calls/debug-forwarding-drops) - [Server events](https://docs.vapi.ai/server-url/events) - [Call ended reasons](https://docs.vapi.ai/calls/call-ended-reason)

I can help you look into why the call ended mid-conversation — this often happens due to timeout limits, end-of-speech detection, or webhook response delays. I’ll review the call logs and configuration settings to trace the exact cause. Once identified, I can adjust the timeout, update the speech settings, or add a fallback handler. Could you confirm if you’re using custom settings or default Vapi parameters? @Mistermanu#000

Im using custom parameters

Thanks for confirming you’re using custom parameters. In that case, the hang-up is likely coming from one of three places: a speech-timeout value that’s too short, an end-of-speech sensitivity that’s triggering early, or a delay in your custom tool response. I can review your parameters, trace the exact trigger, and help you adjust the settings so the agent never drops the call while someone is still talking. If you’d like, we can go over the details privately and fine-tune everything together. @Mistermanu#000

the ended reason is assistant ended the call

It looks like the call is ending because the combination of wait_seconds, voice_seconds, and number_of_words is too aggressive, making the assistant think the speaker finished too early. Adjusting these values slightly and adding a small backoff can help prevent premature hang-ups. I can review your current smart endpoint settings in detail. We can discuss privately and fine-tune everything for smoother call handling. @Mistermanu#000

so what setting do you recommend. No need to discuss privately...nothing really private

Its because this instruction of the prompt is vague:

json
* When the conversation should end, you MUST call the endCall function directly without explaining that you're doing so. Do not say \"I'll end the call now\" or similar phrases - just call the function.

Remove this and it shouldn't hang up randomly anymore.

So true