jp
10/15/2025, 12:36 PMVapi Support Bot
10/15/2025, 12:37 PMjs
navigator.mediaDevices.getUserMedia({ audio: true })
.then(() => console.log("Microphone access granted"))
.catch(err => console.error("Microphone access denied:", err));
If you see an error, resolve the permission or device issue—this is often the root cause when the AI model is not triggered in web calls, even if the dashboard test works.
Source:
- [Debugging voice agents](https://docs.vapi.ai/debugging)Chiranjeet Mishra
10/16/2025, 7:08 PMfirstMessageMode is correctly set to "assistant-speaks-first" both in the assistant settings and assistantOverrides.
- Ensure the assistantId used in web calls matches the one configured for successful dashboard tests.
2. Review Call Logs:
- Navigate to Observe > Call Logs in your [dashboard](https://dashboard.vapi.ai/) to check call statuses and error messages.
- Look for any discrepancies in tool execution results or model invocation errors.
3. Test in Dashboard:
- Use the "Talk to Assistant" and "Call" functions on the [dashboard](https://dashboard.vapi.ai/) to isolate network issues.
4. Provider Status:
- Check the [OpenAI Status Page](https://status.openai.com/) for ongoing issues that might affect voice model operations.
5. Diagnostics:
- If no improvements, refer to [this guide](https://docs.vapi.ai/debugging) for common debugging procedures and ensure your AI service providers aren't reporting issues.
By following these steps, you can identify and resolve the issue causing the web calls to not trigger the AI model.