We are having issues using elevenlabs scribe transcription model with custom LLM. When someone speaks, our server endpoint is never hit. However, issue does not occur with deepgram.

Hi there! I understand you're having issues with the ElevenLabs Scribe Transcription Model not triggering your server endpoint when someone speaks, while Deepgram works fine. Let me help you troubleshoot this! 🔍 Based on what I've found, there are several key differences between ElevenLabs and Deepgram that might be causing this issue: ### Potential Causes 1. Endpointing Configuration - Deepgram has an aggressive default endpointing value (10ms) that can be adjusted to 300ms for better results - If ElevenLabs doesn't have similar endpointing configuration options, this could affect when server endpoints get triggered 2. Confidence Thresholds - Deepgram filters transcriptions based on confidence scores, which might differ from how ElevenLabs handles low-confidence transcriptions ### Troubleshooting Steps 1. Check Server URL and Endpoint Configuration - Ensure your server URL in the Vapi Dashboard is correct and accessible from outside your network - Verify that your endpoint is set up to receive POST requests 2. Test Manually - Use tools like Postman or cURL to manually send a POST request to your endpoint to confirm it's functioning correctly 3. Review SSE Response Format - Make sure your LLM is sending a streaming response in SSE format to Vapi - Verify the response is properly structured and terminates with "data: [DONE]\n\n" 4. Check Custom LLM Configuration - Ensure your custom LLM includes the necessary tools for interacting with ElevenLabs - Verify your model configuration includes the correct provider, URL, model, and tools 5. Examine Server Logs - Look for any errors in your server logs that might prevent the webhook from being called ### Additional Considerations - Network issues or firewalls might be blocking the required traffic - Recent configuration changes could have affected webhook functionality

Hey man, mind providing me a call ID? I'll make sure we look into it.

@Mason that's callId for you 3d8fb205-9a29-4b1c-a2eb-09751dc5226c

Hey Meran, taking a look at the logs it seems like the websocket connection with elevenlabs was experiencing issues. Could you test again and let me know how it goes?