I'm having this issue where intermittently, the voice agent will hang. I can see in logs that we get the deepgram transcript for user speech. Then no openAI azure calls. ElevenLabs websocket just opens and closes, I suppose waiting for model response that doesn't come. It's a very frustrating user experience because it takes multiple "Hello" "Are you there" responses to get back on track 1. What is the root cause of this? Why does Vapi not consistently call our model 2. How should we handle this when it does happen? Trigger retries on our side and not rely on the user to respond to the silence

Hi there, Thank you for your message. Our team is currently out of the office. We operate Monday through Friday, from 9:00 AM to 8:00 PM Pacific Standard Time (PST). We’ll get back to you as soon as possible during our normal business hours. If your message is urgent, please mark it accordingly or include “URGENT” in the subject line, and we’ll do our best to respond promptly. Warm regards, Vapi Customer Support Team

here is a call ID where this happens: 0199d92b-e9dc-7aa6-883b-ff079a6c06cc

Hi Jessica, here are the key points to help diagnose and resolve this issue: For intermittent hanging issues, you should: 1. Check the provider status: - Visit Vapi Status Page - Check OpenAI/Azure status - Verify ElevenLabs status - Monitor Deepgram status 2. Check logs to diagnose the issue: - Navigate to "Observe > API Logs" to verify request payloads and response codes - Review "Call Logs" to identify where calls failed - Look for tool execution results and errors Common error patterns that match your symptoms: - Long silences are often caused by model processing delays - Solution: Use faster models or reduce response length - When calls drop or hang, check all required fields in assistant/workflow settings To handle these situations: 1. Set appropriate timeouts in your configuration 2. Add validation prompts to handle missing responses 3. Use the dashboard debugging features to monitor call flow and identify bottlenecks For calls that hang after transcription: - Verify model configuration in the "Model" section - Ensure "Max Tokens" is set appropriately - Check that necessary tools are enabled and configured correctly