Hi the last week or so I have been getting a lot of errors when trying to connect the call from the dashboard from my desktop and when having the assistant i created make a call to phone numbers. I noticed that the assistant is usually able to say the "first message" and then the connections drops (image attached) I first noticed this happening with claude 3.5 sonnet, but now I am also noticing it happening with OpenAI GPT-4o. Oddly, with claude 3.5 sonnet I was not being able to get my function to run to capture he structured data I required. I am really worried becuse I am using VAPI to prototype and launch a service with hundreds of users a day, and right now I cant event validade this service with customers because of calls being dropped really often. https://cdn.discordapp.com/attachments/1260402868993785926/1260402869119484024/image.png?ex=668f3125&is=668ddfa5&hm=a073f5d867395f6c1328e34b14994d32c1d84be27f5c5864e70bf4d8dc8b2653&

@Marlos I'm sorry to hear that you're experiencing issues with your calls. It seems like there might be a few different factors at play here, including the assistant's performance and the call connection stability. Let's tackle these issues one by one to ensure that your service runs smoothly. For the call connection issues, it's important to check if there's a pattern to the dropped calls, such as a specific time of day or after a certain duration. Additionally, verifying your network stability and the system requirements for running VAPI from your desktop could provide some insights. Regarding the assistant's performance, since you've experienced issues with both Claude 3.5 Sonnet and OpenAI GPT-4o, it might be related to how the assistant is integrated with VAPI. Ensuring that the assistant is correctly configured to handle calls and that the scripts are error-free is crucial. For the function that captures structured data, it's essential to debug the function to identify why it's not running as expected. This could involve checking the code for errors, ensuring that the function is triggered correctly, and that the data it's trying to capture is being passed to it in the right format. Since you're prototyping a service for a significant number of users, it's also a good idea to consider scalability and reliability. This might involve looking into any limitations of the current setup and planning for redundancy to handle any unexpected issues. If you could provide more specific error messages or logs, I could give you more targeted advice. In the meantime, if there's anything else I can assist you with, please let me know.

I noticed that the problem doesnt occur when I am using groq models. But they dont work with function calls which is essencial for my use case.

@Marlos please record a loom video.