I 'm using Vapi's python-server-sdk to create a transient assistant for use in a phone call. I requested the server, and it went through successfully. I used the webhook config to specify a webhook call to my local machine (using ngrok tunnel) so we could process the call transcript after the call was over. It seems the assistant instead relied on the URL pointing to our staging ENV which I configured in the dashboard. Since then, this agent has sent hundreds of requests to our staging ENV (which results in a 404). I've sent call deletion requests (which return 200) and even gone as far as to delete the entire user organization associated with the account. It has made no difference, and we're still getting incessant POST requests to our backend system. There are also severe mismatches between the Python SDK and documentation in terms of proper typing, which has made working with the SDK implausible.

Set the serverUrl in the SDK to your ngrok webhook, limit events via assistant hooks or the dashboard, delete or update the assistant, and use raw HTTP or GitHub reference to avoid SDK type issues..