Hi. We are making a voice assistant in our startup. We came to the conclusion not to use prompts in vapi, since we need to change prompts often, take answers from the cache, create custom answers when the client clicks on buttons on our site. We use sending text for voice from our server in this form body = { "type": "say", "message": text } send to the URL that we receive when creating a call. How to make it so that vapi does not come up with an answer itself. "model": { "provider": "openai", "model": "gpt-4o", "messages": [ { "role": "system", "content": "You should appear for any question, just a space ''.", }, ], }, We send such a prompt when creating a call. But when the conversation is long, vapi still starts to come up with answers without waiting for a message from our server.

@windownik On a serious note, we never manipulate your prompt or cache the responses. Therefore, you may be misunderstanding something.

To help track down this issue, could you share: - The call ID - What response you expected to get - What response you actually got instead This would really help us figure out what went wrong!

Is it possible to somehow configure the vapi assistant to disable its neural network and use it as a service for calls and TTS and STT? There are no problems in the call logs. But during the call itself, one message can stop and another one can start. Especially if the conversation is longer than 5 minutes.

@windownik that's not possible as of now for us, but you can raise an feature request for it over here: https://roadmap.vapi.ai