Under advanced settings on your assistant you’ll see a “Messaging” section. There you can choose what client or server messages get sent, for your use case you’ll want to select model-output