So for custom tools we can provide messages for the AI to say during different stages of execution. I've found that in Request Start we can use the AI's own message but for other stages we have to put a hardcoded one. I'd like it if the AI would provide it's own response based on the returned results from the tool call. I was wondering if Vapi supports this by default or if I have to make a custom solution and send a request to tell it to generate a response once the tool call is complete?