Hello, does anyone know how to send responses similar to Tool calling in VAPI but with a custom language model? I have my custom LLM, and I want to send messages like "Wait a minute while I search for you." The 'request-failed' and 'request-complete' responses are pretty straightforward due to their result nature, but I don't know how to send the initial one that corresponds to 'request-start'. This is an example configuration of a VAPI tool. My question specifically pertains to how to send the 'request-start' message but when functions are only configured in my custom LLM :

[
    {
        "type": "request-start",
        "content": "Wait a minute while I search for you."
    },
    {
        "type": "request-complete",
        "content": "The weather in the location is..."
    },
    {
        "type": "request-failed",
        "content": "I couldn't retrieve the weather information at the moment."
    },
    {
        "type": "request-response-delayed",
        "content": "There seems to be a delay in communication with the weather API.",
        "timingMilliseconds": 2000
    }
]