Hi!
It’s pretty common practice to insert the call’s date (and perhaps time) dynamically into the system prompt or something.
Regarding multiple functions:
I guess we need to distinguish between parallel and sequential function calling.
There’s nothing inherently preventing multiple function calls being sent off in one go. If you use a more limited function interface, you might need to make use of the command/dispatcher pattern, but there’s nothing really stopping you.
But if the result of one function is needed as the input of another function, you can’t call stuff in parallel.
But there’s also nothing stopping you from having functions called in sequence. If you insert the function response into your chat, assuming proper prompting, the model will then call the next logical function in your workflow.
3.*“Find me a room for 2 ppl, 2 weeks from now for a month near NY university”**
I don’t know if you need multiple function calls for this.
If your function signature looks like
{
start: string, // accepts "MMMM D, YYYY", "x days from today"
duration?: string, // accepts "x days", "x weeks", must be set if end isn't set
end?: string, // accepts "MMMM D, YYYY", "x days from today", must be set if duration isn't set
location: string // where the user wants to go
}
you can just pluck the details apart in your program. The model should just be instructed to gather all necessary information before sending it off.
Does that help?