Handling real-time dynamic complex data with function-calling

Dear community.

I’m reading the API documents and examples, but there is something what I cannot understand.

As far as I understand, function-calling is the best way to get an answer about dynamic real-time data. However all the example I saw was, just for a simple short answer like “Temperature of Berlin is 25 degree.”

What I want is;
I have some structured JSON(or whatever) data about the time-table, schedules, jobs of people. And the data is refreshed randomly many times even in a day.
For example;

- John / 09:00 - 13:00 / Job A
- John / 14:00 - 15:30 / Job B
- Michael / 11:00 - 14:00 / Job A

And I want to ask various questions about this data;

  • “Who will work for Job A today?”
  • “When could Michael be free?”
  • “How many hours John will work?”
  • … (Or any other many possible questions)

With prompt I can get the results, but it would consume too many tokens.

If I use function-calling, I have no idea how to design proper function to output. Should I prepare all possible question and answers? It sounds weird and non-sense.
Is it possible to deliver whole data as an output of one function-call? Could AI assistant understand the output and could answer for various question with only one function-call?

Or File Search would be the solution? But I’m not sure the frequent updated data would be suit for File Search or not.|

What is the better practics?