I’ve written here many times that people need to use the right tool for the job.
An LLM is not the right tool for this particular job.
There already exists several open source projects that can do this sort of thing.
I’m not overly familiar with them, as I didn’t have any reason to do any scheduling, but one you might consider is this,
Like I said, this isn’t an area I’m experienced with, so this one might not be the right tool for the job either, but somewhere out there there will be something much better suited for your task than a language model.
Edit: I’m just in the middle of putting a kid to bed. If no one else chimes in by the time I can circle back, I’ll look into it a little bit more for you.
There’s probably a work flow where a GPT takes natural language, converts it into a set of rules one of these schedules can understand, calls an API endpoint to get the schedule, and reports it back in a well-formatted manner.