Possibility of inputting new files into each new message by creating a new tool to process the file and allow the assistant to give feedback on the contents

Hey everyone.

I am working on a movement tracking project and am looking into how making my own assistant could play a role.

Currently I have code that uses mediapipe to get frame by frame body part positioning data and puts all of that into a civ file. When I give the data from one frame, gpt-4 has no problem understanding the person’s positioning, so now I am trying to find a way to give an assistant the full csv file so it can understand the movement over time (and eventually provide a summary of the movement).

I was wondering, to make an assistant which can take in a new csv and interpret it every time, would I need to create a new tool/function call that can help it interpret the csv? And what would be the framework for something like this?

Or is there a better way to do all of this?

Thanks.