Assistants - Do file contents fully count as context tokens every call you make a call?

So I created an assistant with a JSON file with a fair bit of data attached to it, and a lengthy description.

I’ve used over 100K Tokens in a short time with only a handful of queries. It’s getting pricey fast.

Is there a better way to approach this? Or even with Assistants does it need to hit the API with ALL the context every time?