I have followed through on a pricing calculator that seems infallible in terms of delivering the patches count and the pricing (when images are sent according to API reference).
Its javascript is procedural and readable instead of minified and obfuscated (like that powering the bottom of the pricing page, hidden under an “expand”). However, it has procedural conditional statements that may not be the exact way it is done on API inputs.
Awesome documentation would be:
We offer this python function
image_dimensions_to_patches(x,y), delivering the exact patches count received for any additional image input, using our exact algorithm. Multiply its integer output by the model’s cost multiplier (and round [up/down]…).
Also as an output could be the new dimensions - for it is practical to receive the downsized image dimensions, to determine if features could still be observed by AI. And simply perform that same resizing client-side where it is observable.