From their site:
You can pass other info in the outputs hashes, like labels, your own ids, or whatever. That extra info will be passed back to you in the response JSON.
So I assume it is a case of emailing them and asking for an exhaustive list of what can be passed to their api, the documentation definitely needs work.
It looks interesting but I found the documentation page to be pretty lacking. The simple example they show just allows you to pass a height and width parameter, but the home page says you can do cropping, watermarking, etc.