[konschubert]

Has to be online since it is proxied via an online backend which does the actual black-and-white conversion etc.

I have some ideas for letting users point the device directly to a local URL, but I don't want to commit to any promises on that front.

You could try using something like https://ngrok.com, that's what I sometimes use for testing.

[mulmen]

Hm, I see. I’m pretty sure I can figure out how to serve a greyscale image. I suggest making that proxy explicit in your docs. You suggest using HTTPS and an obscure URL to keep our data “secret” but then you’re literally processing our content and could be doing anything with it.

[konschubert]

That's a valid point. Image data is never stored, only hashes. But it passes though the server and I don't like that either.

It's not quite greyscale, it has to be in a specific binary black-and-white representation, but definitely do-able if documented. Or I could publish a little python library.

Just thinking out loud here.

[mulmen]

Yeah don’t mean to diminish the project, it’s very cool! I’m sure you already have a lot of work into it.

This thing is almost exactly what I was looking for less than a week ago. It’s actually even simpler and more clever than I imagined. For me less is more. Just serve an image at a URL. EZPZ.

If you want to make it hackable I suggest docs first because everyone is going to want to use their own preferred language. Maintaining a library is just more work and image processing libraries are already a dime a dozen.

[konschubert]

That's a good point about documentation-first. Sometimes a little reference code can say more than 1000 words, though :)