On the official site, images are parsed when the body contains them.

It seems like the open source version doesn't. Or am I misconfiguring something?

Thanks,
Sam