PlantUML works pretty good in Asciidoc? If you don't want to set up a path to the local jar, you could use Kroki if you don't mind using a service (or you could run Kroki locally).
But I get what you're saying.
There's lots of speedbumps setting up graphing renderers, and the non-Ruby PDF engines aren't optimal (asciidoctor-web-pdf, which is the Paged.JS/CSS PDF toolchain) or they're too arcane for casual users (the DocBook stuff, i.e. DocToolChain/DBXSL/FOPUB). As for the core processor, though, AsciidoctorJS is pacing more or less hand in hand with core Asciidoctor.
Speaking for myself, it was worth the hit, but I'm an edge case. I work primarily with hardware/manufacturing, so I needed the transclusion, conditionals, XML equivalence[1], decent tables, front matter/indices/running content, etc, all in one core spec, because it takes eight months to get a new piece of software. Running back and forth to get permission for every Markdown extension and variant was a non-starter :D
[1] To a nice "relatively sane" XML spec like DocBook
