Despite this thread being very old (same as the related KVR thread), it‘s still an interesting idea. @tsenkov have you made any progress on this idea since then?
As a programmer I love to modularize things, and that would mean to separate the audio processing part and the UI part. The audio processing part should ideally be written as a JS library that can be used as a Web Audio Node in other projects (like a web game). The UI should ideally also be written in JS using web technologies so it could also be used in the browser for interactively creating and tweaking presets.
That means for the audio processing part, something should be used that interfaces directly with the Web Audio API standard. There is an (unfortunately not well maintained) port of the Web Audio API for node: https://github.com/audiojs/web-audio-api
The node and webview still need to be wrapped as VSTi as @eshikafe pointed out. Any suggestions in that direction?