@sun I would be sure and engineer whatever ingestion pipeline you come up with to be basically implemented as site-specific plugins, for extensibility and future proofing against hostile adversaries.
I don't know how much I would trust youtube not to eventually patch out the rss feed of videos, but you might start with basically a glorified rss reader with the ability to call yt-dlp and ffmpeg
@sun Are you planning to include support for some kind of feed of videos that is automatically processed? even if you don't download everything, perhaps you want to be able to see the list and click, "yes go get that." The list would be site-specific and necessitate some sort of (in this case) plugin to translate site page -> feed of videos that you can then either manually fetch from, or automatically sync.
Basically no matter what downloader you use, you probably want some extensible method to distill some source of videos into a url that yt-dlp (or whatever) can fetch. Then the pipeline is not site-specific, just the distillation.