It makes a request to the thing that searches through lists of stuff (torrents-, Usenet-indexers), which then requests some downloader (like a torrenting client) to download the stuff you want.

The whole servarr stack is pretty complicated but that also means it’s not a messy monolith.