We have relational data and I've had to spend a lot of time making Mongo do the things that we need. It's a bad fit for this type of application but it's always seemed more trouble than it's worth to change. However there are more issues than just fit for purpose. Mongo requires an annoying amount of maintenance to keep up-to-date, which we've needed to do because they are still adding stuff that lets you simulate relational models every now and then. There also aren't many affordable managed options for small applications.

If we switch to anything SQL-based, we get many more guarantees, and probably much better performance too. As it is we only have a few 100K medias at most on wlk.yt but loading 1 page of a playlist can already take 500ms because of all the relations we need to look up using fancy mongodb pipelines and whatnot. We could also improve it by changing the data model (we really didn't design a data model for mongo back in the day), but at that point we are already doing a complicated migration, and we might as well make a similar-effort migration for much larger gains.

Maybe 2023 is the year for this. I would like to avoid using an ORM. I think it would be nice if we could even work with sqlite (though i would probably recommend postgres in docs).