Yes, modifying the value is going to break the mappings (see https://github.com/jellyfin/jellyfin/blob/master/Emby.Server.Implementations/Localization/Ratings/us.csv). Anywho, I think we’ve discovered the root of your problem. How you choose to rectify it I leave to you! Personally, I’d recommend suffixing your filenames with [] as per https://jellyfin.org/docs/general/server/media/movies/ and letting themoviedb.org handle it all for you.
https://github.com/jellyfin/jellyfin/blob/31aa44d23d12b5dbb5f9a131242cc82c9ef98f24/Emby.Server.Implementations/Data/SqliteItemRepository.cs#L2279 is what’s discovering similar content. If the InheritedParentalRatingValue is considered zero, it’s only going to match other content with the same value. Can you elaborate on “I did change the name of the key for the rating variable in the metadata to be ‘MPAA rating’ instead of the default which I think was ‘rating’ before since I found it confusing.”? I suspect we’re zeroing (ha ha) in on the problem.
Hey, I’ve worked in the recommendations/similarity calculations. Could you post a screenshot of the detail page for Inside Out? I suspect your media doesn’t have associated metadata (e.g. tmdb tags) that are used to power similarity calculations.
Getting the code running, easy. Getting the pull requests moved forward, a lot more frustrating than expected.
https://lemmy.ca/post/6420647 summarizes my feelings on the latter.
I got a bunch of commits in around searching and similarity. https://github.com/jellyfin/jellyfin/pulls?q=is%3Apr+author%3Abradbeattie+is%3Amerged, https://github.com/jellyfin/jellyfin-web/pulls?q=is%3Apr+author%3Abradbeattie+is%3Amerged.
Totally possible to go overboard in either spectrum of complexity. But yeah, take Prometheus for example. It’s super easy to set up and does a great job of metrics. Reimplementing this in bash would require… a lot of work.
Oh man, you read my mind: https://github.com/jellyfin/jellyfin/pull/7663#issuecomment-1758314971. Thanks for jumping on this!