[Feature] AnimeTosho stopped adding new items in May 2026 and its feed is shutting down, please add AnimeTosho.xyz #220

Activity

1. Svnow132

   added

   enhancementNew feature or requestNew feature or request

   on Oct 2, 2026
2. Abrechen2 commented on Oct 9, 2026

   

   Abrechen2

   on Oct 9, 2026

   Owner

   More actions

   Thanks for the detailed comparison. Sublarr has already been searching feed.animetosho.xyz since 1.14.1 (#208), and downloads come from its storage hosts — so on your 1.15.0 install the provider is not limited to entries before May, and the .org shutdown will not take it down. While checking, I found that the .xyz feed returns a bare file name instead of a storage address for a few attachments; those candidates could win and then fail with a 404. That is fixed for 1.16.0.

   On AvistaZ / CinemaZ: 1.16.0 will take over the providers from the Bazarr+ provider catalog you linked, AvistaZ and CinemaZ included. I'd gladly take you up on your offer to test them with your accounts — I'll post here when there is a build to try.
3. Abrechen2 commented on Oct 10, 2026

   

   Abrechen2

   on Oct 10, 2026

   Owner

   More actions

   Shipped in 1.16.0, now running in production.

   AvistaZ and CinemaZ are among the 43 catalog providers — switch on Catalog providers in the provider settings, enable them and add your session cookies. I'd be glad to hear whether they work with your accounts.
4. Svnow132 commented on Oct 11, 2026

   

   Svnow132

   on Oct 11, 2026

   Author

   More actions

   Thanks for shipping this. I went through the 1.16.0 catalog before turning anything on and ran a few bundles' search() directly with a video dict shaped like video_from_query builds it. I think TsukiHime can't return a result for anyone in 1.16.0, because of what the adapter passes it.

   Episodes

   bundles/tsukihime/provider.py returns [] unless all three are set:

   episode_id = _positive_id(_scalar(video.get("series_anidb_episode_id")))
   episode_no = _positive_id(_scalar(video.get("series_anidb_episode_no")))
   if anidb_id is None or episode_id is None or episode_no is None:
       return []

   providers/hub/adapter.py video_from_query sets series_anidb_id and series_anidb_episode_id, but nothing in backend/ sets series_anidb_episode_no (same on master). So every episode search ends at that return [] before any request goes out.

   There's a second gate on top of that: query.anidb_episode_id is only filled in _resolve_anidb_from_shoko (wanted_search/metadata_enrichers.py). Installs without Shoko, like mine, never get an AniDB episode id, so the bundle would still bail even with the episode number passed.

   Movies

   The movie path needs anilist_id, but the adapter only puts anilist_id in the episode branch of video_from_query, not the movie branch. So movies return [] too.

   What I ran

   The bundle on its own with Frieren S01E01 (AniDB 17617, AniList 154587) and the fields the adapter sends: 0 results, and it returns before any request. Adding series_anidb_episode_no gets it past the gate and querying the API (still 0 for Frieren, but only because of the index, see below).

   Possible fix

   • Pass series_anidb_episode_no from the episode number you already have (absolute_episode for absolute-numbered anime, else episode).
   • Add anilist_id to the movie branch.
   • The bundle's own comment says the episodes endpoint is keyed by episode number and the episode id only acts as a gate. If you change the pinned copy, dropping the episode_id requirement would let it work without Shoko.

   Separate from the bug, TsukiHime's index looks thin for now. I sampled 11 titles straight from its API, and 4 had a completed torrent with English subs. Frieren's only entry there is skipped. So it's not urgent for me, I just wanted you to know it can't fire as wired.
5. Svnow132 commented on Oct 11, 2026

   

   Svnow132

   on Oct 11, 2026

   Author

   More actions

   One more from the same pass, then I'm done. I diffed every key the 43 bundles read off video against what video_from_query sets (on master, 47a2b65), so this should be the full list.

   CinemaZ and AvistaZ can't return anything either

   Both bundles start search() with:

   info_url = str((video or {}).get("info_url") or "")
   if not _is_cinemaz_url(info_url):
       return []

   video_from_query never sets info_url, so both stop there on every search, before the cookie check or any request. The manifest description says the release page comes from Sonarr or Radarr history, so I think the adapter needs to pass the info URL of the grab that produced the file. I haven't checked where Sublarr would read that from. This is the one you asked me to test with my accounts, and as wired it can't find anything yet.

   Minor, no search is blocked

   • fps is never passed. subssabbz and subsunacs can't add their fps match, and titulky's skip_wrong_fps option has nothing to compare against.
   • Several bundles return an encoding next to content_b64 (subtitlecat sets latin-1 when the body isn't UTF-8). _unpack ignores it and keeps the raw bytes. I haven't checked whether a later step re-encodes provider downloads, so this might already be covered.
   • The other unset keys (episode_title, audio_codec, streaming_service, edition, original_title, HDBits' imdb/tvdb) are only extra match hints or have a fallback that is set, so nothing breaks on those.

   Checked and fine

   • Every third-party import a bundle uses is either in backend/requirements.txt (ua-generator, requests, rarfile) or shipped in the bundle and listed in its manifest (subhd/captcha_templates.py).
   • subtitlecat only adds episode to matches when the episode actually matches, so the wrong-episode hits I saw in a standalone run should score lower rather than win.

   So the two that need a fix are TsukiHime (previous comment) and the info_url gap for CinemaZ and AvistaZ.
6. Abrechen2 commented on Oct 24, 2026

   

   Abrechen2

   on Oct 24, 2026

   Owner

   More actions

   Thanks for going through every bundle — both gaps you found are fixed in 1.17.0.

   • TsukiHime now gets series_anidb_episode_no, and movies get anilist_id. One limit remains: the AniDB episode id and number still come only from Shoko. Without Shoko, TsukiHime still skips episodes. I left the bundle's episode_id gate in the pinned copy, so it stays identical to the catalog.
   • CinemaZ / AvistaZ now get info_url. The wanted search takes it from the Sonarr/Radarr history of the grab that produced the file. So only files grabbed from those trackers are searched there. The history is read only while one of them is enabled, and at most once an hour per episode or movie.
   • fps is passed now, taken from Sonarr/Radarr, so SubsSabBz, SubsUnacs and Titulky's skip_wrong_fps have something to compare against.
   • encoding next to content_b64 is now used as a hint when the file is converted to UTF-8 (this matters for ktuvit's windows-1255, for example). The latin-1 that several bundles return for "not UTF-8" is ignored, and the file's own detection applies.

   Your earlier offer would help a lot here: nobody has run CinemaZ or AvistaZ with a real account yet. If you try them on 1.17.0, I'd be glad to hear whether a grab from either tracker finds its subtitle.