Add Webhook notifications and option to prune orphan strm files.

[mcortt]

No description provided.

[mcortt]

Also added an option to prune orphan strm files

[R3XCHRIS]

Sorry for the long silence on this @mcortt — I was away and this deserved a proper read rather than a quick skim. I've now gone through it carefully.

The quality here is genuinely high. Some specific things I want to call out, because they're the things people usually get wrong:

• _collect_expected_movie_files deliberately ignoring batch_size — with the reasoning in the docstring. That's exactly right, and getting it wrong would have deleted most of a library on anyone running the default batch of 250.
• The (keep, ok, count) return with ok=False meaning "MUST NOT prune" — failing closed on a DB/import error is the correct instinct.
• The "0 active but files on disk → refuse" guardrail. A provider outage wiping the library is the obvious catastrophic scenario and you closed it.
• Keeping the .nfo in the keep set regardless of the current NFO toggle, so toggling NFO off doesn't orphan good files.
• realpath for keep-set identity, protected subtrees, and startswith(pref + os.sep) — you even anticipated the /Movies vs /Movies2 prefix bug, which has a test.
• ~40 tests including test_user_files_preserved, test_protected_subtree_untouched, test_root_never_removed_even_when_emptied.

The webhook half I have no concerns about: opt-in, 10s timeout, failures logged not raised, and — importantly — no URLs or credentials in the payloads, only counts.

---

One blocker, and it's not your fault

This PR is from 2 July; main has since had v1.16.1 and v1.17.0, and one of those changes breaks a core assumption here.

v1.16.1 added a tmdb_tag_format setting (issue #9 — Jellyfin/Emby want [tmdbid-123], Plex wants {tmdb-123}). _movie_target_paths / _series_target_folder now take a tmdb_tag_format argument that defaults to "plex".

Your collectors call them without it:

movie_folder, strm_filename, _name, _year = self._movie_target_paths(
    movie, root_folder, cat_name, nest_by_cat, append_tmdb_id,
)   # <-- tmdb_tag_format defaults to "plex"

So for a user with Append TMDB ID ON and format jellyfin:

• Generate writes Cool Hand Luke (1967) [tmdbid-378]/…
• The keep set computes Cool Hand Luke (1967) {tmdb-378}/…
• Nothing on disk matches the keep set → every file is an "orphan" → the whole library is deleted.

The 0-active guardrail doesn't catch it, because count is non-zero — the catalogue resolves fine, it's the paths that are wrong.

That's not a corner case: most users of this plugin are on Jellyfin/Emby (Plex can't play .strm at all), and v1.16.1's release notes actively tell them to switch to the jellyfin format.

Fix is one argument in each collector — pass tmdb_tag_format through, read the same way _generate_movies does:

tmdb_tag_format = (settings.get("tmdb_tag_format") or "plex").strip().lower()

Two smaller items

1. category_exclude (new in v1.17.0) isn't applied in the collectors. This one fails safe — the keep set becomes a superset, so prune under-deletes rather than over-deletes — but it should be applied for correctness, right after _apply_category_filter.
2. Worth a rebase regardless — 4 commits have touched plugin.py since your base, so this will need conflict resolution anyway. (Also: plugin.py is CRLF — worth checking your editor preserves that, or the diff balloons to the whole file.)

How do you want to play it?

Happy either way:

• You rebase + fix, I re-review and merge; or
• I merge and fix it in a follow-up commit on main (same as I did for the _build_proxy_url refactor after Add option to omit stream_id from generated .strm URLs & closes #5 #6) — just say the word and I'll credit you in the release notes.

One last thought, no strong feelings: webhooks and prune are two unrelated features in one PR. If you'd rather land them separately I'd take the webhook half immediately — it has no blockers at all. But I'm equally happy to take both together once the tag-format bug is closed.

Really nice work — the guardrails in the prune path are better than most of what I'd have written first time.

[mcortt]

You are absolutely right. It should've been two separate PRs. My bad. I am fine with whatever you want to do. If you want to rebase/fix and push it go for it. That way you can go ahead and implement the webhook portion, and then once the follow-up commit is processed the prune can be added at your convenience.