Contributing Lidarr, Readarr, Bazarr, Seerr and qBittorrent integrations

[vbrzezina]

Hi! I've been running Logarr on my NAS and built five integrations from the "coming soon" list, plus a fix for *arr file ingestion. They're open as PRs and ready for review. This discussion gives an overview in one place, so you can see how they fit together.

PR	What it adds
#57	Fix: *arr log-file ingestion produced no entries (see #63). Also adds small shared helpers in provider-arr used by #60/#61.
#58	Bazarr: subtitle history, health and failed jobs; tails bazarr.log
#59	Seerr: requests and issues; tails seerr-*.log (also covers Overseerr and Jellyseerr, which share the same API)
#60	Lidarr: v1 API on provider-arr (stacked on #57)
#61	Readarr: v1 API on provider-arr, labelled best-effort since Readarr was retired upstream (stacked on #57)
#62	qBittorrent: cookie login, auth bypass, or the 5.3+ API key; errored torrents, connection status and warning log feed

Suggested merge order: #57 first, because #60 and #61 build on it. The others are independent. They all touch the same registration and README/compose lines, so whichever goes second needs a trivial rebase, which I'm happy to do.

Design points worth your opinion:

• feat(providers): add qBittorrent integration with username/password auth #62 adds a nullable servers.username column (migration 0010) and two optional ProviderCapabilities flags, supportsUsername and apiKeyOptional, so that username/password auth isn't hard-coded to one provider.
• Seerr replaces the separate overseerr and jellyseerr "coming soon" entries with one seerr entry, which search still finds under both names.
• Every provider follows the same conventions as the existing ones: snake_case activity types, [id] Failed to get … logging, and fetch-mocked specs. Each PR has around 80–160 unit tests, and build/lint/typecheck/test pass.

I've been running all of them merged together on my own instance, and #57 has already been through real logs. I'm happy to split, rename or rework anything. If you'd prefer a different order or scope, let me know.

[vbrzezina]

Thanks for merging #57 and #60, and for the detailed reviews. I'm working through the requested changes on #58, #61 and #62 and rebasing #59; I'll reply on each PR when it's updated.

There's one new PR, which came out of running the integrations on a real instance: #65, which fixes log_entries deadlocks and an issue-creation race in file ingestion. It's ready for review.

• Deadlocks: with several log files tailed at once, concurrent batch inserts deadlocked on the (server_id, deduplication_key) unique index (hundreds of deadlock detected errors). The deadlocked batches fell back to row-by-row inserts that skipped issue detection.
• Issue race: concurrent tailers raced on issues_fingerprint_unique and lost occurrence counts.
• Fix: sorted, bounded batches; a retry on deadlock; issue creation and occurrence recording in one transaction; and logging of the real database error, which nestjs-pino was dropping.

I verified it against a real Postgres 17. The old insert pattern deadlocked on about 670 of 800 concurrent statements, the new one on none, and issue counts stayed exact under concurrency. It's been running on my instance with zero deadlocks since. Details are in the PR.

[vbrzezina]

Update on the open PRs. Every review comment so far has been addressed, and I replied on each PR with what changed. Everything is rebased on current master, and each PR is mergeable.

Integrations

• feat(providers): add Seerr integration (Overseerr/Jellyseerr compatible) #59 Seerr/Overseerr and feat(providers): add Readarr integration #61 Readarr: review feedback addressed and pushed.
• feat(providers): add Bazarr integration #58 Bazarr: the history paging was reworked so that a capped poll always makes progress, and day-first dates are now read correctly.
• feat(providers): add qBittorrent integration with username/password auth #62 qBittorrent: login accepts qBittorrent 5.2's 204 response, API keys are sent as a Bearer token, and server responses no longer return the stored API key or password.

Fixes and shared pieces

• fix(ingestion): prevent log_entries deadlocks and issue-creation races #65: fixes log_entries deadlocks and an issue-creation race in file ingestion.
• fix(ingestion): keep synthetic activity timestamps out of the sync cursor #67: keeps synthetic timestamps (health checks, queue rows) out of the sync cursor. The Bazarr, qBittorrent and Readarr PRs build on it, so I'd suggest merging it first.
• chore(provider-arr): address #57 review follow-ups #66: the non-blocking follow-ups from the fix: *arr log file ingestion emits no entries (timestamp regex + context parser contract) #57 review.

Suggested order: #67 and #65 first, then the integrations in any order. I'm happy to adjust anything you'd rather have differently.