fix(crossseed): drop resolution token from TV search queries

[nitrobass24]

Summary

TV cross-seed searches were silently failing on indexers whose free-text search matches only the series name (notably BTN and IPT): the search returned 0 results, so cross-seeds to those trackers were never created — while the same release cross-seeded fine elsewhere.

Reported on Discord by multiple users. The tell-tale: a manual Prowlarr search for tvshow S01E01 returns results on BTN, but tvshow S01E01 1080 returns zero.

Root cause

For TV content, qui appended a bare resolution token to the free-text Torznab query, producing q = "<title> S01E01 1080". Indexers that match q against the series name find nothing for the 1080 token and return zero results. (Sonarr works because it issues a structured tvsearch with IDs + season/ep and no resolution in q; RSS automation works because it matches the full feed title with no q search at all.)

The resolution was injected at two sites, both added in #1871 ("refine TV based searching", which also raised the result limit to 100 + ReturnAllResults):

1. The cross-seed query builder — IncludeResolution for TV (internal/services/crossseed/search_query.go).
2. The jackett Prowlarr ID-driven path — idDrivenTVQuery re-extracted the resolution from the release name even when the query itself was clean (internal/services/jackett/service.go).

Fix

Drop the resolution token from the TV q at both sites:

• Remove the SearchQueryOptions{IncludeResolution} apparatus from the cross-seed query builder.
• ID-driven TV searches now use the bare S01E01 token; removed the now-dead idDrivenTVQuery / firstResolutionToken / searchResolutionToken helpers.

Kept unchanged: the S01E01 season/episode narrowing, the generic appendSearchToken season-token repositioning, and the 100-result / ReturnAllResults behavior.

Resulting query: "<title> S01E01" (non-ID) / "S01E01" (ID-driven) — the form confirmed to return results on BTN.

Why this is safe

Resolution in q was only ever a result-narrowing heuristic, never required for correctness: internal/services/crossseed/matching.go still strictly rejects resolution mismatches after the search (1080p ≠ 2160p are different files). Dropping the token changes which candidates an indexer returns, not which ones we accept.

Testing

• make test — full -race -count=1 ./...: pass (0 failures / 0 races).
• Targeted internal/services/crossseed + internal/services/jackett pass; updated the unit tests that asserted the old … 1080 / S17 1080p / S22 720 queries.
• make precommit (lint): 0 errors. make build: frontend + Go binary compile.
• No schema/DB changes (Postgres / migration CI jobs unaffected).
• Pending: live confirmation against real BTN/IPT by the reporting users.

Notes

Net −75 lines. One residual tradeoff (what #1871 was guarding against): without resolution narrowing, a very popular single episode could return more candidates and, worst case, risk truncation at the 100-result cap — small for episode-level searches. If it ever bites, a 0-result fallback retry (search with resolution, retry the indexer without it) would recover the narrowing without breaking series-name-only indexers; deliberately left out here to keep the change minimal.

Summary by CodeRabbit

• Bug Fixes
  • Removed automatic resolution tokens from generated search queries (including ID-driven TV searches), producing cleaner, more consistent searches and avoiding duplicated resolution tags.
• Tests
  • Updated and extended tests to reflect the new query behavior and added a fallback to ensure queries are never empty.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 41eed64e-f43a-4330-abff-66ec7a993bee

📥 Commits

Reviewing files that changed from the base of the PR and between b2db987 and 1ff6146.

📒 Files selected for processing (2)

• internal/services/crossseed/search_query.go
• internal/services/crossseed/search_query_test.go

🚧 Files skipped from review as they are similar to previous changes (1)

• internal/services/crossseed/search_query_test.go

---

Walkthrough

The PR removes resolution-token handling from CrossSeed and Jackett search query generation. buildSafeSearchQuery no longer accepts options and always uses trimmed/cleaned titles or fallbacks for the query. Jackett’s ID-driven Prowlarr TV workaround now sets q to the season token only and removes season/ep params.

Changes

Resolution token removal from search query builders

Layer / File(s)	Summary
CrossSeed query builder simplification internal/services/crossseed/search_query.go, internal/services/crossseed/service.go, internal/services/crossseed/source_release_inference_test.go	Deletes exported SearchQueryOptions and resolution-token handling; buildSafeSearchQuery(name, release, baseQuery) now sets SearchQuery.Query from trimmed baseQuery or cleaned title with fallbacks; call sites updated to the new signature.
CrossSeed tests and callsite updates internal/services/crossseed/search_query_test.go	Updates tests to remove per-case SearchQueryOptions, asserts parsed titles remain unmodified (no appended resolution), and adds a fallback test for empty cleaned titles.
Jackett Prowlarr TV query simplification internal/services/jackett/service.go, internal/services/jackett/service_test.go	Removes idDrivenTVQuery and firstResolutionToken helpers and one resolution regex; applyProwlarrTVTokenWorkaround now sets q to the SxxExx token only when ID params are present and prunes season/ep; tests updated to expect season-only q.

Sequence Diagram(s)

sequenceDiagram
  participant Caller
  participant buildSafeSearchQuery
  participant SearchQuery
  Caller->>buildSafeSearchQuery: name, release, baseQuery
  alt baseQuery not empty
    buildSafeSearchQuery->>SearchQuery: Query = TrimSpace(baseQuery)
  else is movie
    buildSafeSearchQuery->>SearchQuery: Query = cleanedTitle
  else other
    alt cleanedTitle empty
      buildSafeSearchQuery->>SearchQuery: Query = TrimSpace(baseQuery) or TrimSpace(name)
    else
      buildSafeSearchQuery->>SearchQuery: Query = cleanedTitle
    end
  end
  buildSafeSearchQuery-->>Caller: SearchQuery (no resolution token)

Loading

sequenceDiagram
  participant Workaround
  participant applyProwlarrTVTokenWorkaround
  participant Request
  Workaround->>applyProwlarrTVTokenWorkaround: ID params present (imdbid/tvdbid/tmdbid/tvmazeid)
  applyProwlarrTVTokenWorkaround->>Request: Set q = TV token only (SxxExx)
  applyProwlarrTVTokenWorkaround->>Request: Remove season and ep params
  applyProwlarrTVTokenWorkaround-->>Workaround: Updated request (no resolution appended)

Loading

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Suggested labels

bugfix, cross-seed, backend, tests, area/backend

Suggested reviewers

• s0up4200
• Audionut

Poem

🐰 Resolution tokens fade and flee,
Queries lean, concise, and free,
CrossSeed trims and Jackett prunes,
Season tokens hum empty tunes,
A rabbit cheers the cleaner spree!

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 14.29% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix(crossseed): drop resolution token from TV search queries' directly describes the main change: removing resolution tokens from TV search queries in the crossseed service.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/crossseed-tv-resolution-query

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@internal/services/crossseed/search_query.go`:
- Around line 67-69: The current fallback for cleanedTitle only reassigns from
baseQuery which is already empty in this branch; change the fallback logic in
the cleanedTitle handling so that if cleanedTitle is empty after
cleanAnimeTitle(baseQuery) you fall back to the original name (trimmed) — e.g.,
in the block referencing cleanedTitle, baseQuery and name ensure you first try
baseQuery, and if that yields an empty string then set cleanedTitle =
strings.TrimSpace(name) so you return a conservative non-empty Torznab query.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: d466ee21-a990-4d0c-9ae4-3076a1c9cf15

📥 Commits

Reviewing files that changed from the base of the PR and between 8e8b6d9 and b2db987.

📒 Files selected for processing (6)

• internal/services/crossseed/search_query.go
• internal/services/crossseed/search_query_test.go
• internal/services/crossseed/service.go
• internal/services/crossseed/source_release_inference_test.go
• internal/services/jackett/service.go
• internal/services/jackett/service_test.go

[Audionut]

Bah, I always have that as a source, so never noticed that resolution was a blocker.

We're trading better matching at poorly curated sites, and probably the occasional super long running shows, against working capacity for the listed indexers, a no brainier.

Ran the bot over this, and it's only concern, funnily enough, was the risk of missing content because of the underlying constraints of jackett/prowlarr.

LGTM