fix(swagger): respect base URL for API docs routes

[s0up4200]

Summary

• Fixes swagger docs not loading when using a custom base URL (e.g., /qui/ on seedboxes)
• Routes were hardcoded at /api/docs and /api/openapi.json, ignoring the configured baseURL
• Now prefixes routes with baseURL so /qui/api/docs works correctly

Test plan

• Verify /api/docs still works with default config (no baseURL)
• Verify /qui/api/docs loads correctly when baseUrl: "/qui/" is configured

Summary by CodeRabbit

• Chores
  • Refactored API documentation routes to use a normalized base path, ensuring consistent doc and OpenAPI endpoints.
  • Adjusted development server routing patterns to more accurately match API and proxy paths, improving request routing during local development.

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

Walkthrough

Routes for Swagger UI and OpenAPI now use a normalized base URL (via httphelpers.NormalizeBasePath) when registered, and the Vite navigateFallback denylist regex for API routes was adjusted to match trailing-slash or end-of-path cases and its order was swapped.

Changes

Cohort / File(s)	Summary
Swagger base‑URL normalization & route registration internal/web/swagger/swagger.go	Added httphelpers import and replaced manual trimming with httphelpers.NormalizeBasePath to normalize baseURL (ensures leading slash, no trailing slash, or "" for root). RegisterRoutes now prepends the normalized base URL to both /api/docs and /api/openapi.json endpoints.
Vite navigateFallback denylist regex web/vite.config.ts	Updated navigateFallbackDenylist by changing the first regex from /^\/api/ to `//api(?:/

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

• Verify httphelpers.NormalizeBasePath behavior for root ("") and nested base paths and ensure both Swagger UI and OpenAPI endpoints resolve correctly.
• Confirm no routing conflicts arise when the normalized base URL is empty or /.
• Validate the updated Vite regex matches intended API routes and doesn't unintentionally exclude valid client routes.

Poem

🐰 I hopped through code with careful paws,
I fixed the slashes, tweaked the laws.
Swagger now walks with baseURL pride,
Vite’s denylist hops in stride.
A little rabbit’s routing guide. 🥕

Pre-merge checks and finishing touches

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 37.50% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The pull request title directly summarizes the main change: making swagger API docs routes respect the configured base URL.

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix/swagger-base-url

---

📜 Recent review details

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 0fe48ae and d0dadb9.

📒 Files selected for processing (1)

• internal/web/swagger/swagger.go (3 hunks)

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

• internal/web/swagger/swagger.go

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)

• GitHub Check: Run tests

---

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

[coderabbitai]

Actionable comments posted: 6

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (6)

internal/web/swagger/openapi.yaml (1)

3743-3750: OpenAPI nullable: true for indexerName doesn’t match the Go shape (string + omitempty).
If the API never returns null (only omits or returns a string), drop nullable: true. Otherwise, make the Go field *string so null is representable.

internal/models/crossseed.go (1)

160-170: IndexerName field has inconsistent null/omit/required representation across Go, OpenAPI, and TypeScript types.

OpenAPI spec defines indexerName as nullable: true, but:

• Go struct uses omitempty (field is omitted if empty, not sent as null)
• TypeScript interface treats it as a required field (no ? optional marker)

Align the three to use consistent representation: either all nullable, all optional, or all required.

internal/services/crossseed/service.go (4)

1619-1666: Use the resolved sourceIndexer for IndexerName (currently uses result.Indexer). You already map result.IndexerID → friendly name into sourceIndexer (Lines 1620-1623), but the run results here still use the raw result.Indexer, which can diverge from the resolved name.

 run.Results = append(run.Results, models.CrossSeedRunResult{
   InstanceName: result.Indexer,
-  IndexerName:  result.Indexer,
+  IndexerName:  sourceIndexer,
   Success:      false,
   Status:       "no_match",
   Message:      fmt.Sprintf("No matching torrents for %s", result.Title),
 })

 run.Results = append(run.Results, models.CrossSeedRunResult{
   InstanceName: result.Indexer,
-  IndexerName:  result.Indexer,
+  IndexerName:  sourceIndexer,
   Success:      true,
   Status:       "dry-run",
   Message:      fmt.Sprintf("Dry run: %d viable candidates", candidateCount),
 })

---

1694-1723: Same IndexerName inconsistency in “exists” pre-check results. These blocks also set IndexerName: result.Indexer but should likely use sourceIndexer for consistency with the resolved indexer name.

 existingResults = append(existingResults, models.CrossSeedRunResult{
   InstanceID:   candidate.InstanceID,
   InstanceName: candidate.InstanceName,
-  IndexerName:  result.Indexer,
+  IndexerName:  sourceIndexer,
   Success:      false,
   Status:       "exists",
   Message:      "Torrent already exists (infohash pre-check)",
   ...
 })

 existingResults = append(existingResults, models.CrossSeedRunResult{
   InstanceID:   candidate.InstanceID,
   InstanceName: candidate.InstanceName,
-  IndexerName:  result.Indexer,
+  IndexerName:  sourceIndexer,
   Success:      false,
   Status:       "exists",
   Message:      "Torrent already exists (comment URL pre-check)",
   ...
 })

Also applies to: 1752-1780

---

1833-1893: Propagate IndexerName consistently for “no_result” and per-instance mapped results. Same issue: IndexerName uses result.Indexer instead of sourceIndexer.

 run.Results = append(run.Results, models.CrossSeedRunResult{
   InstanceName: result.Indexer,
-  IndexerName:  result.Indexer,
+  IndexerName:  sourceIndexer,
   Success:      false,
   Status:       "no_result",
   Message:      fmt.Sprintf("Cross-seed returned no actionable instances for %s", result.Title),
 })

 mapped := models.CrossSeedRunResult{
   InstanceID:   instanceResult.InstanceID,
   InstanceName: instanceResult.InstanceName,
-  IndexerName:  result.Indexer,
+  IndexerName:  sourceIndexer,
   Success:      instanceResult.Success,
   Status:       instanceResult.Status,
   Message:      instanceResult.Message,
 }

---

6521-6545: Tag matching uses exact case-sensitive comparison, creating a breaking change if configs historically used different casing. The code at lines 6521-6545 and 6547-6554 matches tags with exact equality (tag == allowed, tag == excluded, tag == "cross-seed") rather than case-insensitive comparison. Other services in the codebase (reannounce, trackerrules) consistently use strings.EqualFold for tag matching, but crossseed does not. Users with tags like "Cross-Seed" or "TAG1" will silently fail to match, since splitTags does not normalize casing before comparison. Confirm this breaking behavior is intentional or apply case-insensitive matching to align with existing patterns.

🧹 Nitpick comments (7)

web/src/types/index.ts (1)

1404-1417: Consider backward-compat: required rssSource* fields can break older API responses.
If the UI can connect to older qui versions, make these optional (or ensure the fetch/normalization layer defaults missing fields to []). If server/UI are always deployed together, current typing is fine.

Also applies to: 1437-1450

internal/api/handlers/crossseed.go (1)

198-210: Patch application for rssSource* is straightforward and correct.
Optional follow-up: consider normalizing (trim/dedupe) before persisting if you want consistent matching behavior.

internal/models/crossseed.go (1)

415-431: Store write-path wiring for rss_source_* looks consistent (columns/placeholders/args).
Optional: normalize (trim/dedupe) RSSSource* before encoding to keep stored settings canonical (there’s already normalizeStringSlice in this file).

Also applies to: 451-494, 507-535

internal/services/crossseed/service.go (1)

1919-2002: RSS source filtering integration looks good; consider normalizing filters to avoid whitespace surprises. splitTags() trims torrent tags, but settings-provided tags/categories may include whitespace and won’t match as expected. A lightweight fix is to normalize the settings slices once when hasRSSSourceFilters is true (either here or at settings-validation time).

web/src/pages/CrossSeedPage.tsx (3)

384-395: Normalize rssSource lists before persisting (avoid dupes / whitespace variants).*
Search filters are typically “set-like”; saving duplicates can lead to confusing UI and non-deterministic backend behavior.

       setAutomationForm({
         enabled: settings.enabled,
         runIntervalMinutes: settings.runIntervalMinutes,
         targetInstanceIds: settings.targetInstanceIds,
         targetIndexerIds: settings.targetIndexerIds,
-        rssSourceCategories: settings.rssSourceCategories ?? [],
-        rssSourceTags: settings.rssSourceTags ?? [],
-        rssSourceExcludeCategories: settings.rssSourceExcludeCategories ?? [],
-        rssSourceExcludeTags: settings.rssSourceExcludeTags ?? [],
+        rssSourceCategories: normalizeStringList(settings.rssSourceCategories ?? []),
+        rssSourceTags: normalizeStringList(settings.rssSourceTags ?? []),
+        rssSourceExcludeCategories: normalizeStringList(settings.rssSourceExcludeCategories ?? []),
+        rssSourceExcludeTags: normalizeStringList(settings.rssSourceExcludeTags ?? []),
       })

     return {
       enabled: automationSource.enabled,
       runIntervalMinutes: automationSource.runIntervalMinutes,
       targetInstanceIds: automationSource.targetInstanceIds,
       targetIndexerIds: automationSource.targetIndexerIds,
-      rssSourceCategories: automationSource.rssSourceCategories,
-      rssSourceTags: automationSource.rssSourceTags,
-      rssSourceExcludeCategories: automationSource.rssSourceExcludeCategories,
-      rssSourceExcludeTags: automationSource.rssSourceExcludeTags,
+      rssSourceCategories: normalizeStringList(automationSource.rssSourceCategories),
+      rssSourceTags: normalizeStringList(automationSource.rssSourceTags),
+      rssSourceExcludeCategories: normalizeStringList(automationSource.rssSourceExcludeCategories),
+      rssSourceExcludeTags: normalizeStringList(automationSource.rssSourceExcludeTags),
     }

Also applies to: 482-507

---

835-874: Select option derivation looks good; consider guarding against include/exclude overlap.
Right now the UI can select the same category/tag in both include and exclude lists; if backend precedence is “exclude wins”, consider enforcing that in UI (or surfacing a warning).

---

989-998: Defensive reduce for stats (avoid NaN if backend ever sends null/undefined).

 return {
-  totalAdded: runs.reduce((sum, run) => sum + run.torrentsAdded, 0),
-  totalFailed: runs.reduce((sum, run) => sum + run.torrentsFailed, 0),
+  totalAdded: runs.reduce((sum, run) => sum + (run.torrentsAdded ?? 0), 0),
+  totalFailed: runs.reduce((sum, run) => sum + (run.torrentsFailed ?? 0), 0),
   totalRuns: runs.length,
 }

📜 Review details

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between d44058f and 3d2f141.

📒 Files selected for processing (8)

• internal/api/handlers/crossseed.go (3 hunks)
• internal/database/migrations/032_add_rss_source_filters.sql (1 hunks)
• internal/models/crossseed.go (11 hunks)
• internal/services/crossseed/service.go (15 hunks)
• internal/web/swagger/openapi.yaml (1 hunks)
• internal/web/swagger/swagger.go (1 hunks)
• web/src/pages/CrossSeedPage.tsx (9 hunks)
• web/src/types/index.ts (3 hunks)

🧰 Additional context used

🧠 Learnings (3)

📚 Learning: 2025-11-28T20:32:30.126Z

Learnt from: s0up4200
Repo: autobrr/qui PR: 641
File: internal/services/crossseed/service.go:209-212
Timestamp: 2025-11-28T20:32:30.126Z
Learning: Repo: autobrr/qui PR: 641
File: internal/services/crossseed/service.go
Learning: The cross-seed recheck-resume worker intentionally runs for the process lifetime and keys pending entries by hash only. This is acceptable under the current constraint that background seeded-search runs operate on a single instance at a time; graceful shutdown and instanceID|hash keying are deferred by design.

Applied to files:

• internal/models/crossseed.go
• internal/services/crossseed/service.go

📚 Learning: 2025-11-28T22:21:20.730Z

Learnt from: s0up4200
Repo: autobrr/qui PR: 641
File: internal/services/crossseed/service.go:2415-2457
Timestamp: 2025-11-28T22:21:20.730Z
Learning: Repo: autobrr/qui PR: 641
File: internal/services/crossseed/service.go
Learning: The determineSavePath function intentionally includes a contentLayout string parameter for future/content-layout branching and API consistency. Its presence is by design even if unused in the current body; do not flag as an issue in reviews.

Applied to files:

• internal/models/crossseed.go
• internal/services/crossseed/service.go

📚 Learning: 2025-11-27T15:33:39.007Z

Learnt from: s0up4200
Repo: autobrr/qui PR: 641
File: internal/services/jackett/service.go:309-330
Timestamp: 2025-11-27T15:33:39.007Z
Learning: In internal/services/jackett/service.go, synchronous execution paths (test executors, interactive searches with explicit indexer selection, and scheduler-unavailable fallback) intentionally pass jobID=0 to result callbacks. This is acceptable because outcome tracking via ReportIndexerOutcome is only used for async cross-seed operations that always use the scheduler and receive unique jobIDs. Interactive searches and tests do not use outcome tracking.

Applied to files:

• internal/services/crossseed/service.go

🧬 Code graph analysis (2)

internal/models/crossseed.go (1)

web/src/types/index.ts (2)

• CrossSeedAutomationSettings (1404-1435)
• Category (427-430)

internal/services/crossseed/service.go (3)

pkg/prowlarr/client.go (1)

• Indexer (76-84)

internal/models/crossseed.go (1)

• CrossSeedAutomationSettings (20-64)

web/src/types/index.ts (3)

• CrossSeedAutomationSettings (1404-1435)
• Torrent (226-283)
• Category (427-430)

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)

• GitHub Check: Run tests

🔇 Additional comments (10)

internal/database/migrations/032_add_rss_source_filters.sql (1)

1-8: LGTM: safe defaults for new filter columns.
Default '[]' is a good compatibility choice for existing rows and for JSON-slice decoding.

web/src/types/index.ts (1)

1375-1384: indexerName?: string aligns with Go omitempty.
No concerns here.

internal/api/handlers/crossseed.go (1)

135-162: isEmpty() correctly accounts for the new RSS source filter fields.

internal/models/crossseed.go (3)

31-36: Good: rssSource fields are explicit and documented on the settings model.*

---

76-92: Good defaults: empty slices avoid nulls and simplify client logic.

---

273-314: Store read-path wiring for rss_source_* looks correct.
Given the migration defaults to '[]', decode should be stable for both new and existing rows.

Also applies to: 357-369

internal/services/crossseed/service.go (2)

1539-1542: Signature update usage looks correct. buildAutomationSnapshots is now called with settings, enabling RSS source filtering during snapshot construction.

---

6434-6463: The original review comment is incorrect. Tags are already normalized upstream before reaching matchesSearchFilters. At line 1374 in validateSearchRunOptions, opts.Tags = normalizeStringSlice(opts.Tags) trims whitespace and removes duplicates from all input tags. This normalization happens during validation (line 1172) before state is created at line 1235, so the concern about spacing mismatches silently filtering torrents is unfounded. Case sensitivity is by design, as noted in comments ("case-sensitive to match qBittorrent behavior"), not a bug.

web/src/pages/CrossSeedPage.tsx (2)

39-66: Nice, cohesive wiring for the new RSS source filter fields (types + icons used later).
The AutomationFormState additions look consistent with the rest of the form model.

---

104-113: Defaults correctly include the new rssSource arrays.*