Plex attaches TMDB GUIDs that identify the individual episode, not the parent series, so using them as `tmdbId` caused incorrect series lookups. Now only movie scrobbles populate `tmdbId` from the TMDB GUID; episode scrobbles rely on IMDB/TVDB IDs (and the title) to resolve the show via `resolveShowTmdbId`.
- Extract IMDB/TMDB/TVDB prefix constants to avoid magic strings
- Add `parsePlexPayload` test asserting `tmdbId` is `undefined` for episode events
- Add `processWebhook` integration test verifying `resolveShowTmdbId` is called without a TMDB ID for Plex episodes
- Hoist `makePlexForm` helper to module scope so it's available across all describe blocks
* feat: reorganize API around operation domains
Restructure the oRPC contract from 14 resource-oriented routers to 8
domain-oriented routers for a cleaner public API surface.
- Consolidate 7 watch procedures into unified tracking.watch/unwatch
with scope + ids input (movie, episode, season, series)
- Split dashboard across tracking (stats, history), library
(continueWatching, upcoming), and discover (recommendations)
- Merge explore + search + discover into single discover router
- Absorb integrations into account.integrations
- Merge system.authConfig into system.publicInfo
- Collapse 6 admin setting endpoints into admin.settings.get/update
- Deduplicate platforms.list + explore.watchProviders into
discover.platforms
- Add symmetric unwatchMovie/unwatchSeries core functions
- Rename titles.detail→get, titles.recommendations→similar,
people.detail→get
BREAKING CHANGE: All client API paths have changed. REST paths now
mirror router structure.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address review feedback on API reorganization
- Fix updateRating invalidation in native app — was only invalidating
title queries, now calls invalidateTitleQueries() to also refresh
tracking.userInfo (drives the rating UI)
- Make unwatchMovie status revert consistent with unwatchSeries — revert
any non-watchlist status, not just "completed"
- Add sync invariant comment on handleWatch/handleUnwatch loops
- Remove unused queryClient import in native use-title-actions
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: updateStatus NOT_FOUND error + rate() invalidation in native
- updateStatus now throws NOT_FOUND when quickAddTitle returns null
(title doesn't exist), instead of silently succeeding
- Add NOT_FOUND error to updateStatus contract definition
- Fix rate() in native title-actions.ts to call invalidateTitleQueries()
instead of only invalidating orpc.titles.key() — mirrors the fix
already applied to the hook-based path in d2ddc0f
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* docs: expand mobile app docs and add Play Store badge
- Add the Google Play badge to the README alongside the App Store badge
- Expand the mobile app docs with the Android/Play Store release details
- Refresh docs site dependencies and related package versions for the updated docs stack
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace `tmdbProviderId` (single int) with a many-to-many `platformTmdbProviders` join table so one platform entry can map to multiple TMDB provider IDs (e.g. a single "Max" platform covers several regional TMDB IDs)
- Add `isSubscription` flag to platforms to distinguish subscription services from transactional ones, replacing the old `displayOrder` field in API responses
- Add `getPlatformTmdbIds` / `getPlatformTmdbIdMap` helpers in `@sofa/core/platforms` and thread them through discover, explore, and platform list procedures
- Replace `providerId: number` with `platformId: string` (UUID) in discover input schema so the client never needs to resolve TMDB IDs
- Add `scripts/sync-tmdb-providers.ts` script to pull live provider data from TMDB and update `platforms.json`
- Split native title detail availability into separate "Stream" and "Buy or Rent" sections matching the web pattern
- Add two new DB migrations for the join table and platform schema changes
* feat: add filtering and sorting across library, explore, and upcoming
Add comprehensive filtering and sorting to all list views on both web and mobile.
Library (new dedicated page):
- New /library route (web) with URL search param persistence
- New Library tab (5th tab, mobile) with FlashList grid
- Filters: status (multi-select), type, genre (library-only), user rating range,
release year (decade presets + custom), content rating, available to stream
- Sort by: title, date added, release date, popularity, user rating, TMDB rating
- Text search within collection
- Filter popover (web) / bottom sheet with Apply (mobile)
- Remove old dashboard.library endpoint in favor of library.list
Explore/Discover:
- New Discover section on Explore page with inline filter controls
- TMDB filters: genre (now optional), year range, min rating, sort order,
original language, streaming provider
- Watch provider list endpoint (WATCH_REGION env var, default "US")
- Default to popular content when no filters selected
Upcoming:
- Type filter (All/Movies/TV Shows) and status filter (All/Watching/Watchlist)
- Toggle chips on both web (URL params) and mobile
- Backend: mediaType and statusFilter params on upcoming endpoint
Navigation:
- Library added to web nav bar (desktop + mobile tab bar)
- Library added as 5th tab on mobile (between Home and Explore)
- Dashboard library section simplified to compact preview with "See all" link
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add user streaming platform subscriptions and redesign filter UI
Replace the denormalized availabilityOffers table with a normalized
platforms + titleAvailability + userPlatforms schema. Users can now
declare which streaming services they subscribe to, enabling
personalized "On your services" availability display and filtered
library/explore results.
Backend: new platforms table seeded from TMDB provider data on startup,
new API endpoints (platforms.list, account.platforms, account.updatePlatforms),
title detail annotates availability with isUserSubscribed flag, library
"on my services" filter checks user's subscribed platforms.
Web UI: post-registration onboarding page for platform selection,
streaming services settings section, title detail split into "On your
services" / "Also available on", explore dropdown uses platforms table
with active filter highlighting. Library filters redesigned from
cluttered popover to clean collapsible inline strip with consolidated
dropdowns.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address race conditions, stale state, and edge cases in platforms feature
- Fix debounced autosave race in streaming services settings: use ref
for latest selection + counter guard so only the most recent save
updates UI. Clean up timers on unmount.
- Fix explore provider dropdown desyncing from query filter when
switching Movie/TV type: reset selectedPlatformId alongside providerId.
- Fix platformIdsExist rejecting duplicate valid IDs by deduplicating
input before comparing against DB results.
- Fix stale platform metadata: always upsert name/logo on TMDB refresh
instead of skipping when platform already exists.
- Fix invisible ratingMax filter: clear ratingMax when user changes
ratingMin since the UI only exposes a single rating dropdown now.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: clear stale saved-state timer and preserve falsy filter values
- Clear previous savedTimerRef timeout before scheduling a new one in
streaming-services-section, preventing an older timer from hiding the
"Saved" indicator too early on rapid successive saves
- Replace `value || undefined` with explicit empty-value check in
library filter handler so numeric 0 (e.g. ratingMin=0) is not
incorrectly dropped
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add TanStack Form and migrate auth + change password forms
Set up TanStack Form v1 composition layer with createFormHook, pre-bound
field components (TextField, TextareaField, CheckboxField, SwitchField,
SubmitButton), and migrate the two traditional submit-based forms:
- AuthForm: replace 4 useState calls with useAppForm, use form.Field
for each input while preserving motion animation layout
- ChangePasswordDialog: replace 6 useState calls with useAppForm + Zod
schema validation (cross-field password match, min length), per-field
error display, form.reset() on dialog close
Also adds @tanstack/react-form to the monorepo catalog and updates both
web and native apps to reference it via catalog:.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Skip recommendations refresh for titles fetched within the last 7 days via `getTitlesWithFreshRecommendations`; only stale titles are re-fetched
- Replace `Promise.allSettled` in `refreshTvChildren` with a `mapWithConcurrency` helper (concurrency = 5) to stay within TMDB's rate limit
- Add 30s request timeout middleware to TMDB client to prevent indefinitely hung fetches
- Expand `optimizeDb` weekly job to prune 30-day-old cron runs, integration events, expired sessions, and expired verifications
- Add `wal_checkpoint(TRUNCATE)` to `optimizeDatabase` so the WAL is flushed before `PRAGMA optimize`
- Add `recoverStaleImportJobs` called on server startup to mark any `pending`/`running` import jobs as errored after a crash/restart
- Return `false` from `triggerJob` immediately if the job is already running
- Add `deleteOrphanedImage(category, path)` helper to `image-cache.ts` for fire-and-forget unlink of stale cached files
- Extend `purgeShellTitlesTransaction` to collect poster, backdrop, still, and profile image paths from deleted titles/persons before cascade deletes, returning them as `orphanedImages` in `PurgeResult`
- Wire orphaned image cleanup into `purgeMetadataCache` (cache purge), `batchUpsertPersons` (profile path changes), `updateTitleWithArtInvalidation` (poster/backdrop changes), and `refreshTvChildren` (season posters and episode stills)
- Remove `getTitlesWithMissingThumbhashes`, `getTitleIdsWithMissingSeasonThumbhashes`, `getTitleIdsWithMissingEpisodeThumbhashes`, and `getTitleIdsWithMissingProfileThumbhashes` from `@sofa/db/queries/cron` — `getThumbhashBackfillTitleIds` now delegates to `getLibraryTitleIds` only
- Fix episode still cleanup to call `deleteOrphanedImage` even when `stillThumbHash` is null (previously skipped)
- Add `@sofa/tsconfig` package with a shared `base.json` (target, lib, strict, moduleResolution, etc.) and migrate all packages/apps to `"extends": "@sofa/tsconfig/base.json"` to eliminate duplicated compiler options
- Move `display-status.ts` (and its `DisplayStatus` type + `getDisplayStatus` fn) from `@sofa/api` to `@sofa/core`; remove the `./display-status` export from `@sofa/api` and update all import sites in `discovery.ts` and `tracking.ts`
- Add `packages/core/test/display-status.test.ts`: 15 tests covering `ONGOING_TMDB_STATUSES` and all `getDisplayStatus` branches (movie, TV watchlist/in_progress/caught_up/completed, edge cases)
- Add `packages/i18n/vitest.config.ts` and two new test files: `date-buckets.test.ts` and `format.test.ts` covering date bucketing and locale formatting utilities
- Bump `@tanstack/react-query` and related packages to 5.95.0
- Add `@orpc/json-schema`, `@orpc/openapi`, `@orpc/server`, `@orpc/zod` to catalog and use catalog refs in `apps/server`
- Move `ORPCError` throws for backup not-found/invalid into `@sofa/core/backup` so the server procedure can simply `await deleteBackup()` without re-wrapping errors
- Add `@orpc/server` dep to `packages/core` to support the above
- Add `expo-updates` to native deps; switch `requireNativeModule` import to `expo` in `WidgetImagesModule`
- Pin native `react`/`react-dom` to catalog to avoid duplicate copies
- Add `@lingui/core` to web and `react` to i18n package deps
- Replace root `eslint-plugin-lingui` jsPlugin with `@e18e/eslint-plugin`; move lingui rules into per-app `.oxlintrc.json` overrides for native and web
- Add `jsx-a11y`, `react-hooks-js` (native), and expanded lingui rule sets to per-app configs
- Add `oxc/no-barrel-file` warning at threshold 0 to root config
- Fix `e18e/prefer-url-canparse`: replace `try { new URL() } catch` with `URL.canParse()` in server-url screen and server lib
- Fix `e18e/prefer-timer-args`: pass callback args directly to `setTimeout` instead of wrapping in arrow functions (use-debounce, integration-card)
- Fix `e18e/prefer-static-regex`: hoist `/\/+$/` to module-level constant in server-url screen
- Fix `react-hooks-js/refs` and `react-hooks-js/set-state-in-effect`: convert `useRef` tracking patterns to `useState` + render-time derived updates in `use-server-connection`, `expandable-text`, and settings screen
- Fix `react-hooks-js/immutability`: extract stable palette sub-values before `useMemo` deps in title detail screen
- Apply `e18e` and other rule fixes across core, tmdb, web components, and i18n packages
- New integrations.test.ts: 13 tests covering CRUD, token generation,
token lookup, regeneration, and multi-user isolation
- New cache.test.ts: 3 tests for purgeMetadataCache (shell title
deletion, fully-fetched title preservation, empty case)
- New cron.test.ts: 5 tests for cron run lifecycle (start, complete,
fail with Error and string)
- New telemetry.test.ts: 7 tests for isTelemetryEnabled toggle and
performTelemetryReport (disabled skip, enabled send, 24h interval
throttle, interval expiry, fetch failure resilience)
345 → 386 tests passing across 27 files
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace bun:test with Vitest across all 22 test files
- Create @sofa/test package with shared setup and DB test helpers
- setup.ts: vi.mock for @sofa/db/client, Bun.randomUUIDv7 polyfill
- db.ts: in-memory SQLite via better-sqlite3, seed helpers
- Add per-project vitest configs (packages/core, apps/web)
- Add root vitest.config.ts with projects and v8 coverage
- Set up @vitest/browser + Playwright for web component tests
- Move validateBackupDatabase from core/backup.ts to @sofa/db/client
- Derive required backup tables from schema instead of hard-coded list
- Update CI workflow with Playwright install and Codecov upload
- Update CLAUDE.md documentation for Vitest
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix `refreshCreditsJob` to treat `null` `lastFetchedAt` as needing refresh (not just missing cast entry)
- Fix `buildBackupCron` 6h expression to correctly fire every 6 hours using explicit hour list instead of `*/6`
- Fix `vacuumDatabase` to pass the path directly as a bound parameter (the manual escape was redundant and incorrect)
- Clean up orphaned `.tmp` files when image cache writes fail in both `downloadAndCacheImage` and `fetchAndMaybeCache`
- Surface import job payload parse/validation errors as job-level `error` status instead of throwing unhandled exceptions; check for cancellation before the processing loop starts
- Replace bare `Error` throws in `readImportJob` with `ORPCError` (`NOT_FOUND` / `FORBIDDEN`) so oRPC maps them to correct HTTP status codes
- Auto-promote title status to `in_progress` when `logEpisodeWatchBatch` is called for titles that have no status or are only on the watchlist
- Refresh seasons from TMDB when a webhook references a season not yet in the DB (handles newly-released seasons)
- Upgrade fire-and-forget enrichment failures and webhook processing errors from `log.debug` → `log.warn` so they surface in production logs
- Extend stale-job auto-cancel to cover `running` jobs (30 min threshold) in addition to `pending` jobs (5 min), so crashed mid-import workers don't permanently block new imports
- Use `startedAt` instead of `createdAt` when calculating age for running jobs
- Scope `readImportJob` in the SSE loop to the authenticated user's ID to prevent cross-user job reads
- Persist progress counters (`processedItems`, `importedCount`, `skippedCount`, `failedCount`) when a job is cancelled mid-run so the UI shows accurate final stats
- Add `ContinueWatching` (systemSmall) and `Upcoming` (systemSmall) SwiftUI widgets via `expo-widgets`, registered in `app.json` under the `group.com.jakejarvis.sofa` app group
- Add `widget-images` local Expo module with a Swift `WidgetImagesModule` that downloads and resizes remote poster images into the shared app group container for use by widgets; also copies bundled assets (sofa icon fallback)
- Add `src/lib/widgets.ts` to fetch dashboard data, cache poster images, and write a JSON timeline to the shared container via `expo-widgets`
- Add `ContinueWatching` and `Upcoming` SwiftUI widget views in `src/widgets/`
- Add `useWidgetRefresh` hook that refreshes widgets on foreground (throttled to once per minute) and on session ready
- Debounce widget refresh in `invalidateTitleQueries` so rapid mutations (e.g. marking multiple episodes watched) batch into a single refresh
- Expose `titles.upcoming` from the dashboard procedure and add `upcomingTitles` field to the dashboard schema
- Bump iOS deployment target to 16.2 (required by WidgetKit timeline APIs)
- Add `GET /api/export/user-data` route that streams a JSON attachment of the authenticated user's full library data, named `sofa-export-<name>-<date>.json`
- Add `generateUserExport` in `@sofa/core/export` and a matching `parseSofaExport` parser so exported files can be re-imported via the existing import pipeline
- Register `sofa` as a new `ImportSource` in the API contract/schemas and wire up `parseSofaExport` in the `parseFile` procedure
- Add `EXPORT_FAILED` error code to the API error registry and surface it in web and native error-message maps
- Update the account settings section with export/import UI (download button, import progress)
- Add tests for `generateUserExport`, `parseSofaExport`, and round-trip fidelity
- Block all non-health API requests with 503 while a database restore is in progress (`withDatabaseAccessBlocked` in `@sofa/db/client`); pause and resume cron jobs around the restore window
- Replace the two-hook first-user admin promotion with an atomic `claimInitialAdmin` query that uses a DB-level unique constraint so concurrent sign-ups during the first-run window can't each see `userCount === 0`
- Fix `refreshAvailability` to always call `replaceAvailabilityTransaction` (clearing stale rows) even when TMDB returns no US providers, instead of returning early and leaving old data in place
- Fix `performUpdateCheck` to read `release_url` (snake_case) from the public API response instead of `releaseUrl`
- Fix `createJob` import handler to catch `SQLITE_CONSTRAINT_UNIQUE` and surface it as an `IMPORT_ALREADY_RUNNING` conflict error instead of a 500
- Relax public API telemetry schema to accept `string | number` for `users` and `titles` fields
- Add tests for availability clearing, import deduplication, and `claimInitialAdmin`
- Rename `watchlist` → `in_watchlist` and `in_progress` → `watching` across the full stack (DB migration, Drizzle schema, core services, API contract, web, native)
- Add a new `caught_up` status for TV shows where all aired episodes are watched but the show is still airing
- Add `titles.watchAll` procedure and `markAllWatched` action to mark every episode of a TV show as watched in one step; replace the old "Mark as Watching" / "Mark as Completed" context-menu actions with "Mark All Watched" (TV) and "Mark as Watched" (movie)
- Extract `display-status.ts` in `@sofa/api` to share status → display label/color logic between web and native
- Add `getDisplayStatusesByTitleIds` to `@sofa/core/tracking` and wire it into the dashboard library feed so clients receive resolved display statuses
- Show a destructive Alert confirmation before removing a title from the library (native)
- Remove "Mark as Completed" from the continue-watching card context menu
- Update i18n catalogs for all 6 locales (de, en, es, fr, it, pt)
The unique constraint on titles is (tmdbId, type), meaning the same
numeric TMDB ID can exist as both a movie and TV show. Several places
were keying maps by tmdbId alone, which caused UNIQUE constraint
violations when an UPDATE inadvertently changed a row's type, and
silently dropped titles when both types existed for the same ID.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Create 18 new query modules under `packages/db/src/queries/` covering every domain (availability, cache, colors, credits, cron, discovery, image-cache, imports, integrations, lists, metadata, person, settings, system-health, thumbhash, title, tracking, webhooks)
- Remove all direct `db` client imports and raw Drizzle expressions from `@sofa/core` services and `apps/server` procedures/cron; replace with named query functions
- Extract cron DB helpers (`startCronRun`, `completeCronRun`, `failCronRun`, `getLibraryTitleIds`, `getStaleLibraryTitles`, etc.) into `packages/core/src/cron.ts`
- Extract integration DB helpers into `packages/core/src/integrations.ts`
- Add `drizzle-orm` as a direct dependency of `@sofa/db` and remove it from `@sofa/core`
- Update `AGENTS.md` to document the strict layered architecture rule
- Enable `perf: warn` in `.oxlintrc.json` and add `builtin` env; disable `no-await-in-loop` and `react/no-array-index-key` globally
- Reorder plugins so `oxc` is first
- Add `.oxfmtrc.json` override to suppress trailing commas in JSON/JSONC files
- Replace spread-into-new-object patterns (`{ ...x, key: val }`) with `Object.assign(x, { key: val })` in discover, explore, integrations, search, credits, and image-cache to satisfy perf rules
- Memoize `ChartContext` and `ToggleGroupContext` provider values to avoid unnecessary re-renders
Migrate the entire monorepo from Biome 2.4.7 to oxlint 1.56.0 (linter)
and oxfmt 0.41.0 (formatter) for faster lint/format and broader rule
coverage.
- Add `.oxlintrc.json` with React, TypeScript, unicorn, import plugins
and correctness/suspicious categories
- Add `.oxfmtrc.json` with 2-space indent, import sorting, and Tailwind
class sorting (all 30+ custom className attributes migrated)
- Add `docs/.oxlintrc.json` and `docs/.oxfmtrc.json` with Next.js plugin
- Update all 12 workspace package.json scripts: `oxlint`, `oxfmt`,
`oxfmt --check`
- Add `format:check` turbo task and CI step
- Update VS Code settings/extensions to use `oxc.oxc-vscode`
- Update CI path triggers from `biome.json` to new config files
- Remove all `biome-ignore` comments and fix shadowed variables
- Delete `biome.json` and `docs/biome.json`
- Reformat entire codebase with oxfmt
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add i18n with LingUI, Crowdin integration, and typed error codes
- Add `@sofa/i18n` shared package with LingUI v5 (v6-ready config using
`defineConfig` + `@lingui/format-po`), eager English + lazy-loaded
fr/de/es/it/pt catalogs, `Intl`-based date/number format utilities,
and test helpers
- Wire `@lingui/vite-plugin` + babel macro plugin for web, and
`@lingui/metro-transformer` + babel config for native
- Add `I18nProvider` to both app roots with locale auto-detection
(navigator.language / expo-localization) and persistence
(localStorage / MMKV)
- Wrap all ~512 user-facing strings across web and native with LingUI
macros (`<Trans>`, `useLingui`/`t`, `i18n._(msg`...`)`, `plural`)
- Add language switcher to Settings in both apps
- Add `@sofa/api/errors` with 13 typed `AppErrorCode` values and
`appErrorData()` helper for contract `.errors()` schemas
- Update all oRPC contract error definitions with typed `data` fields;
convert import procedure `throw new Error()` to `ORPCError` with codes
- Add per-app `error-messages.ts` utilities that map error codes to
localized strings; update global `QueryCache.onError` handlers to stop
leaking raw `error.message` to users
- Add `crowdin.yml` config and `.github/workflows/crowdin.yml` for
automated source upload on merge and translation pull via dispatch
- Add `lingui.config.ts` at repo root with `bun run i18n:extract` and
`bun run i18n:compile` convenience scripts
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* chore: add Crowdin language mapping and ignore context file
* fix(i18n): address 20 localization quality issues
- Stop leaking raw error.message to users in change-password and
account-section; use localized fallbacks instead
- Add typed error data to discover contract's .errors() schema
- Replace all manual English plural suffixes with LingUI plural() macro
in episode counts, backup counts, star ratings, title/image counts
- Replace date-fns formatDistanceToNow and hardcoded date patterns with
locale-aware formatDate/formatRelativeTime from @sofa/i18n/format
- Convert integration-configs from module-scope translations (frozen at
import time) to lazy getIntegrationConfigs(i18n) function
- Combine split Trans fragments into single translatable units in
backup-schedule retention sentence and stats-display period labels
- Localize TV media type badge and full episode accessibilityLabel
- Replace Android-incompatible Alert.alert language picker (max 3
buttons) with zeego DropdownMenu
- Await async locale initialization before hiding splash screen so
non-English users don't see English flash on cold start
- Add @lingui/core as direct native app dependency
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(i18n): add Intl polyfills for Hermes, replace DropdownMenu language picker, and remove date-fns
- Add `@formatjs/intl-*` polyfills loaded at app startup in `intl-polyfills.ts` so Hermes on Android has full `Intl` support for locale-aware formatting
- Alias `@formatjs/icu-messageformat-parser` to its `no-parser` variant in metro config to reduce bundle size
- Replace the zeego `DropdownMenu` language picker in Settings with a new `SelectModal` component that works reliably on all platforms without the Android 3-button limit
- Replace `date-fns` `format`/`parseISO` in `person/[id].tsx` with `formatDate` from `@sofa/i18n/format`; remove `date-fns` from native and web `package.json`
- Refactor `StatusActionButton` to use a `StatusLabel` component with `<Trans>` rather than a `getStatusConfig(t)` factory so labels are always reactive to locale changes
- Fix crowdin workflow to use root `bun run i18n:compile` script instead of `cd packages/i18n && bun run compile`
- Wrap root layout `GestureHandlerRootView` in `SafeAreaProvider` and add `.catch` to locale-ready promise to prevent unhandled rejections blocking the splash screen
* fix(i18n): localize integration status helpers, fix stats-display Trans wrapping, and clean up SelectModal
- Localize `webhookStatus` and `listStatus` in `integration-card.tsx` using `i18n._(msg`...`)` so status strings are translated instead of always rendering in English
- Wrap "Movies {select}" and "Episodes {select}" in a single `<Trans>` in `stats-display.tsx` so the period selector element is embedded inside the translatable unit rather than concatenated outside it
- Fix locale activation order in Settings: close the modal first, then `activateLocale`, and only call `setPersistedLocale` on success to avoid persisting a locale that failed to load
- Remove the redundant `SafeAreaProvider`/`SafeAreaView` wrapper from `SelectModal` — the root layout already provides `SafeAreaProvider`
- Recompile all six `.po`/`.ts` catalogs to pick up new and updated message strings
* chore(i18n): crowdin sync
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All explore, discover, search, recommendation, and person-credit listing procedures now include the internal database `id` on every item so clients can navigate and act without a separate resolve round-trip.
- Remove `titles.resolve` and `people.resolve` mutation calls from the native search screen, hero banners, poster rows, and cast cards; replace with direct `Link` navigation using the pre-returned `id`.
- Change `titles.quickAdd` to accept `{ id }` instead of `{ tmdbId, type }` and update every call site on native and web.
- Key all user-status and episode-progress lookups by `id` instead of `tmdbId-type` composite strings across `PosterCard`, `HorizontalPosterRow`, `FilterableTitleRow`, and `usePosterActions`; drop the `tmdbId` prop from `PosterCard` entirely.
- Remove the `titles.hydrateSeasons` auto-trigger from the title detail screen; season hydration now happens server-side on resolve.
- Delete the `browse-thumbhashes` and `browse-title-ids` server procedures and remove them from the router.
- Extend `@sofa/api` schemas with an `id` field on all listing-item types; update `packages/core` services and add a DB migration accordingly.
Convert trending, search, people.detail, and dashboard.library to page-based responses (`page` / `totalPages`) so clients can load results incrementally.
Update `@sofa/core` discovery and person services to accept a `page` / `limit` input and slice results accordingly. Add a new DB migration to persist the data needed to back paginated filmography queries.
On the web, introduce a `useInfiniteScroll` hook and wire up `useInfiniteQuery` in the explore, person detail, and filterable title row components with an intersection-observer sentinel. On native, switch the same screens from `useQuery` to `useInfiniteQuery` with `onEndReached` / `ListFooterComponent` loading indicators. Also replace remaining `FlatList` usages with `FlashList` in the home and title detail screens.
The people.detail endpoint now fetches a person's complete filmography
from TMDB's combined_credits API (cast + crew) instead of only returning
locally-enriched titles.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Switch `mock.module("../src/image-cache", ...)` stubs in thumbhash, person-detail, colors, and credits tests to `spyOn(globalThis, "fetch")` instead. This avoids the cross-file persistence problem with `mock.module` and tests the real code paths more faithfully.
Update `image-cache.ts` to call `globalThis.fetch` so the spy intercepts it, add `mock.restore()` to the global `afterEach` in `preload.ts`, and set `IMAGE_CACHE_ENABLED=false` in affected tests to avoid disk writes during test runs.
Add `purgeMetadataCache` and `purgeImageCache` procedures that let admins free disk space on demand. Metadata purge removes un-enriched stub titles not in any user's library plus orphaned person records; image purge deletes all cached TMDB files from disk.
Expose both as POST endpoints under `/admin/cache/` via the oRPC contract, implement the core logic in a new `@sofa/core/cache` module, and surface them in a new "Danger Zone" section on the Settings page with individual confirmation dialogs and a combined "Purge all" action.
Add an opt-in telemetry system that sends anonymised instance
statistics to the public API once per day, which proxies them to
PostHog when `POSTHOG_API_KEY` is configured.
- Add `packages/core/src/telemetry.ts` — `performTelemetryReport`
and `isTelemetryEnabled`; user and title counts are bucketed
before sending to avoid exposing exact figures
- Add `getInstanceId()` to `settings.ts` — generates and persists a
stable UUIDv7 for the instance
- Schedule a daily `telemetryReport` cron job (00:30)
- Add `POST /v1/telemetry` to `apps/public-api`; forwards payload
to PostHog or returns 204 silently if key is absent
- Add `admin.telemetry` and `admin.toggleTelemetry` oRPC procedures
for inspecting and toggling the setting
- Expose `instanceId` on `system.publicInfo`
- Add `apps/public-api` (`@sofa/public-api`) — minimal Hono app
deployable to Vercel; `GET /v1/version` fetches the latest GitHub
release and returns `{ version, releaseUrl }` with a 15-minute
CDN cache (`s-maxage=900, stale-while-revalidate=3600`)
- Update `packages/core/src/update-check.ts` to call
`PUBLIC_API_URL/v1/version` instead of the GitHub API directly;
`PUBLIC_API_URL` defaults to `https://public-api.sofa.watch`
* Convert to Turborepo monorepo with shared API contract package
Restructure the repository as a monorepo in preparation for adding
future clients (mobile app, CLI). Extract the oRPC contract and Zod
schemas into `@sofa/api` (packages/api/) as a JIT internal package,
and relocate the Next.js app to `@sofa/web` (apps/web/).
- Add Turborepo with Bun workspaces for task orchestration and caching
- Extract `contract.ts` and `schemas.ts` into `@sofa/api` package
- Move all app code, configs, tests, and migrations to `apps/web/`
- Update 17 import paths from `@/lib/orpc/schemas` to `@sofa/api/schemas`
- Add `outputFileTracingRoot` and `transpilePackages` to next.config.ts
- Rewrite Dockerfile with `turbo prune --docker` for efficient builds
- Update CI workflows to use `turbo run` for lint/check-types/test
- Update CLAUDE.md with monorepo structure and commands
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Extract standalone Hono API server and split shared packages
Separate all server-side concerns from the Next.js frontend into a new
`apps/server/` Hono app and dedicated shared packages, making `@sofa/web`
a frontend-only app with no direct DB or service access.
- Add `@sofa/server` (`apps/server/`) — Hono API on port 3001 hosting
oRPC procedures, Better Auth, cron jobs, and non-RPC routes
- Add `@sofa/core` (`packages/core/`) — All 15 business logic services
moved from `apps/web/lib/services/`; tests moved to `packages/core/test/`
- Add `@sofa/db` (`packages/db/`) — DB client, schema, migrations,
constants, and logger extracted from `apps/web/lib/db/` and `lib/`
- Add `@sofa/tmdb` (`packages/tmdb/`) — TMDB client and image helpers
moved from `apps/web/lib/tmdb/`
- Add `@sofa/auth` (`packages/auth/`) — Better Auth server config moved
from `apps/web/lib/auth/`
- Move oRPC procedures, handler, router, middleware to `apps/server/src/orpc/`
- Move Hono route handlers (avatars, backups, images, lists, webhooks,
health) to `apps/server/src/routes/`; delete equivalent Next.js API routes
- Strip `apps/web` to frontend-only: no DB imports, no service imports,
all data via oRPC client calls to the API server
- Add `entrypoint.sh` to start API server, wait for health, then Next.js
- Update `next.config.ts` rewrites to proxy `/rpc/*` and `/api/*` to
`INTERNAL_API_URL` (default `http://localhost:3001`)
- Update Dockerfile and CLAUDE.md for the new structure
* Migrate web app from Next.js to Vite + TanStack Router SPA and add workspace catalog
Replace Next.js with a pure Vite SPA using TanStack Router for file-based routing,
removing all SSR complexity. The API server (Hono) now serves both API routes and
SPA static files in production, simplifying Docker to a single-process container.
Key changes:
- Vite 7 + @tanstack/react-router with file-based routing via plugin
- Route guards via beforeLoad + authClient.getSession() (replaces server-side auth)
- Route loaders with queryClient.ensureQueryData() (replaces SSR data fetching)
- Self-hosted fonts via @fontsource (replaces next/font/google)
- Tailwind v4 via @tailwindcss/vite (replaces @tailwindcss/postcss)
- Single oRPC client (removed SSR client and server-side session helper)
- Hono serves SPA static files in production (single port 3000)
- Single-process Dockerfile (removed entrypoint.sh)
- Bun workspace catalog for centralized dependency version management
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Extract @sofa/logger and @sofa/config shared packages
- Add `@sofa/logger` (`packages/logger/`) — standalone logger package
extracted from `@sofa/db/logger`; update all imports across server,
core, auth, db, and tmdb packages
- Add `@sofa/config` (`packages/config/`) — standalone config/constants
package extracted from `@sofa/db/constants`; exports `DATA_DIR`,
`DATABASE_URL`, `CACHE_DIR`, `AVATAR_DIR`, `BACKUP_DIR`
- Move `.env.example` from `apps/web/` to repo root; update server dev
scripts to load it via `--env-file=../../.env`
- Move image serving from `/api/images` to `/images`; add `serveStatic`
fast path in `index.ts` for cached files before falling back to the
TMDB fetch route; add `/images` proxy to Vite dev config
- Fix `Sparkline` component: replace `ResponsiveContainer` with
`ResizeObserver` to avoid SSR/hydration issues with recharts
- Replace `VITE_SERVER_URL` env var with `window.location.origin` in
the oRPC client (always same-origin in both dev and production)
* Fix asset caching, SPA 404 fallback, and DATA_DIR resolution
- Add `Cache-Control: immutable` header for hashed `/assets/*` files;
return 404 for missing asset paths instead of falling back to
`index.html` (prevents serving stale chunks after deploy)
- Wrap `query.invalidate` in an arrow function in the oRPC QueryClient
error handler to avoid illegal invocation errors
- Resolve `DATA_DIR` to an absolute path via `path.resolve()` so
relative paths work regardless of the process working directory
* Migrate @sofa/logger to pino for structured logging
- Replace custom logger implementation in `packages/logger/` with pino
+ pino-pretty; add both as workspace catalog dependencies
- Add `pino` and `pino-pretty` to the workspace catalog in `package.json`
- Fix `log.error()` calls in oRPC and OpenAPI handlers to pass the
error directly instead of wrapping it in `{ error }` to match pino's
serializer expectations
* Rename discoverProcedure/statsProcedure exports to discover/stats
* Add TanStackDevtools unified panel and VS Code workspace config
- Replace separate Router/Query devtools with unified `TanStackDevtools`
from `@tanstack/react-devtools` + `@tanstack/devtools-vite` plugin
- Wrap app in `<StrictMode>` in `main.tsx`
- Add `.vscode/settings.json` (Biome formatter, format-on-save, readonly
`routeTree.gen.ts`) and `.vscode/extensions.json` (recommended extensions)
* Move test DB helpers to @sofa/db/test-utils and add root bunfig.toml
Extract in-memory SQLite setup and fixture helpers (insertUser, insertTitle,
etc.) from packages/core/test/sqlite.ts into packages/db/src/test-utils.ts
so DB test utilities live alongside the schema they depend on. Use
import.meta.dir for CWD-independent migration path resolution.
Add root bunfig.toml so `bun test` works from the repo root in addition
to `bun run test` (turbo).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Fix devtools plugin order and whitespace-only TMDB token check
Move devtools() to first position in Vite plugins array per TanStack
docs, and trim TMDB_API_READ_ACCESS_TOKEN before boolean coercion so
whitespace-only values are treated as unconfigured.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
jakeandGitHub
