Fix search source picker defaulting to Spotify regardless of config

Enhanced search + global search popover always opened with the
Spotify icon active even when the user's primary metadata source
was Deezer / iTunes / Discogs / etc.

Trace: shared-helpers.js createSearchController reads
/status.metadata_source to pick the initial active icon, then
gates with SOURCE_LABELS[src]. Backend returns metadata_source
as a dict ({source, connected, response_time, ...}) — used
elsewhere for connection-state display — so SOURCE_LABELS[<dict>]
was always undefined, the guard never fired, and activeSource
silently stayed at the hardcoded 'spotify' default.

Fix reads .source off the dict (with fallback to plain-string for
forward compat). Other consumers already used ?.source — this was
the only stale call site.

webui/static/helper.js

@@ -3416,6 +3416,7 @@ const WHATS_NEW = {
     '2.5.2': [
         // --- May 13, 2026 — 2.5.2 release ---
         { date: 'May 13, 2026 — 2.5.2 release' },
+        { title: 'Search Source Picker: Fix Default Always Sticking To Spotify', desc: 'enhanced search + global search source picker always defaulted to spotify even when the user\'s primary metadata source was deezer / itunes / discogs / etc. trace: `shared-helpers.js:createSearchController` reads `/status.metadata_source` to pick the initial active icon, then checks `SOURCE_LABELS[src]` to validate. backend was returning `metadata_source` as a dict (`{source, connected, response_time, ...}` — used elsewhere for connection-state display), so `SOURCE_LABELS[<dict>]` was always undefined, the `if` guard never fired, and `state.activeSource` silently stayed at the hardcoded `\'spotify\'` default. fix: read `.source` off the dict (with forward-compat fallback to plain-string in case any older /status response shape predates the dict change). other consumers (core.js sidebar tile, helper.js status checker, search.js display) already used `?.source` correctly — this was the only stale call site.', page: 'search' },
         { title: 'Download Discography: No Longer Caps Prolific Artists At 50 Releases', desc: 'discord report: clicking "download discography" on an artist with a deep catalogue (bach, beatles complete box, dance / electronic artists with hundreds of remixes) only showed ~50 albums in the modal. trace: `MetadataLookupOptions(limit=50, max_pages=0)` was hardcoded at the discography endpoint and the artist-detail discography view. spotify\'s `max_pages=0` already paginates through everything (per-page is clamped to 10 internally) so spotify-primary users were unaffected. but deezer / itunes / discogs / hydrabase all honor the outer `limit` as a hard cap. fix: bump `limit` from 50 to 200 at all three call sites (`web_server.py` discography endpoint + artist-detail view + `core/artist_source_detail.py`). 200 matches iTunes\'s and Discogs\'s own internal caps and covers near-everyone\'s full catalogue. spotify behavior unchanged.', page: 'library' },
         { title: 'Artist Page: "Write Artist Image" Button (Real Artist Photos For Navidrome)', desc: 'github issue #572 (rhwc): navidrome shows album-art-derived thumbnails as artist photos because navidrome has no api for setting an artist image — it only reads `artist.jpg` from the artist folder during library scans. soulsync\'s `update_artist_poster` for navidrome was a no-op. new button on the artist detail page header writes `artist.jpg` to the artist\'s folder on disk: looks up any album track, resolves it through the path resolver (handles docker mount translation like #558 settled on), goes up one level to the artist folder, fetches the artist photo from the configured metadata source priority chain (spotify primary, fallback to deezer / discogs / etc), downloads with content-type validation + atomic write via `<filename>.tmp + os.replace`. when active server is navidrome, triggers a library scan immediately so the new file gets indexed. respects existing `artist.jpg` files (asks before overwriting) so user-supplied photos aren\'t clobbered. works for plex / jellyfin too as a fallback layer — both servers also read `artist.jpg` from disk. 26 tests pin the pure helpers in `core/library/artist_image.py`: folder derivation (trailing slash / backslash / empty / non-string), image url picking (missing attr / whitespace strip / non-string), download (non-image content-type / 404 / timeout / empty body), and write (atomic replace / temp-cleanup-on-failure / overwrite guard / missing folder).', page: 'library' },
         { title: 'Library History: Per-Download Audit Trail Modal', desc: 'each download row in library history now has an "audit" button that opens a second modal visualizing the download lifecycle as a vertical chain of decision blocks: request → source selected → source match → verification → post processing → final placement. each step has a status (complete / partial / unknown / error) with a color-coded node, plus a card showing what was decided and the supporting metadata. post-processing step infers observable changes from source-vs-final state (format conversion, file rename via tag template, title/artist rewrite, folder template). new "embedded tags" section below the flow reads the audio file live via mutagen at audit-open time and surfaces every tag actually on the file — title / artist / album / album artist / date / genre / track # / disc # / bpm / mood / style / copyright / publisher / release type+status+country / barcode / catalog # / asin / isrc / replaygain values / cover-art status / lyrics / every source id (spotify, tidal, deezer, musicbrainz, audiodb, lastfm, genius, itunes, beatport ...). file is the single source of truth — a persisted snapshot would drift the moment a background enrichment worker writes more tags. clean fallback when file is missing or unreadable. 19 tests pin the pure mutagen reader: id3 path (TIT2/TPE1/TALB + TXXX user-defined frames + USLT + APIC cover-art), vorbis path (FLAC dict-style + pass-through for unknown _id / _url keys), mp4 stub, format+bitrate+duration metadata, defensive paths (empty path, missing file, mutagen returns None, mutagen raises), stringify edge cases (list / tuple / int / frame-with-text / whitespace). files: core/library/file_tags.py (new mutagen reader), web_server.py (new GET /api/library/history/<id>/file-tags endpoint), webui/index.html (audit-overlay modal), webui/static/wishlist-tools.js (renderer + async fetch + tag-grid render), webui/static/style.css (flow + tags section + lyrics block styles).', page: 'wishlist' },

webui/static/shared-helpers.js

@@ -255,11 +255,20 @@ function createSearchController({
         // /status is public; /api/settings is admin-only and returns 403 for
         // non-admin profiles, which previously caused them to silently fall
         // back to 'spotify' regardless of what admin had configured.
+        //
+        // /status returns `metadata_source` as a dict (`{source, connected,
+        // response_time, ...}`) — pre-fix this code treated it as a string,
+        // so `SOURCE_LABELS[<dict>]` was always undefined and activeSource
+        // silently stayed at the hardcoded 'spotify' default regardless of
+        // what the user had configured. Read `.source` off the dict; fall
+        // back to the legacy string shape for forward-compat with any older
+        // /status response that might predate the dict change.
         try {
             const resp = await fetch('/status');
             if (resp.ok) {
                 const status = await resp.json();
-                const src = status && status.metadata_source;
+                const ms = status && status.metadata_source;
+                const src = (ms && typeof ms === 'object') ? ms.source : ms;
                 if (src && SOURCE_LABELS[src]) state.activeSource = src;
             }
         } catch (_) { /* best-effort */ }