fix: repair Japanese song download pipeline

Five separate root causes fixed:

1. buildTitleKeywords (ytdlp.js): For Japanese titles \w strips all
   kana/kanji leaving no keywords. Now uses the first whitespace-
   delimited segment as a literal regex keyword so --match-filter
   title ~= requires the YTM video to contain the Japanese text.

2. Tokenizer wait (ytdlp.js): If kuromoji hasn't finished loading when
   a Japanese download starts, romajiForStorage returns the original
   Japanese text unchanged — producing Japanese folder/filenames and
   skipping sort tags. Now waits up to 30 s for the tokenizer.

3. titleOnlyMatch guard (metadata.js): iTunes titleOnlyMatch (no artist
   check) was used as a fallback for all searches. For Japanese
   titles/artists this matched wrong US songs with identical titles
   (common in anime OSTs). Now titleOnlyMatch is skipped when the
   search title or artist contains Japanese characters.

4. squid.js rawArtist: rawArtist was not in squid's signature and
   never forwarded to embedMetadata, so Qobuz-downloaded J-Pop tracks
   stored the iTunes-resolved English name as artist instead of the
   Japanese original. Fixed by accepting and forwarding rawArtist.

5. squid.js romaji paths: Folder and filename used raw artist/title
   with no romaji conversion, creating Japanese-named paths
   inconsistent with the yt-dlp path. Now uses romajiForStorage same
   as ytdlp.js.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/lib/metadata.js

@@ -128,7 +128,12 @@ async function lookupItunesTrackMeta(artist, title) {
     // Fetch both in parallel — JP for correct Japanese titles/artwork, US for English genre names
     const [jpResults, usResults] = await Promise.all([fetchItunes('jp'), fetchItunes('us')]);
     let match = strictMatch(jpResults);
-    let usMatch = strictMatch(usResults) || titleOnlyMatch(usResults);
+    // titleOnlyMatch (no artist check) is too permissive for Japanese titles/artists —
+    // common anime/OST titles exist on both JP and US stores under different artists.
+    // Only use it as a fallback when neither title nor artist contains Japanese characters.
+    const JP_RE = /[぀-ヿ㐀-䶿一-鿿]/;
+    const isJpSearch = JP_RE.test(normT) || JP_RE.test(normA);
+    let usMatch = strictMatch(usResults) || (!isJpSearch && titleOnlyMatch(usResults));
     if (!match) match = usMatch;
     if (!match) return {};
     // JP genre names are localized (e.g. イージーリスニング); prefer US genre, map JP→EN as fallback

backend/lib/squid.js

@@ -11,6 +11,7 @@ const {
   downloadArtwork, fetchLyrics, findExistingTrack,
 } = require('./metadata');
 const { embedMetadata } = require('./ytdlp');
+const { romajiForStorage } = require('./romaji');
 
 const SQUID_BASE = 'https://qobuz.squid.wtf';
 
@@ -222,14 +223,18 @@ async function downloadUrlToFile(url, destPath) {
 
 // Download a single track via squid.wtf / Qobuz.
 // Returns { success: true, dir, skipped? } on success, null if track not found on Qobuz.
-async function downloadTrack({ jobId, artist, albumArtist, title, album, year, trackNum, totalTracks, discNum, discTotal, compilation, genre, artworkUrl, explicit, outputDir, durationMs }) {
+async function downloadTrack({ jobId, artist, rawArtist, albumArtist, title, album, year, trackNum, totalTracks, discNum, discTotal, compilation, genre, artworkUrl, explicit, outputDir, durationMs }) {
   const mainArtist = albumArtist || artist;
-  const safeArtist = sanitize(mainArtist.split(',')[0].trim());
-  const safeAlbum  = sanitize(album || 'Singles');
+  // Use romaji for folder/filename — same as ytdlp path so paths are consistent
+  const mainArtistRomaji = romajiForStorage(mainArtist) || mainArtist;
+  const titleRomaji      = romajiForStorage(title)      || title;
+  const albumRomaji      = album ? (romajiForStorage(album) || album) : null;
+  const safeArtist = sanitize(mainArtistRomaji.split(/[,;&]/)[0].trim());
+  const safeAlbum  = sanitize(albumRomaji || 'Singles');
   const dir        = path.join(outputDir || CONFIG.musicPath, safeArtist, safeAlbum);
   fs.mkdirSync(dir, { recursive: true });
 
-  const safeTitle   = sanitize(title);
+  const safeTitle   = sanitize(titleRomaji);
   const trackPrefix = trackNum ? String(trackNum).padStart(2, '0') + ' - ' : '';
   const ext         = squidExt();
 
@@ -278,7 +283,7 @@ async function downloadTrack({ jobId, artist, albumArtist, title, album, year, t
 
   return embedMetadata({
     jobId, audioFile, dir, trackPrefix, safeTitle,
-    artist, mainArtist, album, year, trackNum, totalTracks,
+    rawArtist, artist, mainArtist, album, year, trackNum, totalTracks,
     discNum, discTotal, compilation, genre,
     artworkUrl: finalArtwork, explicit, title,
   });

backend/lib/ytdlp.js

@@ -7,9 +7,20 @@ const {
   sanitize, navidromeArtist,
   downloadArtwork, findExistingTrack,
 } = require('./metadata');
-const { romajiForStorage } = require('./romaji');
+const { romajiForStorage, isTokenizerReady } = require('./romaji');
+
+// Matches hiragana, katakana, CJK — used to detect Japanese content
+const JP_RE = /[ぁ-んァ-ン一-龯]/;
 
 function buildTitleKeywords(title) {
+  // For Japanese titles: \w strips all kana/kanji, leaving nothing useful.
+  // Instead use the first whitespace-delimited segment as a literal keyword so
+  // the YTM --match-filter requires the video title to contain it.
+  if (JP_RE.test(title)) {
+    const firstWord = (title.split(/[\s　]/)[0] || '').slice(0, 12);
+    const escaped   = firstWord.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+    return escaped ? [escaped] : [];
+  }
   const VARIANT_WORDS = new Set(['sped','speedup','slowed','reverb','nightcore','karaoke','parody','remix','cover','acoustic','instrumental']);
   const STOPWORDS = new Set(['the','a','an','and','or','of','in','on','at','to','for','feat','ft','with']);
   const MIX_QUALIFIERS = ['radio', 'extended', 'club', 'vip', 'original', 'edit'];
@@ -51,6 +62,21 @@ function buildVariantFilter(title) {
 async function downloadTrack({ jobId, artist, rawArtist, albumArtist, title, album, year, trackNum, totalTracks, discNum, discTotal, compilation, genre, artworkUrl, explicit, outputDir, durationMs, sourceUrl }) {
   const mainArtist = albumArtist || artist;
 
+  // If the title or artist contains Japanese characters, wait for kuromoji to finish loading.
+  // Without it romajiForStorage returns the original text unchanged, producing Japanese
+  // folder/filenames and preventing sort tags from being written.
+  if (JP_RE.test(title) || JP_RE.test(rawArtist || artist || '')) {
+    if (!isTokenizerReady()) {
+      log(jobId, '⏳ Warte auf kuromoji-Tokenizer…');
+      await new Promise(resolve => {
+        const t0 = Date.now();
+        const iv = setInterval(() => {
+          if (isTokenizerReady() || Date.now() - t0 > 30000) { clearInterval(iv); resolve(); }
+        }, 300);
+      });
+    }
+  }
+
   // Folder / filename: romaji of the resolved artist (iTunes name if available — already in English,
   // otherwise kuromoji). rawArtist (original Japanese) is only used for tags + YTM search.
   const mainArtistRomaji = romajiForStorage(mainArtist) || mainArtist;
@@ -83,7 +109,6 @@ async function downloadTrack({ jobId, artist, rawArtist, albumArtist, title, alb
   const searchArtist = ytmArtist;
 
   const escRe = s => s.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
-  const JP_CHARS = /[ぁ-んァ-ン一-龯]/;
 
   const variantFilter = buildVariantFilter(title);
   const titleKws      = buildTitleKeywords(title);
@@ -95,14 +120,14 @@ async function downloadTrack({ jobId, artist, rawArtist, albumArtist, title, alb
   // Skip when artistFirstWord is non-ASCII (Japanese kanji/kana): YouTube channel names are
   // almost never in kanji, so the filter would reject every valid result.
   const artistFirstWord = (rawArtist || mainArtist).split(/[\s,/]+/)[0];
-  const artistChannelPart = (artistFirstWord.length >= 3 && !JP_CHARS.test(artistFirstWord))
+  const artistChannelPart = (artistFirstWord.length >= 3 && !JP_RE.test(artistFirstWord))
     ? ` & channel ~= (?i)${escRe(artistFirstWord)}` : '';
 
   // When the title is Japanese (keywords stripped to nothing) OR when we have an album that
   // provides better search context, include the album's first two words in the search query.
   // This helps YTM rank the official OST/album track above fan covers and romaji-subtitle uploads.
   // Only use album words that are ASCII (romaji/English) — Japanese album words don't help YTM ranking.
-  const hasJpTitle = JP_CHARS.test(title);
+  const hasJpTitle = JP_RE.test(title);
   const albumWords = (album && album !== 'Singles' && album !== 'Unknown')
     ? album.split(/\s+/).filter(w => /^[A-Za-z0-9]/.test(w)).slice(0, 2).join(' ')
     : '';