fix: playlist matching for romaji-stored Japanese tracks + punctuation cleanup

- Clean Japanese punctuation (、。・) in romajiForStorage so it doesn't pass through
- Add POST /api/romaji-convert endpoint for batch title conversion
- Add romajiConvert() to api.js
- Add third fallback in _createNdPlaylist: convert Japanese title to romaji and
  search Navidrome with it (for songs already downloaded with romaji tags)

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

backend/lib/romaji.js

@@ -90,8 +90,14 @@ function romajiForStorage(text) {
     const needsSpace = /[a-zA-Z0-9]$/.test(parts[i - 1]) && /^[a-zA-Z]/.test(part);
     return acc + (needsSpace ? ' ' : '') + part;
   }, '');
+  // Clean up Japanese punctuation that passes through as-is
+  const cleaned = joined
+    .replace(/[、。]/g, ' ')        // Japanese comma/period → space
+    .replace(/[・･]/g, ' ')        // middle dot → space
+    .replace(/[「」『』【】〔〕〈〉《》]/g, '') // Japanese brackets → remove
+    .replace(/（/g, '(').replace(/）/g, ')');  // fullwidth parens → ASCII
   // Title-case: capitalize first lowercase letter of each word, preserve existing uppercase
-  return joined.replace(/\b([a-z])/g, c => c.toUpperCase()).replace(/\s+/g, ' ').trim();
+  return cleaned.replace(/\b([a-z])/g, c => c.toUpperCase()).replace(/\s+/g, ' ').trim();
 }
 
 // ─── INDEX BUILDER ────────────────────────────────────────────────────────────

backend/novasonic-server.js

@@ -109,6 +109,14 @@ app.delete('/api/romaji-index', (req, res) => {
   res.json({ ok: true });
 });
 
+// POST /api/romaji-convert — convert an array of strings to romaji
+app.post('/api/romaji-convert', (req, res) => {
+  const { texts } = req.body;
+  if (!Array.isArray(texts)) return res.status(400).json({ error: 'texts must be array' });
+  if (!isTokenizerReady()) return res.status(503).json({ error: 'tokenizer not ready' });
+  res.json({ results: texts.map(t => romajiForStorage(t) || t) });
+});
+
 // GET /api/genre-playcounts?ndUrl=...&user=...&pass=...
 // Returns { [genre]: totalPlayCount } sorted by plays desc, cached for 20 min.
 const _genrePlayCache = new Map();

src/api.js

@@ -68,6 +68,20 @@ export async function fetchRomajiIndex() {
 
 export function invalidateRomajiIndex() { _romajiIndex = null; }
 
+export async function romajiConvert(texts) {
+  if (!S.cfg.backendUrl || !texts.length) return texts;
+  try {
+    const r = await fetch(`${S.cfg.backendUrl}/api/romaji-convert`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({ texts }),
+    });
+    if (!r.ok) return texts;
+    const { results } = await r.json();
+    return results;
+  } catch { return texts; }
+}
+
 // ─── ITUNES ───
 // Route through backend proxy to avoid CORS; fall back to direct if no backend configured.
 function _itunesBase() {

src/ui.js

@@ -1,6 +1,6 @@
 import { S } from './state.js';
 import { esc, formatDur, toast } from './utils.js';
-import { api, coverUrl, ndAuth, fetchRomajiIndex } from './api.js';
+import { api, coverUrl, ndAuth, fetchRomajiIndex, romajiConvert } from './api.js';
 import { THEMES, applyTheme, getCurrentThemeId } from './themes.js';
 // Circular imports — safe (only used inside function bodies)
 import { playQueue, dlSongLocal, toggleStar, togglePlay, renderQueue, syncNpStar, updateStarBtn, npToggleStar, playCurrent, _syncNpDirect, setPlayIcon, nativeInsertAt } from './player.js';
@@ -565,7 +565,12 @@ async function _createNdPlaylist(tracks, name) {
     // doesn't handle Japanese text, so we need the index as a fallback.
     const hasJp = tracks.some(t => JP_RE.test(t.trackName) || JP_RE.test(t.artistName));
     const romajiIdx = hasJp ? await fetchRomajiIndex().catch(() => null) : null;
-    for (const t of tracks) {
+    // Pre-convert all Japanese titles to romaji in one batch — needed for songs that were
+    // already downloaded with romaji tags (they're no longer in the romaji index).
+    const jpTitles = tracks.map(t => (hasJp && JP_RE.test(t.trackName)) ? t.trackName : '');
+    const romajiTitles = hasJp ? await romajiConvert(jpTitles).catch(() => jpTitles) : jpTitles;
+    const normR = s => (s || '').toLowerCase().replace(/[^\w\s]/g, ' ').replace(/\s+/g, ' ').trim();
+    for (const [idx, t] of tracks.entries()) {
       if (!t.trackName) continue;
       const missingLabel = t.artistName ? t.artistName + ' – ' + t.trackName : t.trackName;
       try {
@@ -586,12 +591,22 @@ async function _createNdPlaylist(tracks, name) {
         const artistMatch = s => firstWord(s.artist) === firstWord(firstArtist || t.artistName);
         let match = songs.find(s => titleMatch(s) && artistMatch(s))
                  || songs.find(s => titleMatch(s));
-        // Fallback: search romaji index by original Japanese title
+        // Fallback 1: search romaji index by original Japanese title
-        // (Navidrome search3 doesn't index Japanese text properly)
+        // (for old songs that still have Japanese in Navidrome)
         if (!match && romajiIdx && JP_RE.test(t.trackName)) {
           const jpEntry = romajiIdx.find(e => norm(e.title) === norm(t.trackName));
           if (jpEntry) match = { id: jpEntry.id };
         }
+        // Fallback 2: search Navidrome with romaji-converted title
+        // (for songs downloaded after romaji changes — stored with romaji tags)
+        if (!match && JP_RE.test(t.trackName) && romajiTitles[idx]) {
+          const rTitle = romajiTitles[idx];
+          try {
+            const r3 = await api('search3', '&query=' + encodeURIComponent(rTitle) + '&songCount=20&albumCount=0&artistCount=0');
+            const s3 = r3.searchResult3?.song || [];
+            match = s3.find(s => normR(s.title) === normR(rTitle)) || null;
+          } catch { /* ignore */ }
+        }
         if (match) songIds.push(match.id);
         else missing.push(missingLabel);
       } catch { missing.push(missingLabel); }