fix(i18n): add language detection mapping for Chinese locales

web/default/src/i18n/config.ts

@@ -20,6 +20,7 @@ import i18n from 'i18next'
 import LanguageDetector from 'i18next-browser-languagedetector'
 import { initReactI18next } from 'react-i18next'
 
+import { convertDetectedLanguage } from './languages'
 import en from './locales/en.json'
 import fr from './locales/fr.json'
 import ja from './locales/ja.json'
@@ -44,8 +45,8 @@ i18n
   .init({
     resources,
     fallbackLng: 'en',
-    supportedLngs: ['en', 'zhCN', 'fr', 'ru', 'ja', 'vi','zhTW'],
-    load: 'currentOnly', // Convert zh-CN -> zh
+    supportedLngs: ['en', 'zhCN', 'fr', 'ru', 'ja', 'vi', 'zhTW'],
+    load: 'currentOnly',
     nsSeparator: false, // Allow literal colons in keys (e.g., URLs, labels)
     debug: import.meta.env.DEV,
     interpolation: {
@@ -54,6 +55,9 @@ i18n
     detection: {
       order: ['localStorage', 'navigator'],
       caches: ['localStorage'],
+      // Browsers report `zh-CN`/`zh-TW`/`zh`; map them onto our `zhCN`/`zhTW`
+      // codes (non-Chinese codes pass through for normal supportedLngs matching).
+      convertDetectedLanguage,
     },
   })
 

web/default/src/i18n/languages.ts

@@ -33,7 +33,7 @@ export type InterfaceLanguageCode =
 export function normalizeInterfaceLanguage(value?: string | null): string {
   if (!value) return 'en'
 
-  var normalized = value.trim().replace(/_/g, '-').toLowerCase()
+  let normalized = value.trim().replaceAll('_', '-').toLowerCase()
   if (value === 'zh-TW' || value === 'zh-HK' || value === 'zh-MO' || value === 'zhTW') {
     normalized = 'zhTW'
   }
@@ -47,6 +47,30 @@ export function normalizeInterfaceLanguage(value?: string | null): string {
 }
 
 /**
+ * Map a browser-detected locale onto the interface language codes this project
+ * uses with i18next (`zhCN` / `zhTW`).
+ *
+ * Browsers report standard BCP-47 tags (`zh-CN`, `zh-TW`, `zh-Hant`, `zh`, ...),
+ * but `supportedLngs`/resources use the non-standard camelCase codes, so without
+ * this mapping a Chinese browser would never match and fall back to English.
+ * Non-Chinese codes are returned unchanged so i18next's own `supportedLngs`
+ * matching still applies (e.g. `fr-FR` -> `fr`, `ja` -> `ja`).
+ */
+export function convertDetectedLanguage(value: string): string {
+  const lower = value.trim().replaceAll('_', '-').toLowerCase()
+  if (!lower.startsWith('zh')) return value
+  if (
+    lower === 'zh-tw' ||
+    lower === 'zh-hk' ||
+    lower === 'zh-mo' ||
+    lower.startsWith('zh-hant')
+  ) {
+    return 'zhTW'
+  }
+  return 'zhCN'
+}
+
+/**
  * Convert an interface language code (the values i18next uses, such as `zhCN` /
  * `zhTW`) into a valid BCP-47 locale tag that the `Intl.*` APIs accept.
  *