dsh-audiogen 0.4.18 → 0.4.20

package/lib/index.js

@@ -1074,6 +1074,40 @@ const SHARED_VOICE_SORT_OPTIONS = [
 	"oldest",
 	"newest"
 ];
+/**
+* ISO 639-1 -> matching substrings across the two vendor vocabularies:
+* ElevenLabs uses ISO codes ("en"), MiniMax system ids use language labels
+* ("Chinese (Mandarin)_..." / "Japanese_..."). The aliases make one dropdown
+* value match both. "zh" also covers Cantonese (Chinese (Cantonese)).
+*/
+const LANGUAGE_ALIASES = {
+	zh: [
+		"zh",
+		"chinese",
+		"mandarin",
+		"cantonese",
+		"yue"
+	],
+	en: ["en", "english"],
+	ja: ["ja", "japanese"],
+	ko: ["ko", "korean"],
+	es: ["es", "spanish"],
+	fr: ["fr", "french"],
+	de: ["de", "german"],
+	ru: ["ru", "russian"],
+	it: ["it", "italian"],
+	pt: ["pt", "portuguese"],
+	ar: ["ar", "arabic"],
+	hi: ["hi", "hindi"]
+};
+function languageMatches(language, locale, needle) {
+	const haystack = [language ?? "", locale ?? ""].join(" ").toLowerCase();
+	const value = needle.trim().toLowerCase();
+	if (value === "") return true;
+	if (haystack.includes(value)) return true;
+	for (const alias of LANGUAGE_ALIASES[value] ?? []) if (haystack.includes(alias)) return true;
+	return false;
+}
 const MINIMAX_LANGUAGE_PREFIXES = [
 	"Chinese (Mandarin)",
 	"Chinese (Cantonese)",
@@ -1238,7 +1272,10 @@ async function listElevenLabs(channel, options) {
 	} catch (error) {
 		failures.push(`共享音色库：${error instanceof Error ? error.message : String(error)}`);
 	}
-	if (entries.length === 0 && failures.length > 0) throw new Error(`ElevenLabs 音色列表拉取失败：${failures.join("；").slice(0, 300)}`);
+	if (entries.length === 0 && failures.length > 0) {
+		const gatewayHint = gatewayVoiceLibraryHint(failures);
+		throw new Error(`ElevenLabs 音色列表拉取失败：${failures.join("；").slice(0, 300)}${gatewayHint}`);
+	}
 	const seen = /* @__PURE__ */ new Set();
 	const deduped = [];
 	for (const entry of entries) {
@@ -1246,13 +1283,23 @@ async function listElevenLabs(channel, options) {
 		seen.add(entry.voice_id);
 		deduped.push(entry);
 	}
+	const note = failures.length === 0 ? void 0 : `部分端点失败（已忽略）：${failures.join("；").slice(0, 300)}${gatewayVoiceLibraryHint(failures)}`;
 	return {
 		vendor: "elevenlabs",
 		voices: deduped,
 		truncated: false,
-		...failures.length === 0 ? {} : { note: `部分端点失败（已忽略）：${failures.join("；").slice(0, 300)}` }
+		...note === void 0 ? {} : { note }
 	};
 }
+/**
+* Gateway (new-api 类) 通常不映射 ElevenLabs 音色库端点，其响应是
+* 404/Invalid URL。把这种失败翻译成可操作的建议，而不是一份原始 JSON。
+*/
+function gatewayVoiceLibraryHint(failures) {
+	const raw = failures.join(" ").toLowerCase();
+	if (!raw.includes("404") && !raw.includes("invalid url") && !raw.includes("not found")) return "";
+	return "。该渠道网关未提供 ElevenLabs 音色库端点（/v1/voices、/v1/shared-voices）：生成可用，但浏览/删除音色需要配置官方 API 地址 https://api.elevenlabs.io 的渠道";
+}
 async function listVendorVoices(channel, options = {}) {
 	if (channel.apiUrl.trim() === "") throw new Error("渠道未配置 API 地址");
 	if (channel.apiKey.trim() === "") throw new Error("渠道未配置 API 密钥");
@@ -1324,7 +1371,7 @@ async function deleteVendorVoice(channel, voiceId) {
 }
 function cap(options) {
 	const limit = typeof options.limit === "number" && Number.isFinite(options.limit) ? Math.floor(options.limit) : 100;
-	return Math.max(1, Math.min(200, limit));
+	return Math.max(1, Math.min(500, limit));
 }
 /** Which official shared-voice filters are set (for the MiniMax "not supported" note). */
 function serverFilterNote(filters) {
@@ -1341,9 +1388,7 @@ function applyFilter(entries, options) {
 	const count = cap(options);
 	const matched = entries.filter((entry) => {
 		if (source !== "" && entry.source !== source) return false;
-		if (language !== "") {
-			if (![entry.language ?? "", entry.locale ?? ""].join(" ").toLowerCase().includes(language)) return false;
-		}
+		if (language !== "" && !languageMatches(entry.language, entry.locale, language)) return false;
 		if (field(filters.accent) !== "" && field(entry.accent) !== field(filters.accent)) return false;
 		if (field(filters.gender) !== "" && field(entry.gender) !== field(filters.gender)) return false;
 		if (field(filters.age) !== "" && field(entry.age) !== field(filters.age)) return false;
@@ -1367,6 +1412,7 @@ function applyFilter(entries, options) {
 		}
 		return true;
 	});
+	matched.sort((a, b) => a.deletable === b.deletable ? 0 : a.deletable ? -1 : 1);
 	return {
 		voices: matched.slice(0, count),
 		truncated: matched.length > count
@@ -2421,6 +2467,7 @@ function makeRoutes(deps) {
 						...str("language") === void 0 ? {} : { language: str("language") },
 						...str("keyword") === void 0 ? {} : { keyword: str("keyword") },
 						...str("source") === void 0 ? {} : { source: str("source") },
+						...typeof body?.limit === "number" && Number.isFinite(body.limit) ? { limit: Math.floor(body.limit) } : {},
 						...Object.keys(filters).length === 0 ? {} : { serverFilters: filters }
 					});
 					writeJson(res, 200, {

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "dsh-audiogen",
   "description": "AI audio generation plugin for the dsh web GUI: multi-vendor TTS/music/sound-effect channels (OpenAI-compatible, ElevenLabs, MiniMax, Stability AI and custom), per-channel model/voice catalogs, Agent tool and a sidebar AI 音频 panel.",
-  "version": "0.4.18",
+  "version": "0.4.20",
   "type": "module",
   "main": "lib/index.js",
   "exports": {

package/src/client/voices-view.tsx

@@ -31,6 +31,23 @@ const SOURCE_LABELS: Record<VoiceEntry['source'], string> = {
   shared: '社区共享',
 }
 
+/** 语言下拉：ISO 639-1 值；后端按两套厂商词汇匹配（ElevenLabs ISO / MiniMax 前缀）。 */
+const LANGUAGE_OPTIONS: Array<{ value: string; label: string }> = [
+  { value: '', label: '全部' },
+  { value: 'zh', label: '中文（普通话 / 粤语）' },
+  { value: 'en', label: '英语 English' },
+  { value: 'ja', label: '日语 Japanese' },
+  { value: 'ko', label: '韩语 Korean' },
+  { value: 'es', label: '西班牙语 Spanish' },
+  { value: 'fr', label: '法语 French' },
+  { value: 'de', label: '德语 German' },
+  { value: 'ru', label: '俄语 Russian' },
+  { value: 'it', label: '意大利语 Italian' },
+  { value: 'pt', label: '葡萄牙语 Portuguese' },
+  { value: 'ar', label: '阿拉伯语 Arabic' },
+  { value: 'hi', label: '印地语 Hindi' },
+]
+
 function useConfig(scope: AudiogenScope) {
   const [value, setValue] = useState(scope.getSnapshot().value)
   useEffect(() => scope.subscribe(() => { setValue(scope.getSnapshot().value) }), [scope])
@@ -76,6 +93,7 @@ export function VoicesView(props: {
   const [freeUsersAllowed, setFreeUsersAllowed] = useState(false)
 
   const [voices, setVoices] = useState<VoiceEntry[]>([])
+  const [truncated, setTruncated] = useState(false)
   const [loading, setLoading] = useState(false)
   const [error, setError] = useState<string | null>(null)
   const [note, setNote] = useState<string | null>(null)
@@ -103,9 +121,11 @@ export function VoicesView(props: {
         ...(sort === '' ? {} : { sort }),
         ...(featured ? { featured: true } : {}),
         ...(freeUsersAllowed ? { free_users_allowed: true } : {}),
+        limit: 500,
       })
       if (response.ok) {
         setVoices(response.voices ?? [])
+        setTruncated(response.truncated === true)
         setNote(response.note ?? null)
       } else {
         setVoices([])
@@ -168,7 +188,11 @@ export function VoicesView(props: {
         </label>
         <label className={css.field}>
           <span className={css.fieldLabel}>语言</span>
-          <input className={css.input} value={language} onChange={event => setLanguage(event.target.value)} placeholder="en / zh / ja…" />
+          <select className={css.select} value={language} onChange={event => setLanguage(event.target.value)}>
+            {LANGUAGE_OPTIONS.map(option => (
+              <option key={option.value} value={option.value}>{option.label}</option>
+            ))}
+          </select>
         </label>
         <label className={css.field}>
           <span className={css.fieldLabel}>关键词</span>
@@ -218,6 +242,7 @@ export function VoicesView(props: {
         {loading ? <span className={css.stateNote}>正在拉取音色列表…</span> : null}
         {!loading && error !== null ? <span className={css.stateNote} data-error>⚠ {error}</span> : null}
         {!loading && note !== null ? <span className={css.stateNote}>{note}</span> : null}
+        {!loading && truncated ? <span className={css.stateNote}>已截断：仅显示前 {voices.length} 个，可用筛选条件缩小范围</span> : null}
         {!loading && error === null ? <span className={css.count}>{voices.length} 个音色</span> : null}
       </div>
 

package/src/protocol.ts

@@ -8,7 +8,7 @@
 export const AUDIOGEN_SETTINGS_NAMESPACE = 'dsh-audiogen'
 
 /** Published package version shared by the host updater and the client UI. */
-export const PLUGIN_VERSION = '0.4.18'
+export const PLUGIN_VERSION = '0.4.20'
 
 /** Same-origin route family (loopback-only, mirroring dsh-imagegen). */
 export const SETTINGS_API = {
@@ -70,6 +70,8 @@ export interface VoicesListRequest {
   language?: string
   keyword?: string
   source?: string
+  /** Max returned entries (default 100, hard cap 500). */
+  limit?: number
   /** Official /v1/shared-voices server-side filters (ElevenLabs only). */
   search?: string
   use_case?: string

package/src/routes.ts

@@ -426,6 +426,7 @@ export function makeRoutes(deps: AudiogenRoutesDeps): WebRoute[] {
             ...(str('language') === undefined ? {} : { language: str('language')! }),
             ...(str('keyword') === undefined ? {} : { keyword: str('keyword')! }),
             ...(str('source') === undefined ? {} : { source: str('source')! }),
+            ...(typeof body?.limit === 'number' && Number.isFinite(body.limit) ? { limit: Math.floor(body.limit) } : {}),
             ...(Object.keys(filters).length === 0 ? {} : { serverFilters: filters }),
           })
           writeJson(res, 200, {

package/src/voice-manager.ts

@@ -70,6 +70,38 @@ export interface SharedVoiceFilters {
 
 const SHARED_VOICE_SORT_OPTIONS = ['most_used', 'random', 'oldest', 'newest'] as const
 
+/**
+ * ISO 639-1 -> matching substrings across the two vendor vocabularies:
+ * ElevenLabs uses ISO codes ("en"), MiniMax system ids use language labels
+ * ("Chinese (Mandarin)_..." / "Japanese_..."). The aliases make one dropdown
+ * value match both. "zh" also covers Cantonese (Chinese (Cantonese)).
+ */
+const LANGUAGE_ALIASES: Record<string, string[]> = {
+  zh: ['zh', 'chinese', 'mandarin', 'cantonese', 'yue'],
+  en: ['en', 'english'],
+  ja: ['ja', 'japanese'],
+  ko: ['ko', 'korean'],
+  es: ['es', 'spanish'],
+  fr: ['fr', 'french'],
+  de: ['de', 'german'],
+  ru: ['ru', 'russian'],
+  it: ['it', 'italian'],
+  pt: ['pt', 'portuguese'],
+  ar: ['ar', 'arabic'],
+  hi: ['hi', 'hindi'],
+}
+
+function languageMatches(language: string | undefined, locale: string | undefined, needle: string): boolean {
+  const haystack = [language ?? '', locale ?? ''].join(' ').toLowerCase()
+  const value = needle.trim().toLowerCase()
+  if (value === '') return true
+  if (haystack.includes(value)) return true
+  for (const alias of LANGUAGE_ALIASES[value] ?? []) {
+    if (haystack.includes(alias)) return true
+  }
+  return false
+}
+
 export interface ListVoicesResult {
   vendor: string
   voices: VendorVoiceEntry[]
@@ -272,7 +304,10 @@ async function listElevenLabs(channel: AudioChannel, options: ListVoicesOptions)
   }
 
   if (entries.length === 0 && failures.length > 0) {
-    throw new Error(`ElevenLabs 音色列表拉取失败：${failures.join('；').slice(0, 300)}`)
+    const gatewayHint = gatewayVoiceLibraryHint(failures)
+    throw new Error(
+      `ElevenLabs 音色列表拉取失败：${failures.join('；').slice(0, 300)}${gatewayHint}`,
+    )
   }
   // Owned voices come first and shadow shared ones with the same id.
   const seen = new Set<string>()
@@ -282,14 +317,25 @@ async function listElevenLabs(channel: AudioChannel, options: ListVoicesOptions)
     seen.add(entry.voice_id)
     deduped.push(entry)
   }
+  const note = failures.length === 0 ? undefined : `部分端点失败（已忽略）：${failures.join('；').slice(0, 300)}${gatewayVoiceLibraryHint(failures)}`
   return {
     vendor: 'elevenlabs',
     voices: deduped,
     truncated: false,
-    ...(failures.length === 0 ? {} : { note: `部分端点失败（已忽略）：${failures.join('；').slice(0, 300)}` }),
+    ...(note === undefined ? {} : { note }),
   }
 }
 
+/**
+ * Gateway (new-api 类) 通常不映射 ElevenLabs 音色库端点，其响应是
+ * 404/Invalid URL。把这种失败翻译成可操作的建议，而不是一份原始 JSON。
+ */
+function gatewayVoiceLibraryHint(failures: string[]): string {
+  const raw = failures.join(' ').toLowerCase()
+  if (!raw.includes('404') && !raw.includes('invalid url') && !raw.includes('not found')) return ''
+  return '。该渠道网关未提供 ElevenLabs 音色库端点（/v1/voices、/v1/shared-voices）：生成可用，但浏览/删除音色需要配置官方 API 地址 https://api.elevenlabs.io 的渠道'
+}
+
 // ------------------------------------------------------ public surface
 
 export async function listVendorVoices(
@@ -387,7 +433,7 @@ function cap(options: ListVoicesOptions): number {
   const limit = typeof options.limit === 'number' && Number.isFinite(options.limit)
     ? Math.floor(options.limit)
     : 100
-  return Math.max(1, Math.min(200, limit))
+  return Math.max(1, Math.min(500, limit))
 }
 
 /** Which official shared-voice filters are set (for the MiniMax "not supported" note). */
@@ -412,10 +458,7 @@ function applyFilter(
   const count = cap(options)
   const matched = entries.filter(entry => {
     if (source !== '' && entry.source !== source) return false
-    if (language !== '') {
-      const haystack = [entry.language ?? '', entry.locale ?? ''].join(' ').toLowerCase()
-      if (!haystack.includes(language)) return false
-    }
+    if (language !== '' && !languageMatches(entry.language, entry.locale, language)) return false
     // Official shared-voice filters double as local fallback filters so they
     // also apply to owned voices and providers without server-side support.
     if (field(filters.accent) !== '' && field(entry.accent) !== field(filters.accent)) return false
@@ -438,5 +481,8 @@ function applyFilter(
     }
     return true
   })
+  // Deletable (account-owned/custom) voices first so narrow default windows
+  // show the manageable ones; stable sort keeps provider order inside groups.
+  matched.sort((a, b) => (a.deletable === b.deletable ? 0 : a.deletable ? -1 : 1))
   return { voices: matched.slice(0, count), truncated: matched.length > count }
 }