dsh-audiogen 0.4.13 → 0.4.15

package/lib/index.js

@@ -175,10 +175,10 @@ function isPreset(channel, id) {
 function isOpenAICompatible(channel, mode) {
 	return isPreset(channel, "openai") || /(^|\/)(v\d+\/)?audio\/speech$/i.test(channel.apiUrl.trim()) || channel.preset === "custom" && mode === "tts";
 }
-function isElevenLabs$1(channel) {
+function isElevenLabs$2(channel) {
 	return isPreset(channel, "elevenlabs") || /elevenlabs/i.test(channel.apiUrl);
 }
-function isMiniMax$1(channel) {
+function isMiniMax$2(channel) {
 	return isPreset(channel, "minimax") || /minimax/i.test(channel.apiUrl);
 }
 function isStability$1(channel) {
@@ -332,7 +332,7 @@ async function openAITTS(channel, request, signal) {
 		fallbackMime: "audio/mpeg"
 	});
 }
-async function elevenLabs(channel, request, signal) {
+async function elevenLabsOfficial(channel, request, signal) {
 	const base = endpointBase(channel.apiUrl);
 	const model = (request.upstream ?? request.model) || "eleven_multilingual_v2";
 	const headers = {
@@ -377,12 +377,15 @@ async function elevenLabs(channel, request, signal) {
 	}
 	if (request.mode === "music") {
 		const endpoint = `${base}/music`;
+		const musicModel = (request.upstream ?? request.model) || "music_v1";
+		const lyrics = request.lyrics?.trim() ?? "";
+		const instrumental = request.isInstrumental === true || lyrics === "";
 		const body = {
-			model_id: (request.upstream ?? request.model) || "music_v1",
+			model_id: musicModel,
 			prompt: request.prompt,
 			...request.duration !== void 0 && Number.isFinite(request.duration) ? { music_length_ms: Math.round(Math.min(6e5, Math.max(3e3, request.duration * 1e3))) } : {},
-			...request.lyrics !== void 0 && request.lyrics.trim() !== "" ? { lyrics_text: request.lyrics.trim() } : {},
-			...request.isInstrumental !== void 0 ? { force_instrumental: request.isInstrumental } : {}
+			...lyrics === "" ? {} : { lyrics_text: lyrics },
+			...instrumental ? { force_instrumental: true } : {}
 		};
 		const response = await fetchWithTimeout(endpoint, {
 			method: "POST",
@@ -450,6 +453,120 @@ async function elevenLabs(channel, request, signal) {
 		fallbackMime: "audio/mpeg"
 	});
 }
+/**
+* 官方 ElevenLabs 请求被网关拒绝的信号:官方路径未映射(404 Invalid URL)或
+* 网关要求 Bearer 认证而非 xi-api-key(401/403 Invalid token / Invalid API key)。
+* New API 类中转(如 ai.farmmx.com)对 ElevenLabs 官方协议通常返回这类错误。
+*/
+function isGatewayRouteMiss(error) {
+	return error instanceof AudioGenError && error.code === "audio-api-error" && /\bHTTP (404|401|403)\b/.test(error.message) && /\bInvalid URL\b|\bInvalid token\b|\bInvalid API key\b/i.test(error.message);
+}
+/** 网关兼容形态的请求头:仅 Bearer(携带 xi-api-key 会被网关按官方协议校验而 401)。 */
+function gatewayHeaders(apiKey) {
+	return {
+		authorization: `Bearer ${apiKey.trim()}`,
+		"content-type": "application/json",
+		accept: "audio/mpeg, application/json"
+	};
+}
+/** /audio/speech 兼容端点:base 已以此结尾时直接复用,否则拼接。 */
+function speechGatewayEndpoint(base) {
+	return /\/audio\/speech(\?|$)/i.test(base) ? base : `${base}/audio/speech`;
+}
+/**
+* ElevenLabs 渠道的网关兼容形态(OpenAI 风格):路径用 /audio/speech(音效/TTS)
+* 或 /music(音乐),认证用 Bearer、模型用 `model` 字段。
+*
+* 适配未映射 ElevenLabs 官方端点(404 Invalid URL)或要求 Bearer 认证
+* (401 Invalid token)的 New API 类中转,如 ai.farmmx.com。
+*/
+async function elevenLabsGatewayCompat(channel, request, signal) {
+	const base = endpointBase(channel.apiUrl);
+	const headers = gatewayHeaders(channel.apiKey);
+	if (request.mode === "music") {
+		const endpoint = /\/music(\?|$)/i.test(base) ? base : `${base}/music`;
+		const musicModel = (request.upstream ?? request.model) || "music_v1";
+		const lyrics = request.lyrics?.trim() ?? "";
+		const instrumental = request.isInstrumental === true || lyrics === "";
+		const body = {
+			model: musicModel,
+			prompt: request.prompt,
+			...request.duration !== void 0 && Number.isFinite(request.duration) ? { music_length_ms: Math.round(Math.min(6e5, Math.max(3e3, request.duration * 1e3))) } : {},
+			...lyrics === "" ? {} : { lyrics_text: lyrics },
+			...instrumental ? { force_instrumental: true } : {}
+		};
+		const response = await fetchWithTimeout(endpoint, {
+			method: "POST",
+			redirect: "follow",
+			headers,
+			body: JSON.stringify(body),
+			signal
+		}, UPSTREAM_TIMEOUT_MS);
+		if (!response.ok) {
+			const detail = await response.text().catch(() => "");
+			throw new AudioGenError(`ElevenLabs music gateway-compatible API error (HTTP ${response.status})${detail === "" ? "" : `: ${detail.slice(0, 300)}`}`, "audio-api-error");
+		}
+		return normalizeAudioResponse(response, {
+			apiKey: channel.apiKey,
+			fallbackMime: "audio/mpeg"
+		});
+	}
+	if (request.mode === "voice_design") throw new AudioGenError("当前网关不支持 ElevenLabs 音色设计端点(POST /v1/text-to-voice/design);该模式请改用 MiniMax 渠道或 ElevenLabs 官方 API。", "voice-design-unsupported");
+	const endpoint = speechGatewayEndpoint(base);
+	const isSfx = request.mode === "sfx";
+	const model = (request.upstream ?? request.model) || (isSfx ? "eleven_text_to_sound_v2" : "eleven_multilingual_v2");
+	if (!isSfx) {
+		if ((request.voice?.trim() ?? "") === "") {
+			const suggestions = (channel.models ?? []).filter((entry) => {
+				const candidate = entry;
+				return candidate.category === "tts" && candidate.id.trim() !== "" && candidate.id !== candidate.alias;
+			}).map((entry) => `${entry.alias}（${entry.id}）`);
+			throw new AudioGenError(`ElevenLabs 网关渠道的 TTS 必须携带音色 voice_id（网关强制校验，缺失会返回 400 voice or voice_id is required）：请在「音色」字段填入官方音色 ID${suggestions.length === 0 ? "" : `，可选用以下音色：${suggestions.slice(0, 4).join("、")}`}。`, "voice-required");
+		}
+	}
+	const body = isSfx ? {
+		model,
+		input: request.prompt,
+		...request.duration !== void 0 && Number.isFinite(request.duration) ? { duration_seconds: Math.min(30, Math.max(.5, request.duration)) } : {},
+		...request.loop !== void 0 ? { loop: request.loop } : {},
+		...request.promptInfluence !== void 0 && Number.isFinite(request.promptInfluence) ? { prompt_influence: Math.min(1, Math.max(0, request.promptInfluence)) } : {}
+	} : {
+		model,
+		input: request.prompt,
+		voice: request.voice.trim(),
+		response_format: request.format ?? "mp3",
+		...request.speed !== void 0 ? { speed: request.speed } : {}
+	};
+	const response = await fetchWithTimeout(endpoint, {
+		method: "POST",
+		redirect: "follow",
+		headers,
+		body: JSON.stringify(body),
+		signal
+	}, UPSTREAM_TIMEOUT_MS);
+	if (!response.ok) {
+		const detail = await response.text().catch(() => "");
+		throw new AudioGenError(`ElevenLabs ${isSfx ? "sound effects" : "TTS"} gateway-compatible API error (HTTP ${response.status})${detail === "" ? "" : `: ${detail.slice(0, 300)}`}`, "audio-api-error");
+	}
+	return normalizeAudioResponse(response, {
+		apiKey: channel.apiKey,
+		fallbackMime: "audio/mpeg"
+	});
+}
+/**
+* ElevenLabs 渠道入口:官方端点优先;官方协议被网关(New API 类中转)拒绝时,
+* 自动改用 OpenAI 兼容形态重试,使同一渠道同时兼容 ElevenLabs 官方 API 与
+* ai.farmmx.com 类中转。官方地址(api.elevenlabs.io)直连不触发回退。
+*/
+async function elevenLabs(channel, request, signal) {
+	if (/elevenlabs\.io/i.test(channel.apiUrl)) return elevenLabsOfficial(channel, request, signal);
+	try {
+		return await elevenLabsOfficial(channel, request, signal);
+	} catch (error) {
+		if (!isGatewayRouteMiss(error)) throw error;
+	}
+	return elevenLabsGatewayCompat(channel, request, signal);
+}
 function minimaxApiBase(base) {
 	const trimmed = endpointBase(base);
 	return /\/v1$/i.test(trimmed) ? trimmed : `${trimmed}/v1`;
@@ -614,13 +731,13 @@ async function minimax(channel, request, signal) {
 			256e3
 		]);
 		const lyrics = request.lyrics?.trim() ?? "";
-		if (lyrics === "" && request.isInstrumental !== true) throw new AudioGenError("MiniMax 音乐生成需要歌词（lyrics 参数），或在「纯音乐」模式（is_instrumental=true）下生成；也可让面板/Agent 先为提示词创作一段歌词。", "lyrics-required");
+		const instrumental = request.isInstrumental === true || lyrics === "";
 		const endpoint = `${base}/music_generation`;
 		const body = {
 			model,
 			prompt: request.prompt,
 			...lyrics === "" ? {} : { lyrics },
-			...request.isInstrumental !== void 0 ? { is_instrumental: request.isInstrumental } : {},
+			...instrumental ? { is_instrumental: true } : {},
 			...request.duration !== void 0 ? { duration: request.duration } : {},
 			audio_setting: {
 				format: MUSIC_FORMATS.has(request.format ?? "mp3") ? request.format ?? "mp3" : "mp3",
@@ -860,9 +977,9 @@ async function generateAudio(channel, request, signal) {
 	if (channel.apiUrl.trim() === "") throw new AudioGenError("channel API URL is not configured", "audio-no-endpoint");
 	if (channel.apiKey.trim() === "") throw new AudioGenError("channel API key is not configured", "audio-no-key");
 	if (request.prompt.trim() === "") throw new AudioGenError("audio prompt/text is required", "audio-empty-prompt");
-	if (request.mode === "voice_design" && !isMiniMax$1(channel) && !isElevenLabs$1(channel)) throw new AudioGenError("音色设计当前仅支持 MiniMax（/v1/voice_design）与 ElevenLabs（/v1/text-to-voice/design）渠道", "voice-design-unsupported");
-	if (isElevenLabs$1(channel)) return elevenLabs(channel, request, signal);
-	if (isMiniMax$1(channel)) return minimax(channel, request, signal);
+	if (request.mode === "voice_design" && !isMiniMax$2(channel) && !isElevenLabs$2(channel)) throw new AudioGenError("音色设计当前仅支持 MiniMax（/v1/voice_design）与 ElevenLabs（/v1/text-to-voice/design）渠道", "voice-design-unsupported");
+	if (isElevenLabs$2(channel)) return elevenLabs(channel, request, signal);
+	if (isMiniMax$2(channel)) return minimax(channel, request, signal);
 	if (isStability$1(channel) || /^stable-audio-/i.test(((request.upstream ?? request.model) || "").trim())) return stabilityAudio(channel, request, signal);
 	if (isOpenAICompatible(channel, request.mode)) return openAITTS(channel, request, signal);
 	return genericAudio(channel, request, signal);
@@ -1101,16 +1218,16 @@ function audioPresetById(id) {
 }
 //#endregion
 //#region src/audio-models.ts
-function isMiniMax(channel) {
+function isMiniMax$1(channel) {
 	return channel.preset === "minimax" || /minimax/i.test(channel.apiUrl);
 }
-function isElevenLabs(channel) {
+function isElevenLabs$1(channel) {
 	return channel.preset === "elevenlabs" || /elevenlabs/i.test(channel.apiUrl);
 }
 function isStability(channel) {
 	return channel.preset === "stability" || /stability\.ai/i.test(channel.apiUrl);
 }
-function baseUrl(url) {
+function baseUrl$1(url) {
 	return url.trim().replace(/\/+$/, "");
 }
 /** Whether an upstream model id is audio-related at all. */
@@ -1120,7 +1237,7 @@ function categoryFor(id) {
 	if (/(music|song|cover|lyrics|audio|melody|beat)/i.test(value)) return "music";
 	if (/(sfx|sound.?effect|effect|foley)/i.test(value)) return "sfx";
 }
-async function fetchJson(url, init) {
+async function fetchJson$1(url, init) {
 	const response = await fetch(url, init);
 	if (!response.ok) {
 		const text = await response.text().catch(() => "");
@@ -1128,8 +1245,8 @@ async function fetchJson(url, init) {
 	}
 	return response.json();
 }
-async function postJson(url, apiKey, body) {
-	return fetchJson(url, {
+async function postJson$1(url, apiKey, body) {
+	return fetchJson$1(url, {
 		method: "POST",
 		headers: {
 			authorization: `Bearer ${apiKey.trim()}`,
@@ -1142,15 +1259,15 @@ async function postJson(url, apiKey, body) {
 async function discoverAudioModels(channel) {
 	if (channel.apiUrl.trim() === "") throw new Error("API URL is not configured");
 	if (channel.apiKey.trim() === "") throw new Error("API key is not configured");
-	if (isMiniMax(channel)) return discoverMiniMax(channel);
-	if (isElevenLabs(channel)) return discoverElevenLabs(channel);
+	if (isMiniMax$1(channel)) return discoverMiniMax(channel);
+	if (isElevenLabs$1(channel)) return discoverElevenLabs(channel);
 	if (isStability(channel)) return discoverStability(channel);
 	return discoverOpenAICompatible(channel);
 }
 async function discoverMiniMax(channel) {
-	const url = `${baseUrl(channel.apiUrl).replace(/\/v1$/i, "")}/v1/get_voice`;
+	const url = `${baseUrl$1(channel.apiUrl).replace(/\/v1$/i, "")}/v1/get_voice`;
 	try {
-		const payload = await postJson(url, channel.apiKey, { voice_type: "all" });
+		const payload = await postJson$1(url, channel.apiKey, { voice_type: "all" });
 		if (payload.base_resp?.status_code !== void 0 && payload.base_resp.status_code !== 0) throw new Error(payload.base_resp.status_msg ?? `MiniMax returned status ${payload.base_resp.status_code}`);
 		const models = [];
 		for (const voice of payload.system_voice ?? []) {
@@ -1202,12 +1319,12 @@ async function discoverMiniMax(channel) {
 	}
 }
 async function discoverElevenLabs(channel) {
-	const base = baseUrl(channel.apiUrl);
+	const base = baseUrl$1(channel.apiUrl);
 	const headers = { "xi-api-key": channel.apiKey.trim() };
 	const failures = [];
 	const models = [];
 	try {
-		const payload = await fetchJson(`${base}/models`, { headers });
+		const payload = await fetchJson$1(`${base}/models`, { headers });
 		for (const item of Array.isArray(payload) ? payload : []) {
 			const id = item.model_id?.trim() ?? "";
 			if (id === "") continue;
@@ -1223,7 +1340,7 @@ async function discoverElevenLabs(channel) {
 		failures.push(`模型列表：${error instanceof Error ? error.message : String(error)}`);
 	}
 	try {
-		const payload = await fetchJson(`${base}/voices`, { headers });
+		const payload = await fetchJson$1(`${base}/voices`, { headers });
 		for (const voice of Array.isArray(payload?.voices) ? payload.voices : []) {
 			const id = voice.voice_id?.trim() ?? "";
 			if (id === "") continue;
@@ -1257,7 +1374,7 @@ async function discoverStability(channel) {
 	};
 }
 async function discoverOpenAICompatible(channel) {
-	const payload = await fetchJson(`${baseUrl(channel.apiUrl)}/models`, { headers: { authorization: `Bearer ${channel.apiKey.trim()}` } });
+	const payload = await fetchJson$1(`${baseUrl$1(channel.apiUrl)}/models`, { headers: { authorization: `Bearer ${channel.apiKey.trim()}` } });
 	const models = [];
 	for (const item of payload.data ?? []) {
 		const id = item.id?.trim() ?? "";
@@ -2100,11 +2217,13 @@ function makeRoutes(deps) {
 							type: entry.type
 						}];
 					} catch {}
+					const note = request.mode === "music" && request.isInstrumental !== true && (request.lyrics === void 0 || request.lyrics.trim() === "") ? "未提供歌词，已按纯音乐生成" : void 0;
 					writeJson(res, 200, {
 						ok: true,
 						outputs: generated,
 						history,
-						...resources === void 0 ? {} : { resources }
+						...resources === void 0 ? {} : { resources },
+						...note === void 0 ? {} : { note }
 					});
 				} catch (error) {
 					writeJson(res, 200, {
@@ -2453,6 +2572,276 @@ function makeRoutes(deps) {
 	];
 }
 //#endregion
+//#region src/voice-manager.ts
+const MINIMAX_LANGUAGE_PREFIXES = [
+	"Chinese (Mandarin)",
+	"Chinese (Cantonese)",
+	"Japanese",
+	"English",
+	"Korean",
+	"Spanish",
+	"French",
+	"German",
+	"Italian",
+	"Russian",
+	"Portuguese",
+	"Arabic",
+	"Hindi"
+];
+function isMiniMax(channel) {
+	return channel.preset === "minimax" || /minimax/i.test(channel.apiUrl);
+}
+function isElevenLabs(channel) {
+	return channel.preset === "elevenlabs" || /elevenlabs/i.test(channel.apiUrl);
+}
+function supportsVoiceManagement(channel) {
+	return isMiniMax(channel) || isElevenLabs(channel);
+}
+function baseUrl(url) {
+	return url.trim().replace(/\/+$/, "");
+}
+/** MiniMax system voice ids carry a language label prefix. */
+function languageFromMiniMaxId(voiceId) {
+	for (const prefix of MINIMAX_LANGUAGE_PREFIXES) if (voiceId.startsWith(`${prefix}_`)) return prefix;
+}
+function asStringList(value) {
+	if (Array.isArray(value)) return value.map((item) => String(item)).filter((item) => item.trim() !== "");
+	if (typeof value === "string" && value.trim() !== "") return [value.trim()];
+	return [];
+}
+async function fetchJson(url, init) {
+	const response = await fetch(url, init);
+	if (!response.ok) {
+		const text = await response.text().catch(() => "");
+		throw new Error(`HTTP ${response.status}${text === "" ? "" : `: ${text.slice(0, 300)}`}`);
+	}
+	return response.json();
+}
+async function postJson(url, apiKey, body) {
+	return fetchJson(url, {
+		method: "POST",
+		headers: {
+			authorization: `Bearer ${apiKey.trim()}`,
+			"content-type": "application/json"
+		},
+		body: JSON.stringify(body)
+	});
+}
+function normalizeMiniMax(voice, source) {
+	const voiceId = String(voice.voice_id ?? "").trim();
+	const description = asStringList(voice.description).join("；") || void 0;
+	const language = languageFromMiniMaxId(voiceId);
+	return {
+		provider: "minimax",
+		voice_id: voiceId,
+		name: String(voice.voice_name ?? voiceId).trim() || voiceId,
+		source,
+		...language === void 0 ? {} : { language },
+		...description === void 0 ? {} : { description },
+		deletable: source === "custom"
+	};
+}
+async function listMiniMax(channel) {
+	const payload = await postJson(`${baseUrl(channel.apiUrl).replace(/\/v1$/i, "")}/v1/get_voice`, channel.apiKey, { voice_type: "all" });
+	if (payload.base_resp?.status_code !== void 0 && payload.base_resp.status_code !== 0) throw new Error(`MiniMax get_voice 失败：${payload.base_resp.status_msg ?? `status ${payload.base_resp.status_code}`}`);
+	const entries = [];
+	for (const voice of Array.isArray(payload.system_voice) ? payload.system_voice : []) {
+		if (typeof voice !== "object" || voice === null) continue;
+		const entry = normalizeMiniMax(voice, "system");
+		if (entry.voice_id !== "") entries.push(entry);
+	}
+	for (const bucket of ["voice_cloning", "voice_generation"]) for (const voice of Array.isArray(payload[bucket]) ? payload[bucket] : []) {
+		if (typeof voice !== "object" || voice === null) continue;
+		const entry = normalizeMiniMax(voice, "custom");
+		if (entry.voice_id !== "") entries.push(entry);
+	}
+	return {
+		vendor: "minimax",
+		voices: entries,
+		truncated: false
+	};
+}
+function normalizeElevenLabs(voice, source) {
+	const voiceId = String(voice.voice_id ?? "").trim();
+	const labels = typeof voice.labels === "object" && voice.labels !== null ? voice.labels : void 0;
+	const pick = (key) => {
+		const owned = labels?.[key];
+		const direct = voice[key];
+		return (typeof owned === "string" && owned.trim() !== "" ? owned : typeof direct === "string" && direct.trim() !== "" ? direct : void 0)?.trim() || void 0;
+	};
+	const description = pick("description");
+	return {
+		provider: "elevenlabs",
+		voice_id: voiceId,
+		name: String(voice.name ?? voiceId).trim() || voiceId,
+		source,
+		...pick("language") === void 0 ? {} : { language: pick("language") },
+		...pick("locale") === void 0 ? {} : { locale: pick("locale") },
+		...pick("accent") === void 0 ? {} : { accent: pick("accent") },
+		...pick("gender") === void 0 ? {} : { gender: pick("gender") },
+		...pick("age") === void 0 ? {} : { age: pick("age") },
+		...pick("use_case") === void 0 ? {} : { use_case: pick("use_case") },
+		...pick("category") === void 0 ? {} : { category: pick("category") },
+		...description === void 0 ? {} : { description },
+		...typeof voice.preview_url === "string" && voice.preview_url.trim() !== "" ? { preview_url: voice.preview_url.trim() } : {},
+		deletable: source === "owned"
+	};
+}
+async function listElevenLabs(channel, options) {
+	const base = baseUrl(channel.apiUrl);
+	const headers = {
+		"xi-api-key": channel.apiKey.trim(),
+		accept: "application/json"
+	};
+	const entries = [];
+	const failures = [];
+	try {
+		const payload = await fetchJson(`${base}/voices`, { headers });
+		for (const voice of Array.isArray(payload.voices) ? payload.voices : []) {
+			if (typeof voice !== "object" || voice === null) continue;
+			const entry = normalizeElevenLabs(voice, "owned");
+			if (entry.voice_id !== "") entries.push(entry);
+		}
+	} catch (error) {
+		failures.push(`自有音色：${error instanceof Error ? error.message : String(error)}`);
+	}
+	const pageSize = 100;
+	try {
+		for (let page = 0; page < 3; page += 1) {
+			const query = new URLSearchParams({
+				page_size: String(pageSize),
+				page: String(page)
+			});
+			if (options.language !== void 0 && options.language.trim() !== "") query.set("language", options.language.trim());
+			const payload = await fetchJson(`${base}/shared-voices?${query.toString()}`, { headers });
+			const voices = Array.isArray(payload.voices) ? payload.voices : [];
+			for (const voice of voices) {
+				if (typeof voice !== "object" || voice === null) continue;
+				const entry = normalizeElevenLabs(voice, "shared");
+				if (entry.voice_id !== "") entries.push(entry);
+			}
+			if (payload.has_more !== true || voices.length === 0) break;
+		}
+	} catch (error) {
+		failures.push(`共享音色库：${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：${failures.join("；").slice(0, 300)}`);
+	const seen = /* @__PURE__ */ new Set();
+	const deduped = [];
+	for (const entry of entries) {
+		if (seen.has(entry.voice_id)) continue;
+		seen.add(entry.voice_id);
+		deduped.push(entry);
+	}
+	return {
+		vendor: "elevenlabs",
+		voices: deduped,
+		truncated: false,
+		...failures.length === 0 ? {} : { note: `部分端点失败（已忽略）：${failures.join("；").slice(0, 300)}` }
+	};
+}
+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");
+	if (!supportsVoiceManagement(channel)) throw new Error(`当前渠道「${channel.name}」不提供厂商音色管理接口：仅 MiniMax 与 ElevenLabs 支持音色浏览/删除`);
+	if (isMiniMax(channel)) {
+		const result = await listMiniMax(channel);
+		return {
+			vendor: result.vendor,
+			...applyFilter(result.voices, options),
+			...result.note === void 0 ? {} : { note: result.note }
+		};
+	}
+	const result = await listElevenLabs(channel, options);
+	return {
+		vendor: result.vendor,
+		...applyFilter(result.voices, options),
+		...result.note === void 0 ? {} : { note: result.note }
+	};
+}
+async function deleteVendorVoice(channel, voiceId) {
+	if (channel.apiUrl.trim() === "") throw new Error("渠道未配置 API 地址");
+	if (channel.apiKey.trim() === "") throw new Error("渠道未配置 API 密钥");
+	const id = voiceId.trim();
+	if (id === "") throw new Error("voice_id 不能为空");
+	if (isMiniMax(channel)) {
+		const known = (await listMiniMax(channel)).voices.find((entry) => entry.voice_id === id);
+		if (known !== void 0 && known.source === "system") throw new Error(`MiniMax 系统预置音色「${id}」为只读，不能删除（仅自定义音色可删）`);
+		const payload = await postJson(`${baseUrl(channel.apiUrl).replace(/\/v1$/i, "")}/v1/delete_voice`, channel.apiKey, {
+			voice_id: id,
+			voice_type: "voice_cloning"
+		});
+		if (payload.base_resp?.status_code !== void 0 && payload.base_resp.status_code !== 0) throw new Error(`MiniMax delete_voice 失败：${payload.base_resp.status_msg ?? `status ${payload.base_resp.status_code}`}`);
+		return {
+			vendor: "minimax",
+			voice_id: id,
+			deleted: true
+		};
+	}
+	if (isElevenLabs(channel)) {
+		const base = baseUrl(channel.apiUrl);
+		const headers = {
+			"xi-api-key": channel.apiKey.trim(),
+			accept: "application/json"
+		};
+		let owned = false;
+		let checkFailed = false;
+		try {
+			const payload = await fetchJson(`${base}/voices`, { headers });
+			if (Array.isArray(payload.voices)) owned = payload.voices.some((voice) => String(voice.voice_id ?? "") === id);
+		} catch {
+			checkFailed = true;
+		}
+		if (!checkFailed && !owned) throw new Error(`ElevenLabs 音色「${id}」不是账户自有音色（共享库/官方音色只读），不能删除`);
+		const response = await fetch(`${base}/voices/${encodeURIComponent(id)}`, {
+			method: "DELETE",
+			headers
+		});
+		if (!response.ok) {
+			const text = await response.text().catch(() => "");
+			throw new Error(`ElevenLabs 删除失败（HTTP ${response.status}）${text === "" ? "" : `：${text.slice(0, 300)}`}`);
+		}
+		return {
+			vendor: "elevenlabs",
+			voice_id: id,
+			deleted: true
+		};
+	}
+	throw new Error(`当前渠道「${channel.name}」不提供厂商音色管理接口：仅 MiniMax 与 ElevenLabs 支持音色浏览/删除`);
+}
+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));
+}
+function applyFilter(entries, options) {
+	const keyword = options.keyword?.trim().toLowerCase() ?? "";
+	const language = options.language?.trim().toLowerCase() ?? "";
+	const source = options.source?.trim().toLowerCase() ?? "";
+	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 (keyword !== "") {
+			const haystack = [
+				entry.name,
+				entry.description ?? "",
+				entry.accent ?? "",
+				entry.use_case ?? "",
+				entry.gender ?? "",
+				entry.age ?? ""
+			].join(" ").toLowerCase();
+			for (const token of keyword.split(/\s+/)) if (token !== "" && !haystack.includes(token)) return false;
+		}
+		return true;
+	});
+	return {
+		voices: matched.slice(0, count),
+		truncated: matched.length > count
+	};
+}
+//#endregion
 //#region src/agent-audio-tools.ts
 const audioRefSchema = {
 	type: "object",
@@ -3141,11 +3530,183 @@ function registerAgentAudioTools(ctx, resolve) {
 			};
 		}
 	}));
+	/** 厂商音色管理：浏览/筛选 + 删除（仅自建）。删除不可逆，必须 confirm=true。 */
+	const managementDisposer = ctx.tools.register(defineTool({
+		name: "manage_audio_voices",
+		description: "Manage vendor voice libraries (MiniMax / ElevenLabs). action=list: browse available TTS voices of a channel — official/shared voices plus voices designed/cloned by the account, with language/keyword/source filtering; returns voice_id/name/source/description/preview_url and whether each voice is deletable. action=delete: delete one OWNED voice (custom/owned only; official/shared/system voices are read-only and refused) — irreversible, so confirm must be true (pass the exact voice_id from action=list). Use the returned voice_id with generate_audio (mode=tts, voice=<voice_id>) to speak with the selected voice.",
+		parameters: {
+			action: {
+				type: "string",
+				enum: ["list", "delete"],
+				required: true,
+				description: "list = browse/filter voices; delete = remove an owned voice."
+			},
+			channel: {
+				type: "string",
+				description: "Channel name or id (e.g. the channel shown in settings). Defaults to the default channel; required when more than one channel is configured."
+			},
+			language: {
+				type: "string",
+				description: "Filter for list: language substring (ISO code like en/zh/ja, or a label like Chinese (Mandarin))."
+			},
+			keyword: {
+				type: "string",
+				description: "Filter for list: free text over voice name/description/accent/use_case."
+			},
+			source: {
+				type: "string",
+				enum: [
+					"system",
+					"custom",
+					"owned",
+					"shared"
+				],
+				description: "Filter for list: MiniMax system/custom; ElevenLabs owned (account) / shared (community)."
+			},
+			voice_id: {
+				type: "string",
+				description: "Required for delete: the exact voice_id from action=list."
+			},
+			confirm: {
+				type: "boolean",
+				description: "Required for delete: must be true; deletion is irreversible."
+			}
+		},
+		output: {
+			schema: {
+				type: "object",
+				additionalProperties: false,
+				properties: {
+					status: {
+						type: "string",
+						required: true,
+						enum: ["ok"]
+					},
+					kind: {
+						type: "string",
+						required: true,
+						enum: ["list", "delete"]
+					},
+					vendor: {
+						type: "string",
+						required: true
+					},
+					channel: {
+						type: "string",
+						required: true
+					},
+					count: { type: "integer" },
+					truncated: { type: "boolean" },
+					voices: {
+						type: "array",
+						items: {
+							type: "object",
+							additionalProperties: false,
+							properties: {
+								voice_id: {
+									type: "string",
+									required: true
+								},
+								name: {
+									type: "string",
+									required: true
+								},
+								source: {
+									type: "string",
+									required: true
+								},
+								deletable: {
+									type: "boolean",
+									required: true
+								},
+								language: { type: "string" },
+								accent: { type: "string" },
+								gender: { type: "string" },
+								age: { type: "string" },
+								use_case: { type: "string" },
+								description: { type: "string" },
+								preview_url: { type: "string" }
+							}
+						}
+					},
+					voice_id: { type: "string" },
+					deleted: { type: "boolean" },
+					message: { type: "string" },
+					note: { type: "string" }
+				}
+			},
+			render: (_args, value) => [{
+				type: "text",
+				text: JSON.stringify(value)
+			}]
+		},
+		timeoutMs: 12e4,
+		isConcurrencySafe: () => true,
+		async execute(args) {
+			const config = resolve();
+			if (!config.enabled) throw new AudioGenError("AI audio is disabled. Open Settings > Plugins > AI Audio and enable it.", "plugin-disabled");
+			const channel = resolveVoiceManagementChannel(config, args.channel);
+			if ((args.action === "delete" ? "delete" : "list") === "list") {
+				const source = typeof args.source === "string" && [
+					"system",
+					"custom",
+					"owned",
+					"shared"
+				].includes(args.source) ? args.source : void 0;
+				const result = await listVendorVoices(channel, {
+					...typeof args.language === "string" && args.language.trim() !== "" ? { language: args.language.trim() } : {},
+					...typeof args.keyword === "string" && args.keyword.trim() !== "" ? { keyword: args.keyword.trim() } : {},
+					...source === void 0 ? {} : { source }
+				});
+				return {
+					status: "ok",
+					kind: "list",
+					vendor: result.vendor,
+					channel: channel.name,
+					count: result.voices.length,
+					...result.truncated ? { truncated: true } : {},
+					voices: result.voices,
+					...result.note === void 0 ? {} : { note: result.note }
+				};
+			}
+			const voiceId = typeof args.voice_id === "string" ? args.voice_id.trim() : "";
+			if (voiceId === "") throw new AudioGenError("delete requires voice_id (the exact voice_id from action=list).", "voice-id-required");
+			if (args.confirm !== true) throw new AudioGenError("Deletion is irreversible: pass confirm=true after verifying the voice_id.", "voice-delete-requires-confirm");
+			const deleted = await deleteVendorVoice(channel, voiceId);
+			return {
+				status: "ok",
+				kind: "delete",
+				vendor: deleted.vendor,
+				channel: channel.name,
+				voice_id: deleted.voice_id,
+				deleted: true,
+				message: `已删除音色 ${deleted.voice_id}（${channel.name}）`
+			};
+		}
+	}));
 	return () => {
 		disposer();
 		searchDisposer();
+		managementDisposer();
 	};
 }
+/** 解析工具目标渠道：默认渠道 > 唯一可用渠道；多渠道未指定时要求显式选择。 */
+function resolveVoiceManagementChannel(config, requested) {
+	const usable = config.channels.filter((channel) => channel.apiUrl.trim() !== "" && channel.apiKey.trim() !== "");
+	if (usable.length === 0) throw new AudioGenError("Audio API credentials are not configured. Open Settings > Plugins > AI Audio, add a channel and fill its API URL and API key.", "audio-api-not-configured");
+	const wanted = typeof requested === "string" ? requested.trim() : "";
+	if (wanted === "") {
+		if (usable.length === 1) return usable[0];
+		const target = usable.find((channel) => channel.id === config.defaultChannelId);
+		if (target !== void 0) return target;
+		throw new AudioGenError(`Multiple audio channels are configured — specify the channel (one of: ${usable.map((channel) => `"${channel.name}"`).join(", ")}).`, "channel-choice-required");
+	}
+	const direct = usable.find((channel) => channel.name === wanted || channel.id === wanted);
+	if (direct !== void 0) return direct;
+	const partial = usable.filter((channel) => channel.name.toLowerCase().includes(wanted.toLowerCase()));
+	if (partial.length === 1) return partial[0];
+	throw new AudioGenError(`Audio channel "${wanted}" is not configured. Choose one of: ${usable.map((channel) => channel.name).join(", ")}.`, "channel-not-configured");
+}
 //#endregion
 //#region src/index.ts
 /** Stable cordis plugin name. */
@@ -3180,7 +3741,7 @@ const DEFAULT_ENABLED = true;
 const DEFAULT_ANNOUNCE = true;
 const DEFAULT_ALLOW_AGENT_AUDIO = true;
 const SECTION_ORDER = 160;
-const AUDIOGEN_GUIDANCE = "本机已安装 dsh-audiogen 插件（DSH AI 音频）：侧边栏「AI 音频」入口。能力：通过「渠道」对接多个音频生成厂商（OpenAI TTS、ElevenLabs、MiniMax、Stability Audio、自定义 OpenAI 兼容接口），支持 TTS 文本转语音、音乐生成和音效生成。API 地址与密钥在 GUI 设置中按渠道配置，密钥仅存于本机设置文档；生成请求由本地宿主代理转发。Agent 可直接调用 `generate_audio` 提交 TTS/音乐/音效任务，默认等待完成并返回同源音频 URL。限制：生成消耗上游 API 额度；音频内容由上游模型生成；模型只能使用用户在各渠道配置目录中的模型。用户提到「音频 / 语音 / TTS / 配乐 / 音效 / AI 音频」时即指本插件，请据此协作。";
+const AUDIOGEN_GUIDANCE = "本机已安装 dsh-audiogen 插件（DSH AI 音频）：侧边栏「AI 音频」入口。能力：通过「渠道」对接多个音频生成厂商（OpenAI TTS、ElevenLabs、MiniMax、Stability Audio、自定义 OpenAI 兼容接口），支持 TTS 文本转语音、音乐生成和音效生成。API 地址与密钥在 GUI 设置中按渠道配置，密钥仅存于本机设置文档；生成请求由本地宿主代理转发。Agent 可直接调用 `generate_audio` 提交 TTS/音乐/音效任务，默认等待完成并返回同源音频 URL；可用 `manage_audio_voices` 浏览/筛选/删除厂商音色（MiniMax、ElevenLabs），再用选定音色的 voice_id 调用 `generate_audio` 生成。限制：生成消耗上游 API 额度；音频内容由上游模型生成；模型只能使用用户在各渠道配置目录中的模型。用户提到「音频 / 语音 / TTS / 配乐 / 音效 / AI 音频」时即指本插件，请据此协作。";
 function guidanceFor(channels, defaultChannelId) {
 	if (channels.length === 0) return `${AUDIOGEN_GUIDANCE} 尚未配置任何渠道：请先在「设置 → 插件 → AI 音频」添加渠道并填写 API 地址与密钥。`;
 	const table = channels.map((channel) => {