dsh-audiogen 0.1.0 → 0.3.0

package/lib/index.js

@@ -22,6 +22,8 @@ const SETTINGS_API = {
 const GENERATE_API = "/api/dsh-audiogen/generate";
 /** Host-mediated built-in provider catalog (channels the user can instantiate). */
 const PRESETS_API = "/api/dsh-audiogen/presets";
+/** Host-mediated model/voice discovery endpoint. */
+const MODEL_API = { discover: "/api/dsh-audiogen/models/discover" };
 /** Loopback-only audio file reader for panel/tool-result previews. */
 const AUDIO_API = { file: "/api/dsh-audiogen/audio" };
 /** Host-persisted generation history routes. */
@@ -91,7 +93,7 @@ function isOpenAICompatible(channel, mode) {
 function isElevenLabs(channel) {
 	return isPreset(channel, "elevenlabs") || /elevenlabs/i.test(channel.apiUrl);
 }
-function isMiniMax(channel) {
+function isMiniMax$1(channel) {
 	return isPreset(channel, "minimax") || /minimax/i.test(channel.apiUrl);
 }
 function isStability(channel) {
@@ -183,13 +185,14 @@ async function normalizeAudioResponse(response, options) {
 		} catch {
 			throw new AudioGenError("audio endpoint returned an unprocessable response body", "audio-bad-response");
 		}
-		const base64 = findBase64Audio(parsed);
-		if (base64 !== void 0 && base64.length > 0) {
+		const encoded = findBase64Audio(parsed);
+		if (encoded !== void 0 && encoded.length > 0) {
 			let data;
 			try {
-				data = new Uint8Array(Buffer.from(base64, "base64"));
+				const isHex = /^[0-9a-fA-F]+$/.test(encoded) && encoded.length % 2 === 0;
+				data = new Uint8Array(Buffer.from(encoded, isHex ? "hex" : "base64"));
 			} catch {
-				throw new AudioGenError("audio endpoint returned invalid base64", "audio-bad-response");
+				throw new AudioGenError("audio endpoint returned invalid audio encoding", "audio-bad-response");
 			}
 			return [{
 				data,
@@ -274,27 +277,79 @@ async function elevenLabs(channel, request, signal) {
 		fallbackMime: "audio/mpeg"
 	});
 }
+function minimaxApiBase(base) {
+	const trimmed = endpointBase(base);
+	return /\/v1$/i.test(trimmed) ? trimmed : `${trimmed}/v1`;
+}
 async function minimax(channel, request, signal) {
-	const base = endpointBase(channel.apiUrl);
-	const endpoint = /\/t2a_v2(\?|$)/i.test(base) ? base : `${base}/t2a_v2`;
-	const model = (request.upstream ?? request.model) || "speech-01-turbo";
+	const base = minimaxApiBase(channel.apiUrl);
+	const model = (request.upstream ?? request.model) || (request.mode === "music" ? "music-3.0" : "speech-2.8-hd");
 	const voice = request.voice ?? request.model ?? "";
-	const body = {
-		model,
-		text: request.prompt,
-		stream: false,
-		...voice === "" ? {} : { voice_setting: {
-			voice_id: voice,
-			...request.speed !== void 0 ? { speed: request.speed } : {},
-			vol: 1,
-			pitch: 0
-		} },
-		audio_setting: {
-			format: request.format ?? "mp3",
-			sample_rate: 32e3,
-			bitrate: 128e3
+	if (request.mode === "voice_design") {
+		const endpoint = `${base}/voice_design`;
+		const body = {
+			prompt: request.prompt,
+			preview_text: request.previewText ?? request.voice ?? "你好，这是新设计的音色试听。"
+		};
+		const response = await fetchWithTimeout(endpoint, {
+			method: "POST",
+			redirect: "error",
+			headers: {
+				authorization: `Bearer ${channel.apiKey.trim()}`,
+				"content-type": "application/json",
+				accept: "application/json"
+			},
+			body: JSON.stringify(body),
+			signal
+		}, UPSTREAM_TIMEOUT_MS);
+		if (!response.ok) {
+			const detail = await response.text().catch(() => "");
+			throw new AudioGenError(`MiniMax voice design API error (HTTP ${response.status})${detail === "" ? "" : `: ${detail.slice(0, 300)}`}`, "audio-api-error");
 		}
-	};
+		const payload = await response.json();
+		if (payload.base_resp?.status_code !== void 0 && payload.base_resp.status_code !== 0) throw new AudioGenError(payload.base_resp.status_msg ?? `MiniMax returned status ${payload.base_resp.status_code}`, "audio-api-error");
+		const encoded = payload.trial_audio ?? "";
+		if (encoded === "") throw new AudioGenError("MiniMax voice design returned no trial audio", "audio-empty-result");
+		const isHex = /^[0-9a-fA-F]+$/.test(encoded) && encoded.length % 2 === 0;
+		return [{
+			data: new Uint8Array(Buffer.from(encoded, isHex ? "hex" : "base64")),
+			mime: "audio/mpeg",
+			...payload.voice_id === void 0 ? {} : { voiceId: payload.voice_id }
+		}];
+	}
+	let endpoint;
+	let body;
+	if (request.mode === "music") {
+		endpoint = `${base}/music_generation`;
+		body = {
+			model,
+			prompt: request.prompt,
+			...request.duration !== void 0 ? { duration: request.duration } : {},
+			audio_setting: {
+				format: request.format ?? "mp3",
+				sample_rate: 44100,
+				bitrate: 256e3
+			}
+		};
+	} else {
+		endpoint = `${base}/t2a_v2`;
+		body = {
+			model,
+			text: request.prompt,
+			stream: false,
+			...voice === "" ? {} : { voice_setting: {
+				voice_id: voice,
+				...request.speed !== void 0 ? { speed: request.speed } : {},
+				vol: 1,
+				pitch: 0
+			} },
+			audio_setting: {
+				format: request.format ?? "mp3",
+				sample_rate: 32e3,
+				bitrate: 128e3
+			}
+		};
+	}
 	return normalizeAudioResponse(await fetchWithTimeout(endpoint, {
 		method: "POST",
 		redirect: "error",
@@ -369,8 +424,9 @@ 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)) throw new AudioGenError("音色设计当前仅支持 MiniMax 渠道", "voice-design-unsupported");
 	if (isElevenLabs(channel)) return elevenLabs(channel, request, signal);
-	if (isMiniMax(channel)) return minimax(channel, request, signal);
+	if (isMiniMax$1(channel)) return minimax(channel, request, signal);
 	if (isStability(channel)) return stabilityAudio(channel, request, signal);
 	if (isOpenAICompatible(channel, request.mode)) return openAITTS(channel, request, signal);
 	return genericAudio(channel, request, signal);
@@ -386,15 +442,18 @@ const AUDIO_PRESETS = [
 		models: [
 			{
 				alias: "tts-1",
-				id: "tts-1"
+				id: "tts-1",
+				category: "tts"
 			},
 			{
 				alias: "tts-1-hd",
-				id: "tts-1-hd"
+				id: "tts-1-hd",
+				category: "tts"
 			},
 			{
 				alias: "gpt-4o-mini-tts",
-				id: "gpt-4o-mini-tts"
+				id: "gpt-4o-mini-tts",
+				category: "tts"
 			}
 		]
 	},
@@ -406,43 +465,86 @@ const AUDIO_PRESETS = [
 		models: [
 			{
 				alias: "Rachel",
-				id: "21m00Tcm4TlvDq8ikWAM"
+				id: "21m00Tcm4TlvDq8ikWAM",
+				category: "tts"
 			},
 			{
 				alias: "Adam",
-				id: "pNInz6obpgDQGcFmaJgB"
+				id: "pNInz6obpgDQGcFmaJgB",
+				category: "tts"
 			},
 			{
 				alias: "Antoni",
-				id: "ErXwobaYiN019PkySvjV"
+				id: "ErXwobaYiN019PkySvjV",
+				category: "tts"
 			},
 			{
 				alias: "Bella",
-				id: "EXAVITQu4vr4xnSDxMaL"
+				id: "EXAVITQu4vr4xnSDxMaL",
+				category: "tts"
 			}
 		]
 	},
 	{
 		id: "minimax",
 		name: "MiniMax",
-		apiUrl: "https://api.minimax.chat/v1",
-		hint: "MiniMax 语音合成（T2A）；需在 API URL 后按官方要求携带 GroupId 或使用完整接口地址",
+		apiUrl: "https://api.minimaxi.com",
+		hint: "MiniMax 音色设计 / TTS / 音乐生成；可使用“获取可用模型”拉取账号音色",
 		models: [
 			{
-				alias: "speech-01-turbo",
-				id: "speech-01-turbo"
+				alias: "speech-2.8-hd",
+				id: "speech-2.8-hd",
+				category: "tts"
 			},
 			{
-				alias: "speech-01-hd",
-				id: "speech-01-hd"
+				alias: "speech-2.8-turbo",
+				id: "speech-2.8-turbo",
+				category: "tts"
 			},
 			{
-				alias: "speech-02-turbo",
-				id: "speech-02-turbo"
+				alias: "speech-2.6-hd",
+				id: "speech-2.6-hd",
+				category: "tts"
+			},
+			{
+				alias: "speech-2.6-turbo",
+				id: "speech-2.6-turbo",
+				category: "tts"
 			},
 			{
 				alias: "speech-02-hd",
-				id: "speech-02-hd"
+				id: "speech-02-hd",
+				category: "tts"
+			},
+			{
+				alias: "speech-02-turbo",
+				id: "speech-02-turbo",
+				category: "tts"
+			},
+			{
+				alias: "speech-01-hd",
+				id: "speech-01-hd",
+				category: "tts"
+			},
+			{
+				alias: "speech-01-turbo",
+				id: "speech-01-turbo",
+				category: "tts"
+			},
+			{
+				alias: "music-3.0",
+				id: "music-3.0",
+				category: "music"
+			},
+			{
+				alias: "music-2.6",
+				id: "music-2.6",
+				category: "music"
+			},
+			{
+				alias: "music-cover",
+				id: "music-cover",
+				category: "music"
 			}
 		]
 	},
@@ -453,10 +555,12 @@ const AUDIO_PRESETS = [
 		hint: "Stability AI 音乐/音效生成（stable-audio 系列）",
 		models: [{
 			alias: "stable-audio-2.0",
-			id: "stable-audio-2.0"
+			id: "stable-audio-2.0",
+			category: "music"
 		}, {
 			alias: "stable-audio-1.0",
-			id: "stable-audio-1.0"
+			id: "stable-audio-1.0",
+			category: "music"
 		}]
 	},
 	{
@@ -472,6 +576,113 @@ function audioPresetById(id) {
 	return AUDIO_PRESETS.find((preset) => preset.id === id);
 }
 //#endregion
+//#region src/audio-models.ts
+function isMiniMax(channel) {
+	return channel.preset === "minimax" || /minimax/i.test(channel.apiUrl);
+}
+function baseUrl(url) {
+	return url.trim().replace(/\/+$/, "");
+}
+function categoryFor(id) {
+	const value = id.toLowerCase();
+	if (/(tts|speech|voice|t2a)/i.test(value)) return "tts";
+	if (/(music|song|cover|lyrics)/i.test(value)) return "music";
+	if (/(sfx|sound.?effect|effect|foley)/i.test(value)) return "sfx";
+}
+async function postJson(url, apiKey, body) {
+	const response = await fetch(url, {
+		method: "POST",
+		headers: {
+			authorization: `Bearer ${apiKey.trim()}`,
+			"content-type": "application/json"
+		},
+		body: JSON.stringify(body)
+	});
+	if (!response.ok) {
+		const text = await response.text().catch(() => "");
+		throw new Error(`HTTP ${response.status}${text === "" ? "" : `: ${text.slice(0, 300)}`}`);
+	}
+	return response.json();
+}
+/** Discover available models/voices for a channel. */
+async function discoverAudioModels(channel) {
+	if (channel.apiUrl.trim() === "") throw new Error("API URL is not configured");
+	if (channel.apiKey.trim() === "") throw new Error("API key is not configured");
+	if (isMiniMax(channel)) {
+		const payload = await postJson(`${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(payload.base_resp.status_msg ?? `MiniMax returned status ${payload.base_resp.status_code}`);
+		const models = [];
+		for (const voice of payload.system_voice ?? []) {
+			const id = voice.voice_id?.trim() ?? "";
+			if (id === "") continue;
+			models.push({
+				alias: voice.voice_name?.trim() || id,
+				id,
+				category: "tts",
+				...voice.description !== void 0 && voice.description.length > 0 ? { description: voice.description.join("；") } : {}
+			});
+		}
+		for (const voice of payload.voice_cloning ?? []) {
+			const id = voice.voice_id?.trim() ?? "";
+			if (id === "") continue;
+			models.push({
+				alias: id,
+				id,
+				category: "tts",
+				...voice.description !== void 0 && voice.description.length > 0 ? { description: voice.description.join("；") } : {}
+			});
+		}
+		for (const voice of payload.voice_generation ?? []) {
+			const id = voice.voice_id?.trim() ?? "";
+			if (id === "") continue;
+			models.push({
+				alias: id,
+				id,
+				category: "tts",
+				...voice.description !== void 0 && voice.description.length > 0 ? { description: voice.description.join("；") } : {}
+			});
+		}
+		const music = (audioPresetById("minimax")?.models ?? []).filter((model) => model.category === "music");
+		for (const model of music) models.push({
+			...model,
+			category: "music"
+		});
+		return {
+			models: dedupe(models),
+			source: "MiniMax get_voice + built-in music catalog"
+		};
+	}
+	const url = `${baseUrl(channel.apiUrl)}/models`;
+	const response = await fetch(url, { headers: { authorization: `Bearer ${channel.apiKey.trim()}` } });
+	if (!response.ok) throw new Error(`model list request failed (HTTP ${response.status}); please add models manually`);
+	const payload = await response.json();
+	const models = [];
+	for (const item of payload.data ?? []) {
+		const id = item.id?.trim() ?? "";
+		if (id === "") continue;
+		const category = categoryFor(id) ?? "tts";
+		models.push({
+			alias: id,
+			id,
+			category
+		});
+	}
+	return {
+		models: dedupe(models),
+		source: "OpenAI-compatible /models"
+	};
+}
+function dedupe(models) {
+	const seen = /* @__PURE__ */ new Set();
+	const out = [];
+	for (const model of models) {
+		if (seen.has(model.id)) continue;
+		seen.add(model.id);
+		out.push(model);
+	}
+	return out;
+}
+//#endregion
 //#region src/audio-store.ts
 /**
 * Host-side persistence for generated audio and generation history.
@@ -553,13 +764,15 @@ async function appendHistory(entry) {
 		model: entry.model,
 		prompt: entry.prompt,
 		...entry.voice === void 0 ? {} : { voice: entry.voice },
+		...entry.voiceId === void 0 ? {} : { voiceId: entry.voiceId },
 		...entry.speed === void 0 ? {} : { speed: entry.speed },
 		...entry.duration === void 0 ? {} : { duration: entry.duration },
 		...entry.format === void 0 ? {} : { format: entry.format },
 		audio: entry.audio.map((audio) => ({
 			url: audio.url,
 			mime: audio.mime,
-			...audio.duration === void 0 ? {} : { duration: audio.duration }
+			...audio.duration === void 0 ? {} : { duration: audio.duration },
+			...audio.voiceId === void 0 ? {} : { voiceId: audio.voiceId }
 		})),
 		...entry.channelId === void 0 ? {} : { channelId: entry.channelId },
 		...entry.channel === void 0 ? {} : { channel: entry.channel }
@@ -631,7 +844,7 @@ function messageOf(error) {
 	return error instanceof Error ? error.message : String(error);
 }
 function parseGenerateRequest(body) {
-	const mode = body.mode === "music" ? "music" : body.mode === "sfx" ? "sfx" : "tts";
+	const mode = body.mode === "music" ? "music" : body.mode === "sfx" ? "sfx" : body.mode === "voice_design" ? "voice_design" : "tts";
 	const prompt = typeof body.prompt === "string" ? body.prompt.trim() : "";
 	if (prompt === "") return void 0;
 	return {
@@ -639,6 +852,7 @@ function parseGenerateRequest(body) {
 		model: typeof body.model === "string" ? body.model.trim() : "",
 		prompt,
 		...typeof body.voice === "string" && body.voice.trim() !== "" ? { voice: body.voice.trim() } : {},
+		...typeof body.previewText === "string" && body.previewText.trim() !== "" ? { previewText: body.previewText.trim() } : {},
 		...typeof body.speed === "number" ? { speed: body.speed } : {},
 		...typeof body.duration === "number" ? { duration: body.duration } : {},
 		...typeof body.format === "string" && body.format.trim() !== "" ? { format: body.format.trim() } : {},
@@ -685,6 +899,21 @@ function resolveChannelRequest(request, view) {
 	const target = explicit ?? defaults;
 	const asked = request.model.trim();
 	if (asked === "") {
+		if (request.mode === "voice_design") {
+			if (target === void 0) return {
+				ok: false,
+				code: "no-channels",
+				message: "尚未配置任何渠道"
+			};
+			return {
+				ok: true,
+				request: {
+					...request,
+					channelId: target.id,
+					channel: target.name
+				}
+			};
+		}
 		const alias = target?.models[0]?.alias ?? "";
 		if (alias === "") return {
 			ok: false,
@@ -758,6 +987,36 @@ function makeRoutes(deps) {
 				});
 			}
 		},
+		{
+			kind: "exact",
+			path: MODEL_API.discover,
+			handler: async (req, res) => {
+				if (!guard(req, res, "POST")) return;
+				const body = await readJsonBody(req);
+				const view = deps.resolveChannels();
+				const stored = view.channels.find((candidate) => candidate.id === (typeof body?.channelId === "string" ? body.channelId : void 0)) ?? view.channels.find((candidate) => candidate.id === view.defaultChannelId) ?? view.channels[0];
+				const channel = {
+					id: stored?.id ?? "preview",
+					preset: stored?.preset ?? "",
+					name: stored?.name ?? "",
+					apiUrl: typeof body?.apiUrl === "string" && body.apiUrl.trim() !== "" ? body.apiUrl.trim() : stored?.apiUrl ?? "",
+					apiKey: typeof body?.apiKey === "string" && body.apiKey.trim() !== "" ? body.apiKey.trim() : stored?.apiKey ?? "",
+					models: stored?.models ?? []
+				};
+				try {
+					writeJson(res, 200, {
+						ok: true,
+						...await discoverAudioModels(channel)
+					});
+				} catch (error) {
+					writeJson(res, 200, {
+						ok: false,
+						code: "model-discovery-failed",
+						message: messageOf(error)
+					});
+				}
+			}
+		},
 		{
 			kind: "exact",
 			path: SETTINGS_API.describe,
@@ -855,7 +1114,8 @@ function makeRoutes(deps) {
 							b64: Buffer.from(output.data).toString("base64"),
 							mime: saved.mime,
 							bytes: saved.bytes,
-							url: `${AUDIO_API.file}/${encodeURIComponent(saved.file)}`
+							url: `${AUDIO_API.file}/${encodeURIComponent(saved.file)}`,
+							...output.voiceId === void 0 ? {} : { voiceId: output.voiceId }
 						});
 					}
 					let history;
@@ -1012,7 +1272,8 @@ const resultSchema = {
 			enum: [
 				"tts",
 				"music",
-				"sfx"
+				"sfx",
+				"voice_design"
 			]
 		},
 		model: {
@@ -1041,7 +1302,8 @@ const resultSchema = {
 					bytes: {
 						type: "integer",
 						required: true
-					}
+					},
+					voiceId: { type: "string" }
 				}
 			}
 		},
@@ -1079,7 +1341,7 @@ function ensureConfigured(config) {
 function registerAgentAudioTools(ctx, resolve) {
 	return ctx.tools.register(defineTool({
 		name: "generate_audio",
-		description: "Generate audio with the configured audio provider. Supports text-to-speech, music generation and sound effects. The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.",
+		description: "Generate audio with the configured audio provider. Supports text-to-speech, music generation, sound effects and MiniMax voice design. The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.",
 		parameters: {
 			prompt: {
 				type: "string",
@@ -1091,7 +1353,8 @@ function registerAgentAudioTools(ctx, resolve) {
 				enum: [
 					"tts",
 					"music",
-					"sfx"
+					"sfx",
+					"voice_design"
 				],
 				description: "Generation mode. Defaults to tts."
 			},
@@ -1103,6 +1366,10 @@ function registerAgentAudioTools(ctx, resolve) {
 				type: "string",
 				description: "Optional voice id/name for TTS providers."
 			},
+			preview_text: {
+				type: "string",
+				description: "Optional preview text for voice_design."
+			},
 			speed: {
 				type: "number",
 				description: "Optional speaking rate / speed multiplier where supported."
@@ -1125,15 +1392,26 @@ function registerAgentAudioTools(ctx, resolve) {
 		async execute(args, exec) {
 			const config = resolve();
 			ensureConfigured(config);
-			const picked = resolveModel(config, args.model);
+			const mode = args.mode === "music" ? "music" : args.mode === "sfx" ? "sfx" : args.mode === "voice_design" ? "voice_design" : "tts";
+			const picked = mode === "voice_design" ? (() => {
+				const usable = config.channels.filter((channel) => channel.apiUrl.trim() !== "" && channel.apiKey.trim() !== "");
+				const target = usable.find((channel) => channel.id === config.defaultChannelId) ?? usable[0];
+				if (target === void 0) throw new AudioGenError("No usable audio channel is configured for voice design.", "no-channel-available");
+				return {
+					channel: target,
+					alias: "",
+					upstream: ""
+				};
+			})() : resolveModel(config, args.model);
 			const request = {
-				mode: args.mode === "music" ? "music" : args.mode === "sfx" ? "sfx" : "tts",
+				mode,
 				model: picked.alias,
 				upstream: picked.upstream,
 				channelId: picked.channel.id,
 				channel: picked.channel.name,
 				prompt: args.prompt.trim(),
 				...typeof args.voice === "string" && args.voice.trim() !== "" ? { voice: args.voice.trim() } : {},
+				...typeof args.preview_text === "string" && args.preview_text.trim() !== "" ? { previewText: args.preview_text.trim() } : {},
 				...typeof args.speed === "number" ? { speed: args.speed } : {},
 				...typeof args.duration === "number" ? { duration: args.duration } : {},
 				...typeof args.format === "string" && args.format.trim() !== "" ? { format: args.format.trim() } : {}
@@ -1147,7 +1425,8 @@ function registerAgentAudioTools(ctx, resolve) {
 						id: saved.id,
 						url: `/api/dsh-audiogen/audio/${encodeURIComponent(saved.file)}`,
 						mime: saved.mime,
-						bytes: saved.bytes
+						bytes: saved.bytes,
+						...output.voiceId === void 0 ? {} : { voiceId: output.voiceId }
 					});
 				}
 				try {
@@ -1166,7 +1445,8 @@ function registerAgentAudioTools(ctx, resolve) {
 							b64: Buffer.from(output.data).toString("base64"),
 							mime: audio[index].mime,
 							bytes: audio[index].bytes,
-							url: audio[index].url
+							url: audio[index].url,
+							...output.voiceId === void 0 ? {} : { voiceId: output.voiceId }
 						})),
 						channelId: picked.channel.id,
 						channel: picked.channel.name

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": "0.1.0",
+  "version": "0.3.0",
   "type": "module",
   "main": "lib/index.js",
   "exports": {

package/src/agent-audio-tools.ts

@@ -24,6 +24,7 @@ interface AgentAudioRef {
   url: string
   mime: string
   bytes: number
+  voiceId?: string
 }
 
 interface AgentAudioResult {
@@ -43,6 +44,7 @@ const audioRefSchema = {
     url: { type: 'string', required: true },
     mime: { type: 'string', required: true },
     bytes: { type: 'integer', required: true },
+    voiceId: { type: 'string' },
   },
 } as const
 
@@ -52,7 +54,7 @@ const resultSchema = {
   properties: {
     status: { type: 'string', required: true },
     message: { type: 'string', required: true },
-    mode: { type: 'string', required: true, enum: ['tts', 'music', 'sfx'] },
+    mode: { type: 'string', required: true, enum: ['tts', 'music', 'sfx', 'voice_design'] },
     model: { type: 'string', required: true },
     audio: { type: 'array', required: true, items: audioRefSchema },
     error: { type: 'string' },
@@ -98,12 +100,13 @@ function ensureConfigured(config: AgentAudioToolConfig): void {
 export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioToolConfig): () => void {
   const disposer = ctx.tools.register(defineTool({
     name: 'generate_audio',
-    description: 'Generate audio with the configured audio provider. Supports text-to-speech, music generation and sound effects. The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.',
+    description: 'Generate audio with the configured audio provider. Supports text-to-speech, music generation, sound effects and MiniMax voice design. The tool call waits for the upstream result and returns same-origin audio URLs; pass those URLs to the user for playback or download. If multiple models are configured, first ask the user which one to use or pass model explicitly.',
     parameters: {
       prompt: { type: 'string', required: true, description: 'For tts, the text to speak. For music/sfx, a descriptive prompt.' },
-      mode: { type: 'string', enum: ['tts', 'music', 'sfx'], description: 'Generation mode. Defaults to tts.' },
+      mode: { type: 'string', enum: ['tts', 'music', 'sfx', 'voice_design'], description: 'Generation mode. Defaults to tts.' },
       model: { type: 'string', description: 'One of the configured audio models/voices. Defaults to the first configured model.' },
       voice: { type: 'string', description: 'Optional voice id/name for TTS providers.' },
+      preview_text: { type: 'string', description: 'Optional preview text for voice_design.' },
       speed: { type: 'number', description: 'Optional speaking rate / speed multiplier where supported.' },
       duration: { type: 'number', description: 'Requested duration in seconds for music/sfx.' },
       format: { type: 'string', description: 'Output format such as mp3 or wav.' },
@@ -117,15 +120,24 @@ export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioT
     async execute(args, exec) {
       const config = resolve()
       ensureConfigured(config)
-      const picked = resolveModel(config, args.model)
+      const mode = args.mode === 'music' ? 'music' : args.mode === 'sfx' ? 'sfx' : args.mode === 'voice_design' ? 'voice_design' : 'tts'
+      const picked = mode === 'voice_design'
+        ? (() => {
+          const usable = config.channels.filter(channel => channel.apiUrl.trim() !== '' && channel.apiKey.trim() !== '')
+          const target = usable.find(channel => channel.id === config.defaultChannelId) ?? usable[0]
+          if (target === undefined) throw new AudioGenError('No usable audio channel is configured for voice design.', 'no-channel-available')
+          return { channel: target, alias: '', upstream: '' }
+        })()
+        : resolveModel(config, args.model)
       const request: GenerateAudioRequest = {
-        mode: args.mode === 'music' ? 'music' : args.mode === 'sfx' ? 'sfx' : 'tts',
+        mode,
         model: picked.alias,
         upstream: picked.upstream,
         channelId: picked.channel.id,
         channel: picked.channel.name,
         prompt: args.prompt.trim(),
         ...(typeof args.voice === 'string' && args.voice.trim() !== '' ? { voice: args.voice.trim() } : {}),
+        ...(typeof args.preview_text === 'string' && args.preview_text.trim() !== '' ? { previewText: args.preview_text.trim() } : {}),
         ...(typeof args.speed === 'number' ? { speed: args.speed } : {}),
         ...(typeof args.duration === 'number' ? { duration: args.duration } : {}),
         ...(typeof args.format === 'string' && args.format.trim() !== '' ? { format: args.format.trim() } : {}),
@@ -140,6 +152,7 @@ export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioT
             url: `/api/dsh-audiogen/audio/${encodeURIComponent(saved.file)}`,
             mime: saved.mime,
             bytes: saved.bytes,
+            ...(output.voiceId === undefined ? {} : { voiceId: output.voiceId }),
           })
         }
         try {
@@ -159,6 +172,7 @@ export function registerAgentAudioTools(ctx: Context, resolve: () => AgentAudioT
               mime: audio[index]!.mime,
               bytes: audio[index]!.bytes,
               url: audio[index]!.url,
+              ...(output.voiceId === undefined ? {} : { voiceId: output.voiceId }),
             })),
             channelId: picked.channel.id,
             channel: picked.channel.name,