dsh-audiogen 0.1.0 → 0.2.0

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 +277,47 @@ 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
-		}
-	};
+	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",
@@ -370,7 +393,7 @@ async function generateAudio(channel, request, signal) {
 	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 (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 +409,18 @@ 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 +432,86 @@ 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 +522,12 @@ 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 +543,113 @@ 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.
@@ -758,6 +936,36 @@ 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,

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": "0.2.0",
   "type": "module",
   "main": "lib/index.js",
   "exports": {

package/src/audio-engine.ts

@@ -187,13 +187,14 @@ async function normalizeAudioResponse(
     } catch {
       throw new AudioGenError('audio endpoint returned an unprocessable response body', 'audio-bad-response')
     }
-    const base64 = findBase64Audio(parsed)
-    if (base64 !== undefined && base64.length > 0) {
+    const encoded = findBase64Audio(parsed)
+    if (encoded !== undefined && encoded.length > 0) {
       let data: Uint8Array
       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, mime: detectAudioMime(data) ?? contentType ?? 'audio/mpeg' }]
     }
@@ -268,27 +269,50 @@ async function elevenLabs(channel: AudioChannel, request: GenerateAudioRequest,
   return normalizeAudioResponse(response, { apiKey: channel.apiKey, fallbackMime: 'audio/mpeg' })
 }
 
+function minimaxApiBase(base: string): string {
+  const trimmed = endpointBase(base)
+  return /\/v1$/i.test(trimmed) ? trimmed : `${trimmed}/v1`
+}
+
 async function minimax(channel: AudioChannel, request: GenerateAudioRequest, signal?: AbortSignal): Promise<Array<{ data: Uint8Array; mime: string }>> {
-  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: Record<string, unknown> = {
-    model,
-    text: request.prompt,
-    stream: false,
-    ...(voice === '' ? {} : { voice_setting: {
-      voice_id: voice,
-      ...(request.speed !== undefined ? { speed: request.speed } : {}),
-      vol: 1,
-      pitch: 0,
-    } }),
-    audio_setting: {
-      format: request.format ?? 'mp3',
-      sample_rate: 32000,
-      bitrate: 128000,
-    },
+
+  let endpoint: string
+  let body: Record<string, unknown>
+  if (request.mode === 'music') {
+    endpoint = `${base}/music_generation`
+    body = {
+      model,
+      prompt: request.prompt,
+      ...(request.duration !== undefined ? { duration: request.duration } : {}),
+      audio_setting: {
+        format: request.format ?? 'mp3',
+        sample_rate: 44100,
+        bitrate: 256000,
+      },
+    }
+  } else {
+    endpoint = `${base}/t2a_v2`
+    body = {
+      model,
+      text: request.prompt,
+      stream: false,
+      ...(voice === '' ? {} : { voice_setting: {
+        voice_id: voice,
+        ...(request.speed !== undefined ? { speed: request.speed } : {}),
+        vol: 1,
+        pitch: 0,
+      } }),
+      audio_setting: {
+        format: request.format ?? 'mp3',
+        sample_rate: 32000,
+        bitrate: 128000,
+      },
+    }
   }
+
   const response = await fetchWithTimeout(endpoint, {
     method: 'POST',
     redirect: 'error',

package/src/audio-models.ts

@@ -0,0 +1,129 @@
+/**
+ * Host-side model/voice discovery.
+ *
+ * MiniMax exposes a voice-management API that returns all available system and
+ * user-generated voice ids; we combine those with the known MiniMax music
+ * models so the settings card can offer a full categorized catalog.
+ */
+
+import type { AudioChannel } from './audio-engine.ts'
+import type { AudioModelCategory, DiscoveredAudioModel } from './protocol.ts'
+import { audioPresetById } from './audio-presets.ts'
+
+function isMiniMax(channel: AudioChannel): boolean {
+  return channel.preset === 'minimax' || /minimax/i.test(channel.apiUrl)
+}
+
+function baseUrl(url: string): string {
+  return url.trim().replace(/\/+$/, '')
+}
+
+function categoryFor(id: string): AudioModelCategory | undefined {
+  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'
+  return undefined
+}
+
+async function postJson(url: string, apiKey: string, body: unknown): Promise<unknown> {
+  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 === '' ? '' : `: ${text.slice(0, 300)}`}`)
+  }
+  return response.json()
+}
+
+/** Discover available models/voices for a channel. */
+export async function discoverAudioModels(channel: AudioChannel): Promise<{ models: DiscoveredAudioModel[]; source: string }> {
+  if (channel.apiUrl.trim() === '') throw new Error('API URL is not configured')
+  if (channel.apiKey.trim() === '') throw new Error('API key is not configured')
+
+  if (isMiniMax(channel)) {
+    const base = baseUrl(channel.apiUrl).replace(/\/v1$/i, '')
+    const url = `${base}/v1/get_voice`
+    const payload = await postJson(url, channel.apiKey, { voice_type: 'all' }) as {
+      system_voice?: Array<{ voice_id?: string; voice_name?: string; description?: string[] }>
+      voice_cloning?: Array<{ voice_id?: string; description?: string[] }>
+      voice_generation?: Array<{ voice_id?: string; description?: string[] }>
+      base_resp?: { status_code?: number; status_msg?: string }
+    }
+    if (payload.base_resp?.status_code !== undefined && payload.base_resp.status_code !== 0) {
+      throw new Error(payload.base_resp.status_msg ?? `MiniMax returned status ${payload.base_resp.status_code}`)
+    }
+    const models: DiscoveredAudioModel[] = []
+    for (const voice of payload.system_voice ?? []) {
+      const id = voice.voice_id?.trim() ?? ''
+      if (id === '') continue
+      models.push({
+        alias: voice.voice_name?.trim() || id,
+        id,
+        category: 'tts',
+        ...(voice.description !== undefined && voice.description.length > 0 ? { description: voice.description.join('；') } : {}),
+      })
+    }
+    for (const voice of payload.voice_cloning ?? []) {
+      const id = voice.voice_id?.trim() ?? ''
+      if (id === '') continue
+      models.push({
+        alias: id,
+        id,
+        category: 'tts',
+        ...(voice.description !== undefined && voice.description.length > 0 ? { description: voice.description.join('；') } : {}),
+      })
+    }
+    for (const voice of payload.voice_generation ?? []) {
+      const id = voice.voice_id?.trim() ?? ''
+      if (id === '') continue
+      models.push({
+        alias: id,
+        id,
+        category: 'tts',
+        ...(voice.description !== undefined && voice.description.length > 0 ? { description: voice.description.join('；') } : {}),
+      })
+    }
+    // MiniMax music models are not returned by get_voice; append the static catalog.
+    const music = (audioPresetById('minimax')?.models ?? []).filter(model => model.category === 'music')
+    for (const model of music) models.push({ ...model, category: 'music' as const })
+    const deduped = dedupe(models)
+    return { models: deduped, source: 'MiniMax get_voice + built-in music catalog' }
+  }
+
+  // Best-effort OpenAI-compatible /models discovery.
+  const base = baseUrl(channel.apiUrl)
+  const url = `${base}/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() as { data?: Array<{ id?: string }> }
+  const models: DiscoveredAudioModel[] = []
+  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: DiscoveredAudioModel[]): DiscoveredAudioModel[] {
+  const seen = new Set<string>()
+  const out: DiscoveredAudioModel[] = []
+  for (const model of models) {
+    if (seen.has(model.id)) continue
+    seen.add(model.id)
+    out.push(model)
+  }
+  return out
+}