@xopcai/xopc 0.0.143 → 0.0.145

package/dist/src/voice/local/runtime-worker.js

@@ -0,0 +1,249 @@
+import { getLocalVoiceModel, resolveLocalVoiceModelDir } from "./models.js";
+import { downloadLocalVoiceModelFiles } from "./model-files.js";
+import { createRuntimeDispatcher } from "./runtime-dispatcher.js";
+import { join, resolve, sep } from "node:path";
+import { mkdir, rm } from "node:fs/promises";
+import { EnvHttpProxyAgent, RetryAgent, setGlobalDispatcher } from "undici";
+import { createInterface } from "node:readline";
+//#region src/voice/local/runtime-worker.ts
+const pipelines = /* @__PURE__ */ new Map();
+const sherpaRecognizers = /* @__PURE__ */ new Map();
+const networkAgent = new RetryAgent(new EnvHttpProxyAgent(), {
+	maxRetries: 4,
+	minTimeout: 500,
+	maxTimeout: 8e3,
+	timeoutFactor: 2,
+	retryAfter: true
+});
+setGlobalDispatcher(networkAgent);
+function normalizeRemoteHost(value) {
+	const url = new URL(value);
+	if (url.username || url.password) throw new Error("Local voice model endpoint must not contain credentials");
+	if (url.protocol !== "https:" && url.hostname !== "127.0.0.1" && url.hostname !== "localhost") throw new Error("Local voice model endpoint must use HTTPS");
+	url.pathname = `${url.pathname.replace(/\/+$/, "")}/`;
+	return url.toString();
+}
+function resolveRemoteHosts() {
+	const configured = process.env.XOPC_VOICE_MODEL_ENDPOINT?.trim() || process.env.HF_ENDPOINT?.trim();
+	if (configured) return [normalizeRemoteHost(configured)];
+	return [normalizeRemoteHost("https://xopc.ai/api/voice/models"), normalizeRemoteHost("https://huggingface.co")];
+}
+function serializeError(error) {
+	if (!(error instanceof Error)) return { message: String(error) };
+	const withCode = error;
+	const nested = withCode.cause;
+	const nestedMessage = nested instanceof Error ? nested.message : void 0;
+	const nestedCode = nested && typeof nested === "object" && "code" in nested ? String(nested.code ?? "") : void 0;
+	return {
+		message: error.message,
+		...typeof withCode.code === "string" ? { code: withCode.code } : {},
+		...nestedCode ? { code: nestedCode } : {},
+		...nestedMessage && nestedMessage !== error.message ? { cause: nestedMessage } : {}
+	};
+}
+function send(message) {
+	process.stdout.write(`${JSON.stringify(message)}\n`);
+}
+async function loadPipeline(modelId, allowRemote, cacheDir = resolveLocalVoiceModelDir(modelId), onProgress) {
+	const key = `${modelId}:${cacheDir}`;
+	const existing = pipelines.get(key);
+	if (existing) return existing;
+	const model = getLocalVoiceModel(modelId);
+	if (model.engine !== "transformers.js") throw new Error(`Model ${model.id} does not use the Transformers.js engine`);
+	const promise = (async () => {
+		const transformers = await import("@huggingface/transformers");
+		transformers.env.cacheDir = cacheDir;
+		transformers.env.allowRemoteModels = allowRemote;
+		transformers.env.allowLocalModels = true;
+		const remoteHosts = allowRemote ? resolveRemoteHosts() : [normalizeRemoteHost("https://xopc.ai")];
+		let lastError;
+		for (const [index, remoteHost] of remoteHosts.entries()) {
+			transformers.env.remoteHost = remoteHost;
+			try {
+				return await transformers.pipeline("automatic-speech-recognition", model.repository, {
+					revision: model.revision,
+					dtype: model.dtype,
+					progress_callback: onProgress
+				});
+			} catch (error) {
+				lastError = error;
+				if (index + 1 < remoteHosts.length) {
+					await rm(cacheDir, {
+						recursive: true,
+						force: true
+					});
+					await mkdir(cacheDir, { recursive: true });
+				}
+			}
+		}
+		throw lastError;
+	})();
+	pipelines.set(key, promise);
+	try {
+		return await promise;
+	} catch (error) {
+		pipelines.delete(key);
+		throw error;
+	}
+}
+function normalizeSenseVoiceLanguage(language) {
+	if (typeof language !== "string") return "auto";
+	const value = language.trim().toLowerCase().replace("_", "-");
+	if (!value || value === "auto") return "auto";
+	if (value === "cmn" || value.startsWith("zh")) return "zh";
+	if (value === "cantonese" || value.startsWith("yue")) return "yue";
+	if (value.startsWith("en")) return "en";
+	if (value.startsWith("ja")) return "ja";
+	if (value.startsWith("ko")) return "ko";
+	return "auto";
+}
+function cleanSenseVoiceTag(value) {
+	return value?.replace(/^<\|/, "").replace(/\|>$/, "").trim() || void 0;
+}
+async function loadSherpaRecognizer(modelId, language) {
+	const key = `${modelId}:${language}`;
+	const existing = sherpaRecognizers.get(key);
+	if (existing) return existing;
+	const model = getLocalVoiceModel(modelId);
+	if (model.engine !== "sherpa-onnx") throw new Error(`Model ${model.id} does not use the sherpa-onnx engine`);
+	const modelDir = resolveLocalVoiceModelDir(model.id);
+	const promise = (async () => {
+		const imported = await import("sherpa-onnx-node");
+		return new (imported.default ?? imported).OfflineRecognizer({
+			featConfig: {
+				sampleRate: 16e3,
+				featureDim: 80
+			},
+			modelConfig: {
+				senseVoice: {
+					model: join(modelDir, "model.int8.onnx"),
+					language,
+					useInverseTextNormalization: 1
+				},
+				tokens: join(modelDir, "tokens.txt"),
+				numThreads: Math.max(1, Math.min(4, Number(process.env.XOPC_VOICE_THREADS) || 2)),
+				provider: "cpu",
+				debug: 0
+			}
+		});
+	})();
+	sherpaRecognizers.set(key, promise);
+	try {
+		return await promise;
+	} catch (error) {
+		sherpaRecognizers.delete(key);
+		throw error;
+	}
+}
+async function transcribeWithSherpa(modelId, audio, language) {
+	const recognizer = await loadSherpaRecognizer(modelId, normalizeSenseVoiceLanguage(language));
+	const stream = recognizer.createStream();
+	stream.acceptWaveform({
+		sampleRate: 16e3,
+		samples: audio
+	});
+	const result = recognizer.decodeAsync ? await recognizer.decodeAsync(stream) : (recognizer.decode(stream), recognizer.getResult(stream));
+	return {
+		text: String(result.text ?? "").trim(),
+		...cleanSenseVoiceTag(result.lang) ? { language: cleanSenseVoiceTag(result.lang) } : {},
+		...cleanSenseVoiceTag(result.emotion) ? { emotion: cleanSenseVoiceTag(result.emotion) } : {},
+		...cleanSenseVoiceTag(result.event) ? { event: cleanSenseVoiceTag(result.event) } : {}
+	};
+}
+async function handle(request) {
+	const params = request.params ?? {};
+	if (request.method === "health") return {
+		ok: true,
+		protocolVersion: 2,
+		engine: "transformers.js+sherpa-onnx",
+		engines: ["transformers.js", "sherpa-onnx"]
+	};
+	if (request.method === "model.install") {
+		const modelId = String(params.modelId ?? "");
+		const modelDir = resolve(resolveLocalVoiceModelDir(modelId));
+		const requestedCacheDir = typeof params.cacheDir === "string" ? resolve(params.cacheDir) : modelDir;
+		const allowedPrefix = `${resolve(modelDir, "..")}${sep}`;
+		if (requestedCacheDir !== modelDir && !requestedCacheDir.startsWith(allowedPrefix)) throw new Error("Local voice model cache directory is outside the managed root");
+		const model = getLocalVoiceModel(modelId);
+		if (model.engine === "sherpa-onnx") await downloadLocalVoiceModelFiles({
+			model,
+			destinationDir: requestedCacheDir,
+			remoteHosts: resolveRemoteHosts(),
+			onProgress: (data) => send({
+				id: request.id,
+				event: "progress",
+				data
+			})
+		});
+		else {
+			await loadPipeline(modelId, true, requestedCacheDir, (data) => send({
+				id: request.id,
+				event: "progress",
+				data
+			}));
+			pipelines.delete(`${modelId}:${requestedCacheDir}`);
+		}
+		return {
+			installed: true,
+			modelId
+		};
+	}
+	if (request.method === "model.unload") {
+		const modelId = String(params.modelId ?? "");
+		for (const key of pipelines.keys()) if (key.startsWith(`${modelId}:`)) pipelines.delete(key);
+		for (const key of sherpaRecognizers.keys()) if (key.startsWith(`${modelId}:`)) sherpaRecognizers.delete(key);
+		return { unloaded: true };
+	}
+	if (request.method === "transcribe") {
+		const modelId = String(params.modelId ?? "");
+		const bytes = Buffer.from(String(params.audioBase64 ?? ""), "base64");
+		const copied = Uint8Array.from(bytes);
+		const audio = new Float32Array(copied.buffer);
+		const language = typeof params.language === "string" && params.language ? params.language : void 0;
+		const model = getLocalVoiceModel(modelId);
+		if (model.engine === "sherpa-onnx") return {
+			...await transcribeWithSherpa(model.id, audio, language),
+			modelId
+		};
+		const output = await (await loadPipeline(modelId, false))(audio, {
+			task: "transcribe",
+			...language ? { language } : {},
+			chunk_length_s: 30,
+			stride_length_s: 5,
+			return_timestamps: false
+		});
+		return {
+			text: (output && typeof output === "object" && "text" in output ? String(output.text ?? "") : "").trim(),
+			modelId
+		};
+	}
+	throw new Error(`Unknown local voice runtime method: ${request.method}`);
+}
+const lines = createInterface({ input: process.stdin });
+const dispatch = createRuntimeDispatcher({
+	execute: handle,
+	sendResult: (request, result) => send({
+		id: request.id,
+		result
+	}),
+	sendError: (request, error) => send({
+		id: request.id,
+		error: serializeError(error)
+	})
+});
+lines.on("line", (line) => {
+	let request;
+	try {
+		request = JSON.parse(line);
+	} catch {
+		return;
+	}
+	dispatch(request);
+});
+lines.once("close", () => {
+	networkAgent.close().finally(() => process.exit(0));
+});
+//#endregion
+export {};
+
+//# sourceMappingURL=runtime-worker.js.map

package/dist/src/voice/local/runtime-worker.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"runtime-worker.js","names":[],"sources":["../../../../src/voice/local/runtime-worker.ts"],"sourcesContent":["import { createInterface } from 'node:readline';\nimport { join, resolve, sep } from 'node:path';\nimport { mkdir, rm } from 'node:fs/promises';\n\nimport { EnvHttpProxyAgent, RetryAgent, setGlobalDispatcher } from 'undici';\n\nimport { getLocalVoiceModel, resolveLocalVoiceModelDir } from './models.js';\nimport { downloadLocalVoiceModelFiles } from './model-files.js';\nimport { createRuntimeDispatcher } from './runtime-dispatcher.js';\n\ntype RuntimeRequest = { id: number; method: string; params?: Record<string, unknown> };\ntype AsrPipeline = (audio: Float32Array, options?: Record<string, unknown>) => Promise<unknown>;\n\nconst pipelines = new Map<string, Promise<AsrPipeline>>();\n\ninterface SherpaOfflineStream {\n  acceptWaveform(input: { sampleRate: number; samples: Float32Array }): void;\n  setOption?(key: string, value: string): void;\n}\n\ninterface SherpaResult {\n  text?: string;\n  lang?: string;\n  emotion?: string;\n  event?: string;\n}\n\ninterface SherpaOfflineRecognizer {\n  createStream(): SherpaOfflineStream;\n  decodeAsync?(stream: SherpaOfflineStream): Promise<SherpaResult>;\n  decode(stream: SherpaOfflineStream): void;\n  getResult(stream: SherpaOfflineStream): SherpaResult;\n}\n\ntype SherpaModule = {\n  OfflineRecognizer: new (config: Record<string, unknown>) => SherpaOfflineRecognizer;\n};\n\nconst sherpaRecognizers = new Map<string, Promise<SherpaOfflineRecognizer>>();\n\nconst networkAgent = new RetryAgent(new EnvHttpProxyAgent(), {\n  maxRetries: 4,\n  minTimeout: 500,\n  maxTimeout: 8_000,\n  timeoutFactor: 2,\n  retryAfter: true,\n});\nsetGlobalDispatcher(networkAgent);\n\nfunction normalizeRemoteHost(value: string): string {\n  const url = new URL(value);\n  if (url.username || url.password) {\n    throw new Error('Local voice model endpoint must not contain credentials');\n  }\n  if (url.protocol !== 'https:' && url.hostname !== '127.0.0.1' && url.hostname !== 'localhost') {\n    throw new Error('Local voice model endpoint must use HTTPS');\n  }\n  url.pathname = `${url.pathname.replace(/\\/+$/, '')}/`;\n  return url.toString();\n}\n\nfunction resolveRemoteHosts(): string[] {\n  const configured = process.env.XOPC_VOICE_MODEL_ENDPOINT?.trim()\n    || process.env.HF_ENDPOINT?.trim();\n  if (configured) return [normalizeRemoteHost(configured)];\n  return [\n    normalizeRemoteHost('https://xopc.ai/api/voice/models'),\n    normalizeRemoteHost('https://huggingface.co'),\n  ];\n}\n\nfunction serializeError(error: unknown): { message: string; code?: string; cause?: string } {\n  if (!(error instanceof Error)) return { message: String(error) };\n  const withCode = error as Error & { code?: unknown; cause?: unknown };\n  const nested = withCode.cause;\n  const nestedMessage = nested instanceof Error ? nested.message : undefined;\n  const nestedCode = nested && typeof nested === 'object' && 'code' in nested\n    ? String((nested as { code?: unknown }).code ?? '')\n    : undefined;\n  return {\n    message: error.message,\n    ...(typeof withCode.code === 'string' ? { code: withCode.code } : {}),\n    ...(nestedCode ? { code: nestedCode } : {}),\n    ...(nestedMessage && nestedMessage !== error.message ? { cause: nestedMessage } : {}),\n  };\n}\n\nfunction send(message: Record<string, unknown>): void {\n  process.stdout.write(`${JSON.stringify(message)}\\n`);\n}\n\nasync function loadPipeline(\n  modelId: string,\n  allowRemote: boolean,\n  cacheDir = resolveLocalVoiceModelDir(modelId),\n  onProgress?: (data: Record<string, unknown>) => void,\n): Promise<AsrPipeline> {\n  const key = `${modelId}:${cacheDir}`;\n  const existing = pipelines.get(key);\n  if (existing) return existing;\n  const model = getLocalVoiceModel(modelId);\n  if (model.engine !== 'transformers.js') {\n    throw new Error(`Model ${model.id} does not use the Transformers.js engine`);\n  }\n  const promise = (async () => {\n    const transformers = await import('@huggingface/transformers');\n    transformers.env.cacheDir = cacheDir;\n    transformers.env.allowRemoteModels = allowRemote;\n    transformers.env.allowLocalModels = true;\n    const remoteHosts = allowRemote ? resolveRemoteHosts() : [normalizeRemoteHost('https://xopc.ai')];\n    let lastError: unknown;\n    for (const [index, remoteHost] of remoteHosts.entries()) {\n      transformers.env.remoteHost = remoteHost;\n      try {\n        const result = await transformers.pipeline('automatic-speech-recognition', model.repository, {\n          revision: model.revision,\n          dtype: model.dtype,\n          progress_callback: onProgress,\n        });\n        return result as unknown as AsrPipeline;\n      } catch (error) {\n        lastError = error;\n        if (index + 1 < remoteHosts.length) {\n          await rm(cacheDir, { recursive: true, force: true });\n          await mkdir(cacheDir, { recursive: true });\n        }\n      }\n    }\n    throw lastError;\n  })();\n  pipelines.set(key, promise);\n  try {\n    return await promise;\n  } catch (error) {\n    pipelines.delete(key);\n    throw error;\n  }\n}\n\nfunction normalizeSenseVoiceLanguage(language: unknown): string {\n  if (typeof language !== 'string') return 'auto';\n  const value = language.trim().toLowerCase().replace('_', '-');\n  if (!value || value === 'auto') return 'auto';\n  if (value === 'cmn' || value.startsWith('zh')) return 'zh';\n  if (value === 'cantonese' || value.startsWith('yue')) return 'yue';\n  if (value.startsWith('en')) return 'en';\n  if (value.startsWith('ja')) return 'ja';\n  if (value.startsWith('ko')) return 'ko';\n  return 'auto';\n}\n\nfunction cleanSenseVoiceTag(value: string | undefined): string | undefined {\n  const cleaned = value?.replace(/^<\\|/, '').replace(/\\|>$/, '').trim();\n  return cleaned || undefined;\n}\n\nasync function loadSherpaRecognizer(\n  modelId: string,\n  language: string,\n): Promise<SherpaOfflineRecognizer> {\n  const key = `${modelId}:${language}`;\n  const existing = sherpaRecognizers.get(key);\n  if (existing) return existing;\n  const model = getLocalVoiceModel(modelId);\n  if (model.engine !== 'sherpa-onnx') {\n    throw new Error(`Model ${model.id} does not use the sherpa-onnx engine`);\n  }\n  const modelDir = resolveLocalVoiceModelDir(model.id);\n  const promise = (async () => {\n    const imported = await import('sherpa-onnx-node');\n    const sherpa = ((imported as { default?: SherpaModule }).default ?? imported) as SherpaModule;\n    return new sherpa.OfflineRecognizer({\n      featConfig: { sampleRate: 16_000, featureDim: 80 },\n      modelConfig: {\n        senseVoice: {\n          model: join(modelDir, 'model.int8.onnx'),\n          language,\n          useInverseTextNormalization: 1,\n        },\n        tokens: join(modelDir, 'tokens.txt'),\n        numThreads: Math.max(1, Math.min(4, Number(process.env.XOPC_VOICE_THREADS) || 2)),\n        provider: 'cpu',\n        debug: 0,\n      },\n    });\n  })();\n  sherpaRecognizers.set(key, promise);\n  try {\n    return await promise;\n  } catch (error) {\n    sherpaRecognizers.delete(key);\n    throw error;\n  }\n}\n\nasync function transcribeWithSherpa(\n  modelId: string,\n  audio: Float32Array,\n  language: unknown,\n): Promise<{ text: string; language?: string; emotion?: string; event?: string }> {\n  const recognizer = await loadSherpaRecognizer(modelId, normalizeSenseVoiceLanguage(language));\n  const stream = recognizer.createStream();\n  stream.acceptWaveform({ sampleRate: 16_000, samples: audio });\n  const result = recognizer.decodeAsync\n    ? await recognizer.decodeAsync(stream)\n    : (recognizer.decode(stream), recognizer.getResult(stream));\n  return {\n    text: String(result.text ?? '').trim(),\n    ...(cleanSenseVoiceTag(result.lang) ? { language: cleanSenseVoiceTag(result.lang) } : {}),\n    ...(cleanSenseVoiceTag(result.emotion) ? { emotion: cleanSenseVoiceTag(result.emotion) } : {}),\n    ...(cleanSenseVoiceTag(result.event) ? { event: cleanSenseVoiceTag(result.event) } : {}),\n  };\n}\n\nasync function handle(request: RuntimeRequest): Promise<unknown> {\n  const params = request.params ?? {};\n  if (request.method === 'health') {\n    return {\n      ok: true,\n      protocolVersion: 2,\n      engine: 'transformers.js+sherpa-onnx',\n      engines: ['transformers.js', 'sherpa-onnx'],\n    };\n  }\n  if (request.method === 'model.install') {\n    const modelId = String(params.modelId ?? '');\n    const modelDir = resolve(resolveLocalVoiceModelDir(modelId));\n    const requestedCacheDir = typeof params.cacheDir === 'string' ? resolve(params.cacheDir) : modelDir;\n    const allowedPrefix = `${resolve(modelDir, '..')}${sep}`;\n    if (requestedCacheDir !== modelDir && !requestedCacheDir.startsWith(allowedPrefix)) {\n      throw new Error('Local voice model cache directory is outside the managed root');\n    }\n    const model = getLocalVoiceModel(modelId);\n    if (model.engine === 'sherpa-onnx') {\n      await downloadLocalVoiceModelFiles({\n        model,\n        destinationDir: requestedCacheDir,\n        remoteHosts: resolveRemoteHosts(),\n        onProgress: (data) => send({ id: request.id, event: 'progress', data }),\n      });\n    } else {\n      await loadPipeline(\n        modelId,\n        true,\n        requestedCacheDir,\n        (data) => send({ id: request.id, event: 'progress', data }),\n      );\n      pipelines.delete(`${modelId}:${requestedCacheDir}`);\n    }\n    return { installed: true, modelId };\n  }\n  if (request.method === 'model.unload') {\n    const modelId = String(params.modelId ?? '');\n    for (const key of pipelines.keys()) {\n      if (key.startsWith(`${modelId}:`)) pipelines.delete(key);\n    }\n    for (const key of sherpaRecognizers.keys()) {\n      if (key.startsWith(`${modelId}:`)) sherpaRecognizers.delete(key);\n    }\n    return { unloaded: true };\n  }\n  if (request.method === 'transcribe') {\n    const modelId = String(params.modelId ?? '');\n    const bytes = Buffer.from(String(params.audioBase64 ?? ''), 'base64');\n    const copied = Uint8Array.from(bytes);\n    const audio = new Float32Array(copied.buffer);\n    const language = typeof params.language === 'string' && params.language ? params.language : undefined;\n    const model = getLocalVoiceModel(modelId);\n    if (model.engine === 'sherpa-onnx') {\n      return { ...(await transcribeWithSherpa(model.id, audio, language)), modelId };\n    }\n    const pipeline = await loadPipeline(modelId, false);\n    const output = await pipeline(audio, {\n      task: 'transcribe',\n      ...(language ? { language } : {}),\n      chunk_length_s: 30,\n      stride_length_s: 5,\n      return_timestamps: false,\n    });\n    const text = output && typeof output === 'object' && 'text' in output\n      ? String((output as { text?: unknown }).text ?? '')\n      : '';\n    return { text: text.trim(), modelId };\n  }\n  throw new Error(`Unknown local voice runtime method: ${request.method}`);\n}\n\nconst lines = createInterface({ input: process.stdin });\nconst dispatch = createRuntimeDispatcher<RuntimeRequest>({\n  execute: handle,\n  sendResult: (request, result) => send({ id: request.id, result }),\n  sendError: (request, error) => send({ id: request.id, error: serializeError(error) }),\n});\nlines.on('line', (line) => {\n  let request: RuntimeRequest;\n  try {\n    request = JSON.parse(line) as RuntimeRequest;\n  } catch {\n    return;\n  }\n  dispatch(request);\n});\nlines.once('close', () => {\n  void networkAgent.close().finally(() => process.exit(0));\n});\n"],"mappings":";;;;;;;;AAaA,MAAM,4BAAY,IAAI,KAAmC;AAyBzD,MAAM,oCAAoB,IAAI,KAA+C;AAE7E,MAAM,eAAe,IAAI,WAAW,IAAI,mBAAmB,EAAE;CAC3D,YAAY;CACZ,YAAY;CACZ,YAAY;CACZ,eAAe;CACf,YAAY;CACb,CAAC;AACF,oBAAoB,aAAa;AAEjC,SAAS,oBAAoB,OAAuB;CAClD,MAAM,MAAM,IAAI,IAAI,MAAM;AAC1B,KAAI,IAAI,YAAY,IAAI,SACtB,OAAM,IAAI,MAAM,0DAA0D;AAE5E,KAAI,IAAI,aAAa,YAAY,IAAI,aAAa,eAAe,IAAI,aAAa,YAChF,OAAM,IAAI,MAAM,4CAA4C;AAE9D,KAAI,WAAW,GAAG,IAAI,SAAS,QAAQ,QAAQ,GAAG,CAAC;AACnD,QAAO,IAAI,UAAU;;AAGvB,SAAS,qBAA+B;CACtC,MAAM,aAAa,QAAQ,IAAI,2BAA2B,MAAM,IAC3D,QAAQ,IAAI,aAAa,MAAM;AACpC,KAAI,WAAY,QAAO,CAAC,oBAAoB,WAAW,CAAC;AACxD,QAAO,CACL,oBAAoB,mCAAmC,EACvD,oBAAoB,yBAAyB,CAC9C;;AAGH,SAAS,eAAe,OAAoE;AAC1F,KAAI,EAAE,iBAAiB,OAAQ,QAAO,EAAE,SAAS,OAAO,MAAM,EAAE;CAChE,MAAM,WAAW;CACjB,MAAM,SAAS,SAAS;CACxB,MAAM,gBAAgB,kBAAkB,QAAQ,OAAO,UAAU,KAAA;CACjE,MAAM,aAAa,UAAU,OAAO,WAAW,YAAY,UAAU,SACjE,OAAQ,OAA8B,QAAQ,GAAG,GACjD,KAAA;AACJ,QAAO;EACL,SAAS,MAAM;EACf,GAAI,OAAO,SAAS,SAAS,WAAW,EAAE,MAAM,SAAS,MAAM,GAAG,EAAE;EACpE,GAAI,aAAa,EAAE,MAAM,YAAY,GAAG,EAAE;EAC1C,GAAI,iBAAiB,kBAAkB,MAAM,UAAU,EAAE,OAAO,eAAe,GAAG,EAAE;EACrF;;AAGH,SAAS,KAAK,SAAwC;AACpD,SAAQ,OAAO,MAAM,GAAG,KAAK,UAAU,QAAQ,CAAC,IAAI;;AAGtD,eAAe,aACb,SACA,aACA,WAAW,0BAA0B,QAAQ,EAC7C,YACsB;CACtB,MAAM,MAAM,GAAG,QAAQ,GAAG;CAC1B,MAAM,WAAW,UAAU,IAAI,IAAI;AACnC,KAAI,SAAU,QAAO;CACrB,MAAM,QAAQ,mBAAmB,QAAQ;AACzC,KAAI,MAAM,WAAW,kBACnB,OAAM,IAAI,MAAM,SAAS,MAAM,GAAG,0CAA0C;CAE9E,MAAM,WAAW,YAAY;EAC3B,MAAM,eAAe,MAAM,OAAO;AAClC,eAAa,IAAI,WAAW;AAC5B,eAAa,IAAI,oBAAoB;AACrC,eAAa,IAAI,mBAAmB;EACpC,MAAM,cAAc,cAAc,oBAAoB,GAAG,CAAC,oBAAoB,kBAAkB,CAAC;EACjG,IAAI;AACJ,OAAK,MAAM,CAAC,OAAO,eAAe,YAAY,SAAS,EAAE;AACvD,gBAAa,IAAI,aAAa;AAC9B,OAAI;AAMF,WAAO,MALc,aAAa,SAAS,gCAAgC,MAAM,YAAY;KAC3F,UAAU,MAAM;KAChB,OAAO,MAAM;KACb,mBAAmB;KACpB,CAAC;YAEK,OAAO;AACd,gBAAY;AACZ,QAAI,QAAQ,IAAI,YAAY,QAAQ;AAClC,WAAM,GAAG,UAAU;MAAE,WAAW;MAAM,OAAO;MAAM,CAAC;AACpD,WAAM,MAAM,UAAU,EAAE,WAAW,MAAM,CAAC;;;;AAIhD,QAAM;KACJ;AACJ,WAAU,IAAI,KAAK,QAAQ;AAC3B,KAAI;AACF,SAAO,MAAM;UACN,OAAO;AACd,YAAU,OAAO,IAAI;AACrB,QAAM;;;AAIV,SAAS,4BAA4B,UAA2B;AAC9D,KAAI,OAAO,aAAa,SAAU,QAAO;CACzC,MAAM,QAAQ,SAAS,MAAM,CAAC,aAAa,CAAC,QAAQ,KAAK,IAAI;AAC7D,KAAI,CAAC,SAAS,UAAU,OAAQ,QAAO;AACvC,KAAI,UAAU,SAAS,MAAM,WAAW,KAAK,CAAE,QAAO;AACtD,KAAI,UAAU,eAAe,MAAM,WAAW,MAAM,CAAE,QAAO;AAC7D,KAAI,MAAM,WAAW,KAAK,CAAE,QAAO;AACnC,KAAI,MAAM,WAAW,KAAK,CAAE,QAAO;AACnC,KAAI,MAAM,WAAW,KAAK,CAAE,QAAO;AACnC,QAAO;;AAGT,SAAS,mBAAmB,OAA+C;AAEzE,QADgB,OAAO,QAAQ,QAAQ,GAAG,CAAC,QAAQ,QAAQ,GAAG,CAAC,MAAM,IACnD,KAAA;;AAGpB,eAAe,qBACb,SACA,UACkC;CAClC,MAAM,MAAM,GAAG,QAAQ,GAAG;CAC1B,MAAM,WAAW,kBAAkB,IAAI,IAAI;AAC3C,KAAI,SAAU,QAAO;CACrB,MAAM,QAAQ,mBAAmB,QAAQ;AACzC,KAAI,MAAM,WAAW,cACnB,OAAM,IAAI,MAAM,SAAS,MAAM,GAAG,sCAAsC;CAE1E,MAAM,WAAW,0BAA0B,MAAM,GAAG;CACpD,MAAM,WAAW,YAAY;EAC3B,MAAM,WAAW,MAAM,OAAO;AAE9B,SAAO,KADU,SAAwC,WAAW,UAClD,kBAAkB;GAClC,YAAY;IAAE,YAAY;IAAQ,YAAY;IAAI;GAClD,aAAa;IACX,YAAY;KACV,OAAO,KAAK,UAAU,kBAAkB;KACxC;KACA,6BAA6B;KAC9B;IACD,QAAQ,KAAK,UAAU,aAAa;IACpC,YAAY,KAAK,IAAI,GAAG,KAAK,IAAI,GAAG,OAAO,QAAQ,IAAI,mBAAmB,IAAI,EAAE,CAAC;IACjF,UAAU;IACV,OAAO;IACR;GACF,CAAC;KACA;AACJ,mBAAkB,IAAI,KAAK,QAAQ;AACnC,KAAI;AACF,SAAO,MAAM;UACN,OAAO;AACd,oBAAkB,OAAO,IAAI;AAC7B,QAAM;;;AAIV,eAAe,qBACb,SACA,OACA,UACgF;CAChF,MAAM,aAAa,MAAM,qBAAqB,SAAS,4BAA4B,SAAS,CAAC;CAC7F,MAAM,SAAS,WAAW,cAAc;AACxC,QAAO,eAAe;EAAE,YAAY;EAAQ,SAAS;EAAO,CAAC;CAC7D,MAAM,SAAS,WAAW,cACtB,MAAM,WAAW,YAAY,OAAO,IACnC,WAAW,OAAO,OAAO,EAAE,WAAW,UAAU,OAAO;AAC5D,QAAO;EACL,MAAM,OAAO,OAAO,QAAQ,GAAG,CAAC,MAAM;EACtC,GAAI,mBAAmB,OAAO,KAAK,GAAG,EAAE,UAAU,mBAAmB,OAAO,KAAK,EAAE,GAAG,EAAE;EACxF,GAAI,mBAAmB,OAAO,QAAQ,GAAG,EAAE,SAAS,mBAAmB,OAAO,QAAQ,EAAE,GAAG,EAAE;EAC7F,GAAI,mBAAmB,OAAO,MAAM,GAAG,EAAE,OAAO,mBAAmB,OAAO,MAAM,EAAE,GAAG,EAAE;EACxF;;AAGH,eAAe,OAAO,SAA2C;CAC/D,MAAM,SAAS,QAAQ,UAAU,EAAE;AACnC,KAAI,QAAQ,WAAW,SACrB,QAAO;EACL,IAAI;EACJ,iBAAiB;EACjB,QAAQ;EACR,SAAS,CAAC,mBAAmB,cAAc;EAC5C;AAEH,KAAI,QAAQ,WAAW,iBAAiB;EACtC,MAAM,UAAU,OAAO,OAAO,WAAW,GAAG;EAC5C,MAAM,WAAW,QAAQ,0BAA0B,QAAQ,CAAC;EAC5D,MAAM,oBAAoB,OAAO,OAAO,aAAa,WAAW,QAAQ,OAAO,SAAS,GAAG;EAC3F,MAAM,gBAAgB,GAAG,QAAQ,UAAU,KAAK,GAAG;AACnD,MAAI,sBAAsB,YAAY,CAAC,kBAAkB,WAAW,cAAc,CAChF,OAAM,IAAI,MAAM,gEAAgE;EAElF,MAAM,QAAQ,mBAAmB,QAAQ;AACzC,MAAI,MAAM,WAAW,cACnB,OAAM,6BAA6B;GACjC;GACA,gBAAgB;GAChB,aAAa,oBAAoB;GACjC,aAAa,SAAS,KAAK;IAAE,IAAI,QAAQ;IAAI,OAAO;IAAY;IAAM,CAAC;GACxE,CAAC;OACG;AACL,SAAM,aACJ,SACA,MACA,oBACC,SAAS,KAAK;IAAE,IAAI,QAAQ;IAAI,OAAO;IAAY;IAAM,CAAC,CAC5D;AACD,aAAU,OAAO,GAAG,QAAQ,GAAG,oBAAoB;;AAErD,SAAO;GAAE,WAAW;GAAM;GAAS;;AAErC,KAAI,QAAQ,WAAW,gBAAgB;EACrC,MAAM,UAAU,OAAO,OAAO,WAAW,GAAG;AAC5C,OAAK,MAAM,OAAO,UAAU,MAAM,CAChC,KAAI,IAAI,WAAW,GAAG,QAAQ,GAAG,CAAE,WAAU,OAAO,IAAI;AAE1D,OAAK,MAAM,OAAO,kBAAkB,MAAM,CACxC,KAAI,IAAI,WAAW,GAAG,QAAQ,GAAG,CAAE,mBAAkB,OAAO,IAAI;AAElE,SAAO,EAAE,UAAU,MAAM;;AAE3B,KAAI,QAAQ,WAAW,cAAc;EACnC,MAAM,UAAU,OAAO,OAAO,WAAW,GAAG;EAC5C,MAAM,QAAQ,OAAO,KAAK,OAAO,OAAO,eAAe,GAAG,EAAE,SAAS;EACrE,MAAM,SAAS,WAAW,KAAK,MAAM;EACrC,MAAM,QAAQ,IAAI,aAAa,OAAO,OAAO;EAC7C,MAAM,WAAW,OAAO,OAAO,aAAa,YAAY,OAAO,WAAW,OAAO,WAAW,KAAA;EAC5F,MAAM,QAAQ,mBAAmB,QAAQ;AACzC,MAAI,MAAM,WAAW,cACnB,QAAO;GAAE,GAAI,MAAM,qBAAqB,MAAM,IAAI,OAAO,SAAS;GAAG;GAAS;EAGhF,MAAM,SAAS,OAAM,MADE,aAAa,SAAS,MAAM,EACrB,OAAO;GACnC,MAAM;GACN,GAAI,WAAW,EAAE,UAAU,GAAG,EAAE;GAChC,gBAAgB;GAChB,iBAAiB;GACjB,mBAAmB;GACpB,CAAC;AAIF,SAAO;GAAE,OAHI,UAAU,OAAO,WAAW,YAAY,UAAU,SAC3D,OAAQ,OAA8B,QAAQ,GAAG,GACjD,IACgB,MAAM;GAAE;GAAS;;AAEvC,OAAM,IAAI,MAAM,uCAAuC,QAAQ,SAAS;;AAG1E,MAAM,QAAQ,gBAAgB,EAAE,OAAO,QAAQ,OAAO,CAAC;AACvD,MAAM,WAAW,wBAAwC;CACvD,SAAS;CACT,aAAa,SAAS,WAAW,KAAK;EAAE,IAAI,QAAQ;EAAI;EAAQ,CAAC;CACjE,YAAY,SAAS,UAAU,KAAK;EAAE,IAAI,QAAQ;EAAI,OAAO,eAAe,MAAM;EAAE,CAAC;CACtF,CAAC;AACF,MAAM,GAAG,SAAS,SAAS;CACzB,IAAI;AACJ,KAAI;AACF,YAAU,KAAK,MAAM,KAAK;SACpB;AACN;;AAEF,UAAS,QAAQ;EACjB;AACF,MAAM,KAAK,eAAe;AACnB,cAAa,OAAO,CAAC,cAAc,QAAQ,KAAK,EAAE,CAAC;EACxD"}

package/dist/src/voice/local/wav.d.ts

@@ -0,0 +1,6 @@
+export interface DecodedPcmAudio {
+    samples: Float32Array;
+    sampleRate: number;
+    durationSeconds: number;
+}
+export declare function decodeWavToMonoFloat32(buffer: Buffer, targetRate?: number): DecodedPcmAudio;

package/dist/src/voice/local/wav.js

@@ -0,0 +1,69 @@
+//#region src/voice/local/wav.ts
+function readChunkId(buffer, offset) {
+	return buffer.toString("ascii", offset, offset + 4);
+}
+function resampleLinear(input, sourceRate, targetRate) {
+	if (sourceRate === targetRate) return input;
+	const outputLength = Math.max(1, Math.round(input.length * targetRate / sourceRate));
+	const output = new Float32Array(outputLength);
+	const ratio = sourceRate / targetRate;
+	for (let i = 0; i < outputLength; i += 1) {
+		const sourcePosition = i * ratio;
+		const left = Math.min(input.length - 1, Math.floor(sourcePosition));
+		const right = Math.min(input.length - 1, left + 1);
+		const fraction = sourcePosition - left;
+		output[i] = input[left] * (1 - fraction) + input[right] * fraction;
+	}
+	return output;
+}
+function decodeWavToMonoFloat32(buffer, targetRate = 16e3) {
+	if (buffer.length < 44 || readChunkId(buffer, 0) !== "RIFF" || readChunkId(buffer, 8) !== "WAVE") throw new Error("Local STT currently requires PCM WAV audio");
+	let offset = 12;
+	let format;
+	let dataOffset = -1;
+	let dataLength = 0;
+	while (offset + 8 <= buffer.length) {
+		const id = readChunkId(buffer, offset);
+		const length = buffer.readUInt32LE(offset + 4);
+		const bodyOffset = offset + 8;
+		if (bodyOffset + length > buffer.length) break;
+		if (id === "fmt " && length >= 16) format = {
+			audioFormat: buffer.readUInt16LE(bodyOffset),
+			channels: buffer.readUInt16LE(bodyOffset + 2),
+			sampleRate: buffer.readUInt32LE(bodyOffset + 4),
+			bitsPerSample: buffer.readUInt16LE(bodyOffset + 14)
+		};
+		else if (id === "data") {
+			dataOffset = bodyOffset;
+			dataLength = length;
+			break;
+		}
+		offset = bodyOffset + length + length % 2;
+	}
+	if (!format || dataOffset < 0 || dataLength === 0) throw new Error("WAV audio is missing format or data chunks");
+	if (format.channels < 1 || format.channels > 8 || format.sampleRate < 8e3) throw new Error("Unsupported WAV channel count or sample rate");
+	const bytesPerSample = format.bitsPerSample / 8;
+	if (!Number.isInteger(bytesPerSample) || bytesPerSample < 2) throw new Error(`Unsupported WAV bit depth: ${format.bitsPerSample}`);
+	const frameBytes = bytesPerSample * format.channels;
+	const frameCount = Math.floor(dataLength / frameBytes);
+	const mono = new Float32Array(frameCount);
+	for (let frame = 0; frame < frameCount; frame += 1) {
+		let sum = 0;
+		for (let channel = 0; channel < format.channels; channel += 1) {
+			const sampleOffset = dataOffset + frame * frameBytes + channel * bytesPerSample;
+			if (format.audioFormat === 1 && format.bitsPerSample === 16) sum += buffer.readInt16LE(sampleOffset) / 32768;
+			else if (format.audioFormat === 3 && format.bitsPerSample === 32) sum += buffer.readFloatLE(sampleOffset);
+			else throw new Error(`Unsupported WAV encoding: format ${format.audioFormat}, ${format.bitsPerSample}-bit`);
+		}
+		mono[frame] = Math.max(-1, Math.min(1, sum / format.channels));
+	}
+	return {
+		samples: resampleLinear(mono, format.sampleRate, targetRate),
+		sampleRate: targetRate,
+		durationSeconds: frameCount / format.sampleRate
+	};
+}
+//#endregion
+export { decodeWavToMonoFloat32 };
+
+//# sourceMappingURL=wav.js.map

package/dist/src/voice/local/wav.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"wav.js","names":[],"sources":["../../../../src/voice/local/wav.ts"],"sourcesContent":["export interface DecodedPcmAudio {\n  samples: Float32Array;\n  sampleRate: number;\n  durationSeconds: number;\n}\n\nfunction readChunkId(buffer: Buffer, offset: number): string {\n  return buffer.toString('ascii', offset, offset + 4);\n}\n\nfunction resampleLinear(input: Float32Array, sourceRate: number, targetRate: number): Float32Array {\n  if (sourceRate === targetRate) return input;\n  const outputLength = Math.max(1, Math.round((input.length * targetRate) / sourceRate));\n  const output = new Float32Array(outputLength);\n  const ratio = sourceRate / targetRate;\n  for (let i = 0; i < outputLength; i += 1) {\n    const sourcePosition = i * ratio;\n    const left = Math.min(input.length - 1, Math.floor(sourcePosition));\n    const right = Math.min(input.length - 1, left + 1);\n    const fraction = sourcePosition - left;\n    output[i] = input[left]! * (1 - fraction) + input[right]! * fraction;\n  }\n  return output;\n}\n\nexport function decodeWavToMonoFloat32(buffer: Buffer, targetRate = 16_000): DecodedPcmAudio {\n  if (buffer.length < 44 || readChunkId(buffer, 0) !== 'RIFF' || readChunkId(buffer, 8) !== 'WAVE') {\n    throw new Error('Local STT currently requires PCM WAV audio');\n  }\n\n  let offset = 12;\n  let format: { audioFormat: number; channels: number; sampleRate: number; bitsPerSample: number } | undefined;\n  let dataOffset = -1;\n  let dataLength = 0;\n  while (offset + 8 <= buffer.length) {\n    const id = readChunkId(buffer, offset);\n    const length = buffer.readUInt32LE(offset + 4);\n    const bodyOffset = offset + 8;\n    if (bodyOffset + length > buffer.length) break;\n    if (id === 'fmt ' && length >= 16) {\n      format = {\n        audioFormat: buffer.readUInt16LE(bodyOffset),\n        channels: buffer.readUInt16LE(bodyOffset + 2),\n        sampleRate: buffer.readUInt32LE(bodyOffset + 4),\n        bitsPerSample: buffer.readUInt16LE(bodyOffset + 14),\n      };\n    } else if (id === 'data') {\n      dataOffset = bodyOffset;\n      dataLength = length;\n      break;\n    }\n    offset = bodyOffset + length + (length % 2);\n  }\n\n  if (!format || dataOffset < 0 || dataLength === 0) {\n    throw new Error('WAV audio is missing format or data chunks');\n  }\n  if (format.channels < 1 || format.channels > 8 || format.sampleRate < 8_000) {\n    throw new Error('Unsupported WAV channel count or sample rate');\n  }\n  const bytesPerSample = format.bitsPerSample / 8;\n  if (!Number.isInteger(bytesPerSample) || bytesPerSample < 2) {\n    throw new Error(`Unsupported WAV bit depth: ${format.bitsPerSample}`);\n  }\n  const frameBytes = bytesPerSample * format.channels;\n  const frameCount = Math.floor(dataLength / frameBytes);\n  const mono = new Float32Array(frameCount);\n  for (let frame = 0; frame < frameCount; frame += 1) {\n    let sum = 0;\n    for (let channel = 0; channel < format.channels; channel += 1) {\n      const sampleOffset = dataOffset + frame * frameBytes + channel * bytesPerSample;\n      if (format.audioFormat === 1 && format.bitsPerSample === 16) {\n        sum += buffer.readInt16LE(sampleOffset) / 32768;\n      } else if (format.audioFormat === 3 && format.bitsPerSample === 32) {\n        sum += buffer.readFloatLE(sampleOffset);\n      } else {\n        throw new Error(`Unsupported WAV encoding: format ${format.audioFormat}, ${format.bitsPerSample}-bit`);\n      }\n    }\n    mono[frame] = Math.max(-1, Math.min(1, sum / format.channels));\n  }\n\n  const samples = resampleLinear(mono, format.sampleRate, targetRate);\n  return {\n    samples,\n    sampleRate: targetRate,\n    durationSeconds: frameCount / format.sampleRate,\n  };\n}\n"],"mappings":";AAMA,SAAS,YAAY,QAAgB,QAAwB;AAC3D,QAAO,OAAO,SAAS,SAAS,QAAQ,SAAS,EAAE;;AAGrD,SAAS,eAAe,OAAqB,YAAoB,YAAkC;AACjG,KAAI,eAAe,WAAY,QAAO;CACtC,MAAM,eAAe,KAAK,IAAI,GAAG,KAAK,MAAO,MAAM,SAAS,aAAc,WAAW,CAAC;CACtF,MAAM,SAAS,IAAI,aAAa,aAAa;CAC7C,MAAM,QAAQ,aAAa;AAC3B,MAAK,IAAI,IAAI,GAAG,IAAI,cAAc,KAAK,GAAG;EACxC,MAAM,iBAAiB,IAAI;EAC3B,MAAM,OAAO,KAAK,IAAI,MAAM,SAAS,GAAG,KAAK,MAAM,eAAe,CAAC;EACnE,MAAM,QAAQ,KAAK,IAAI,MAAM,SAAS,GAAG,OAAO,EAAE;EAClD,MAAM,WAAW,iBAAiB;AAClC,SAAO,KAAK,MAAM,SAAU,IAAI,YAAY,MAAM,SAAU;;AAE9D,QAAO;;AAGT,SAAgB,uBAAuB,QAAgB,aAAa,MAAyB;AAC3F,KAAI,OAAO,SAAS,MAAM,YAAY,QAAQ,EAAE,KAAK,UAAU,YAAY,QAAQ,EAAE,KAAK,OACxF,OAAM,IAAI,MAAM,6CAA6C;CAG/D,IAAI,SAAS;CACb,IAAI;CACJ,IAAI,aAAa;CACjB,IAAI,aAAa;AACjB,QAAO,SAAS,KAAK,OAAO,QAAQ;EAClC,MAAM,KAAK,YAAY,QAAQ,OAAO;EACtC,MAAM,SAAS,OAAO,aAAa,SAAS,EAAE;EAC9C,MAAM,aAAa,SAAS;AAC5B,MAAI,aAAa,SAAS,OAAO,OAAQ;AACzC,MAAI,OAAO,UAAU,UAAU,GAC7B,UAAS;GACP,aAAa,OAAO,aAAa,WAAW;GAC5C,UAAU,OAAO,aAAa,aAAa,EAAE;GAC7C,YAAY,OAAO,aAAa,aAAa,EAAE;GAC/C,eAAe,OAAO,aAAa,aAAa,GAAG;GACpD;WACQ,OAAO,QAAQ;AACxB,gBAAa;AACb,gBAAa;AACb;;AAEF,WAAS,aAAa,SAAU,SAAS;;AAG3C,KAAI,CAAC,UAAU,aAAa,KAAK,eAAe,EAC9C,OAAM,IAAI,MAAM,6CAA6C;AAE/D,KAAI,OAAO,WAAW,KAAK,OAAO,WAAW,KAAK,OAAO,aAAa,IACpE,OAAM,IAAI,MAAM,+CAA+C;CAEjE,MAAM,iBAAiB,OAAO,gBAAgB;AAC9C,KAAI,CAAC,OAAO,UAAU,eAAe,IAAI,iBAAiB,EACxD,OAAM,IAAI,MAAM,8BAA8B,OAAO,gBAAgB;CAEvE,MAAM,aAAa,iBAAiB,OAAO;CAC3C,MAAM,aAAa,KAAK,MAAM,aAAa,WAAW;CACtD,MAAM,OAAO,IAAI,aAAa,WAAW;AACzC,MAAK,IAAI,QAAQ,GAAG,QAAQ,YAAY,SAAS,GAAG;EAClD,IAAI,MAAM;AACV,OAAK,IAAI,UAAU,GAAG,UAAU,OAAO,UAAU,WAAW,GAAG;GAC7D,MAAM,eAAe,aAAa,QAAQ,aAAa,UAAU;AACjE,OAAI,OAAO,gBAAgB,KAAK,OAAO,kBAAkB,GACvD,QAAO,OAAO,YAAY,aAAa,GAAG;YACjC,OAAO,gBAAgB,KAAK,OAAO,kBAAkB,GAC9D,QAAO,OAAO,YAAY,aAAa;OAEvC,OAAM,IAAI,MAAM,oCAAoC,OAAO,YAAY,IAAI,OAAO,cAAc,MAAM;;AAG1G,OAAK,SAAS,KAAK,IAAI,IAAI,KAAK,IAAI,GAAG,MAAM,OAAO,SAAS,CAAC;;AAIhE,QAAO;EACL,SAFc,eAAe,MAAM,OAAO,YAAY,WAE/C;EACP,YAAY;EACZ,iBAAiB,aAAa,OAAO;EACtC"}

package/dist/src/voice/metadata/builtin.js

@@ -30,6 +30,28 @@ const alibabaSttModels = [
 		name: "Paraformer MTL v1 (Multilingual)"
 	}
 ];
+const localSttModels = [
+	{
+		id: "sensevoice-small",
+		name: "SenseVoice Small 中文（推荐）",
+		description: "约 230 MB · 普通话、粤语、中英混说"
+	},
+	{
+		id: "tiny",
+		name: "Whisper Tiny (Fastest)",
+		description: "About 45 MB"
+	},
+	{
+		id: "base",
+		name: "Whisper Base (Multilingual)",
+		description: "About 80 MB"
+	},
+	{
+		id: "small",
+		name: "Whisper Small (Higher quality)",
+		description: "About 245 MB"
+	}
+];
 const alibabaTtsModels = [
 	{
 		id: "qwen-tts",
@@ -133,6 +155,30 @@ const edgeVoices = [
 	}
 ];
 const builtinVoiceProviderMetadata = [
+	{
+		id: "xopc-local",
+		capability: "stt",
+		displayName: "xopc Local Voice",
+		description: "Private on-device Chinese and multilingual transcription. Audio never leaves this device.",
+		aliases: ["local"],
+		models: localSttModels,
+		fields: [{
+			key: "model",
+			label: "Model",
+			type: "select",
+			options: localSttModels,
+			defaultValue: "sensevoice-small"
+		}, {
+			key: "language",
+			label: "Language hint",
+			type: "string",
+			placeholder: "zh"
+		}],
+		diagnostics: {
+			requiresApiKey: false,
+			configPath: "tools.media.audio.providers.xopc-local"
+		}
+	},
 	{
 		id: "alibaba",
 		capability: "stt",

package/dist/src/voice/metadata/builtin.js.map

@@ -1 +1 @@
-{"version":3,"file":"builtin.js","names":[],"sources":["../../../../src/voice/metadata/builtin.ts"],"sourcesContent":["import { registerVoiceProviderMetadata } from './registry.js';\nimport type { VoiceOptionMetadata, VoiceProviderMetadata } from './types.js';\nimport { MINIMAX_TTS_MODELS, MINIMAX_TTS_VOICES, OPENAI_TTS_MODELS, OPENAI_TTS_VOICES } from '../tts/providers/index.js';\n\nfunction titleCase(value: string): string {\n  return value.charAt(0).toUpperCase() + value.slice(1);\n}\n\nfunction options(values: readonly string[], names?: Record<string, string>): VoiceOptionMetadata[] {\n  return values.map((id) => ({ id, name: names?.[id] ?? titleCase(id) }));\n}\n\nconst alibabaSttModels = [\n  { id: 'paraformer-v2', name: 'Paraformer v2 (Recommended)' },\n  { id: 'paraformer-v1', name: 'Paraformer v1' },\n  { id: 'paraformer-8k-v1', name: 'Paraformer 8k v1 (Phone)' },\n  { id: 'paraformer-mtl-v1', name: 'Paraformer MTL v1 (Multilingual)' },\n];\n\nconst alibabaTtsModels = [\n  { id: 'qwen-tts', name: 'Qwen TTS (Recommended)' },\n  { id: 'qwen-tts-realtime', name: 'Qwen TTS Realtime' },\n  { id: 'qwen3-tts-flash', name: 'Qwen3 TTS Flash' },\n  { id: 'qwen3-tts-instruct-flash', name: 'Qwen3 TTS Instruct Flash' },\n];\n\nconst alibabaTtsVoices = [\n  { id: 'Cherry', name: 'Cherry' },\n  { id: 'Ethan', name: 'Ethan' },\n  { id: 'Serena', name: 'Serena' },\n  { id: 'Chelsie', name: 'Chelsie' },\n  { id: 'longxiaochun', name: 'Long Xiao Chun (龙小春)' },\n  { id: 'longxiaobai', name: 'Long Xiao Bai (龙小白)' },\n  { id: 'longwan', name: 'Long Wan (龙婉)' },\n  { id: 'longcheng', name: 'Long Cheng (龙呈)' },\n];\n\nconst edgeVoices = [\n  { id: 'en-US-MichelleNeural', name: 'Michelle (US English, Female)' },\n  { id: 'en-US-JennyNeural', name: 'Jenny (US English, Female)' },\n  { id: 'en-US-GuyNeural', name: 'Guy (US English, Male)' },\n  { id: 'en-GB-SoniaNeural', name: 'Sonia (UK English, Female)' },\n  { id: 'en-GB-RyanNeural', name: 'Ryan (UK English, Male)' },\n  { id: 'zh-CN-XiaoxiaoNeural', name: '晓晓 (中文, 女声)' },\n  { id: 'zh-CN-YunyangNeural', name: '云扬 (中文, 男声)' },\n  { id: 'zh-CN-XiaoyiNeural', name: '晓伊 (中文, 女声)' },\n  { id: 'ja-JP-NanamiNeural', name: '七海 (日本語, 女性)' },\n  { id: 'de-DE-KatjaNeural', name: 'Katja (Deutsch, Weiblich)' },\n  { id: 'fr-FR-DeniseNeural', name: 'Denise (Français, Féminin)' },\n  { id: 'es-ES-ElviraNeural', name: 'Elvira (Español, Femenino)' },\n];\n\nexport const builtinVoiceProviderMetadata: VoiceProviderMetadata[] = [\n  {\n    id: 'alibaba',\n    capability: 'stt',\n    displayName: 'Alibaba DashScope',\n    description: 'DashScope Paraformer speech-to-text, strong Chinese recognition.',\n    aliases: ['dashscope', 'paraformer'],\n    models: alibabaSttModels,\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'DASHSCOPE_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: alibabaSttModels, defaultValue: 'paraformer-v2' },\n      { key: 'language', label: 'Language hint', type: 'string', placeholder: 'zh' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['DASHSCOPE_API_KEY'], configPath: 'tools.media.audio.providers.alibaba' },\n  },\n  {\n    id: 'openai',\n    capability: 'stt',\n    displayName: 'OpenAI Whisper',\n    description: 'OpenAI Whisper transcription endpoint.',\n    models: [{ id: 'whisper-1', name: 'Whisper-1' }],\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'OPENAI_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: [{ id: 'whisper-1', name: 'Whisper-1' }], defaultValue: 'whisper-1' },\n      { key: 'baseUrl', label: 'Base URL', type: 'string', placeholder: 'https://api.openai.com/v1' },\n      { key: 'language', label: 'Language hint', type: 'string', placeholder: 'en' },\n      { key: 'prompt', label: 'Prompt', type: 'textarea' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['OPENAI_API_KEY'], configPath: 'tools.media.audio.providers.openai' },\n  },\n  {\n    id: 'openai',\n    capability: 'tts',\n    displayName: 'OpenAI',\n    description: 'OpenAI speech synthesis with multiple voices.',\n    models: options(OPENAI_TTS_MODELS, {\n      'gpt-4o-mini-tts': 'GPT-4o Mini TTS (Latest)',\n      'tts-1': 'TTS-1 (Fast)',\n      'tts-1-hd': 'TTS-1 HD (High Quality)',\n    }),\n    voices: options(OPENAI_TTS_VOICES),\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'OPENAI_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: options(OPENAI_TTS_MODELS), defaultValue: 'tts-1' },\n      { key: 'voice', label: 'Voice', type: 'select', options: options(OPENAI_TTS_VOICES), defaultValue: 'alloy' },\n      { key: 'baseUrl', label: 'Base URL', type: 'string', placeholder: 'https://api.openai.com/v1' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['OPENAI_API_KEY'], configPath: 'messages.tts.providers.openai' },\n  },\n  {\n    id: 'alibaba',\n    capability: 'tts',\n    displayName: 'Alibaba DashScope',\n    description: 'DashScope Qwen TTS, strong Chinese voice quality.',\n    models: alibabaTtsModels,\n    voices: alibabaTtsVoices,\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'DASHSCOPE_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: alibabaTtsModels, defaultValue: 'qwen-tts' },\n      { key: 'voice', label: 'Voice', type: 'select', options: alibabaTtsVoices, defaultValue: 'longxiaochun' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['DASHSCOPE_API_KEY'], configPath: 'messages.tts.providers.alibaba' },\n  },\n  {\n    id: 'minimax',\n    capability: 'tts',\n    displayName: 'MiniMax',\n    description: 'MiniMax high-quality async text-to-speech.',\n    models: options(MINIMAX_TTS_MODELS),\n    voices: options(MINIMAX_TTS_VOICES),\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'eyJ...', description: 'MINIMAX_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: options(MINIMAX_TTS_MODELS), defaultValue: 'speech-2.8-hd' },\n      { key: 'voice', label: 'Voice', type: 'select', options: options(MINIMAX_TTS_VOICES), defaultValue: 'male-qn-qingse' },\n      { key: 'baseUrl', label: 'Base URL', type: 'string', placeholder: 'https://api.minimaxi.com/v1' },\n      { key: 'groupId', label: 'Group ID', type: 'string' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['MINIMAX_API_KEY'], configPath: 'messages.tts.providers.minimax' },\n  },\n  {\n    id: 'edge',\n    capability: 'tts',\n    displayName: 'Microsoft Edge',\n    description: 'Free Edge TTS endpoint, no API key required.',\n    models: [{ id: 'edge-default', name: 'Edge TTS (Free)' }],\n    voices: edgeVoices,\n    fields: [\n      { key: 'voice', label: 'Voice', type: 'select', options: edgeVoices, defaultValue: 'en-US-MichelleNeural' },\n      { key: 'lang', label: 'Language', type: 'string', placeholder: 'en-US' },\n      { key: 'outputFormat', label: 'Output format', type: 'string', placeholder: 'audio-24khz-48kbitrate-mono-mp3' },\n      { key: 'rate', label: 'Rate', type: 'string', placeholder: '+0%' },\n      { key: 'pitch', label: 'Pitch', type: 'string', placeholder: '+0Hz' },\n      { key: 'volume', label: 'Volume', type: 'string', placeholder: '+0%' },\n      { key: 'proxy', label: 'Proxy', type: 'string' },\n    ],\n    diagnostics: { requiresApiKey: false, configPath: 'messages.tts.providers.edge' },\n  },\n];\n\nfor (const metadata of builtinVoiceProviderMetadata) {\n  registerVoiceProviderMetadata(metadata);\n}\n"],"mappings":";;;;;AAIA,SAAS,UAAU,OAAuB;AACxC,QAAO,MAAM,OAAO,EAAE,CAAC,aAAa,GAAG,MAAM,MAAM,EAAE;;AAGvD,SAAS,QAAQ,QAA2B,OAAuD;AACjG,QAAO,OAAO,KAAK,QAAQ;EAAE;EAAI,MAAM,QAAQ,OAAO,UAAU,GAAG;EAAE,EAAE;;AAGzE,MAAM,mBAAmB;CACvB;EAAE,IAAI;EAAiB,MAAM;EAA+B;CAC5D;EAAE,IAAI;EAAiB,MAAM;EAAiB;CAC9C;EAAE,IAAI;EAAoB,MAAM;EAA4B;CAC5D;EAAE,IAAI;EAAqB,MAAM;EAAoC;CACtE;AAED,MAAM,mBAAmB;CACvB;EAAE,IAAI;EAAY,MAAM;EAA0B;CAClD;EAAE,IAAI;EAAqB,MAAM;EAAqB;CACtD;EAAE,IAAI;EAAmB,MAAM;EAAmB;CAClD;EAAE,IAAI;EAA4B,MAAM;EAA4B;CACrE;AAED,MAAM,mBAAmB;CACvB;EAAE,IAAI;EAAU,MAAM;EAAU;CAChC;EAAE,IAAI;EAAS,MAAM;EAAS;CAC9B;EAAE,IAAI;EAAU,MAAM;EAAU;CAChC;EAAE,IAAI;EAAW,MAAM;EAAW;CAClC;EAAE,IAAI;EAAgB,MAAM;EAAwB;CACpD;EAAE,IAAI;EAAe,MAAM;EAAuB;CAClD;EAAE,IAAI;EAAW,MAAM;EAAiB;CACxC;EAAE,IAAI;EAAa,MAAM;EAAmB;CAC7C;AAED,MAAM,aAAa;CACjB;EAAE,IAAI;EAAwB,MAAM;EAAiC;CACrE;EAAE,IAAI;EAAqB,MAAM;EAA8B;CAC/D;EAAE,IAAI;EAAmB,MAAM;EAA0B;CACzD;EAAE,IAAI;EAAqB,MAAM;EAA8B;CAC/D;EAAE,IAAI;EAAoB,MAAM;EAA2B;CAC3D;EAAE,IAAI;EAAwB,MAAM;EAAe;CACnD;EAAE,IAAI;EAAuB,MAAM;EAAe;CAClD;EAAE,IAAI;EAAsB,MAAM;EAAe;CACjD;EAAE,IAAI;EAAsB,MAAM;EAAgB;CAClD;EAAE,IAAI;EAAqB,MAAM;EAA6B;CAC9D;EAAE,IAAI;EAAsB,MAAM;EAA8B;CAChE;EAAE,IAAI;EAAsB,MAAM;EAA8B;CACjE;AAED,MAAa,+BAAwD;CACnE;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,SAAS,CAAC,aAAa,aAAa;EACpC,QAAQ;EACR,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAqB;GAC5H;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAkB,cAAc;IAAiB;GAC1G;IAAE,KAAK;IAAY,OAAO;IAAiB,MAAM;IAAU,aAAa;IAAM;GAC/E;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,oBAAoB;GAAE,YAAY;GAAuC;EACzH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,CAAC;GAAE,IAAI;GAAa,MAAM;GAAa,CAAC;EAChD,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAkB;GACzH;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,CAAC;KAAE,IAAI;KAAa,MAAM;KAAa,CAAC;IAAE,cAAc;IAAa;GAC9H;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU,aAAa;IAA6B;GAC/F;IAAE,KAAK;IAAY,OAAO;IAAiB,MAAM;IAAU,aAAa;IAAM;GAC9E;IAAE,KAAK;IAAU,OAAO;IAAU,MAAM;IAAY;GACrD;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,iBAAiB;GAAE,YAAY;GAAsC;EACrH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,QAAQ,mBAAmB;GACjC,mBAAmB;GACnB,SAAS;GACT,YAAY;GACb,CAAC;EACF,QAAQ,QAAQ,kBAAkB;EAClC,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAkB;GACzH;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,kBAAkB;IAAE,cAAc;IAAS;GAC5G;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,kBAAkB;IAAE,cAAc;IAAS;GAC5G;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU,aAAa;IAA6B;GAChG;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,iBAAiB;GAAE,YAAY;GAAiC;EAChH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ;EACR,QAAQ;EACR,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAqB;GAC5H;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAkB,cAAc;IAAY;GACrG;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAkB,cAAc;IAAgB;GAC1G;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,oBAAoB;GAAE,YAAY;GAAkC;EACpH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,QAAQ,mBAAmB;EACnC,QAAQ,QAAQ,mBAAmB;EACnC,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAmB;GAC1H;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,mBAAmB;IAAE,cAAc;IAAiB;GACrH;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,mBAAmB;IAAE,cAAc;IAAkB;GACtH;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU,aAAa;IAA+B;GACjG;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU;GACtD;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,kBAAkB;GAAE,YAAY;GAAkC;EAClH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,CAAC;GAAE,IAAI;GAAgB,MAAM;GAAmB,CAAC;EACzD,QAAQ;EACR,QAAQ;GACN;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAY,cAAc;IAAwB;GAC3G;IAAE,KAAK;IAAQ,OAAO;IAAY,MAAM;IAAU,aAAa;IAAS;GACxE;IAAE,KAAK;IAAgB,OAAO;IAAiB,MAAM;IAAU,aAAa;IAAmC;GAC/G;IAAE,KAAK;IAAQ,OAAO;IAAQ,MAAM;IAAU,aAAa;IAAO;GAClE;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,aAAa;IAAQ;GACrE;IAAE,KAAK;IAAU,OAAO;IAAU,MAAM;IAAU,aAAa;IAAO;GACtE;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU;GACjD;EACD,aAAa;GAAE,gBAAgB;GAAO,YAAY;GAA+B;EAClF;CACF;AAED,KAAK,MAAM,YAAY,6BACrB,+BAA8B,SAAS"}
+{"version":3,"file":"builtin.js","names":[],"sources":["../../../../src/voice/metadata/builtin.ts"],"sourcesContent":["import { registerVoiceProviderMetadata } from './registry.js';\nimport type { VoiceOptionMetadata, VoiceProviderMetadata } from './types.js';\nimport { MINIMAX_TTS_MODELS, MINIMAX_TTS_VOICES, OPENAI_TTS_MODELS, OPENAI_TTS_VOICES } from '../tts/providers/index.js';\n\nfunction titleCase(value: string): string {\n  return value.charAt(0).toUpperCase() + value.slice(1);\n}\n\nfunction options(values: readonly string[], names?: Record<string, string>): VoiceOptionMetadata[] {\n  return values.map((id) => ({ id, name: names?.[id] ?? titleCase(id) }));\n}\n\nconst alibabaSttModels = [\n  { id: 'paraformer-v2', name: 'Paraformer v2 (Recommended)' },\n  { id: 'paraformer-v1', name: 'Paraformer v1' },\n  { id: 'paraformer-8k-v1', name: 'Paraformer 8k v1 (Phone)' },\n  { id: 'paraformer-mtl-v1', name: 'Paraformer MTL v1 (Multilingual)' },\n];\n\nconst localSttModels = [\n  {\n    id: 'sensevoice-small',\n    name: 'SenseVoice Small 中文（推荐）',\n    description: '约 230 MB · 普通话、粤语、中英混说',\n  },\n  { id: 'tiny', name: 'Whisper Tiny (Fastest)', description: 'About 45 MB' },\n  { id: 'base', name: 'Whisper Base (Multilingual)', description: 'About 80 MB' },\n  { id: 'small', name: 'Whisper Small (Higher quality)', description: 'About 245 MB' },\n];\n\nconst alibabaTtsModels = [\n  { id: 'qwen-tts', name: 'Qwen TTS (Recommended)' },\n  { id: 'qwen-tts-realtime', name: 'Qwen TTS Realtime' },\n  { id: 'qwen3-tts-flash', name: 'Qwen3 TTS Flash' },\n  { id: 'qwen3-tts-instruct-flash', name: 'Qwen3 TTS Instruct Flash' },\n];\n\nconst alibabaTtsVoices = [\n  { id: 'Cherry', name: 'Cherry' },\n  { id: 'Ethan', name: 'Ethan' },\n  { id: 'Serena', name: 'Serena' },\n  { id: 'Chelsie', name: 'Chelsie' },\n  { id: 'longxiaochun', name: 'Long Xiao Chun (龙小春)' },\n  { id: 'longxiaobai', name: 'Long Xiao Bai (龙小白)' },\n  { id: 'longwan', name: 'Long Wan (龙婉)' },\n  { id: 'longcheng', name: 'Long Cheng (龙呈)' },\n];\n\nconst edgeVoices = [\n  { id: 'en-US-MichelleNeural', name: 'Michelle (US English, Female)' },\n  { id: 'en-US-JennyNeural', name: 'Jenny (US English, Female)' },\n  { id: 'en-US-GuyNeural', name: 'Guy (US English, Male)' },\n  { id: 'en-GB-SoniaNeural', name: 'Sonia (UK English, Female)' },\n  { id: 'en-GB-RyanNeural', name: 'Ryan (UK English, Male)' },\n  { id: 'zh-CN-XiaoxiaoNeural', name: '晓晓 (中文, 女声)' },\n  { id: 'zh-CN-YunyangNeural', name: '云扬 (中文, 男声)' },\n  { id: 'zh-CN-XiaoyiNeural', name: '晓伊 (中文, 女声)' },\n  { id: 'ja-JP-NanamiNeural', name: '七海 (日本語, 女性)' },\n  { id: 'de-DE-KatjaNeural', name: 'Katja (Deutsch, Weiblich)' },\n  { id: 'fr-FR-DeniseNeural', name: 'Denise (Français, Féminin)' },\n  { id: 'es-ES-ElviraNeural', name: 'Elvira (Español, Femenino)' },\n];\n\nexport const builtinVoiceProviderMetadata: VoiceProviderMetadata[] = [\n  {\n    id: 'xopc-local',\n    capability: 'stt',\n    displayName: 'xopc Local Voice',\n    description: 'Private on-device Chinese and multilingual transcription. Audio never leaves this device.',\n    aliases: ['local'],\n    models: localSttModels,\n    fields: [\n      { key: 'model', label: 'Model', type: 'select', options: localSttModels, defaultValue: 'sensevoice-small' },\n      { key: 'language', label: 'Language hint', type: 'string', placeholder: 'zh' },\n    ],\n    diagnostics: { requiresApiKey: false, configPath: 'tools.media.audio.providers.xopc-local' },\n  },\n  {\n    id: 'alibaba',\n    capability: 'stt',\n    displayName: 'Alibaba DashScope',\n    description: 'DashScope Paraformer speech-to-text, strong Chinese recognition.',\n    aliases: ['dashscope', 'paraformer'],\n    models: alibabaSttModels,\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'DASHSCOPE_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: alibabaSttModels, defaultValue: 'paraformer-v2' },\n      { key: 'language', label: 'Language hint', type: 'string', placeholder: 'zh' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['DASHSCOPE_API_KEY'], configPath: 'tools.media.audio.providers.alibaba' },\n  },\n  {\n    id: 'openai',\n    capability: 'stt',\n    displayName: 'OpenAI Whisper',\n    description: 'OpenAI Whisper transcription endpoint.',\n    models: [{ id: 'whisper-1', name: 'Whisper-1' }],\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'OPENAI_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: [{ id: 'whisper-1', name: 'Whisper-1' }], defaultValue: 'whisper-1' },\n      { key: 'baseUrl', label: 'Base URL', type: 'string', placeholder: 'https://api.openai.com/v1' },\n      { key: 'language', label: 'Language hint', type: 'string', placeholder: 'en' },\n      { key: 'prompt', label: 'Prompt', type: 'textarea' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['OPENAI_API_KEY'], configPath: 'tools.media.audio.providers.openai' },\n  },\n  {\n    id: 'openai',\n    capability: 'tts',\n    displayName: 'OpenAI',\n    description: 'OpenAI speech synthesis with multiple voices.',\n    models: options(OPENAI_TTS_MODELS, {\n      'gpt-4o-mini-tts': 'GPT-4o Mini TTS (Latest)',\n      'tts-1': 'TTS-1 (Fast)',\n      'tts-1-hd': 'TTS-1 HD (High Quality)',\n    }),\n    voices: options(OPENAI_TTS_VOICES),\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'OPENAI_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: options(OPENAI_TTS_MODELS), defaultValue: 'tts-1' },\n      { key: 'voice', label: 'Voice', type: 'select', options: options(OPENAI_TTS_VOICES), defaultValue: 'alloy' },\n      { key: 'baseUrl', label: 'Base URL', type: 'string', placeholder: 'https://api.openai.com/v1' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['OPENAI_API_KEY'], configPath: 'messages.tts.providers.openai' },\n  },\n  {\n    id: 'alibaba',\n    capability: 'tts',\n    displayName: 'Alibaba DashScope',\n    description: 'DashScope Qwen TTS, strong Chinese voice quality.',\n    models: alibabaTtsModels,\n    voices: alibabaTtsVoices,\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'sk-...', description: 'DASHSCOPE_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: alibabaTtsModels, defaultValue: 'qwen-tts' },\n      { key: 'voice', label: 'Voice', type: 'select', options: alibabaTtsVoices, defaultValue: 'longxiaochun' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['DASHSCOPE_API_KEY'], configPath: 'messages.tts.providers.alibaba' },\n  },\n  {\n    id: 'minimax',\n    capability: 'tts',\n    displayName: 'MiniMax',\n    description: 'MiniMax high-quality async text-to-speech.',\n    models: options(MINIMAX_TTS_MODELS),\n    voices: options(MINIMAX_TTS_VOICES),\n    fields: [\n      { key: 'apiKey', label: 'API Key', type: 'password', secret: true, placeholder: 'eyJ...', description: 'MINIMAX_API_KEY' },\n      { key: 'model', label: 'Model', type: 'select', options: options(MINIMAX_TTS_MODELS), defaultValue: 'speech-2.8-hd' },\n      { key: 'voice', label: 'Voice', type: 'select', options: options(MINIMAX_TTS_VOICES), defaultValue: 'male-qn-qingse' },\n      { key: 'baseUrl', label: 'Base URL', type: 'string', placeholder: 'https://api.minimaxi.com/v1' },\n      { key: 'groupId', label: 'Group ID', type: 'string' },\n    ],\n    diagnostics: { requiresApiKey: true, envKeys: ['MINIMAX_API_KEY'], configPath: 'messages.tts.providers.minimax' },\n  },\n  {\n    id: 'edge',\n    capability: 'tts',\n    displayName: 'Microsoft Edge',\n    description: 'Free Edge TTS endpoint, no API key required.',\n    models: [{ id: 'edge-default', name: 'Edge TTS (Free)' }],\n    voices: edgeVoices,\n    fields: [\n      { key: 'voice', label: 'Voice', type: 'select', options: edgeVoices, defaultValue: 'en-US-MichelleNeural' },\n      { key: 'lang', label: 'Language', type: 'string', placeholder: 'en-US' },\n      { key: 'outputFormat', label: 'Output format', type: 'string', placeholder: 'audio-24khz-48kbitrate-mono-mp3' },\n      { key: 'rate', label: 'Rate', type: 'string', placeholder: '+0%' },\n      { key: 'pitch', label: 'Pitch', type: 'string', placeholder: '+0Hz' },\n      { key: 'volume', label: 'Volume', type: 'string', placeholder: '+0%' },\n      { key: 'proxy', label: 'Proxy', type: 'string' },\n    ],\n    diagnostics: { requiresApiKey: false, configPath: 'messages.tts.providers.edge' },\n  },\n];\n\nfor (const metadata of builtinVoiceProviderMetadata) {\n  registerVoiceProviderMetadata(metadata);\n}\n"],"mappings":";;;;;AAIA,SAAS,UAAU,OAAuB;AACxC,QAAO,MAAM,OAAO,EAAE,CAAC,aAAa,GAAG,MAAM,MAAM,EAAE;;AAGvD,SAAS,QAAQ,QAA2B,OAAuD;AACjG,QAAO,OAAO,KAAK,QAAQ;EAAE;EAAI,MAAM,QAAQ,OAAO,UAAU,GAAG;EAAE,EAAE;;AAGzE,MAAM,mBAAmB;CACvB;EAAE,IAAI;EAAiB,MAAM;EAA+B;CAC5D;EAAE,IAAI;EAAiB,MAAM;EAAiB;CAC9C;EAAE,IAAI;EAAoB,MAAM;EAA4B;CAC5D;EAAE,IAAI;EAAqB,MAAM;EAAoC;CACtE;AAED,MAAM,iBAAiB;CACrB;EACE,IAAI;EACJ,MAAM;EACN,aAAa;EACd;CACD;EAAE,IAAI;EAAQ,MAAM;EAA0B,aAAa;EAAe;CAC1E;EAAE,IAAI;EAAQ,MAAM;EAA+B,aAAa;EAAe;CAC/E;EAAE,IAAI;EAAS,MAAM;EAAkC,aAAa;EAAgB;CACrF;AAED,MAAM,mBAAmB;CACvB;EAAE,IAAI;EAAY,MAAM;EAA0B;CAClD;EAAE,IAAI;EAAqB,MAAM;EAAqB;CACtD;EAAE,IAAI;EAAmB,MAAM;EAAmB;CAClD;EAAE,IAAI;EAA4B,MAAM;EAA4B;CACrE;AAED,MAAM,mBAAmB;CACvB;EAAE,IAAI;EAAU,MAAM;EAAU;CAChC;EAAE,IAAI;EAAS,MAAM;EAAS;CAC9B;EAAE,IAAI;EAAU,MAAM;EAAU;CAChC;EAAE,IAAI;EAAW,MAAM;EAAW;CAClC;EAAE,IAAI;EAAgB,MAAM;EAAwB;CACpD;EAAE,IAAI;EAAe,MAAM;EAAuB;CAClD;EAAE,IAAI;EAAW,MAAM;EAAiB;CACxC;EAAE,IAAI;EAAa,MAAM;EAAmB;CAC7C;AAED,MAAM,aAAa;CACjB;EAAE,IAAI;EAAwB,MAAM;EAAiC;CACrE;EAAE,IAAI;EAAqB,MAAM;EAA8B;CAC/D;EAAE,IAAI;EAAmB,MAAM;EAA0B;CACzD;EAAE,IAAI;EAAqB,MAAM;EAA8B;CAC/D;EAAE,IAAI;EAAoB,MAAM;EAA2B;CAC3D;EAAE,IAAI;EAAwB,MAAM;EAAe;CACnD;EAAE,IAAI;EAAuB,MAAM;EAAe;CAClD;EAAE,IAAI;EAAsB,MAAM;EAAe;CACjD;EAAE,IAAI;EAAsB,MAAM;EAAgB;CAClD;EAAE,IAAI;EAAqB,MAAM;EAA6B;CAC9D;EAAE,IAAI;EAAsB,MAAM;EAA8B;CAChE;EAAE,IAAI;EAAsB,MAAM;EAA8B;CACjE;AAED,MAAa,+BAAwD;CACnE;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,SAAS,CAAC,QAAQ;EAClB,QAAQ;EACR,QAAQ,CACN;GAAE,KAAK;GAAS,OAAO;GAAS,MAAM;GAAU,SAAS;GAAgB,cAAc;GAAoB,EAC3G;GAAE,KAAK;GAAY,OAAO;GAAiB,MAAM;GAAU,aAAa;GAAM,CAC/E;EACD,aAAa;GAAE,gBAAgB;GAAO,YAAY;GAA0C;EAC7F;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,SAAS,CAAC,aAAa,aAAa;EACpC,QAAQ;EACR,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAqB;GAC5H;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAkB,cAAc;IAAiB;GAC1G;IAAE,KAAK;IAAY,OAAO;IAAiB,MAAM;IAAU,aAAa;IAAM;GAC/E;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,oBAAoB;GAAE,YAAY;GAAuC;EACzH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,CAAC;GAAE,IAAI;GAAa,MAAM;GAAa,CAAC;EAChD,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAkB;GACzH;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,CAAC;KAAE,IAAI;KAAa,MAAM;KAAa,CAAC;IAAE,cAAc;IAAa;GAC9H;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU,aAAa;IAA6B;GAC/F;IAAE,KAAK;IAAY,OAAO;IAAiB,MAAM;IAAU,aAAa;IAAM;GAC9E;IAAE,KAAK;IAAU,OAAO;IAAU,MAAM;IAAY;GACrD;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,iBAAiB;GAAE,YAAY;GAAsC;EACrH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,QAAQ,mBAAmB;GACjC,mBAAmB;GACnB,SAAS;GACT,YAAY;GACb,CAAC;EACF,QAAQ,QAAQ,kBAAkB;EAClC,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAkB;GACzH;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,kBAAkB;IAAE,cAAc;IAAS;GAC5G;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,kBAAkB;IAAE,cAAc;IAAS;GAC5G;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU,aAAa;IAA6B;GAChG;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,iBAAiB;GAAE,YAAY;GAAiC;EAChH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ;EACR,QAAQ;EACR,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAqB;GAC5H;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAkB,cAAc;IAAY;GACrG;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAkB,cAAc;IAAgB;GAC1G;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,oBAAoB;GAAE,YAAY;GAAkC;EACpH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,QAAQ,mBAAmB;EACnC,QAAQ,QAAQ,mBAAmB;EACnC,QAAQ;GACN;IAAE,KAAK;IAAU,OAAO;IAAW,MAAM;IAAY,QAAQ;IAAM,aAAa;IAAU,aAAa;IAAmB;GAC1H;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,mBAAmB;IAAE,cAAc;IAAiB;GACrH;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS,QAAQ,mBAAmB;IAAE,cAAc;IAAkB;GACtH;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU,aAAa;IAA+B;GACjG;IAAE,KAAK;IAAW,OAAO;IAAY,MAAM;IAAU;GACtD;EACD,aAAa;GAAE,gBAAgB;GAAM,SAAS,CAAC,kBAAkB;GAAE,YAAY;GAAkC;EAClH;CACD;EACE,IAAI;EACJ,YAAY;EACZ,aAAa;EACb,aAAa;EACb,QAAQ,CAAC;GAAE,IAAI;GAAgB,MAAM;GAAmB,CAAC;EACzD,QAAQ;EACR,QAAQ;GACN;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,SAAS;IAAY,cAAc;IAAwB;GAC3G;IAAE,KAAK;IAAQ,OAAO;IAAY,MAAM;IAAU,aAAa;IAAS;GACxE;IAAE,KAAK;IAAgB,OAAO;IAAiB,MAAM;IAAU,aAAa;IAAmC;GAC/G;IAAE,KAAK;IAAQ,OAAO;IAAQ,MAAM;IAAU,aAAa;IAAO;GAClE;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU,aAAa;IAAQ;GACrE;IAAE,KAAK;IAAU,OAAO;IAAU,MAAM;IAAU,aAAa;IAAO;GACtE;IAAE,KAAK;IAAS,OAAO;IAAS,MAAM;IAAU;GACjD;EACD,aAAa;GAAE,gBAAgB;GAAO,YAAY;GAA+B;EAClF;CACF;AAED,KAAK,MAAM,YAAY,6BACrB,+BAA8B,SAAS"}

package/dist/src/voice/stt/factory.js

@@ -19,7 +19,13 @@ function resolveSTTProviderConfig(providerId, config, entry) {
 	}
 	const fields = readSttProviderFields(resolveSttProviderConfigSlice(providerId, config), entry);
 	const apiKey = resolveApiKey(providerId, fields, plugin.envKey);
-	if (plugin.requiresApiKey !== false && !apiKey) {
+	const requiresApiKey = plugin.requiresApiKey !== false;
+	const model = fields.model ?? entry?.model ?? plugin.defaultModels?.audio;
+	if (plugin.isConfigured && !plugin.isConfigured({ model })) {
+		log.debug({ providerId }, `STT provider "${providerId}" runtime is not ready; skipping`);
+		return null;
+	}
+	if (requiresApiKey && !apiKey) {
 		log.debug({ providerId }, `STT provider "${providerId}" missing API key; skipping`);
 		return null;
 	}
@@ -28,7 +34,7 @@ function resolveSTTProviderConfig(providerId, config, entry) {
 		...apiKey ? { apiKey } : {},
 		...fields.baseUrl ? { baseUrl: fields.baseUrl } : {},
 		...fields.headers ? { headers: fields.headers } : {},
-		...fields.model ?? entry?.model ?? plugin.defaultModels?.audio ? { model: fields.model ?? entry?.model ?? plugin.defaultModels?.audio } : {},
+		...model ? { model } : {},
 		...fields.language ? { language: fields.language } : {},
 		...fields.prompt ? { prompt: fields.prompt } : {}
 	};

package/dist/src/voice/stt/factory.js.map

@@ -1 +1 @@
-{"version":3,"file":"factory.js","names":[],"sources":["../../../../src/voice/stt/factory.ts"],"sourcesContent":["/**\n * STT factory — registry-driven provider chain construction.\n *\n * Resolves STTConfig slices into `AudioProviderResolvedConfig` arrays that are\n * consumed by `runAudioTranscription`. Side-effect import of `./providers/index.js`\n * ensures all built-in STT providers self-register with the media-understanding\n * registry before the first lookup.\n */\n\nimport { createLogger } from '../../utils/logger.js';\n\nimport './providers/index.js'; // side-effect: register built-in STT providers\nimport { resolveModelEntries } from '../../media-understanding/resolve-entries.js';\nimport {\n  getMediaUnderstandingProvider,\n  listMediaUnderstandingProviders,\n} from '../../media-understanding/registry.js';\nimport type { AudioProviderResolvedConfig } from '../../media-understanding/audio-transcription-runner.js';\nimport {\n  STT_LEGACY_ENV_KEYS,\n  readSttProviderFields,\n  resolveSttProviderConfigSlice,\n} from './config-slice.js';\nimport type { MediaUnderstandingModelEntry, STTConfig } from './types.js';\n\nconst log = createLogger('STT:Factory');\n\nfunction resolveApiKey(\n  providerId: string,\n  fields: ReturnType<typeof readSttProviderFields>,\n  envKey?: string,\n): string | undefined {\n  return (\n    fields.apiKey ??\n    (envKey ? process.env[envKey] : undefined) ??\n    process.env[STT_LEGACY_ENV_KEYS[providerId] ?? '']\n  );\n}\n\n/** Resolve a single STT provider config slice → runner-shaped resolved config, or null when unavailable. */\nexport function resolveSTTProviderConfig(\n  providerId: string,\n  config: STTConfig,\n  entry?: MediaUnderstandingModelEntry,\n): AudioProviderResolvedConfig | null {\n  const plugin = getMediaUnderstandingProvider(providerId);\n  if (!plugin || typeof plugin.transcribeAudio !== 'function') {\n    log.warn(\n      { providerId },\n      `STT provider \"${providerId}\" is not registered or does not implement transcribeAudio`,\n    );\n    return null;\n  }\n\n  const slice = resolveSttProviderConfigSlice(providerId, config);\n  const fields = readSttProviderFields(slice, entry);\n  const apiKey = resolveApiKey(providerId, fields, plugin.envKey);\n  const requiresApiKey = plugin.requiresApiKey !== false;\n  if (requiresApiKey && !apiKey) {\n    log.debug({ providerId }, `STT provider \"${providerId}\" missing API key; skipping`);\n    return null;\n  }\n\n  return {\n    id: providerId,\n    ...(apiKey ? { apiKey } : {}),\n    ...(fields.baseUrl ? { baseUrl: fields.baseUrl } : {}),\n    ...(fields.headers ? { headers: fields.headers } : {}),\n    ...(fields.model ?? entry?.model ?? plugin.defaultModels?.audio\n      ? { model: fields.model ?? entry?.model ?? plugin.defaultModels?.audio }\n      : {}),\n    ...(fields.language ? { language: fields.language } : {}),\n    ...(fields.prompt ? { prompt: fields.prompt } : {}),\n  };\n}\n\nfunction sortAudioProvidersByAutoPriority(\n  plugins: ReturnType<typeof listMediaUnderstandingProviders>,\n): string[] {\n  return [...plugins]\n    .filter(\n      (plugin) =>\n        plugin.capabilities?.includes('audio') && typeof plugin.transcribeAudio === 'function',\n    )\n    .sort((left, right) => {\n      const leftOrder = left.autoPriority?.audio ?? Number.MAX_SAFE_INTEGER;\n      const rightOrder = right.autoPriority?.audio ?? Number.MAX_SAFE_INTEGER;\n      if (leftOrder !== rightOrder) {\n        return leftOrder - rightOrder;\n      }\n      return left.id.localeCompare(right.id);\n    })\n    .map((plugin) => plugin.id);\n}\n\n/** Resolve provider order: explicit fallback when enabled, otherwise auto-select by registry order. */\nexport function resolveSTTProviderOrder(\n  primary: string,\n  fallback: STTConfig['fallback'] | undefined,\n  config: STTConfig,\n): string[] {\n  if (fallback?.enabled && fallback.order.length > 0) {\n    const order: string[] = [primary];\n    for (const providerId of fallback.order) {\n      if (providerId !== primary && !order.includes(providerId)) {\n        order.push(providerId);\n      }\n    }\n    return order;\n  }\n\n  const configured = sortAudioProvidersByAutoPriority(listMediaUnderstandingProviders()).filter(\n    (providerId) => resolveSTTProviderConfig(providerId, config) !== null,\n  );\n\n  const order: string[] = [primary];\n  for (const providerId of configured) {\n    if (providerId !== primary && !order.includes(providerId)) {\n      order.push(providerId);\n    }\n  }\n  return order;\n}\n\nfunction resolveModelEntryChain(config: STTConfig): AudioProviderResolvedConfig[] {\n  const entries = resolveModelEntries({\n    capability: 'audio',\n    capabilityModels: config.models,\n    sharedModels: config.sharedModels,\n  });\n  const chain: AudioProviderResolvedConfig[] = [];\n  for (const entry of entries) {\n    const providerId = entry.provider?.trim();\n    if (!providerId) {\n      continue;\n    }\n    const resolved = resolveSTTProviderConfig(providerId, config, entry);\n    if (resolved) {\n      chain.push(resolved);\n    }\n  }\n  return chain;\n}\n\n/** Build the full chain of resolved provider configs in priority order. */\nexport function resolveSTTProviderChain(config: STTConfig): AudioProviderResolvedConfig[] {\n  if (!config.enabled) {\n    return [];\n  }\n\n  const modelChain = resolveModelEntryChain(config);\n  if (modelChain.length > 0) {\n    return modelChain;\n  }\n\n  const order = resolveSTTProviderOrder(config.provider, config.fallback, config);\n  const chain: AudioProviderResolvedConfig[] = [];\n  for (const providerId of order) {\n    const resolved = resolveSTTProviderConfig(providerId, config);\n    if (resolved) {\n      chain.push(resolved);\n    }\n  }\n  return chain;\n}\n"],"mappings":";;;;;;;aASqD;AAgBrD,MAAM,MAAM,aAAa,cAAc;AAEvC,SAAS,cACP,YACA,QACA,QACoB;AACpB,QACE,OAAO,WACN,SAAS,QAAQ,IAAI,UAAU,KAAA,MAChC,QAAQ,IAAI,oBAAoB,eAAe;;;AAKnD,SAAgB,yBACd,YACA,QACA,OACoC;CACpC,MAAM,SAAS,8BAA8B,WAAW;AACxD,KAAI,CAAC,UAAU,OAAO,OAAO,oBAAoB,YAAY;AAC3D,MAAI,KACF,EAAE,YAAY,EACd,iBAAiB,WAAW,2DAC7B;AACD,SAAO;;CAIT,MAAM,SAAS,sBADD,8BAA8B,YAAY,OACd,EAAE,MAAM;CAClD,MAAM,SAAS,cAAc,YAAY,QAAQ,OAAO,OAAO;AAE/D,KADuB,OAAO,mBAAmB,SAC3B,CAAC,QAAQ;AAC7B,MAAI,MAAM,EAAE,YAAY,EAAE,iBAAiB,WAAW,6BAA6B;AACnF,SAAO;;AAGT,QAAO;EACL,IAAI;EACJ,GAAI,SAAS,EAAE,QAAQ,GAAG,EAAE;EAC5B,GAAI,OAAO,UAAU,EAAE,SAAS,OAAO,SAAS,GAAG,EAAE;EACrD,GAAI,OAAO,UAAU,EAAE,SAAS,OAAO,SAAS,GAAG,EAAE;EACrD,GAAI,OAAO,SAAS,OAAO,SAAS,OAAO,eAAe,QACtD,EAAE,OAAO,OAAO,SAAS,OAAO,SAAS,OAAO,eAAe,OAAO,GACtE,EAAE;EACN,GAAI,OAAO,WAAW,EAAE,UAAU,OAAO,UAAU,GAAG,EAAE;EACxD,GAAI,OAAO,SAAS,EAAE,QAAQ,OAAO,QAAQ,GAAG,EAAE;EACnD;;AAGH,SAAS,iCACP,SACU;AACV,QAAO,CAAC,GAAG,QAAQ,CAChB,QACE,WACC,OAAO,cAAc,SAAS,QAAQ,IAAI,OAAO,OAAO,oBAAoB,WAC/E,CACA,MAAM,MAAM,UAAU;EACrB,MAAM,YAAY,KAAK,cAAc,SAAS,OAAO;EACrD,MAAM,aAAa,MAAM,cAAc,SAAS,OAAO;AACvD,MAAI,cAAc,WAChB,QAAO,YAAY;AAErB,SAAO,KAAK,GAAG,cAAc,MAAM,GAAG;GACtC,CACD,KAAK,WAAW,OAAO,GAAG;;;AAI/B,SAAgB,wBACd,SACA,UACA,QACU;AACV,KAAI,UAAU,WAAW,SAAS,MAAM,SAAS,GAAG;EAClD,MAAM,QAAkB,CAAC,QAAQ;AACjC,OAAK,MAAM,cAAc,SAAS,MAChC,KAAI,eAAe,WAAW,CAAC,MAAM,SAAS,WAAW,CACvD,OAAM,KAAK,WAAW;AAG1B,SAAO;;CAGT,MAAM,aAAa,iCAAiC,iCAAiC,CAAC,CAAC,QACpF,eAAe,yBAAyB,YAAY,OAAO,KAAK,KAClE;CAED,MAAM,QAAkB,CAAC,QAAQ;AACjC,MAAK,MAAM,cAAc,WACvB,KAAI,eAAe,WAAW,CAAC,MAAM,SAAS,WAAW,CACvD,OAAM,KAAK,WAAW;AAG1B,QAAO;;AAGT,SAAS,uBAAuB,QAAkD;CAChF,MAAM,UAAU,oBAAoB;EAClC,YAAY;EACZ,kBAAkB,OAAO;EACzB,cAAc,OAAO;EACtB,CAAC;CACF,MAAM,QAAuC,EAAE;AAC/C,MAAK,MAAM,SAAS,SAAS;EAC3B,MAAM,aAAa,MAAM,UAAU,MAAM;AACzC,MAAI,CAAC,WACH;EAEF,MAAM,WAAW,yBAAyB,YAAY,QAAQ,MAAM;AACpE,MAAI,SACF,OAAM,KAAK,SAAS;;AAGxB,QAAO;;;AAIT,SAAgB,wBAAwB,QAAkD;AACxF,KAAI,CAAC,OAAO,QACV,QAAO,EAAE;CAGX,MAAM,aAAa,uBAAuB,OAAO;AACjD,KAAI,WAAW,SAAS,EACtB,QAAO;CAGT,MAAM,QAAQ,wBAAwB,OAAO,UAAU,OAAO,UAAU,OAAO;CAC/E,MAAM,QAAuC,EAAE;AAC/C,MAAK,MAAM,cAAc,OAAO;EAC9B,MAAM,WAAW,yBAAyB,YAAY,OAAO;AAC7D,MAAI,SACF,OAAM,KAAK,SAAS;;AAGxB,QAAO"}
+{"version":3,"file":"factory.js","names":[],"sources":["../../../../src/voice/stt/factory.ts"],"sourcesContent":["/**\n * STT factory — registry-driven provider chain construction.\n *\n * Resolves STTConfig slices into `AudioProviderResolvedConfig` arrays that are\n * consumed by `runAudioTranscription`. Side-effect import of `./providers/index.js`\n * ensures all built-in STT providers self-register with the media-understanding\n * registry before the first lookup.\n */\n\nimport { createLogger } from '../../utils/logger.js';\n\nimport './providers/index.js'; // side-effect: register built-in STT providers\nimport { resolveModelEntries } from '../../media-understanding/resolve-entries.js';\nimport {\n  getMediaUnderstandingProvider,\n  listMediaUnderstandingProviders,\n} from '../../media-understanding/registry.js';\nimport type { AudioProviderResolvedConfig } from '../../media-understanding/audio-transcription-runner.js';\nimport {\n  STT_LEGACY_ENV_KEYS,\n  readSttProviderFields,\n  resolveSttProviderConfigSlice,\n} from './config-slice.js';\nimport type { MediaUnderstandingModelEntry, STTConfig } from './types.js';\n\nconst log = createLogger('STT:Factory');\n\nfunction resolveApiKey(\n  providerId: string,\n  fields: ReturnType<typeof readSttProviderFields>,\n  envKey?: string,\n): string | undefined {\n  return (\n    fields.apiKey ??\n    (envKey ? process.env[envKey] : undefined) ??\n    process.env[STT_LEGACY_ENV_KEYS[providerId] ?? '']\n  );\n}\n\n/** Resolve a single STT provider config slice → runner-shaped resolved config, or null when unavailable. */\nexport function resolveSTTProviderConfig(\n  providerId: string,\n  config: STTConfig,\n  entry?: MediaUnderstandingModelEntry,\n): AudioProviderResolvedConfig | null {\n  const plugin = getMediaUnderstandingProvider(providerId);\n  if (!plugin || typeof plugin.transcribeAudio !== 'function') {\n    log.warn(\n      { providerId },\n      `STT provider \"${providerId}\" is not registered or does not implement transcribeAudio`,\n    );\n    return null;\n  }\n\n  const slice = resolveSttProviderConfigSlice(providerId, config);\n  const fields = readSttProviderFields(slice, entry);\n  const apiKey = resolveApiKey(providerId, fields, plugin.envKey);\n  const requiresApiKey = plugin.requiresApiKey !== false;\n  const model = fields.model ?? entry?.model ?? plugin.defaultModels?.audio;\n  if (plugin.isConfigured && !plugin.isConfigured({ model })) {\n    log.debug({ providerId }, `STT provider \"${providerId}\" runtime is not ready; skipping`);\n    return null;\n  }\n  if (requiresApiKey && !apiKey) {\n    log.debug({ providerId }, `STT provider \"${providerId}\" missing API key; skipping`);\n    return null;\n  }\n\n  return {\n    id: providerId,\n    ...(apiKey ? { apiKey } : {}),\n    ...(fields.baseUrl ? { baseUrl: fields.baseUrl } : {}),\n    ...(fields.headers ? { headers: fields.headers } : {}),\n    ...(model ? { model } : {}),\n    ...(fields.language ? { language: fields.language } : {}),\n    ...(fields.prompt ? { prompt: fields.prompt } : {}),\n  };\n}\n\nfunction sortAudioProvidersByAutoPriority(\n  plugins: ReturnType<typeof listMediaUnderstandingProviders>,\n): string[] {\n  return [...plugins]\n    .filter(\n      (plugin) =>\n        plugin.capabilities?.includes('audio') && typeof plugin.transcribeAudio === 'function',\n    )\n    .sort((left, right) => {\n      const leftOrder = left.autoPriority?.audio ?? Number.MAX_SAFE_INTEGER;\n      const rightOrder = right.autoPriority?.audio ?? Number.MAX_SAFE_INTEGER;\n      if (leftOrder !== rightOrder) {\n        return leftOrder - rightOrder;\n      }\n      return left.id.localeCompare(right.id);\n    })\n    .map((plugin) => plugin.id);\n}\n\n/** Resolve provider order: explicit fallback when enabled, otherwise auto-select by registry order. */\nexport function resolveSTTProviderOrder(\n  primary: string,\n  fallback: STTConfig['fallback'] | undefined,\n  config: STTConfig,\n): string[] {\n  if (fallback?.enabled && fallback.order.length > 0) {\n    const order: string[] = [primary];\n    for (const providerId of fallback.order) {\n      if (providerId !== primary && !order.includes(providerId)) {\n        order.push(providerId);\n      }\n    }\n    return order;\n  }\n\n  const configured = sortAudioProvidersByAutoPriority(listMediaUnderstandingProviders()).filter(\n    (providerId) => resolveSTTProviderConfig(providerId, config) !== null,\n  );\n\n  const order: string[] = [primary];\n  for (const providerId of configured) {\n    if (providerId !== primary && !order.includes(providerId)) {\n      order.push(providerId);\n    }\n  }\n  return order;\n}\n\nfunction resolveModelEntryChain(config: STTConfig): AudioProviderResolvedConfig[] {\n  const entries = resolveModelEntries({\n    capability: 'audio',\n    capabilityModels: config.models,\n    sharedModels: config.sharedModels,\n  });\n  const chain: AudioProviderResolvedConfig[] = [];\n  for (const entry of entries) {\n    const providerId = entry.provider?.trim();\n    if (!providerId) {\n      continue;\n    }\n    const resolved = resolveSTTProviderConfig(providerId, config, entry);\n    if (resolved) {\n      chain.push(resolved);\n    }\n  }\n  return chain;\n}\n\n/** Build the full chain of resolved provider configs in priority order. */\nexport function resolveSTTProviderChain(config: STTConfig): AudioProviderResolvedConfig[] {\n  if (!config.enabled) {\n    return [];\n  }\n\n  const modelChain = resolveModelEntryChain(config);\n  if (modelChain.length > 0) {\n    return modelChain;\n  }\n\n  const order = resolveSTTProviderOrder(config.provider, config.fallback, config);\n  const chain: AudioProviderResolvedConfig[] = [];\n  for (const providerId of order) {\n    const resolved = resolveSTTProviderConfig(providerId, config);\n    if (resolved) {\n      chain.push(resolved);\n    }\n  }\n  return chain;\n}\n"],"mappings":";;;;;;;aASqD;AAgBrD,MAAM,MAAM,aAAa,cAAc;AAEvC,SAAS,cACP,YACA,QACA,QACoB;AACpB,QACE,OAAO,WACN,SAAS,QAAQ,IAAI,UAAU,KAAA,MAChC,QAAQ,IAAI,oBAAoB,eAAe;;;AAKnD,SAAgB,yBACd,YACA,QACA,OACoC;CACpC,MAAM,SAAS,8BAA8B,WAAW;AACxD,KAAI,CAAC,UAAU,OAAO,OAAO,oBAAoB,YAAY;AAC3D,MAAI,KACF,EAAE,YAAY,EACd,iBAAiB,WAAW,2DAC7B;AACD,SAAO;;CAIT,MAAM,SAAS,sBADD,8BAA8B,YAAY,OACd,EAAE,MAAM;CAClD,MAAM,SAAS,cAAc,YAAY,QAAQ,OAAO,OAAO;CAC/D,MAAM,iBAAiB,OAAO,mBAAmB;CACjD,MAAM,QAAQ,OAAO,SAAS,OAAO,SAAS,OAAO,eAAe;AACpE,KAAI,OAAO,gBAAgB,CAAC,OAAO,aAAa,EAAE,OAAO,CAAC,EAAE;AAC1D,MAAI,MAAM,EAAE,YAAY,EAAE,iBAAiB,WAAW,kCAAkC;AACxF,SAAO;;AAET,KAAI,kBAAkB,CAAC,QAAQ;AAC7B,MAAI,MAAM,EAAE,YAAY,EAAE,iBAAiB,WAAW,6BAA6B;AACnF,SAAO;;AAGT,QAAO;EACL,IAAI;EACJ,GAAI,SAAS,EAAE,QAAQ,GAAG,EAAE;EAC5B,GAAI,OAAO,UAAU,EAAE,SAAS,OAAO,SAAS,GAAG,EAAE;EACrD,GAAI,OAAO,UAAU,EAAE,SAAS,OAAO,SAAS,GAAG,EAAE;EACrD,GAAI,QAAQ,EAAE,OAAO,GAAG,EAAE;EAC1B,GAAI,OAAO,WAAW,EAAE,UAAU,OAAO,UAAU,GAAG,EAAE;EACxD,GAAI,OAAO,SAAS,EAAE,QAAQ,OAAO,QAAQ,GAAG,EAAE;EACnD;;AAGH,SAAS,iCACP,SACU;AACV,QAAO,CAAC,GAAG,QAAQ,CAChB,QACE,WACC,OAAO,cAAc,SAAS,QAAQ,IAAI,OAAO,OAAO,oBAAoB,WAC/E,CACA,MAAM,MAAM,UAAU;EACrB,MAAM,YAAY,KAAK,cAAc,SAAS,OAAO;EACrD,MAAM,aAAa,MAAM,cAAc,SAAS,OAAO;AACvD,MAAI,cAAc,WAChB,QAAO,YAAY;AAErB,SAAO,KAAK,GAAG,cAAc,MAAM,GAAG;GACtC,CACD,KAAK,WAAW,OAAO,GAAG;;;AAI/B,SAAgB,wBACd,SACA,UACA,QACU;AACV,KAAI,UAAU,WAAW,SAAS,MAAM,SAAS,GAAG;EAClD,MAAM,QAAkB,CAAC,QAAQ;AACjC,OAAK,MAAM,cAAc,SAAS,MAChC,KAAI,eAAe,WAAW,CAAC,MAAM,SAAS,WAAW,CACvD,OAAM,KAAK,WAAW;AAG1B,SAAO;;CAGT,MAAM,aAAa,iCAAiC,iCAAiC,CAAC,CAAC,QACpF,eAAe,yBAAyB,YAAY,OAAO,KAAK,KAClE;CAED,MAAM,QAAkB,CAAC,QAAQ;AACjC,MAAK,MAAM,cAAc,WACvB,KAAI,eAAe,WAAW,CAAC,MAAM,SAAS,WAAW,CACvD,OAAM,KAAK,WAAW;AAG1B,QAAO;;AAGT,SAAS,uBAAuB,QAAkD;CAChF,MAAM,UAAU,oBAAoB;EAClC,YAAY;EACZ,kBAAkB,OAAO;EACzB,cAAc,OAAO;EACtB,CAAC;CACF,MAAM,QAAuC,EAAE;AAC/C,MAAK,MAAM,SAAS,SAAS;EAC3B,MAAM,aAAa,MAAM,UAAU,MAAM;AACzC,MAAI,CAAC,WACH;EAEF,MAAM,WAAW,yBAAyB,YAAY,QAAQ,MAAM;AACpE,MAAI,SACF,OAAM,KAAK,SAAS;;AAGxB,QAAO;;;AAIT,SAAgB,wBAAwB,QAAkD;AACxF,KAAI,CAAC,OAAO,QACV,QAAO,EAAE;CAGX,MAAM,aAAa,uBAAuB,OAAO;AACjD,KAAI,WAAW,SAAS,EACtB,QAAO;CAGT,MAAM,QAAQ,wBAAwB,OAAO,UAAU,OAAO,UAAU,OAAO;CAC/E,MAAM,QAAuC,EAAE;AAC/C,MAAK,MAAM,cAAc,OAAO;EAC9B,MAAM,WAAW,yBAAyB,YAAY,OAAO;AAC7D,MAAI,SACF,OAAM,KAAK,SAAS;;AAGxB,QAAO"}