@ck123pm/paseo-server 0.1.80

package/dist/server/server/speech/providers/local/runtime.js

@@ -0,0 +1,278 @@
+import { PocketTtsOnnxTTS } from "./pocket/pocket-tts-onnx.js";
+import { getLocalSpeechModelDir, DEFAULT_LOCAL_STT_MODEL, DEFAULT_LOCAL_TTS_MODEL, LocalSttModelIdSchema, LocalTtsModelIdSchema, } from "./models.js";
+import { SherpaOfflineRecognizerEngine } from "./sherpa/sherpa-offline-recognizer.js";
+import { SherpaOnlineRecognizerEngine } from "./sherpa/sherpa-online-recognizer.js";
+import { SherpaOnnxParakeetSTT } from "./sherpa/sherpa-parakeet-stt.js";
+import { SherpaParakeetRealtimeTranscriptionSession } from "./sherpa/sherpa-parakeet-realtime-session.js";
+import { SherpaRealtimeTranscriptionSession } from "./sherpa/sherpa-realtime-session.js";
+import { SherpaOnnxSTT } from "./sherpa/sherpa-stt.js";
+import { SherpaOnnxTTS } from "./sherpa/sherpa-tts.js";
+import { ensureSileroVadModel, SherpaSileroTurnDetectionProvider, } from "./sherpa/silero-vad-provider.js";
+function buildModelDownloadHint(modelId) {
+    return `Use 'paseo speech download --model ${modelId}' to download this model.`;
+}
+function resolveConfiguredLocalModels(speechConfig) {
+    return {
+        dictationLocalSttModel: LocalSttModelIdSchema.parse(speechConfig?.local?.models.dictationStt ?? DEFAULT_LOCAL_STT_MODEL),
+        voiceLocalSttModel: LocalSttModelIdSchema.parse(speechConfig?.local?.models.voiceStt ?? DEFAULT_LOCAL_STT_MODEL),
+        voiceLocalTtsModel: LocalTtsModelIdSchema.parse(speechConfig?.local?.models.voiceTts ?? DEFAULT_LOCAL_TTS_MODEL),
+    };
+}
+export function getLocalSpeechAvailability(speechConfig) {
+    const localConfig = speechConfig?.local ?? null;
+    return {
+        configured: Boolean(localConfig),
+        modelsDir: localConfig?.modelsDir ?? null,
+    };
+}
+function computeRequiredLocalModelIds(params) {
+    const ids = new Set();
+    if (params.providers.dictationStt.enabled !== false &&
+        params.providers.dictationStt.provider === "local") {
+        ids.add(params.models.dictationLocalSttModel);
+    }
+    if (params.providers.voiceStt.enabled !== false &&
+        params.providers.voiceStt.provider === "local") {
+        ids.add(params.models.voiceLocalSttModel);
+    }
+    if (params.providers.voiceTts.enabled !== false &&
+        params.providers.voiceTts.provider === "local") {
+        ids.add(params.models.voiceLocalTtsModel);
+    }
+    return Array.from(ids);
+}
+async function createLocalSttEngine(params) {
+    const { modelId, modelsDir, logger } = params;
+    if (modelId === "parakeet-tdt-0.6b-v3-int8" || modelId === "parakeet-tdt-0.6b-v2-int8") {
+        const modelDir = getLocalSpeechModelDir(modelsDir, modelId);
+        return {
+            kind: "offline",
+            engine: new SherpaOfflineRecognizerEngine({
+                model: {
+                    kind: "nemo_transducer",
+                    encoder: `${modelDir}/encoder.int8.onnx`,
+                    decoder: `${modelDir}/decoder.int8.onnx`,
+                    joiner: `${modelDir}/joiner.int8.onnx`,
+                    tokens: `${modelDir}/tokens.txt`,
+                },
+                numThreads: 2,
+                debug: 0,
+            }, logger),
+        };
+    }
+    if (modelId === "paraformer-bilingual-zh-en") {
+        const modelDir = getLocalSpeechModelDir(modelsDir, modelId);
+        return {
+            kind: "online",
+            engine: new SherpaOnlineRecognizerEngine({
+                model: {
+                    kind: "paraformer",
+                    encoder: `${modelDir}/encoder.int8.onnx`,
+                    decoder: `${modelDir}/decoder.int8.onnx`,
+                    tokens: `${modelDir}/tokens.txt`,
+                },
+                numThreads: 1,
+                debug: 0,
+            }, logger),
+        };
+    }
+    if (modelId === "zipformer-bilingual-zh-en-2023-02-20") {
+        const modelDir = getLocalSpeechModelDir(modelsDir, modelId);
+        return {
+            kind: "online",
+            engine: new SherpaOnlineRecognizerEngine({
+                model: {
+                    kind: "transducer",
+                    encoder: `${modelDir}/encoder-epoch-99-avg-1.onnx`,
+                    decoder: `${modelDir}/decoder-epoch-99-avg-1.onnx`,
+                    joiner: `${modelDir}/joiner-epoch-99-avg-1.onnx`,
+                    tokens: `${modelDir}/tokens.txt`,
+                    modelType: "zipformer",
+                },
+                numThreads: 1,
+                debug: 0,
+            }, logger),
+        };
+    }
+    throw new Error(`Unsupported local STT model '${modelId}'`);
+}
+function isLocalProviderEnabled(provider) {
+    return provider.enabled !== false && provider.provider === "local";
+}
+async function initializeLocalTurnDetection(localConfig, logger) {
+    let vadModelPath;
+    if (localConfig) {
+        try {
+            vadModelPath = await ensureSileroVadModel(localConfig.modelsDir, logger);
+        }
+        catch (err) {
+            logger.warn({ err }, "Failed to provision Silero VAD model, falling back to bundled");
+        }
+    }
+    return new SherpaSileroTurnDetectionProvider({ modelPath: vadModelPath }, logger);
+}
+async function initializeLocalVoiceStt(params) {
+    const { localConfig, modelId, logger, getLocalSttEngine } = params;
+    if (!localConfig) {
+        logger.warn({ configured: false }, "Local STT selected for voice but local provider config is missing; STT will be unavailable");
+        return null;
+    }
+    const voiceEngine = await getLocalSttEngine(modelId);
+    if (voiceEngine?.kind === "offline") {
+        return new SherpaOnnxParakeetSTT({ engine: voiceEngine.engine }, logger);
+    }
+    if (voiceEngine?.kind === "online") {
+        return new SherpaOnnxSTT({ engine: voiceEngine.engine }, logger);
+    }
+    return null;
+}
+async function initializeLocalDictationStt(params) {
+    const { localConfig, modelId, logger, getLocalSttEngine } = params;
+    if (!localConfig) {
+        logger.warn({ configured: false }, "Local STT selected for dictation but local provider config is missing; dictation STT will be unavailable");
+        return null;
+    }
+    const dictationEngine = await getLocalSttEngine(modelId);
+    if (dictationEngine?.kind === "offline") {
+        return {
+            id: "local",
+            createSession: () => new SherpaParakeetRealtimeTranscriptionSession({ engine: dictationEngine.engine }),
+        };
+    }
+    if (dictationEngine?.kind === "online") {
+        return {
+            id: "local",
+            createSession: () => new SherpaRealtimeTranscriptionSession({ engine: dictationEngine.engine }),
+        };
+    }
+    return null;
+}
+async function initializeLocalVoiceTts(params) {
+    const { localConfig, speechConfig, localModels, logger } = params;
+    if (!localConfig) {
+        logger.warn({ configured: false }, "Local TTS selected for voice but local provider config is missing; TTS will be unavailable");
+        return null;
+    }
+    try {
+        const modelDir = getLocalSpeechModelDir(localConfig.modelsDir, localModels.voiceLocalTtsModel);
+        if (localModels.voiceLocalTtsModel === "pocket-tts-onnx-int8") {
+            return await PocketTtsOnnxTTS.create({
+                modelDir,
+                precision: "int8",
+                targetChunkMs: 50,
+            }, logger);
+        }
+        return new SherpaOnnxTTS({
+            preset: localModels.voiceLocalTtsModel,
+            modelDir,
+            speakerId: speechConfig?.local?.models.voiceTtsSpeakerId,
+            speed: speechConfig?.local?.models.voiceTtsSpeed,
+        }, logger);
+    }
+    catch (err) {
+        logger.warn({
+            err,
+            modelsDir: localConfig.modelsDir,
+            modelId: localModels.voiceLocalTtsModel,
+            hint: buildModelDownloadHint(localModels.voiceLocalTtsModel),
+        }, "Local TTS engine unavailable");
+        return null;
+    }
+}
+export async function initializeLocalSpeechServices(params) {
+    const { providers, logger, speechConfig } = params;
+    const localConfig = speechConfig?.local ?? null;
+    const localModels = resolveConfiguredLocalModels(speechConfig);
+    let sttService = null;
+    let ttsService = null;
+    let dictationSttService = null;
+    let turnDetectionService = null;
+    let localVoiceTtsProvider = null;
+    const requiredLocalModelIds = computeRequiredLocalModelIds({
+        providers,
+        models: localModels,
+    });
+    const localSttEngines = new Map();
+    const getLocalSttEngine = async (modelId) => {
+        const existing = localSttEngines.get(modelId);
+        if (existing) {
+            return existing;
+        }
+        if (!localConfig) {
+            return null;
+        }
+        try {
+            const created = await createLocalSttEngine({
+                modelId,
+                modelsDir: localConfig.modelsDir,
+                logger,
+            });
+            localSttEngines.set(modelId, created);
+            return created;
+        }
+        catch (err) {
+            logger.warn({
+                err,
+                modelsDir: localConfig.modelsDir,
+                modelId,
+                hint: buildModelDownloadHint(modelId),
+            }, "Local STT engine unavailable");
+            return null;
+        }
+    };
+    if (isLocalProviderEnabled(providers.voiceTurnDetection)) {
+        turnDetectionService = await initializeLocalTurnDetection(localConfig, logger);
+    }
+    if (isLocalProviderEnabled(providers.voiceStt)) {
+        sttService = await initializeLocalVoiceStt({
+            localConfig,
+            modelId: localModels.voiceLocalSttModel,
+            logger,
+            getLocalSttEngine,
+        });
+    }
+    if (isLocalProviderEnabled(providers.dictationStt)) {
+        dictationSttService = await initializeLocalDictationStt({
+            localConfig,
+            modelId: localModels.dictationLocalSttModel,
+            logger,
+            getLocalSttEngine,
+        });
+    }
+    if (isLocalProviderEnabled(providers.voiceTts)) {
+        localVoiceTtsProvider = await initializeLocalVoiceTts({
+            localConfig,
+            speechConfig,
+            localModels,
+            logger,
+        });
+        if (localVoiceTtsProvider) {
+            ttsService = localVoiceTtsProvider;
+        }
+    }
+    const cleanup = () => {
+        const maybeFreeable = localVoiceTtsProvider;
+        if (typeof maybeFreeable?.free === "function") {
+            maybeFreeable.free();
+        }
+        for (const engine of localSttEngines.values()) {
+            engine.engine.free();
+        }
+    };
+    return {
+        turnDetectionService,
+        sttService,
+        ttsService,
+        dictationSttService,
+        localVoiceTtsProvider,
+        localModelConfig: localConfig
+            ? {
+                modelsDir: localConfig.modelsDir,
+                defaultModelIds: requiredLocalModelIds,
+            }
+            : null,
+        availability: getLocalSpeechAvailability(speechConfig),
+        cleanup,
+    };
+}
+//# sourceMappingURL=runtime.js.map

package/dist/server/server/speech/providers/local/runtime.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"runtime.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/local/runtime.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,gBAAgB,EAAE,MAAM,6BAA6B,CAAC;AAC/D,OAAO,EACL,sBAAsB,EACtB,uBAAuB,EACvB,uBAAuB,EACvB,qBAAqB,EACrB,qBAAqB,GAItB,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,6BAA6B,EAAE,MAAM,uCAAuC,CAAC;AACtF,OAAO,EAAE,4BAA4B,EAAE,MAAM,sCAAsC,CAAC;AACpF,OAAO,EAAE,qBAAqB,EAAE,MAAM,iCAAiC,CAAC;AACxE,OAAO,EAAE,0CAA0C,EAAE,MAAM,8CAA8C,CAAC;AAC1G,OAAO,EAAE,kCAAkC,EAAE,MAAM,qCAAqC,CAAC;AACzF,OAAO,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AACvD,OAAO,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AACvD,OAAO,EACL,oBAAoB,EACpB,iCAAiC,GAClC,MAAM,iCAAiC,CAAC;AA+BzC,SAAS,sBAAsB,CAAC,OAA2B;IACzD,OAAO,sCAAsC,OAAO,2BAA2B,CAAC;AAClF,CAAC;AAED,SAAS,4BAA4B,CAAC,YAAsC;IAC1E,OAAO;QACL,sBAAsB,EAAE,qBAAqB,CAAC,KAAK,CACjD,YAAY,EAAE,KAAK,EAAE,MAAM,CAAC,YAAY,IAAI,uBAAuB,CACpE;QACD,kBAAkB,EAAE,qBAAqB,CAAC,KAAK,CAC7C,YAAY,EAAE,KAAK,EAAE,MAAM,CAAC,QAAQ,IAAI,uBAAuB,CAChE;QACD,kBAAkB,EAAE,qBAAqB,CAAC,KAAK,CAC7C,YAAY,EAAE,KAAK,EAAE,MAAM,CAAC,QAAQ,IAAI,uBAAuB,CAChE;KACF,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,0BAA0B,CACxC,YAAsC;IAEtC,MAAM,WAAW,GAAG,YAAY,EAAE,KAAK,IAAI,IAAI,CAAC;IAChD,OAAO;QACL,UAAU,EAAE,OAAO,CAAC,WAAW,CAAC;QAChC,SAAS,EAAE,WAAW,EAAE,SAAS,IAAI,IAAI;KAC1C,CAAC;AACJ,CAAC;AAED,SAAS,4BAA4B,CAAC,MAGrC;IACC,MAAM,GAAG,GAAG,IAAI,GAAG,EAAsB,CAAC;IAC1C,IACE,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,OAAO,KAAK,KAAK;QAC/C,MAAM,CAAC,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,OAAO,EAClD,CAAC;QACD,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC,sBAAsB,CAAC,CAAC;IAChD,CAAC;IACD,IACE,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;QAC3C,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,OAAO,EAC9C,CAAC;QACD,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC,kBAAkB,CAAC,CAAC;IAC5C,CAAC;IACD,IACE,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;QAC3C,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,OAAO,EAC9C,CAAC;QACD,GAAG,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC,kBAAkB,CAAC,CAAC;IAC5C,CAAC;IACD,OAAO,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;AACzB,CAAC;AAED,KAAK,UAAU,oBAAoB,CAAC,MAInC;IACC,MAAM,EAAE,OAAO,EAAE,SAAS,EAAE,MAAM,EAAE,GAAG,MAAM,CAAC;IAE9C,IAAI,OAAO,KAAK,2BAA2B,IAAI,OAAO,KAAK,2BAA2B,EAAE,CAAC;QACvF,MAAM,QAAQ,GAAG,sBAAsB,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;QAC5D,OAAO;YACL,IAAI,EAAE,SAAS;YACf,MAAM,EAAE,IAAI,6BAA6B,CACvC;gBACE,KAAK,EAAE;oBACL,IAAI,EAAE,iBAAiB;oBACvB,OAAO,EAAE,GAAG,QAAQ,oBAAoB;oBACxC,OAAO,EAAE,GAAG,QAAQ,oBAAoB;oBACxC,MAAM,EAAE,GAAG,QAAQ,mBAAmB;oBACtC,MAAM,EAAE,GAAG,QAAQ,aAAa;iBACjC;gBACD,UAAU,EAAE,CAAC;gBACb,KAAK,EAAE,CAAC;aACT,EACD,MAAM,CACP;SACF,CAAC;IACJ,CAAC;IAED,IAAI,OAAO,KAAK,4BAA4B,EAAE,CAAC;QAC7C,MAAM,QAAQ,GAAG,sBAAsB,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;QAC5D,OAAO;YACL,IAAI,EAAE,QAAQ;YACd,MAAM,EAAE,IAAI,4BAA4B,CACtC;gBACE,KAAK,EAAE;oBACL,IAAI,EAAE,YAAY;oBAClB,OAAO,EAAE,GAAG,QAAQ,oBAAoB;oBACxC,OAAO,EAAE,GAAG,QAAQ,oBAAoB;oBACxC,MAAM,EAAE,GAAG,QAAQ,aAAa;iBACjC;gBACD,UAAU,EAAE,CAAC;gBACb,KAAK,EAAE,CAAC;aACT,EACD,MAAM,CACP;SACF,CAAC;IACJ,CAAC;IAED,IAAI,OAAO,KAAK,sCAAsC,EAAE,CAAC;QACvD,MAAM,QAAQ,GAAG,sBAAsB,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;QAC5D,OAAO;YACL,IAAI,EAAE,QAAQ;YACd,MAAM,EAAE,IAAI,4BAA4B,CACtC;gBACE,KAAK,EAAE;oBACL,IAAI,EAAE,YAAY;oBAClB,OAAO,EAAE,GAAG,QAAQ,8BAA8B;oBAClD,OAAO,EAAE,GAAG,QAAQ,8BAA8B;oBAClD,MAAM,EAAE,GAAG,QAAQ,6BAA6B;oBAChD,MAAM,EAAE,GAAG,QAAQ,aAAa;oBAChC,SAAS,EAAE,WAAW;iBACvB;gBACD,UAAU,EAAE,CAAC;gBACb,KAAK,EAAE,CAAC;aACT,EACD,MAAM,CACP;SACF,CAAC;IACJ,CAAC;IAED,MAAM,IAAI,KAAK,CAAC,gCAAgC,OAAO,GAAG,CAAC,CAAC;AAC9D,CAAC;AAID,SAAS,sBAAsB,CAAC,QAAiD;IAC/E,OAAO,QAAQ,CAAC,OAAO,KAAK,KAAK,IAAI,QAAQ,CAAC,QAAQ,KAAK,OAAO,CAAC;AACrE,CAAC;AAED,KAAK,UAAU,4BAA4B,CACzC,WAA+B,EAC/B,MAAc;IAEd,IAAI,YAAgC,CAAC;IACrC,IAAI,WAAW,EAAE,CAAC;QAChB,IAAI,CAAC;YACH,YAAY,GAAG,MAAM,oBAAoB,CAAC,WAAW,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;QAC3E,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,CAAC,IAAI,CAAC,EAAE,GAAG,EAAE,EAAE,+DAA+D,CAAC,CAAC;QACxF,CAAC;IACH,CAAC;IACD,OAAO,IAAI,iCAAiC,CAAC,EAAE,SAAS,EAAE,YAAY,EAAE,EAAE,MAAM,CAAC,CAAC;AACpF,CAAC;AAED,KAAK,UAAU,uBAAuB,CAAC,MAKtC;IACC,MAAM,EAAE,WAAW,EAAE,OAAO,EAAE,MAAM,EAAE,iBAAiB,EAAE,GAAG,MAAM,CAAC;IACnE,IAAI,CAAC,WAAW,EAAE,CAAC;QACjB,MAAM,CAAC,IAAI,CACT,EAAE,UAAU,EAAE,KAAK,EAAE,EACrB,4FAA4F,CAC7F,CAAC;QACF,OAAO,IAAI,CAAC;IACd,CAAC;IACD,MAAM,WAAW,GAAG,MAAM,iBAAiB,CAAC,OAAO,CAAC,CAAC;IACrD,IAAI,WAAW,EAAE,IAAI,KAAK,SAAS,EAAE,CAAC;QACpC,OAAO,IAAI,qBAAqB,CAAC,EAAE,MAAM,EAAE,WAAW,CAAC,MAAM,EAAE,EAAE,MAAM,CAAC,CAAC;IAC3E,CAAC;IACD,IAAI,WAAW,EAAE,IAAI,KAAK,QAAQ,EAAE,CAAC;QACnC,OAAO,IAAI,aAAa,CAAC,EAAE,MAAM,EAAE,WAAW,CAAC,MAAM,EAAE,EAAE,MAAM,CAAC,CAAC;IACnE,CAAC;IACD,OAAO,IAAI,CAAC;AACd,CAAC;AAED,KAAK,UAAU,2BAA2B,CAAC,MAK1C;IACC,MAAM,EAAE,WAAW,EAAE,OAAO,EAAE,MAAM,EAAE,iBAAiB,EAAE,GAAG,MAAM,CAAC;IACnE,IAAI,CAAC,WAAW,EAAE,CAAC;QACjB,MAAM,CAAC,IAAI,CACT,EAAE,UAAU,EAAE,KAAK,EAAE,EACrB,0GAA0G,CAC3G,CAAC;QACF,OAAO,IAAI,CAAC;IACd,CAAC;IACD,MAAM,eAAe,GAAG,MAAM,iBAAiB,CAAC,OAAO,CAAC,CAAC;IACzD,IAAI,eAAe,EAAE,IAAI,KAAK,SAAS,EAAE,CAAC;QACxC,OAAO;YACL,EAAE,EAAE,OAAO;YACX,aAAa,EAAE,GAAG,EAAE,CAClB,IAAI,0CAA0C,CAAC,EAAE,MAAM,EAAE,eAAe,CAAC,MAAM,EAAE,CAAC;SACrF,CAAC;IACJ,CAAC;IACD,IAAI,eAAe,EAAE,IAAI,KAAK,QAAQ,EAAE,CAAC;QACvC,OAAO;YACL,EAAE,EAAE,OAAO;YACX,aAAa,EAAE,GAAG,EAAE,CAClB,IAAI,kCAAkC,CAAC,EAAE,MAAM,EAAE,eAAe,CAAC,MAAM,EAAE,CAAC;SAC7E,CAAC;IACJ,CAAC;IACD,OAAO,IAAI,CAAC;AACd,CAAC;AAED,KAAK,UAAU,uBAAuB,CAAC,MAKtC;IACC,MAAM,EAAE,WAAW,EAAE,YAAY,EAAE,WAAW,EAAE,MAAM,EAAE,GAAG,MAAM,CAAC;IAClE,IAAI,CAAC,WAAW,EAAE,CAAC;QACjB,MAAM,CAAC,IAAI,CACT,EAAE,UAAU,EAAE,KAAK,EAAE,EACrB,4FAA4F,CAC7F,CAAC;QACF,OAAO,IAAI,CAAC;IACd,CAAC;IACD,IAAI,CAAC;QACH,MAAM,QAAQ,GAAG,sBAAsB,CAAC,WAAW,CAAC,SAAS,EAAE,WAAW,CAAC,kBAAkB,CAAC,CAAC;QAC/F,IAAI,WAAW,CAAC,kBAAkB,KAAK,sBAAsB,EAAE,CAAC;YAC9D,OAAO,MAAM,gBAAgB,CAAC,MAAM,CAClC;gBACE,QAAQ;gBACR,SAAS,EAAE,MAAM;gBACjB,aAAa,EAAE,EAAE;aAClB,EACD,MAAM,CACP,CAAC;QACJ,CAAC;QACD,OAAO,IAAI,aAAa,CACtB;YACE,MAAM,EAAE,WAAW,CAAC,kBAAkB;YACtC,QAAQ;YACR,SAAS,EAAE,YAAY,EAAE,KAAK,EAAE,MAAM,CAAC,iBAAiB;YACxD,KAAK,EAAE,YAAY,EAAE,KAAK,EAAE,MAAM,CAAC,aAAa;SACjD,EACD,MAAM,CACP,CAAC;IACJ,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACb,MAAM,CAAC,IAAI,CACT;YACE,GAAG;YACH,SAAS,EAAE,WAAW,CAAC,SAAS;YAChC,OAAO,EAAE,WAAW,CAAC,kBAAkB;YACvC,IAAI,EAAE,sBAAsB,CAAC,WAAW,CAAC,kBAAkB,CAAC;SAC7D,EACD,8BAA8B,CAC/B,CAAC;QACF,OAAO,IAAI,CAAC;IACd,CAAC;AACH,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,6BAA6B,CAAC,MAInD;IACC,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,CAAC;IACnD,MAAM,WAAW,GAAG,YAAY,EAAE,KAAK,IAAI,IAAI,CAAC;IAChD,MAAM,WAAW,GAAG,4BAA4B,CAAC,YAAY,CAAC,CAAC;IAE/D,IAAI,UAAU,GAAgC,IAAI,CAAC;IACnD,IAAI,UAAU,GAAgC,IAAI,CAAC;IACnD,IAAI,mBAAmB,GAAgC,IAAI,CAAC;IAC5D,IAAI,oBAAoB,GAAiC,IAAI,CAAC;IAC9D,IAAI,qBAAqB,GAAgC,IAAI,CAAC;IAE9D,MAAM,qBAAqB,GAAG,4BAA4B,CAAC;QACzD,SAAS;QACT,MAAM,EAAE,WAAW;KACpB,CAAC,CAAC;IAEH,MAAM,eAAe,GAAG,IAAI,GAAG,EAAmC,CAAC;IAEnE,MAAM,iBAAiB,GAAG,KAAK,EAAE,OAAwB,EAAkC,EAAE;QAC3F,MAAM,QAAQ,GAAG,eAAe,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC;QAC9C,IAAI,QAAQ,EAAE,CAAC;YACb,OAAO,QAAQ,CAAC;QAClB,CAAC;QACD,IAAI,CAAC,WAAW,EAAE,CAAC;YACjB,OAAO,IAAI,CAAC;QACd,CAAC;QACD,IAAI,CAAC;YACH,MAAM,OAAO,GAAG,MAAM,oBAAoB,CAAC;gBACzC,OAAO;gBACP,SAAS,EAAE,WAAW,CAAC,SAAS;gBAChC,MAAM;aACP,CAAC,CAAC;YACH,eAAe,CAAC,GAAG,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;YACtC,OAAO,OAAO,CAAC;QACjB,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,CAAC,IAAI,CACT;gBACE,GAAG;gBACH,SAAS,EAAE,WAAW,CAAC,SAAS;gBAChC,OAAO;gBACP,IAAI,EAAE,sBAAsB,CAAC,OAAO,CAAC;aACtC,EACD,8BAA8B,CAC/B,CAAC;YACF,OAAO,IAAI,CAAC;QACd,CAAC;IACH,CAAC,CAAC;IAEF,IAAI,sBAAsB,CAAC,SAAS,CAAC,kBAAkB,CAAC,EAAE,CAAC;QACzD,oBAAoB,GAAG,MAAM,4BAA4B,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC;IACjF,CAAC;IAED,IAAI,sBAAsB,CAAC,SAAS,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC/C,UAAU,GAAG,MAAM,uBAAuB,CAAC;YACzC,WAAW;YACX,OAAO,EAAE,WAAW,CAAC,kBAAkB;YACvC,MAAM;YACN,iBAAiB;SAClB,CAAC,CAAC;IACL,CAAC;IAED,IAAI,sBAAsB,CAAC,SAAS,CAAC,YAAY,CAAC,EAAE,CAAC;QACnD,mBAAmB,GAAG,MAAM,2BAA2B,CAAC;YACtD,WAAW;YACX,OAAO,EAAE,WAAW,CAAC,sBAAsB;YAC3C,MAAM;YACN,iBAAiB;SAClB,CAAC,CAAC;IACL,CAAC;IAED,IAAI,sBAAsB,CAAC,SAAS,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC/C,qBAAqB,GAAG,MAAM,uBAAuB,CAAC;YACpD,WAAW;YACX,YAAY;YACZ,WAAW;YACX,MAAM;SACP,CAAC,CAAC;QACH,IAAI,qBAAqB,EAAE,CAAC;YAC1B,UAAU,GAAG,qBAAqB,CAAC;QACrC,CAAC;IACH,CAAC;IAED,MAAM,OAAO,GAAG,GAAG,EAAE;QACnB,MAAM,aAAa,GAAG,qBAAgE,CAAC;QACvF,IAAI,OAAO,aAAa,EAAE,IAAI,KAAK,UAAU,EAAE,CAAC;YAC9C,aAAa,CAAC,IAAI,EAAE,CAAC;QACvB,CAAC;QACD,KAAK,MAAM,MAAM,IAAI,eAAe,CAAC,MAAM,EAAE,EAAE,CAAC;YAC9C,MAAM,CAAC,MAAM,CAAC,IAAI,EAAE,CAAC;QACvB,CAAC;IACH,CAAC,CAAC;IAEF,OAAO;QACL,oBAAoB;QACpB,UAAU;QACV,UAAU;QACV,mBAAmB;QACnB,qBAAqB;QACrB,gBAAgB,EAAE,WAAW;YAC3B,CAAC,CAAC;gBACE,SAAS,EAAE,WAAW,CAAC,SAAS;gBAChC,eAAe,EAAE,qBAAqB;aACvC;YACH,CAAC,CAAC,IAAI;QACR,YAAY,EAAE,0BAA0B,CAAC,YAAY,CAAC;QACtD,OAAO;KACR,CAAC;AACJ,CAAC"}

package/dist/server/server/speech/providers/local/sherpa/model-catalog.d.ts

@@ -0,0 +1,117 @@
+import { z } from "zod";
+export type SherpaOnnxModelKind = "stt-online" | "stt-offline" | "tts";
+type DefaultModelRole = "stt" | "tts";
+interface SherpaOnnxCatalogEntry {
+    kind: SherpaOnnxModelKind;
+    archiveUrl?: string;
+    downloadFiles?: Array<{
+        url: string;
+        relPath: string;
+    }>;
+    extractedDir: string;
+    requiredFiles: string[];
+    description: string;
+    aliases?: readonly string[];
+    defaultFor?: DefaultModelRole;
+}
+export declare const SHERPA_ONNX_MODEL_CATALOG: {
+    readonly "zipformer-bilingual-zh-en-2023-02-20": {
+        readonly kind: "stt-online";
+        readonly archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2";
+        readonly extractedDir: "sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20";
+        readonly requiredFiles: ["encoder-epoch-99-avg-1.onnx", "decoder-epoch-99-avg-1.onnx", "joiner-epoch-99-avg-1.onnx", "tokens.txt"];
+        readonly description: "Streaming Zipformer transducer (fast, good accuracy).";
+        readonly aliases: readonly ["zipformer", "zipformer-bilingual"];
+    };
+    readonly "paraformer-bilingual-zh-en": {
+        readonly kind: "stt-online";
+        readonly archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-paraformer-bilingual-zh-en.tar.bz2";
+        readonly extractedDir: "sherpa-onnx-streaming-paraformer-bilingual-zh-en";
+        readonly requiredFiles: ["encoder.int8.onnx", "decoder.int8.onnx", "tokens.txt"];
+        readonly description: "Streaming Paraformer (often strong accuracy; heavier).";
+        readonly aliases: readonly ["paraformer"];
+    };
+    readonly "parakeet-tdt-0.6b-v2-int8": {
+        readonly kind: "stt-offline";
+        readonly archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2";
+        readonly extractedDir: "sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8";
+        readonly requiredFiles: ["encoder.int8.onnx", "decoder.int8.onnx", "joiner.int8.onnx", "tokens.txt"];
+        readonly description: "NVIDIA Parakeet TDT v2 (offline NeMo transducer, English).";
+        readonly aliases: readonly ["parakeet-v2", "parakeet-tdt-v2"];
+        readonly defaultFor: "stt";
+    };
+    readonly "parakeet-tdt-0.6b-v3-int8": {
+        readonly kind: "stt-offline";
+        readonly archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v3-int8.tar.bz2";
+        readonly extractedDir: "sherpa-onnx-nemo-parakeet-tdt-0.6b-v3-int8";
+        readonly requiredFiles: ["encoder.int8.onnx", "decoder.int8.onnx", "joiner.int8.onnx", "tokens.txt"];
+        readonly description: "NVIDIA Parakeet TDT v3 (offline NeMo transducer, multilingual).";
+        readonly aliases: readonly ["parakeet", "parakeet-v3", "parakeet-tdt"];
+    };
+    readonly "kitten-nano-en-v0_1-fp16": {
+        readonly kind: "tts";
+        readonly archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kitten-nano-en-v0_1-fp16.tar.bz2";
+        readonly extractedDir: "kitten-nano-en-v0_1-fp16";
+        readonly requiredFiles: ["model.fp16.onnx", "voices.bin", "tokens.txt", "espeak-ng-data"];
+        readonly description: "KittenTTS (small, fast English TTS).";
+        readonly aliases: readonly ["kitten"];
+    };
+    readonly "kokoro-en-v0_19": {
+        readonly kind: "tts";
+        readonly archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kokoro-en-v0_19.tar.bz2";
+        readonly extractedDir: "kokoro-en-v0_19";
+        readonly requiredFiles: ["model.onnx", "voices.bin", "tokens.txt", "espeak-ng-data"];
+        readonly description: "Kokoro TTS (higher quality; larger).";
+        readonly aliases: readonly ["kokoro"];
+        readonly defaultFor: "tts";
+    };
+    readonly "pocket-tts-onnx-int8": {
+        readonly kind: "tts";
+        readonly extractedDir: "pocket-tts-onnx-int8";
+        readonly downloadFiles: [{
+            readonly url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/mimi_encoder.onnx";
+            readonly relPath: "onnx/mimi_encoder.onnx";
+        }, {
+            readonly url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/text_conditioner.onnx";
+            readonly relPath: "onnx/text_conditioner.onnx";
+        }, {
+            readonly url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/flow_lm_main_int8.onnx";
+            readonly relPath: "onnx/flow_lm_main_int8.onnx";
+        }, {
+            readonly url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/flow_lm_flow_int8.onnx";
+            readonly relPath: "onnx/flow_lm_flow_int8.onnx";
+        }, {
+            readonly url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/mimi_decoder_int8.onnx";
+            readonly relPath: "onnx/mimi_decoder_int8.onnx";
+        }, {
+            readonly url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/tokenizer.model";
+            readonly relPath: "tokenizer.model";
+        }, {
+            readonly url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/reference_sample.wav";
+            readonly relPath: "reference_sample.wav";
+        }];
+        readonly requiredFiles: ["onnx/mimi_encoder.onnx", "onnx/text_conditioner.onnx", "onnx/flow_lm_main_int8.onnx", "onnx/flow_lm_flow_int8.onnx", "onnx/mimi_decoder_int8.onnx", "tokenizer.model", "reference_sample.wav"];
+        readonly description: "Pocket TTS ONNX (INT8) with streaming decode support (via onnxruntime).";
+        readonly aliases: readonly ["pocket", "pocket-tts"];
+    };
+};
+export type SherpaOnnxModelId = keyof typeof SHERPA_ONNX_MODEL_CATALOG;
+export type LocalSpeechModelId = SherpaOnnxModelId;
+type ModelIdByKind<K extends SherpaOnnxModelKind> = {
+    [Id in SherpaOnnxModelId]: (typeof SHERPA_ONNX_MODEL_CATALOG)[Id]["kind"] extends K ? Id : never;
+}[SherpaOnnxModelId];
+export type LocalSttModelId = ModelIdByKind<"stt-online"> | ModelIdByKind<"stt-offline">;
+export type LocalTtsModelId = ModelIdByKind<"tts">;
+export declare const LOCAL_STT_MODEL_IDS: LocalSttModelId[];
+export declare const LOCAL_TTS_MODEL_IDS: LocalTtsModelId[];
+export declare const DEFAULT_LOCAL_STT_MODEL: LocalSttModelId;
+export declare const DEFAULT_LOCAL_TTS_MODEL: LocalTtsModelId;
+export declare const LocalSttModelIdSchema: z.ZodType<LocalSttModelId, z.ZodTypeDef, string>;
+export declare const LocalTtsModelIdSchema: z.ZodType<LocalTtsModelId, z.ZodTypeDef, string>;
+export type SherpaOnnxModelSpec = SherpaOnnxCatalogEntry & {
+    id: SherpaOnnxModelId;
+};
+export declare function listSherpaOnnxModels(): SherpaOnnxModelSpec[];
+export declare function getSherpaOnnxModelSpec(id: SherpaOnnxModelId): SherpaOnnxModelSpec;
+export {};
+//# sourceMappingURL=model-catalog.d.ts.map

package/dist/server/server/speech/providers/local/sherpa/model-catalog.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"model-catalog.d.ts","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/model-catalog.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAExB,MAAM,MAAM,mBAAmB,GAAG,YAAY,GAAG,aAAa,GAAG,KAAK,CAAC;AAEvE,KAAK,gBAAgB,GAAG,KAAK,GAAG,KAAK,CAAC;AAEtC,UAAU,sBAAsB;IAC9B,IAAI,EAAE,mBAAmB,CAAC;IAC1B,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,aAAa,CAAC,EAAE,KAAK,CAAC;QAAE,GAAG,EAAE,MAAM,CAAC;QAAC,OAAO,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IACxD,YAAY,EAAE,MAAM,CAAC;IACrB,aAAa,EAAE,MAAM,EAAE,CAAC;IACxB,WAAW,EAAE,MAAM,CAAC;IACpB,OAAO,CAAC,EAAE,SAAS,MAAM,EAAE,CAAC;IAC5B,UAAU,CAAC,EAAE,gBAAgB,CAAC;CAC/B;AAED,eAAO,MAAM,yBAAyB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA2GqB,CAAC;AAE5D,MAAM,MAAM,iBAAiB,GAAG,MAAM,OAAO,yBAAyB,CAAC;AACvE,MAAM,MAAM,kBAAkB,GAAG,iBAAiB,CAAC;AAEnD,KAAK,aAAa,CAAC,CAAC,SAAS,mBAAmB,IAAI;KACjD,EAAE,IAAI,iBAAiB,GAAG,CAAC,OAAO,yBAAyB,CAAC,CAAC,EAAE,CAAC,CAAC,MAAM,CAAC,SAAS,CAAC,GAAG,EAAE,GAAG,KAAK;CACjG,CAAC,iBAAiB,CAAC,CAAC;AAErB,MAAM,MAAM,eAAe,GAAG,aAAa,CAAC,YAAY,CAAC,GAAG,aAAa,CAAC,aAAa,CAAC,CAAC;AACzF,MAAM,MAAM,eAAe,GAAG,aAAa,CAAC,KAAK,CAAC,CAAC;AAcnD,eAAO,MAAM,mBAAmB,EAAE,eAAe,EAA4C,CAAC;AAE9F,eAAO,MAAM,mBAAmB,EAAE,eAAe,EAA4C,CAAC;AAe9F,eAAO,MAAM,uBAAuB,iBAA+B,CAAC;AACpE,eAAO,MAAM,uBAAuB,iBAA+B,CAAC;AAmCpE,eAAO,MAAM,qBAAqB,kDAGhC,CAAC;AAEH,eAAO,MAAM,qBAAqB,kDAGhC,CAAC;AAEH,MAAM,MAAM,mBAAmB,GAAG,sBAAsB,GAAG;IACzD,EAAE,EAAE,iBAAiB,CAAC;CACvB,CAAC;AAEF,wBAAgB,oBAAoB,IAAI,mBAAmB,EAAE,CAE5D;AAED,wBAAgB,sBAAsB,CAAC,EAAE,EAAE,iBAAiB,GAAG,mBAAmB,CASjF"}

package/dist/server/server/speech/providers/local/sherpa/model-catalog.js

@@ -0,0 +1,169 @@
+import { z } from "zod";
+export const SHERPA_ONNX_MODEL_CATALOG = {
+    "zipformer-bilingual-zh-en-2023-02-20": {
+        kind: "stt-online",
+        archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2",
+        extractedDir: "sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20",
+        requiredFiles: [
+            "encoder-epoch-99-avg-1.onnx",
+            "decoder-epoch-99-avg-1.onnx",
+            "joiner-epoch-99-avg-1.onnx",
+            "tokens.txt",
+        ],
+        description: "Streaming Zipformer transducer (fast, good accuracy).",
+        aliases: ["zipformer", "zipformer-bilingual"],
+    },
+    "paraformer-bilingual-zh-en": {
+        kind: "stt-online",
+        archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-paraformer-bilingual-zh-en.tar.bz2",
+        extractedDir: "sherpa-onnx-streaming-paraformer-bilingual-zh-en",
+        requiredFiles: ["encoder.int8.onnx", "decoder.int8.onnx", "tokens.txt"],
+        description: "Streaming Paraformer (often strong accuracy; heavier).",
+        aliases: ["paraformer"],
+    },
+    "parakeet-tdt-0.6b-v2-int8": {
+        kind: "stt-offline",
+        archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2",
+        extractedDir: "sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8",
+        requiredFiles: ["encoder.int8.onnx", "decoder.int8.onnx", "joiner.int8.onnx", "tokens.txt"],
+        description: "NVIDIA Parakeet TDT v2 (offline NeMo transducer, English).",
+        aliases: ["parakeet-v2", "parakeet-tdt-v2"],
+        defaultFor: "stt",
+    },
+    "parakeet-tdt-0.6b-v3-int8": {
+        kind: "stt-offline",
+        archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v3-int8.tar.bz2",
+        extractedDir: "sherpa-onnx-nemo-parakeet-tdt-0.6b-v3-int8",
+        requiredFiles: ["encoder.int8.onnx", "decoder.int8.onnx", "joiner.int8.onnx", "tokens.txt"],
+        description: "NVIDIA Parakeet TDT v3 (offline NeMo transducer, multilingual).",
+        aliases: ["parakeet", "parakeet-v3", "parakeet-tdt"],
+    },
+    "kitten-nano-en-v0_1-fp16": {
+        kind: "tts",
+        archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kitten-nano-en-v0_1-fp16.tar.bz2",
+        extractedDir: "kitten-nano-en-v0_1-fp16",
+        requiredFiles: ["model.fp16.onnx", "voices.bin", "tokens.txt", "espeak-ng-data"],
+        description: "KittenTTS (small, fast English TTS).",
+        aliases: ["kitten"],
+    },
+    "kokoro-en-v0_19": {
+        kind: "tts",
+        archiveUrl: "https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kokoro-en-v0_19.tar.bz2",
+        extractedDir: "kokoro-en-v0_19",
+        requiredFiles: ["model.onnx", "voices.bin", "tokens.txt", "espeak-ng-data"],
+        description: "Kokoro TTS (higher quality; larger).",
+        aliases: ["kokoro"],
+        defaultFor: "tts",
+    },
+    "pocket-tts-onnx-int8": {
+        kind: "tts",
+        extractedDir: "pocket-tts-onnx-int8",
+        downloadFiles: [
+            {
+                url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/mimi_encoder.onnx",
+                relPath: "onnx/mimi_encoder.onnx",
+            },
+            {
+                url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/text_conditioner.onnx",
+                relPath: "onnx/text_conditioner.onnx",
+            },
+            {
+                url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/flow_lm_main_int8.onnx",
+                relPath: "onnx/flow_lm_main_int8.onnx",
+            },
+            {
+                url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/flow_lm_flow_int8.onnx",
+                relPath: "onnx/flow_lm_flow_int8.onnx",
+            },
+            {
+                url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/onnx/mimi_decoder_int8.onnx",
+                relPath: "onnx/mimi_decoder_int8.onnx",
+            },
+            {
+                url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/tokenizer.model",
+                relPath: "tokenizer.model",
+            },
+            {
+                url: "https://huggingface.co/KevinAHM/pocket-tts-onnx/resolve/main/reference_sample.wav",
+                relPath: "reference_sample.wav",
+            },
+        ],
+        requiredFiles: [
+            "onnx/mimi_encoder.onnx",
+            "onnx/text_conditioner.onnx",
+            "onnx/flow_lm_main_int8.onnx",
+            "onnx/flow_lm_flow_int8.onnx",
+            "onnx/mimi_decoder_int8.onnx",
+            "tokenizer.model",
+            "reference_sample.wav",
+        ],
+        description: "Pocket TTS ONNX (INT8) with streaming decode support (via onnxruntime).",
+        aliases: ["pocket", "pocket-tts"],
+    },
+};
+const ALL_MODEL_IDS = Object.keys(SHERPA_ONNX_MODEL_CATALOG).filter((k) => k in SHERPA_ONNX_MODEL_CATALOG);
+function isLocalSttModelId(id) {
+    return SHERPA_ONNX_MODEL_CATALOG[id].kind !== "tts";
+}
+function isLocalTtsModelId(id) {
+    return SHERPA_ONNX_MODEL_CATALOG[id].kind === "tts";
+}
+export const LOCAL_STT_MODEL_IDS = ALL_MODEL_IDS.filter(isLocalSttModelId);
+export const LOCAL_TTS_MODEL_IDS = ALL_MODEL_IDS.filter(isLocalTtsModelId);
+function resolveDefaultModelId(role) {
+    const match = ALL_MODEL_IDS.find((id) => {
+        const entry = SHERPA_ONNX_MODEL_CATALOG[id];
+        return entry.defaultFor === role;
+    });
+    if (!match) {
+        throw new Error(`No default model configured for role '${role}'`);
+    }
+    return match;
+}
+export const DEFAULT_LOCAL_STT_MODEL = resolveDefaultModelId("stt");
+export const DEFAULT_LOCAL_TTS_MODEL = resolveDefaultModelId("tts");
+function buildAliasMap(modelIds) {
+    const aliasMap = {};
+    for (const modelId of modelIds) {
+        const aliases = SHERPA_ONNX_MODEL_CATALOG[modelId].aliases ?? [];
+        for (const alias of aliases) {
+            aliasMap[alias.trim().toLowerCase()] = modelId;
+        }
+    }
+    return aliasMap;
+}
+function createAliasedModelIdSchema(params) {
+    const validIds = new Set(params.modelIds);
+    return z
+        .string()
+        .trim()
+        .toLowerCase()
+        .refine((value) => validIds.has(value) || Object.prototype.hasOwnProperty.call(params.aliases, value), {
+        message: "Invalid model id",
+    })
+        .transform((value) => params.aliases[value] ?? value);
+}
+const STT_MODEL_ALIASES = buildAliasMap(LOCAL_STT_MODEL_IDS);
+const TTS_MODEL_ALIASES = buildAliasMap(LOCAL_TTS_MODEL_IDS);
+export const LocalSttModelIdSchema = createAliasedModelIdSchema({
+    modelIds: LOCAL_STT_MODEL_IDS,
+    aliases: STT_MODEL_ALIASES,
+});
+export const LocalTtsModelIdSchema = createAliasedModelIdSchema({
+    modelIds: LOCAL_TTS_MODEL_IDS,
+    aliases: TTS_MODEL_ALIASES,
+});
+export function listSherpaOnnxModels() {
+    return ALL_MODEL_IDS.map((id) => Object.assign({ id }, SHERPA_ONNX_MODEL_CATALOG[id]));
+}
+export function getSherpaOnnxModelSpec(id) {
+    const spec = SHERPA_ONNX_MODEL_CATALOG[id];
+    if (!spec) {
+        throw new Error(`Unknown local speech model id: ${id}`);
+    }
+    return {
+        id,
+        ...spec,
+    };
+}
+//# sourceMappingURL=model-catalog.js.map

package/dist/server/server/speech/providers/local/sherpa/model-catalog.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"model-catalog.js","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/model-catalog.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAiBxB,MAAM,CAAC,MAAM,yBAAyB,GAAG;IACvC,sCAAsC,EAAE;QACtC,IAAI,EAAE,YAAY;QAClB,UAAU,EACR,uIAAuI;QACzI,YAAY,EAAE,4DAA4D;QAC1E,aAAa,EAAE;YACb,6BAA6B;YAC7B,6BAA6B;YAC7B,4BAA4B;YAC5B,YAAY;SACb;QACD,WAAW,EAAE,uDAAuD;QACpE,OAAO,EAAE,CAAC,WAAW,EAAE,qBAAqB,CAAC;KAC9C;IACD,4BAA4B,EAAE;QAC5B,IAAI,EAAE,YAAY;QAClB,UAAU,EACR,6HAA6H;QAC/H,YAAY,EAAE,kDAAkD;QAChE,aAAa,EAAE,CAAC,mBAAmB,EAAE,mBAAmB,EAAE,YAAY,CAAC;QACvE,WAAW,EAAE,wDAAwD;QACrE,OAAO,EAAE,CAAC,YAAY,CAAC;KACxB;IACD,2BAA2B,EAAE;QAC3B,IAAI,EAAE,aAAa;QACnB,UAAU,EACR,uHAAuH;QACzH,YAAY,EAAE,4CAA4C;QAC1D,aAAa,EAAE,CAAC,mBAAmB,EAAE,mBAAmB,EAAE,kBAAkB,EAAE,YAAY,CAAC;QAC3F,WAAW,EAAE,4DAA4D;QACzE,OAAO,EAAE,CAAC,aAAa,EAAE,iBAAiB,CAAC;QAC3C,UAAU,EAAE,KAAK;KAClB;IACD,2BAA2B,EAAE;QAC3B,IAAI,EAAE,aAAa;QACnB,UAAU,EACR,uHAAuH;QACzH,YAAY,EAAE,4CAA4C;QAC1D,aAAa,EAAE,CAAC,mBAAmB,EAAE,mBAAmB,EAAE,kBAAkB,EAAE,YAAY,CAAC;QAC3F,WAAW,EAAE,iEAAiE;QAC9E,OAAO,EAAE,CAAC,UAAU,EAAE,aAAa,EAAE,cAAc,CAAC;KACrD;IACD,0BAA0B,EAAE;QAC1B,IAAI,EAAE,KAAK;QACX,UAAU,EACR,qGAAqG;QACvG,YAAY,EAAE,0BAA0B;QACxC,aAAa,EAAE,CAAC,iBAAiB,EAAE,YAAY,EAAE,YAAY,EAAE,gBAAgB,CAAC;QAChF,WAAW,EAAE,sCAAsC;QACnD,OAAO,EAAE,CAAC,QAAQ,CAAC;KACpB;IACD,iBAAiB,EAAE;QACjB,IAAI,EAAE,KAAK;QACX,UAAU,EACR,4FAA4F;QAC9F,YAAY,EAAE,iBAAiB;QAC/B,aAAa,EAAE,CAAC,YAAY,EAAE,YAAY,EAAE,YAAY,EAAE,gBAAgB,CAAC;QAC3E,WAAW,EAAE,sCAAsC;QACnD,OAAO,EAAE,CAAC,QAAQ,CAAC;QACnB,UAAU,EAAE,KAAK;KAClB;IACD,sBAAsB,EAAE;QACtB,IAAI,EAAE,KAAK;QACX,YAAY,EAAE,sBAAsB;QACpC,aAAa,EAAE;YACb;gBACE,GAAG,EAAE,qFAAqF;gBAC1F,OAAO,EAAE,wBAAwB;aAClC;YACD;gBACE,GAAG,EAAE,yFAAyF;gBAC9F,OAAO,EAAE,4BAA4B;aACtC;YACD;gBACE,GAAG,EAAE,0FAA0F;gBAC/F,OAAO,EAAE,6BAA6B;aACvC;YACD;gBACE,GAAG,EAAE,0FAA0F;gBAC/F,OAAO,EAAE,6BAA6B;aACvC;YACD;gBACE,GAAG,EAAE,0FAA0F;gBAC/F,OAAO,EAAE,6BAA6B;aACvC;YACD;gBACE,GAAG,EAAE,8EAA8E;gBACnF,OAAO,EAAE,iBAAiB;aAC3B;YACD;gBACE,GAAG,EAAE,mFAAmF;gBACxF,OAAO,EAAE,sBAAsB;aAChC;SACF;QACD,aAAa,EAAE;YACb,wBAAwB;YACxB,4BAA4B;YAC5B,6BAA6B;YAC7B,6BAA6B;YAC7B,6BAA6B;YAC7B,iBAAiB;YACjB,sBAAsB;SACvB;QACD,WAAW,EAAE,yEAAyE;QACtF,OAAO,EAAE,CAAC,QAAQ,EAAE,YAAY,CAAC;KAClC;CACwD,CAAC;AAY5D,MAAM,aAAa,GAAwB,MAAM,CAAC,IAAI,CAAC,yBAAyB,CAAC,CAAC,MAAM,CACtF,CAAC,CAAC,EAA0B,EAAE,CAAC,CAAC,IAAI,yBAAyB,CAC9D,CAAC;AAEF,SAAS,iBAAiB,CAAC,EAAqB;IAC9C,OAAO,yBAAyB,CAAC,EAAE,CAAC,CAAC,IAAI,KAAK,KAAK,CAAC;AACtD,CAAC;AAED,SAAS,iBAAiB,CAAC,EAAqB;IAC9C,OAAO,yBAAyB,CAAC,EAAE,CAAC,CAAC,IAAI,KAAK,KAAK,CAAC;AACtD,CAAC;AAED,MAAM,CAAC,MAAM,mBAAmB,GAAsB,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,CAAC;AAE9F,MAAM,CAAC,MAAM,mBAAmB,GAAsB,aAAa,CAAC,MAAM,CAAC,iBAAiB,CAAC,CAAC;AAI9F,SAAS,qBAAqB,CAAC,IAAsB;IACnD,MAAM,KAAK,GAAG,aAAa,CAAC,IAAI,CAAC,CAAC,EAAE,EAAE,EAAE;QACtC,MAAM,KAAK,GAA2B,yBAAyB,CAAC,EAAE,CAAC,CAAC;QACpE,OAAO,KAAK,CAAC,UAAU,KAAK,IAAI,CAAC;IACnC,CAAC,CAAC,CAAC;IACH,IAAI,CAAC,KAAK,EAAE,CAAC;QACX,MAAM,IAAI,KAAK,CAAC,yCAAyC,IAAI,GAAG,CAAC,CAAC;IACpE,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,MAAM,CAAC,MAAM,uBAAuB,GAAG,qBAAqB,CAAC,KAAK,CAAC,CAAC;AACpE,MAAM,CAAC,MAAM,uBAAuB,GAAG,qBAAqB,CAAC,KAAK,CAAC,CAAC;AAEpE,SAAS,aAAa,CAA8B,QAAsB;IACxE,MAAM,QAAQ,GAAsB,EAAE,CAAC;IACvC,KAAK,MAAM,OAAO,IAAI,QAAQ,EAAE,CAAC;QAC/B,MAAM,OAAO,GAAG,yBAAyB,CAAC,OAAO,CAAC,CAAC,OAAO,IAAI,EAAE,CAAC;QACjE,KAAK,MAAM,KAAK,IAAI,OAAO,EAAE,CAAC;YAC5B,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC,GAAG,OAAO,CAAC;QACjD,CAAC;IACH,CAAC;IACD,OAAO,QAAQ,CAAC;AAClB,CAAC;AAED,SAAS,0BAA0B,CAAmB,MAGrD;IACC,MAAM,QAAQ,GAAG,IAAI,GAAG,CAAS,MAAM,CAAC,QAAQ,CAAC,CAAC;IAClD,OAAO,CAAC;SACL,MAAM,EAAE;SACR,IAAI,EAAE;SACN,WAAW,EAAE;SACb,MAAM,CACL,CAAC,KAAK,EAAc,EAAE,CACpB,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,MAAM,CAAC,SAAS,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,EAAE,KAAK,CAAC,EACpF;QACE,OAAO,EAAE,kBAAkB;KAC5B,CACF;SACA,SAAS,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,MAAM,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,KAAK,CAAC,CAAC;AAC1D,CAAC;AAED,MAAM,iBAAiB,GAAG,aAAa,CAAC,mBAAmB,CAAC,CAAC;AAC7D,MAAM,iBAAiB,GAAG,aAAa,CAAC,mBAAmB,CAAC,CAAC;AAE7D,MAAM,CAAC,MAAM,qBAAqB,GAAG,0BAA0B,CAAC;IAC9D,QAAQ,EAAE,mBAAmB;IAC7B,OAAO,EAAE,iBAAiB;CAC3B,CAAC,CAAC;AAEH,MAAM,CAAC,MAAM,qBAAqB,GAAG,0BAA0B,CAAC;IAC9D,QAAQ,EAAE,mBAAmB;IAC7B,OAAO,EAAE,iBAAiB;CAC3B,CAAC,CAAC;AAMH,MAAM,UAAU,oBAAoB;IAClC,OAAO,aAAa,CAAC,GAAG,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC,EAAE,EAAE,EAAE,EAAE,yBAAyB,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;AACzF,CAAC;AAED,MAAM,UAAU,sBAAsB,CAAC,EAAqB;IAC1D,MAAM,IAAI,GAAG,yBAAyB,CAAC,EAAE,CAAC,CAAC;IAC3C,IAAI,CAAC,IAAI,EAAE,CAAC;QACV,MAAM,IAAI,KAAK,CAAC,kCAAkC,EAAE,EAAE,CAAC,CAAC;IAC1D,CAAC;IACD,OAAO;QACL,EAAE;QACF,GAAG,IAAI;KACR,CAAC;AACJ,CAAC"}

package/dist/server/server/speech/providers/local/sherpa/model-downloader.d.ts

@@ -0,0 +1,15 @@
+import type pino from "pino";
+import { type SherpaOnnxModelId } from "./model-catalog.js";
+export interface EnsureSherpaOnnxModelOptions {
+    modelsDir: string;
+    modelId: SherpaOnnxModelId;
+    logger: pino.Logger;
+}
+export declare function getSherpaOnnxModelDir(modelsDir: string, modelId: SherpaOnnxModelId): string;
+export declare function ensureSherpaOnnxModel(options: EnsureSherpaOnnxModelOptions): Promise<string>;
+export declare function ensureSherpaOnnxModels(options: {
+    modelsDir: string;
+    modelIds: SherpaOnnxModelId[];
+    logger: pino.Logger;
+}): Promise<Record<SherpaOnnxModelId, string>>;
+//# sourceMappingURL=model-downloader.d.ts.map

package/dist/server/server/speech/providers/local/sherpa/model-downloader.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"model-downloader.d.ts","sourceRoot":"","sources":["../../../../../../../src/server/speech/providers/local/sherpa/model-downloader.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,IAAI,MAAM,MAAM,CAAC;AAE7B,OAAO,EAA0B,KAAK,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAGpF,MAAM,WAAW,4BAA4B;IAC3C,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,iBAAiB,CAAC;IAC3B,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC;CACrB;AAED,wBAAgB,qBAAqB,CAAC,SAAS,EAAE,MAAM,EAAE,OAAO,EAAE,iBAAiB,GAAG,MAAM,CAG3F;AA2ED,wBAAsB,qBAAqB,CACzC,OAAO,EAAE,4BAA4B,GACpC,OAAO,CAAC,MAAM,CAAC,CA2GjB;AAED,wBAAsB,sBAAsB,CAAC,OAAO,EAAE;IACpD,SAAS,EAAE,MAAM,CAAC;IAClB,QAAQ,EAAE,iBAAiB,EAAE,CAAC;IAC9B,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC;CACrB,GAAG,OAAO,CAAC,MAAM,CAAC,iBAAiB,EAAE,MAAM,CAAC,CAAC,CAc7C"}