@livekit/agents-plugin-azure 1.5.0

package/dist/stt.d.ts

@@ -0,0 +1,106 @@
+import { type APIConnectOptions, type AudioBuffer, stt } from '@livekit/agents';
+import * as speechsdk from 'microsoft-cognitiveservices-speech-sdk';
+export { speechsdk };
+/** @public */
+export interface STTOptions {
+    /** Azure Speech subscription key. Defaults to `AZURE_SPEECH_KEY`. */
+    speechKey?: string;
+    /** Azure Speech region. Defaults to `AZURE_SPEECH_REGION`. */
+    speechRegion?: string;
+    /** Azure Speech container host. Defaults to `AZURE_SPEECH_HOST`. */
+    speechHost?: string;
+    /** Ephemeral Microsoft Entra auth token. */
+    speechAuthToken?: string;
+    /** Azure Speech endpoint URL. */
+    speechEndpoint?: string;
+    sampleRate: number;
+    numChannels: number;
+    segmentationSilenceTimeoutMs?: number;
+    segmentationMaxTimeMs?: number;
+    segmentationStrategy?: string;
+    language: string[];
+    profanity?: speechsdk.ProfanityOption;
+    phraseList?: string[] | null;
+    explicitPunctuation: boolean;
+    trueTextPostProcessing: boolean;
+}
+/** @public */
+export type STTUpdateOptions = Partial<Pick<STTOptions, 'segmentationSilenceTimeoutMs' | 'segmentationMaxTimeMs' | 'segmentationStrategy'>> & {
+    language?: string | string[];
+};
+/** @public */
+export interface STTConstructorOptions extends STTUpdateOptions {
+    speechKey?: string;
+    speechRegion?: string;
+    speechHost?: string;
+    speechAuthToken?: string;
+    speechEndpoint?: string;
+    sampleRate?: number;
+    numChannels?: number;
+    profanity?: speechsdk.ProfanityOption;
+    phraseList?: string[] | null;
+    explicitPunctuation?: boolean;
+    trueTextPostProcessing?: boolean;
+}
+/** @internal */
+export interface _CanceledEvent {
+    errorDetails?: string;
+    reason: speechsdk.CancellationReason;
+    errorCode?: speechsdk.CancellationErrorCode;
+}
+/** @internal */
+export interface _WaitableEvent {
+    isSet: boolean;
+    wait(): Promise<void>;
+    set(): void;
+    clear(): void;
+}
+/** @public */
+export declare class STT extends stt.STT {
+    #private;
+    label: string;
+    get model(): string;
+    get provider(): string;
+    constructor(opts?: STTConstructorOptions);
+    _recognize(_frame: AudioBuffer): Promise<stt.SpeechEvent>;
+    updateOptions(opts: STTUpdateOptions): void;
+    stream(options?: {
+        language?: string;
+        connOptions?: APIConnectOptions;
+    }): SpeechStream;
+}
+/** @public */
+export declare class SpeechStream extends stt.SpeechStream {
+    #private;
+    /** @internal */
+    _opts: STTOptions;
+    /** @internal */
+    _speaking: boolean;
+    /** @internal */
+    _sessionStoppedEvent: _WaitableEvent;
+    /** @internal */
+    _sessionStartedEvent: _WaitableEvent;
+    /** @internal */
+    _reconnectEvent: _WaitableEvent;
+    /** @internal */
+    _cancellationError: _CanceledEvent | null;
+    label: string;
+    constructor(stt: STT, opts: STTOptions, connOptions: APIConnectOptions);
+    updateOptions(opts: STTUpdateOptions): void;
+    protected run(): Promise<void>;
+    /** @internal */
+    _onRecognized(evt: speechsdk.SpeechRecognitionEventArgs): void;
+    /** @internal */
+    _onRecognizing(evt: speechsdk.SpeechRecognitionEventArgs): void;
+    /** @internal */
+    _onSpeechStart(_evt: speechsdk.RecognitionEventArgs): void;
+    /** @internal */
+    _onSpeechEnd(_evt: speechsdk.RecognitionEventArgs): void;
+    /** @internal */
+    _onSessionStarted(_evt: speechsdk.SessionEventArgs): void;
+    /** @internal */
+    _onSessionStopped(_evt: speechsdk.SessionEventArgs): void;
+    /** @internal */
+    _onCanceled(evt: _CanceledEvent): void;
+}
+//# sourceMappingURL=stt.d.ts.map

package/dist/stt.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,iBAAiB,EAEtB,KAAK,WAAW,EAIhB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAEzB,OAAO,KAAK,SAAS,MAAM,wCAAwC,CAAC;AAEpE,OAAO,EAAE,SAAS,EAAE,CAAC;AAErB,cAAc;AACd,MAAM,WAAW,UAAU;IACzB,qEAAqE;IACrE,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,8DAA8D;IAC9D,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,oEAAoE;IACpE,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,4CAA4C;IAC5C,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,iCAAiC;IACjC,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,4BAA4B,CAAC,EAAE,MAAM,CAAC;IACtC,qBAAqB,CAAC,EAAE,MAAM,CAAC;IAC/B,oBAAoB,CAAC,EAAE,MAAM,CAAC;IAC9B,QAAQ,EAAE,MAAM,EAAE,CAAC;IACnB,SAAS,CAAC,EAAE,SAAS,CAAC,eAAe,CAAC;IACtC,UAAU,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;IAC7B,mBAAmB,EAAE,OAAO,CAAC;IAC7B,sBAAsB,EAAE,OAAO,CAAC;CACjC;AAED,cAAc;AACd,MAAM,MAAM,gBAAgB,GAAG,OAAO,CACpC,IAAI,CACF,UAAU,EACV,8BAA8B,GAAG,uBAAuB,GAAG,sBAAsB,CAClF,CACF,GAAG;IACF,QAAQ,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC9B,CAAC;AAEF,cAAc;AACd,MAAM,WAAW,qBAAsB,SAAQ,gBAAgB;IAC7D,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,CAAC,EAAE,SAAS,CAAC,eAAe,CAAC;IACtC,UAAU,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;IAC7B,mBAAmB,CAAC,EAAE,OAAO,CAAC;IAC9B,sBAAsB,CAAC,EAAE,OAAO,CAAC;CAClC;AAED,gBAAgB;AAChB,MAAM,WAAW,cAAc;IAC7B,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,MAAM,EAAE,SAAS,CAAC,kBAAkB,CAAC;IACrC,SAAS,CAAC,EAAE,SAAS,CAAC,qBAAqB,CAAC;CAC7C;AAED,gBAAgB;AAChB,MAAM,WAAW,cAAc;IAC7B,KAAK,EAAE,OAAO,CAAC;IACf,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;IACtB,GAAG,IAAI,IAAI,CAAC;IACZ,KAAK,IAAI,IAAI,CAAC;CACf;AA2CD,cAAc;AACd,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAe;IAEpB,IAAI,KAAK,IAAI,MAAM,CAElB;IAED,IAAI,QAAQ,IAAI,MAAM,CAErB;gBAEW,IAAI,GAAE,qBAA0B;IAoCtC,UAAU,CAAC,MAAM,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI/D,aAAa,CAAC,IAAI,EAAE,gBAAgB,GAAG,IAAI;IAkB3C,MAAM,CAAC,OAAO,GAAE;QAAE,QAAQ,CAAC,EAAE,MAAM,CAAC;QAAC,WAAW,CAAC,EAAE,iBAAiB,CAAA;KAAO,GAAG,YAAY;CAS3F;AAED,cAAc;AACd,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAChD,gBAAgB;IAChB,KAAK,EAAE,UAAU,CAAC;IAClB,gBAAgB;IAChB,SAAS,UAAS;IAClB,gBAAgB;IAChB,oBAAoB,EAAE,cAAc,CAAuB;IAC3D,gBAAgB;IAChB,oBAAoB,EAAE,cAAc,CAAuB;IAC3D,gBAAgB;IAChB,eAAe,EAAE,cAAc,CAAuB;IACtD,gBAAgB;IAChB,kBAAkB,EAAE,cAAc,GAAG,IAAI,CAAQ;IAIjD,KAAK,SAAwB;gBAEjB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU,EAAE,WAAW,EAAE,iBAAiB;IAMtE,aAAa,CAAC,IAAI,EAAE,gBAAgB,GAAG,IAAI;cAU3B,GAAG,IAAI,OAAO,CAAC,IAAI,CAAC;IA8GpC,gBAAgB;IAChB,aAAa,CAAC,GAAG,EAAE,SAAS,CAAC,0BAA0B,GAAG,IAAI;IAU9D,gBAAgB;IAChB,cAAc,CAAC,GAAG,EAAE,SAAS,CAAC,0BAA0B,GAAG,IAAI;IAU/D,gBAAgB;IAChB,cAAc,CAAC,IAAI,EAAE,SAAS,CAAC,oBAAoB,GAAG,IAAI;IAM1D,gBAAgB;IAChB,YAAY,CAAC,IAAI,EAAE,SAAS,CAAC,oBAAoB,GAAG,IAAI;IAMxD,gBAAgB;IAChB,iBAAiB,CAAC,IAAI,EAAE,SAAS,CAAC,gBAAgB,GAAG,IAAI;IAIzD,gBAAgB;IAChB,iBAAiB,CAAC,IAAI,EAAE,SAAS,CAAC,gBAAgB,GAAG,IAAI;IAIzD,gBAAgB;IAChB,WAAW,CAAC,GAAG,EAAE,cAAc,GAAG,IAAI;CA4CvC"}

package/dist/stt.js

@@ -0,0 +1,435 @@
+import {
+  APIConnectionError,
+  DEFAULT_API_CONNECT_OPTIONS,
+  asLanguageCode,
+  log,
+  stt
+} from "@livekit/agents";
+import * as speechsdk from "microsoft-cognitiveservices-speech-sdk";
+class DeferredEvent {
+  #resolve;
+  #promise;
+  isSet = false;
+  constructor() {
+    this.#promise = new Promise((resolve) => {
+      this.#resolve = resolve;
+    });
+  }
+  wait() {
+    return this.#promise;
+  }
+  set() {
+    var _a;
+    if (this.isSet) return;
+    this.isSet = true;
+    (_a = this.#resolve) == null ? void 0 : _a.call(this);
+  }
+  clear() {
+    if (!this.isSet) return;
+    this.isSet = false;
+    this.#promise = new Promise((resolve) => {
+      this.#resolve = resolve;
+    });
+  }
+}
+const defaultSTTOptions = {
+  sampleRate: 16e3,
+  numChannels: 1,
+  language: ["en-US"],
+  explicitPunctuation: false,
+  trueTextPostProcessing: false
+};
+class STT extends stt.STT {
+  #opts;
+  #streams = /* @__PURE__ */ new Set();
+  label = "azure.STT";
+  get model() {
+    return "unknown";
+  }
+  get provider() {
+    return "Azure STT";
+  }
+  constructor(opts = {}) {
+    super({ streaming: true, interimResults: true, alignedTranscript: "chunk" });
+    const speechHost = opts.speechHost ?? process.env.AZURE_SPEECH_HOST;
+    const speechKey = opts.speechKey ?? process.env.AZURE_SPEECH_KEY;
+    const speechRegion = opts.speechRegion ?? process.env.AZURE_SPEECH_REGION;
+    const speechAuthToken = opts.speechAuthToken;
+    const speechEndpoint = opts.speechEndpoint;
+    if (!speechHost && !(speechKey && speechRegion) && !(speechAuthToken && speechRegion) && !(speechKey && speechEndpoint)) {
+      throw new Error(
+        "AZURE_SPEECH_HOST or AZURE_SPEECH_KEY and AZURE_SPEECH_REGION or speechAuthToken and AZURE_SPEECH_REGION or AZURE_SPEECH_KEY and speechEndpoint must be set"
+      );
+    }
+    if (speechRegion && speechEndpoint) {
+      log().warn("speechRegion and speechEndpoint both are set, using speechEndpoint");
+    }
+    this.#opts = {
+      ...defaultSTTOptions,
+      ...opts,
+      speechHost,
+      speechKey,
+      speechRegion: speechEndpoint ? void 0 : speechRegion,
+      speechAuthToken,
+      speechEndpoint,
+      language: normalizeLanguages(opts.language)
+    };
+  }
+  async _recognize(_frame) {
+    throw new Error("Azure STT does not support single frame recognition");
+  }
+  updateOptions(opts) {
+    this.#opts = {
+      ...this.#opts,
+      ...opts,
+      language: opts.language === void 0 ? this.#opts.language : normalizeLanguages(opts.language)
+    };
+    for (const ref of this.#streams) {
+      const stream = ref.deref();
+      if (stream) {
+        stream.updateOptions(opts);
+      } else {
+        this.#streams.delete(ref);
+      }
+    }
+  }
+  stream(options = {}) {
+    const opts = {
+      ...this.#opts,
+      language: options.language ? [options.language] : [...this.#opts.language]
+    };
+    const stream = new SpeechStream(this, opts, options.connOptions ?? DEFAULT_API_CONNECT_OPTIONS);
+    this.#streams.add(new WeakRef(stream));
+    return stream;
+  }
+}
+class SpeechStream extends stt.SpeechStream {
+  /** @internal */
+  _opts;
+  /** @internal */
+  _speaking = false;
+  /** @internal */
+  _sessionStoppedEvent = new DeferredEvent();
+  /** @internal */
+  _sessionStartedEvent = new DeferredEvent();
+  /** @internal */
+  _reconnectEvent = new DeferredEvent();
+  /** @internal */
+  _cancellationError = null;
+  #connOptions;
+  #audioDuration = 0;
+  #lastAudioDurationReportTime = performance.now();
+  label = "azure.SpeechStream";
+  constructor(stt2, opts, connOptions) {
+    super(stt2, opts.sampleRate, connOptions);
+    this._opts = opts;
+    this.#connOptions = connOptions;
+  }
+  updateOptions(opts) {
+    this._opts = {
+      ...this._opts,
+      ...opts,
+      language: opts.language === void 0 ? this._opts.language : normalizeLanguages(opts.language)
+    };
+    this._reconnectEvent.set();
+  }
+  async run() {
+    while (!this.input.closed && !this.closed) {
+      this._sessionStoppedEvent.clear();
+      this._sessionStartedEvent.clear();
+      this._cancellationError = null;
+      this._speaking = false;
+      const pushStream = speechsdk.AudioInputStream.createPushStream(
+        speechsdk.AudioStreamFormat.getWaveFormatPCM(
+          this._opts.sampleRate,
+          16,
+          this._opts.numChannels
+        )
+      );
+      const recognizer = createSpeechRecognizer(this._opts, pushStream);
+      this.#connectRecognizerEvents(recognizer);
+      await new Promise((resolve, reject) => {
+        recognizer.startContinuousRecognitionAsync(resolve, (error) => reject(new Error(error)));
+      });
+      try {
+        await withTimeout(this._sessionStartedEvent.wait(), this.#connOptions.timeoutMs);
+        const inputAbortController = new AbortController();
+        const inputTask = this.#processInput(pushStream, inputAbortController.signal);
+        let inputEnded = false;
+        try {
+          const completed = await Promise.race([
+            inputTask.then(() => "input"),
+            this._reconnectEvent.wait().then(() => "reconnect"),
+            this._sessionStoppedEvent.wait().then(() => "stopped")
+          ]);
+          if (completed === "stopped") {
+            const details = this._cancellationError;
+            if (details !== null) {
+              throw new APIConnectionError({
+                message: `Azure STT canceled: ${details.errorDetails || details.reason} (${details.errorCode})`
+              });
+            }
+            throw new APIConnectionError({ message: "SpeechRecognition session stopped" });
+          }
+          if (completed === "reconnect") {
+            this._reconnectEvent.clear();
+          }
+          inputEnded = completed === "input";
+        } finally {
+          inputAbortController.abort();
+          await inputTask;
+          pushStream.close();
+        }
+        if (inputEnded) {
+          await this._sessionStoppedEvent.wait();
+          break;
+        }
+      } finally {
+        await new Promise((resolve) => {
+          recognizer.stopContinuousRecognitionAsync(resolve, () => resolve());
+        });
+        recognizer.close();
+      }
+    }
+  }
+  async #processInput(pushStream, abortSignal) {
+    try {
+      while (!this.closed) {
+        let result;
+        try {
+          result = await this.input.next({ signal: abortSignal });
+        } catch (error) {
+          if (abortSignal.aborted) break;
+          throw error;
+        }
+        if (result.done) break;
+        const input = result.value;
+        if (input === SpeechStream.FLUSH_SENTINEL) {
+          this.#emitRecognitionUsage();
+          continue;
+        }
+        this.#audioDuration += input.samplesPerChannel / input.sampleRate;
+        this.#maybeEmitRecognitionUsage();
+        pushStream.write(toArrayBuffer(input));
+      }
+    } finally {
+      this.#emitRecognitionUsage();
+    }
+  }
+  #connectRecognizerEvents(recognizer) {
+    recognizer.recognizing = (_sender, evt) => this._onRecognizing(evt);
+    recognizer.recognized = (_sender, evt) => this._onRecognized(evt);
+    recognizer.speechStartDetected = (_sender, evt) => this._onSpeechStart(evt);
+    recognizer.speechEndDetected = (_sender, evt) => this._onSpeechEnd(evt);
+    recognizer.sessionStarted = (_sender, evt) => this._onSessionStarted(evt);
+    recognizer.sessionStopped = (_sender, evt) => this._onSessionStopped(evt);
+    recognizer.canceled = (_sender, evt) => this._onCanceled(evt);
+  }
+  /** @internal */
+  _onRecognized(evt) {
+    const text = evt.result.text.trim();
+    if (!text) return;
+    this.queue.put({
+      type: stt.SpeechEventType.FINAL_TRANSCRIPT,
+      alternatives: [this.#speechData(evt, 1)]
+    });
+  }
+  /** @internal */
+  _onRecognizing(evt) {
+    const text = evt.result.text.trim();
+    if (!text) return;
+    this.queue.put({
+      type: stt.SpeechEventType.INTERIM_TRANSCRIPT,
+      alternatives: [this.#speechData(evt, 0)]
+    });
+  }
+  /** @internal */
+  _onSpeechStart(_evt) {
+    if (this._speaking) return;
+    this._speaking = true;
+    this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });
+  }
+  /** @internal */
+  _onSpeechEnd(_evt) {
+    if (!this._speaking) return;
+    this._speaking = false;
+    this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });
+  }
+  /** @internal */
+  _onSessionStarted(_evt) {
+    this._sessionStartedEvent.set();
+  }
+  /** @internal */
+  _onSessionStopped(_evt) {
+    this._sessionStoppedEvent.set();
+  }
+  /** @internal */
+  _onCanceled(evt) {
+    if (evt.reason === speechsdk.CancellationReason.Error) {
+      log().warn(
+        {
+          code: evt.errorCode,
+          reason: evt.reason,
+          errorDetails: evt.errorDetails
+        },
+        `Speech recognition canceled: ${evt.errorDetails || evt.reason}`
+      );
+      this._cancellationError = evt;
+      this._sessionStoppedEvent.set();
+    }
+  }
+  #speechData(evt, confidence) {
+    const result = speechsdk.AutoDetectSourceLanguageResult.fromResult(evt.result);
+    const language = result.language || this._opts.language[0] || "";
+    return {
+      language: asLanguageCode(language),
+      confidence,
+      text: evt.result.text,
+      startTime: evt.result.offset / 10 ** 7 + this.startTimeOffset,
+      endTime: (evt.result.offset + evt.result.duration) / 10 ** 7 + this.startTimeOffset
+    };
+  }
+  #maybeEmitRecognitionUsage() {
+    if (performance.now() - this.#lastAudioDurationReportTime >= 5e3) {
+      this.#emitRecognitionUsage();
+    }
+  }
+  #emitRecognitionUsage() {
+    if (this.#audioDuration <= 0) return;
+    const audioDuration = this.#audioDuration;
+    this.#audioDuration = 0;
+    this.#lastAudioDurationReportTime = performance.now();
+    this.queue.put({
+      type: stt.SpeechEventType.RECOGNITION_USAGE,
+      recognitionUsage: { audioDuration }
+    });
+  }
+}
+function normalizeLanguages(language) {
+  if (language === void 0) return [...defaultSTTOptions.language];
+  return Array.isArray(language) ? [...language] : [language];
+}
+function createSpeechRecognizer(config, stream) {
+  var _a;
+  const speechConfig = createSpeechConfig(config);
+  if (config.segmentationSilenceTimeoutMs !== void 0) {
+    speechConfig.setProperty(
+      speechsdk.PropertyId.Speech_SegmentationSilenceTimeoutMs,
+      String(config.segmentationSilenceTimeoutMs)
+    );
+  }
+  if (config.segmentationMaxTimeMs !== void 0) {
+    speechConfig.setProperty(
+      speechsdk.PropertyId.Speech_SegmentationMaximumTimeMs,
+      String(config.segmentationMaxTimeMs)
+    );
+  }
+  if (config.segmentationStrategy !== void 0) {
+    speechConfig.setProperty(
+      speechsdk.PropertyId.Speech_SegmentationStrategy,
+      config.segmentationStrategy
+    );
+  }
+  if (config.profanity !== void 0) {
+    speechConfig.setProfanity(config.profanity);
+  }
+  if (config.explicitPunctuation) {
+    speechConfig.setServiceProperty(
+      "punctuation",
+      "explicit",
+      speechsdk.ServicePropertyChannel.UriQueryParameter
+    );
+  }
+  if (config.trueTextPostProcessing) {
+    speechConfig.setProperty(
+      speechsdk.PropertyId.SpeechServiceResponse_PostProcessingOption,
+      "TrueText"
+    );
+  }
+  const audioConfig = speechsdk.AudioConfig.fromStreamInput(stream);
+  const recognizer = config.language.length > 1 ? createMultiLanguageRecognizer(speechConfig, audioConfig, config.language) : new speechsdk.SpeechRecognizer(speechConfig, audioConfig);
+  if ((_a = config.phraseList) == null ? void 0 : _a.length) {
+    const phraseListGrammar = speechsdk.PhraseListGrammar.fromRecognizer(recognizer);
+    for (const phrase of config.phraseList) {
+      phraseListGrammar.addPhrase(phrase);
+    }
+  }
+  return recognizer;
+}
+function createSpeechConfig(config) {
+  if (config.speechEndpoint) {
+    const speechConfig2 = speechsdk.SpeechConfig.fromEndpoint(
+      new URL(config.speechEndpoint),
+      config.speechKey
+    );
+    if (config.speechAuthToken) speechConfig2.authorizationToken = config.speechAuthToken;
+    speechConfig2.speechRecognitionLanguage = firstLanguage(config);
+    return speechConfig2;
+  }
+  if (config.speechHost) {
+    const speechConfig2 = speechsdk.SpeechConfig.fromHost(
+      new URL(config.speechHost),
+      config.speechKey
+    );
+    if (config.speechAuthToken) speechConfig2.authorizationToken = config.speechAuthToken;
+    speechConfig2.speechRecognitionLanguage = firstLanguage(config);
+    return speechConfig2;
+  }
+  if (config.speechAuthToken) {
+    const speechConfig2 = speechsdk.SpeechConfig.fromAuthorizationToken(
+      config.speechAuthToken,
+      config.speechRegion
+    );
+    speechConfig2.speechRecognitionLanguage = firstLanguage(config);
+    return speechConfig2;
+  }
+  const speechConfig = speechsdk.SpeechConfig.fromSubscription(
+    config.speechKey,
+    config.speechRegion
+  );
+  speechConfig.speechRecognitionLanguage = firstLanguage(config);
+  return speechConfig;
+}
+function firstLanguage(config) {
+  return config.language[0] ?? "en-US";
+}
+function createMultiLanguageRecognizer(speechConfig, audioConfig, languages) {
+  speechConfig.setProperty(
+    speechsdk.PropertyId.SpeechServiceConnection_LanguageIdMode,
+    "Continuous"
+  );
+  const autoDetectSourceLanguageConfig = speechsdk.AutoDetectSourceLanguageConfig.fromLanguages(languages);
+  return speechsdk.SpeechRecognizer.FromConfig(
+    speechConfig,
+    autoDetectSourceLanguageConfig,
+    audioConfig
+  );
+}
+function toArrayBuffer(frame) {
+  const view = new Uint8Array(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);
+  const buffer = new ArrayBuffer(view.byteLength);
+  new Uint8Array(buffer).set(view);
+  return buffer;
+}
+async function withTimeout(promise, timeoutMs) {
+  let timeout;
+  try {
+    await Promise.race([
+      promise,
+      new Promise((_, reject) => {
+        timeout = setTimeout(
+          () => reject(new APIConnectionError({ message: "Request timed out." })),
+          timeoutMs
+        );
+      })
+    ]);
+  } finally {
+    if (timeout) clearTimeout(timeout);
+  }
+}
+export {
+  STT,
+  SpeechStream,
+  speechsdk
+};
+//# sourceMappingURL=stt.js.map

package/dist/stt.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2026 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n  type APIConnectOptions,\n  APIConnectionError,\n  type AudioBuffer,\n  DEFAULT_API_CONNECT_OPTIONS,\n  asLanguageCode,\n  log,\n  stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport * as speechsdk from 'microsoft-cognitiveservices-speech-sdk';\n\nexport { speechsdk };\n\n/** @public */\nexport interface STTOptions {\n  /** Azure Speech subscription key. Defaults to `AZURE_SPEECH_KEY`. */\n  speechKey?: string;\n  /** Azure Speech region. Defaults to `AZURE_SPEECH_REGION`. */\n  speechRegion?: string;\n  /** Azure Speech container host. Defaults to `AZURE_SPEECH_HOST`. */\n  speechHost?: string;\n  /** Ephemeral Microsoft Entra auth token. */\n  speechAuthToken?: string;\n  /** Azure Speech endpoint URL. */\n  speechEndpoint?: string;\n  sampleRate: number;\n  numChannels: number;\n  segmentationSilenceTimeoutMs?: number;\n  segmentationMaxTimeMs?: number;\n  segmentationStrategy?: string;\n  language: string[];\n  profanity?: speechsdk.ProfanityOption;\n  phraseList?: string[] | null;\n  explicitPunctuation: boolean;\n  trueTextPostProcessing: boolean;\n}\n\n/** @public */\nexport type STTUpdateOptions = Partial<\n  Pick<\n    STTOptions,\n    'segmentationSilenceTimeoutMs' | 'segmentationMaxTimeMs' | 'segmentationStrategy'\n  >\n> & {\n  language?: string | string[];\n};\n\n/** @public */\nexport interface STTConstructorOptions extends STTUpdateOptions {\n  speechKey?: string;\n  speechRegion?: string;\n  speechHost?: string;\n  speechAuthToken?: string;\n  speechEndpoint?: string;\n  sampleRate?: number;\n  numChannels?: number;\n  profanity?: speechsdk.ProfanityOption;\n  phraseList?: string[] | null;\n  explicitPunctuation?: boolean;\n  trueTextPostProcessing?: boolean;\n}\n\n/** @internal */\nexport interface _CanceledEvent {\n  errorDetails?: string;\n  reason: speechsdk.CancellationReason;\n  errorCode?: speechsdk.CancellationErrorCode;\n}\n\n/** @internal */\nexport interface _WaitableEvent {\n  isSet: boolean;\n  wait(): Promise<void>;\n  set(): void;\n  clear(): void;\n}\n\nclass DeferredEvent implements _WaitableEvent {\n  #resolve?: () => void;\n  #promise: Promise<void>;\n  isSet = false;\n\n  constructor() {\n    this.#promise = new Promise<void>((resolve) => {\n      this.#resolve = resolve;\n    });\n  }\n\n  wait(): Promise<void> {\n    return this.#promise;\n  }\n\n  set(): void {\n    if (this.isSet) return;\n    this.isSet = true;\n    this.#resolve?.();\n  }\n\n  clear(): void {\n    if (!this.isSet) return;\n    this.isSet = false;\n    this.#promise = new Promise<void>((resolve) => {\n      this.#resolve = resolve;\n    });\n  }\n}\n\nconst defaultSTTOptions = {\n  sampleRate: 16000,\n  numChannels: 1,\n  language: ['en-US'],\n  explicitPunctuation: false,\n  trueTextPostProcessing: false,\n} satisfies Pick<\n  STTOptions,\n  'sampleRate' | 'numChannels' | 'language' | 'explicitPunctuation' | 'trueTextPostProcessing'\n>;\n\n/** @public */\nexport class STT extends stt.STT {\n  #opts: STTOptions;\n  #streams = new Set<WeakRef<SpeechStream>>();\n  label = 'azure.STT';\n\n  get model(): string {\n    return 'unknown';\n  }\n\n  get provider(): string {\n    return 'Azure STT';\n  }\n\n  constructor(opts: STTConstructorOptions = {}) {\n    super({ streaming: true, interimResults: true, alignedTranscript: 'chunk' });\n\n    const speechHost = opts.speechHost ?? process.env.AZURE_SPEECH_HOST;\n    const speechKey = opts.speechKey ?? process.env.AZURE_SPEECH_KEY;\n    const speechRegion = opts.speechRegion ?? process.env.AZURE_SPEECH_REGION;\n    const speechAuthToken = opts.speechAuthToken;\n    const speechEndpoint = opts.speechEndpoint;\n\n    if (\n      !speechHost &&\n      !(speechKey && speechRegion) &&\n      !(speechAuthToken && speechRegion) &&\n      !(speechKey && speechEndpoint)\n    ) {\n      throw new Error(\n        'AZURE_SPEECH_HOST or AZURE_SPEECH_KEY and AZURE_SPEECH_REGION or speechAuthToken and AZURE_SPEECH_REGION or AZURE_SPEECH_KEY and speechEndpoint must be set',\n      );\n    }\n\n    if (speechRegion && speechEndpoint) {\n      log().warn('speechRegion and speechEndpoint both are set, using speechEndpoint');\n    }\n\n    this.#opts = {\n      ...defaultSTTOptions,\n      ...opts,\n      speechHost,\n      speechKey,\n      speechRegion: speechEndpoint ? undefined : speechRegion,\n      speechAuthToken,\n      speechEndpoint,\n      language: normalizeLanguages(opts.language),\n    };\n  }\n\n  async _recognize(_frame: AudioBuffer): Promise<stt.SpeechEvent> {\n    throw new Error('Azure STT does not support single frame recognition');\n  }\n\n  updateOptions(opts: STTUpdateOptions): void {\n    this.#opts = {\n      ...this.#opts,\n      ...opts,\n      language:\n        opts.language === undefined ? this.#opts.language : normalizeLanguages(opts.language),\n    };\n\n    for (const ref of this.#streams) {\n      const stream = ref.deref();\n      if (stream) {\n        stream.updateOptions(opts);\n      } else {\n        this.#streams.delete(ref);\n      }\n    }\n  }\n\n  stream(options: { language?: string; connOptions?: APIConnectOptions } = {}): SpeechStream {\n    const opts = {\n      ...this.#opts,\n      language: options.language ? [options.language] : [...this.#opts.language],\n    };\n    const stream = new SpeechStream(this, opts, options.connOptions ?? DEFAULT_API_CONNECT_OPTIONS);\n    this.#streams.add(new WeakRef(stream));\n    return stream;\n  }\n}\n\n/** @public */\nexport class SpeechStream extends stt.SpeechStream {\n  /** @internal */\n  _opts: STTOptions;\n  /** @internal */\n  _speaking = false;\n  /** @internal */\n  _sessionStoppedEvent: _WaitableEvent = new DeferredEvent();\n  /** @internal */\n  _sessionStartedEvent: _WaitableEvent = new DeferredEvent();\n  /** @internal */\n  _reconnectEvent: _WaitableEvent = new DeferredEvent();\n  /** @internal */\n  _cancellationError: _CanceledEvent | null = null;\n  #connOptions: APIConnectOptions;\n  #audioDuration = 0;\n  #lastAudioDurationReportTime = performance.now();\n  label = 'azure.SpeechStream';\n\n  constructor(stt: STT, opts: STTOptions, connOptions: APIConnectOptions) {\n    super(stt, opts.sampleRate, connOptions);\n    this._opts = opts;\n    this.#connOptions = connOptions;\n  }\n\n  updateOptions(opts: STTUpdateOptions): void {\n    this._opts = {\n      ...this._opts,\n      ...opts,\n      language:\n        opts.language === undefined ? this._opts.language : normalizeLanguages(opts.language),\n    };\n    this._reconnectEvent.set();\n  }\n\n  protected async run(): Promise<void> {\n    while (!this.input.closed && !this.closed) {\n      this._sessionStoppedEvent.clear();\n      this._sessionStartedEvent.clear();\n      this._cancellationError = null;\n      this._speaking = false;\n\n      const pushStream = speechsdk.AudioInputStream.createPushStream(\n        speechsdk.AudioStreamFormat.getWaveFormatPCM(\n          this._opts.sampleRate,\n          16,\n          this._opts.numChannels,\n        ),\n      );\n      const recognizer = createSpeechRecognizer(this._opts, pushStream);\n      this.#connectRecognizerEvents(recognizer);\n\n      await new Promise<void>((resolve, reject) => {\n        recognizer.startContinuousRecognitionAsync(resolve, (error) => reject(new Error(error)));\n      });\n\n      try {\n        await withTimeout(this._sessionStartedEvent.wait(), this.#connOptions.timeoutMs);\n\n        const inputAbortController = new AbortController();\n        const inputTask = this.#processInput(pushStream, inputAbortController.signal);\n        let inputEnded = false;\n        try {\n          const completed = await Promise.race([\n            inputTask.then(() => 'input' as const),\n            this._reconnectEvent.wait().then(() => 'reconnect' as const),\n            this._sessionStoppedEvent.wait().then(() => 'stopped' as const),\n          ]);\n\n          if (completed === 'stopped') {\n            const details = this._cancellationError as _CanceledEvent | null;\n            if (details !== null) {\n              throw new APIConnectionError({\n                message:\n                  `Azure STT canceled: ${details.errorDetails || details.reason} ` +\n                  `(${details.errorCode})`,\n              });\n            }\n            throw new APIConnectionError({ message: 'SpeechRecognition session stopped' });\n          }\n\n          if (completed === 'reconnect') {\n            this._reconnectEvent.clear();\n          }\n\n          inputEnded = completed === 'input';\n        } finally {\n          inputAbortController.abort();\n          await inputTask;\n          pushStream.close();\n        }\n\n        if (inputEnded) {\n          await this._sessionStoppedEvent.wait();\n          break;\n        }\n      } finally {\n        await new Promise<void>((resolve) => {\n          recognizer.stopContinuousRecognitionAsync(resolve, () => resolve());\n        });\n        recognizer.close();\n      }\n    }\n  }\n\n  async #processInput(\n    pushStream: speechsdk.PushAudioInputStream,\n    abortSignal: AbortSignal,\n  ): Promise<void> {\n    try {\n      while (!this.closed) {\n        let result: IteratorResult<AudioFrame | typeof SpeechStream.FLUSH_SENTINEL>;\n        try {\n          result = await this.input.next({ signal: abortSignal });\n        } catch (error) {\n          if (abortSignal.aborted) break;\n          throw error;\n        }\n        if (result.done) break;\n\n        const input = result.value;\n        if (input === SpeechStream.FLUSH_SENTINEL) {\n          this.#emitRecognitionUsage();\n          continue;\n        }\n\n        this.#audioDuration += input.samplesPerChannel / input.sampleRate;\n        this.#maybeEmitRecognitionUsage();\n        pushStream.write(toArrayBuffer(input));\n      }\n    } finally {\n      this.#emitRecognitionUsage();\n    }\n  }\n\n  #connectRecognizerEvents(recognizer: speechsdk.SpeechRecognizer): void {\n    recognizer.recognizing = (_sender, evt) => this._onRecognizing(evt);\n    recognizer.recognized = (_sender, evt) => this._onRecognized(evt);\n    recognizer.speechStartDetected = (_sender, evt) => this._onSpeechStart(evt);\n    recognizer.speechEndDetected = (_sender, evt) => this._onSpeechEnd(evt);\n    recognizer.sessionStarted = (_sender, evt) => this._onSessionStarted(evt);\n    recognizer.sessionStopped = (_sender, evt) => this._onSessionStopped(evt);\n    recognizer.canceled = (_sender, evt) => this._onCanceled(evt);\n  }\n\n  /** @internal */\n  _onRecognized(evt: speechsdk.SpeechRecognitionEventArgs): void {\n    const text = evt.result.text.trim();\n    if (!text) return;\n\n    this.queue.put({\n      type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n      alternatives: [this.#speechData(evt, 1)],\n    });\n  }\n\n  /** @internal */\n  _onRecognizing(evt: speechsdk.SpeechRecognitionEventArgs): void {\n    const text = evt.result.text.trim();\n    if (!text) return;\n\n    this.queue.put({\n      type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n      alternatives: [this.#speechData(evt, 0)],\n    });\n  }\n\n  /** @internal */\n  _onSpeechStart(_evt: speechsdk.RecognitionEventArgs): void {\n    if (this._speaking) return;\n    this._speaking = true;\n    this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n  }\n\n  /** @internal */\n  _onSpeechEnd(_evt: speechsdk.RecognitionEventArgs): void {\n    if (!this._speaking) return;\n    this._speaking = false;\n    this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n  }\n\n  /** @internal */\n  _onSessionStarted(_evt: speechsdk.SessionEventArgs): void {\n    this._sessionStartedEvent.set();\n  }\n\n  /** @internal */\n  _onSessionStopped(_evt: speechsdk.SessionEventArgs): void {\n    this._sessionStoppedEvent.set();\n  }\n\n  /** @internal */\n  _onCanceled(evt: _CanceledEvent): void {\n    if (evt.reason === speechsdk.CancellationReason.Error) {\n      log().warn(\n        {\n          code: evt.errorCode,\n          reason: evt.reason,\n          errorDetails: evt.errorDetails,\n        },\n        `Speech recognition canceled: ${evt.errorDetails || evt.reason}`,\n      );\n      this._cancellationError = evt;\n      this._sessionStoppedEvent.set();\n    }\n  }\n\n  #speechData(evt: speechsdk.SpeechRecognitionEventArgs, confidence: number): stt.SpeechData {\n    const result = speechsdk.AutoDetectSourceLanguageResult.fromResult(evt.result);\n    const language = result.language || this._opts.language[0] || '';\n    return {\n      language: asLanguageCode(language),\n      confidence,\n      text: evt.result.text,\n      startTime: evt.result.offset / 10 ** 7 + this.startTimeOffset,\n      endTime: (evt.result.offset + evt.result.duration) / 10 ** 7 + this.startTimeOffset,\n    };\n  }\n\n  #maybeEmitRecognitionUsage(): void {\n    if (performance.now() - this.#lastAudioDurationReportTime >= 5000) {\n      this.#emitRecognitionUsage();\n    }\n  }\n\n  #emitRecognitionUsage(): void {\n    if (this.#audioDuration <= 0) return;\n\n    const audioDuration = this.#audioDuration;\n    this.#audioDuration = 0;\n    this.#lastAudioDurationReportTime = performance.now();\n    this.queue.put({\n      type: stt.SpeechEventType.RECOGNITION_USAGE,\n      recognitionUsage: { audioDuration },\n    });\n  }\n}\n\nfunction normalizeLanguages(language?: string | string[]): string[] {\n  if (language === undefined) return [...defaultSTTOptions.language];\n  return Array.isArray(language) ? [...language] : [language];\n}\n\nfunction createSpeechRecognizer(\n  config: STTOptions,\n  stream: speechsdk.PushAudioInputStream,\n): speechsdk.SpeechRecognizer {\n  const speechConfig = createSpeechConfig(config);\n\n  if (config.segmentationSilenceTimeoutMs !== undefined) {\n    speechConfig.setProperty(\n      speechsdk.PropertyId.Speech_SegmentationSilenceTimeoutMs,\n      String(config.segmentationSilenceTimeoutMs),\n    );\n  }\n  if (config.segmentationMaxTimeMs !== undefined) {\n    speechConfig.setProperty(\n      speechsdk.PropertyId.Speech_SegmentationMaximumTimeMs,\n      String(config.segmentationMaxTimeMs),\n    );\n  }\n  if (config.segmentationStrategy !== undefined) {\n    speechConfig.setProperty(\n      speechsdk.PropertyId.Speech_SegmentationStrategy,\n      config.segmentationStrategy,\n    );\n  }\n  if (config.profanity !== undefined) {\n    speechConfig.setProfanity(config.profanity);\n  }\n  if (config.explicitPunctuation) {\n    speechConfig.setServiceProperty(\n      'punctuation',\n      'explicit',\n      speechsdk.ServicePropertyChannel.UriQueryParameter,\n    );\n  }\n  if (config.trueTextPostProcessing) {\n    speechConfig.setProperty(\n      speechsdk.PropertyId.SpeechServiceResponse_PostProcessingOption,\n      'TrueText',\n    );\n  }\n\n  const audioConfig = speechsdk.AudioConfig.fromStreamInput(stream);\n  const recognizer =\n    config.language.length > 1\n      ? createMultiLanguageRecognizer(speechConfig, audioConfig, config.language)\n      : new speechsdk.SpeechRecognizer(speechConfig, audioConfig);\n\n  if (config.phraseList?.length) {\n    const phraseListGrammar = speechsdk.PhraseListGrammar.fromRecognizer(recognizer);\n    for (const phrase of config.phraseList) {\n      phraseListGrammar.addPhrase(phrase);\n    }\n  }\n\n  return recognizer;\n}\n\nfunction createSpeechConfig(config: STTOptions): speechsdk.SpeechConfig {\n  if (config.speechEndpoint) {\n    const speechConfig = speechsdk.SpeechConfig.fromEndpoint(\n      new URL(config.speechEndpoint),\n      config.speechKey,\n    );\n    if (config.speechAuthToken) speechConfig.authorizationToken = config.speechAuthToken;\n    speechConfig.speechRecognitionLanguage = firstLanguage(config);\n    return speechConfig;\n  }\n  if (config.speechHost) {\n    const speechConfig = speechsdk.SpeechConfig.fromHost(\n      new URL(config.speechHost),\n      config.speechKey,\n    );\n    if (config.speechAuthToken) speechConfig.authorizationToken = config.speechAuthToken;\n    speechConfig.speechRecognitionLanguage = firstLanguage(config);\n    return speechConfig;\n  }\n  if (config.speechAuthToken) {\n    const speechConfig = speechsdk.SpeechConfig.fromAuthorizationToken(\n      config.speechAuthToken,\n      config.speechRegion!,\n    );\n    speechConfig.speechRecognitionLanguage = firstLanguage(config);\n    return speechConfig;\n  }\n  const speechConfig = speechsdk.SpeechConfig.fromSubscription(\n    config.speechKey!,\n    config.speechRegion!,\n  );\n  speechConfig.speechRecognitionLanguage = firstLanguage(config);\n  return speechConfig;\n}\n\nfunction firstLanguage(config: STTOptions): string {\n  return config.language[0] ?? 'en-US';\n}\n\nfunction createMultiLanguageRecognizer(\n  speechConfig: speechsdk.SpeechConfig,\n  audioConfig: speechsdk.AudioConfig,\n  languages: string[],\n): speechsdk.SpeechRecognizer {\n  speechConfig.setProperty(\n    speechsdk.PropertyId.SpeechServiceConnection_LanguageIdMode,\n    'Continuous',\n  );\n  const autoDetectSourceLanguageConfig =\n    speechsdk.AutoDetectSourceLanguageConfig.fromLanguages(languages);\n  return speechsdk.SpeechRecognizer.FromConfig(\n    speechConfig,\n    autoDetectSourceLanguageConfig,\n    audioConfig,\n  );\n}\n\nfunction toArrayBuffer(frame: AudioFrame): ArrayBuffer {\n  const view = new Uint8Array(frame.data.buffer, frame.data.byteOffset, frame.data.byteLength);\n  const buffer = new ArrayBuffer(view.byteLength);\n  new Uint8Array(buffer).set(view);\n  return buffer;\n}\n\nasync function withTimeout(promise: Promise<void>, timeoutMs: number): Promise<void> {\n  let timeout: NodeJS.Timeout | undefined;\n  try {\n    await Promise.race([\n      promise,\n      new Promise<void>((_, reject) => {\n        timeout = setTimeout(\n          () => reject(new APIConnectionError({ message: 'Request timed out.' })),\n          timeoutMs,\n        );\n      }),\n    ]);\n  } finally {\n    if (timeout) clearTimeout(timeout);\n  }\n}\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,YAAY,eAAe;AAoE3B,MAAM,cAAwC;AAAA,EAC5C;AAAA,EACA;AAAA,EACA,QAAQ;AAAA,EAER,cAAc;AACZ,SAAK,WAAW,IAAI,QAAc,CAAC,YAAY;AAC7C,WAAK,WAAW;AAAA,IAClB,CAAC;AAAA,EACH;AAAA,EAEA,OAAsB;AACpB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAY;AAhGd;AAiGI,QAAI,KAAK,MAAO;AAChB,SAAK,QAAQ;AACb,eAAK,aAAL;AAAA,EACF;AAAA,EAEA,QAAc;AACZ,QAAI,CAAC,KAAK,MAAO;AACjB,SAAK,QAAQ;AACb,SAAK,WAAW,IAAI,QAAc,CAAC,YAAY;AAC7C,WAAK,WAAW;AAAA,IAClB,CAAC;AAAA,EACH;AACF;AAEA,MAAM,oBAAoB;AAAA,EACxB,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC,OAAO;AAAA,EAClB,qBAAqB;AAAA,EACrB,wBAAwB;AAC1B;AAMO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,WAAW,oBAAI,IAA2B;AAAA,EAC1C,QAAQ;AAAA,EAER,IAAI,QAAgB;AAClB,WAAO;AAAA,EACT;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO;AAAA,EACT;AAAA,EAEA,YAAY,OAA8B,CAAC,GAAG;AAC5C,UAAM,EAAE,WAAW,MAAM,gBAAgB,MAAM,mBAAmB,QAAQ,CAAC;AAE3E,UAAM,aAAa,KAAK,cAAc,QAAQ,IAAI;AAClD,UAAM,YAAY,KAAK,aAAa,QAAQ,IAAI;AAChD,UAAM,eAAe,KAAK,gBAAgB,QAAQ,IAAI;AACtD,UAAM,kBAAkB,KAAK;AAC7B,UAAM,iBAAiB,KAAK;AAE5B,QACE,CAAC,cACD,EAAE,aAAa,iBACf,EAAE,mBAAmB,iBACrB,EAAE,aAAa,iBACf;AACA,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,QAAI,gBAAgB,gBAAgB;AAClC,UAAI,EAAE,KAAK,oEAAoE;AAAA,IACjF;AAEA,SAAK,QAAQ;AAAA,MACX,GAAG;AAAA,MACH,GAAG;AAAA,MACH;AAAA,MACA;AAAA,MACA,cAAc,iBAAiB,SAAY;AAAA,MAC3C;AAAA,MACA;AAAA,MACA,UAAU,mBAAmB,KAAK,QAAQ;AAAA,IAC5C;AAAA,EACF;AAAA,EAEA,MAAM,WAAW,QAA+C;AAC9D,UAAM,IAAI,MAAM,qDAAqD;AAAA,EACvE;AAAA,EAEA,cAAc,MAA8B;AAC1C,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,KAAK,MAAM,WAAW,mBAAmB,KAAK,QAAQ;AAAA,IACxF;AAEA,eAAW,OAAO,KAAK,UAAU;AAC/B,YAAM,SAAS,IAAI,MAAM;AACzB,UAAI,QAAQ;AACV,eAAO,cAAc,IAAI;AAAA,MAC3B,OAAO;AACL,aAAK,SAAS,OAAO,GAAG;AAAA,MAC1B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,OAAO,UAAkE,CAAC,GAAiB;AACzF,UAAM,OAAO;AAAA,MACX,GAAG,KAAK;AAAA,MACR,UAAU,QAAQ,WAAW,CAAC,QAAQ,QAAQ,IAAI,CAAC,GAAG,KAAK,MAAM,QAAQ;AAAA,IAC3E;AACA,UAAM,SAAS,IAAI,aAAa,MAAM,MAAM,QAAQ,eAAe,2BAA2B;AAC9F,SAAK,SAAS,IAAI,IAAI,QAAQ,MAAM,CAAC;AACrC,WAAO;AAAA,EACT;AACF;AAGO,MAAM,qBAAqB,IAAI,aAAa;AAAA;AAAA,EAEjD;AAAA;AAAA,EAEA,YAAY;AAAA;AAAA,EAEZ,uBAAuC,IAAI,cAAc;AAAA;AAAA,EAEzD,uBAAuC,IAAI,cAAc;AAAA;AAAA,EAEzD,kBAAkC,IAAI,cAAc;AAAA;AAAA,EAEpD,qBAA4C;AAAA,EAC5C;AAAA,EACA,iBAAiB;AAAA,EACjB,+BAA+B,YAAY,IAAI;AAAA,EAC/C,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB,aAAgC;AACtE,UAAMA,MAAK,KAAK,YAAY,WAAW;AACvC,SAAK,QAAQ;AACb,SAAK,eAAe;AAAA,EACtB;AAAA,EAEA,cAAc,MAA8B;AAC1C,SAAK,QAAQ;AAAA,MACX,GAAG,KAAK;AAAA,MACR,GAAG;AAAA,MACH,UACE,KAAK,aAAa,SAAY,KAAK,MAAM,WAAW,mBAAmB,KAAK,QAAQ;AAAA,IACxF;AACA,SAAK,gBAAgB,IAAI;AAAA,EAC3B;AAAA,EAEA,MAAgB,MAAqB;AACnC,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,WAAK,qBAAqB,MAAM;AAChC,WAAK,qBAAqB,MAAM;AAChC,WAAK,qBAAqB;AAC1B,WAAK,YAAY;AAEjB,YAAM,aAAa,UAAU,iBAAiB;AAAA,QAC5C,UAAU,kBAAkB;AAAA,UAC1B,KAAK,MAAM;AAAA,UACX;AAAA,UACA,KAAK,MAAM;AAAA,QACb;AAAA,MACF;AACA,YAAM,aAAa,uBAAuB,KAAK,OAAO,UAAU;AAChE,WAAK,yBAAyB,UAAU;AAExC,YAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,mBAAW,gCAAgC,SAAS,CAAC,UAAU,OAAO,IAAI,MAAM,KAAK,CAAC,CAAC;AAAA,MACzF,CAAC;AAED,UAAI;AACF,cAAM,YAAY,KAAK,qBAAqB,KAAK,GAAG,KAAK,aAAa,SAAS;AAE/E,cAAM,uBAAuB,IAAI,gBAAgB;AACjD,cAAM,YAAY,KAAK,cAAc,YAAY,qBAAqB,MAAM;AAC5E,YAAI,aAAa;AACjB,YAAI;AACF,gBAAM,YAAY,MAAM,QAAQ,KAAK;AAAA,YACnC,UAAU,KAAK,MAAM,OAAgB;AAAA,YACrC,KAAK,gBAAgB,KAAK,EAAE,KAAK,MAAM,WAAoB;AAAA,YAC3D,KAAK,qBAAqB,KAAK,EAAE,KAAK,MAAM,SAAkB;AAAA,UAChE,CAAC;AAED,cAAI,cAAc,WAAW;AAC3B,kBAAM,UAAU,KAAK;AACrB,gBAAI,YAAY,MAAM;AACpB,oBAAM,IAAI,mBAAmB;AAAA,gBAC3B,SACE,uBAAuB,QAAQ,gBAAgB,QAAQ,MAAM,KACzD,QAAQ,SAAS;AAAA,cACzB,CAAC;AAAA,YACH;AACA,kBAAM,IAAI,mBAAmB,EAAE,SAAS,oCAAoC,CAAC;AAAA,UAC/E;AAEA,cAAI,cAAc,aAAa;AAC7B,iBAAK,gBAAgB,MAAM;AAAA,UAC7B;AAEA,uBAAa,cAAc;AAAA,QAC7B,UAAE;AACA,+BAAqB,MAAM;AAC3B,gBAAM;AACN,qBAAW,MAAM;AAAA,QACnB;AAEA,YAAI,YAAY;AACd,gBAAM,KAAK,qBAAqB,KAAK;AACrC;AAAA,QACF;AAAA,MACF,UAAE;AACA,cAAM,IAAI,QAAc,CAAC,YAAY;AACnC,qBAAW,+BAA+B,SAAS,MAAM,QAAQ,CAAC;AAAA,QACpE,CAAC;AACD,mBAAW,MAAM;AAAA,MACnB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAM,cACJ,YACA,aACe;AACf,QAAI;AACF,aAAO,CAAC,KAAK,QAAQ;AACnB,YAAI;AACJ,YAAI;AACF,mBAAS,MAAM,KAAK,MAAM,KAAK,EAAE,QAAQ,YAAY,CAAC;AAAA,QACxD,SAAS,OAAO;AACd,cAAI,YAAY,QAAS;AACzB,gBAAM;AAAA,QACR;AACA,YAAI,OAAO,KAAM;AAEjB,cAAM,QAAQ,OAAO;AACrB,YAAI,UAAU,aAAa,gBAAgB;AACzC,eAAK,sBAAsB;AAC3B;AAAA,QACF;AAEA,aAAK,kBAAkB,MAAM,oBAAoB,MAAM;AACvD,aAAK,2BAA2B;AAChC,mBAAW,MAAM,cAAc,KAAK,CAAC;AAAA,MACvC;AAAA,IACF,UAAE;AACA,WAAK,sBAAsB;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,yBAAyB,YAA8C;AACrE,eAAW,cAAc,CAAC,SAAS,QAAQ,KAAK,eAAe,GAAG;AAClE,eAAW,aAAa,CAAC,SAAS,QAAQ,KAAK,cAAc,GAAG;AAChE,eAAW,sBAAsB,CAAC,SAAS,QAAQ,KAAK,eAAe,GAAG;AAC1E,eAAW,oBAAoB,CAAC,SAAS,QAAQ,KAAK,aAAa,GAAG;AACtE,eAAW,iBAAiB,CAAC,SAAS,QAAQ,KAAK,kBAAkB,GAAG;AACxE,eAAW,iBAAiB,CAAC,SAAS,QAAQ,KAAK,kBAAkB,GAAG;AACxE,eAAW,WAAW,CAAC,SAAS,QAAQ,KAAK,YAAY,GAAG;AAAA,EAC9D;AAAA;AAAA,EAGA,cAAc,KAAiD;AAC7D,UAAM,OAAO,IAAI,OAAO,KAAK,KAAK;AAClC,QAAI,CAAC,KAAM;AAEX,SAAK,MAAM,IAAI;AAAA,MACb,MAAM,IAAI,gBAAgB;AAAA,MAC1B,cAAc,CAAC,KAAK,YAAY,KAAK,CAAC,CAAC;AAAA,IACzC,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,eAAe,KAAiD;AAC9D,UAAM,OAAO,IAAI,OAAO,KAAK,KAAK;AAClC,QAAI,CAAC,KAAM;AAEX,SAAK,MAAM,IAAI;AAAA,MACb,MAAM,IAAI,gBAAgB;AAAA,MAC1B,cAAc,CAAC,KAAK,YAAY,KAAK,CAAC,CAAC;AAAA,IACzC,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,eAAe,MAA4C;AACzD,QAAI,KAAK,UAAW;AACpB,SAAK,YAAY;AACjB,SAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAAA,EAC9D;AAAA;AAAA,EAGA,aAAa,MAA4C;AACvD,QAAI,CAAC,KAAK,UAAW;AACrB,SAAK,YAAY;AACjB,SAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,EAC5D;AAAA;AAAA,EAGA,kBAAkB,MAAwC;AACxD,SAAK,qBAAqB,IAAI;AAAA,EAChC;AAAA;AAAA,EAGA,kBAAkB,MAAwC;AACxD,SAAK,qBAAqB,IAAI;AAAA,EAChC;AAAA;AAAA,EAGA,YAAY,KAA2B;AACrC,QAAI,IAAI,WAAW,UAAU,mBAAmB,OAAO;AACrD,UAAI,EAAE;AAAA,QACJ;AAAA,UACE,MAAM,IAAI;AAAA,UACV,QAAQ,IAAI;AAAA,UACZ,cAAc,IAAI;AAAA,QACpB;AAAA,QACA,gCAAgC,IAAI,gBAAgB,IAAI,MAAM;AAAA,MAChE;AACA,WAAK,qBAAqB;AAC1B,WAAK,qBAAqB,IAAI;AAAA,IAChC;AAAA,EACF;AAAA,EAEA,YAAY,KAA2C,YAAoC;AACzF,UAAM,SAAS,UAAU,+BAA+B,WAAW,IAAI,MAAM;AAC7E,UAAM,WAAW,OAAO,YAAY,KAAK,MAAM,SAAS,CAAC,KAAK;AAC9D,WAAO;AAAA,MACL,UAAU,eAAe,QAAQ;AAAA,MACjC;AAAA,MACA,MAAM,IAAI,OAAO;AAAA,MACjB,WAAW,IAAI,OAAO,SAAS,MAAM,IAAI,KAAK;AAAA,MAC9C,UAAU,IAAI,OAAO,SAAS,IAAI,OAAO,YAAY,MAAM,IAAI,KAAK;AAAA,IACtE;AAAA,EACF;AAAA,EAEA,6BAAmC;AACjC,QAAI,YAAY,IAAI,IAAI,KAAK,gCAAgC,KAAM;AACjE,WAAK,sBAAsB;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,wBAA8B;AAC5B,QAAI,KAAK,kBAAkB,EAAG;AAE9B,UAAM,gBAAgB,KAAK;AAC3B,SAAK,iBAAiB;AACtB,SAAK,+BAA+B,YAAY,IAAI;AACpD,SAAK,MAAM,IAAI;AAAA,MACb,MAAM,IAAI,gBAAgB;AAAA,MAC1B,kBAAkB,EAAE,cAAc;AAAA,IACpC,CAAC;AAAA,EACH;AACF;AAEA,SAAS,mBAAmB,UAAwC;AAClE,MAAI,aAAa,OAAW,QAAO,CAAC,GAAG,kBAAkB,QAAQ;AACjE,SAAO,MAAM,QAAQ,QAAQ,IAAI,CAAC,GAAG,QAAQ,IAAI,CAAC,QAAQ;AAC5D;AAEA,SAAS,uBACP,QACA,QAC4B;AAnc9B;AAocE,QAAM,eAAe,mBAAmB,MAAM;AAE9C,MAAI,OAAO,iCAAiC,QAAW;AACrD,iBAAa;AAAA,MACX,UAAU,WAAW;AAAA,MACrB,OAAO,OAAO,4BAA4B;AAAA,IAC5C;AAAA,EACF;AACA,MAAI,OAAO,0BAA0B,QAAW;AAC9C,iBAAa;AAAA,MACX,UAAU,WAAW;AAAA,MACrB,OAAO,OAAO,qBAAqB;AAAA,IACrC;AAAA,EACF;AACA,MAAI,OAAO,yBAAyB,QAAW;AAC7C,iBAAa;AAAA,MACX,UAAU,WAAW;AAAA,MACrB,OAAO;AAAA,IACT;AAAA,EACF;AACA,MAAI,OAAO,cAAc,QAAW;AAClC,iBAAa,aAAa,OAAO,SAAS;AAAA,EAC5C;AACA,MAAI,OAAO,qBAAqB;AAC9B,iBAAa;AAAA,MACX;AAAA,MACA;AAAA,MACA,UAAU,uBAAuB;AAAA,IACnC;AAAA,EACF;AACA,MAAI,OAAO,wBAAwB;AACjC,iBAAa;AAAA,MACX,UAAU,WAAW;AAAA,MACrB;AAAA,IACF;AAAA,EACF;AAEA,QAAM,cAAc,UAAU,YAAY,gBAAgB,MAAM;AAChE,QAAM,aACJ,OAAO,SAAS,SAAS,IACrB,8BAA8B,cAAc,aAAa,OAAO,QAAQ,IACxE,IAAI,UAAU,iBAAiB,cAAc,WAAW;AAE9D,OAAI,YAAO,eAAP,mBAAmB,QAAQ;AAC7B,UAAM,oBAAoB,UAAU,kBAAkB,eAAe,UAAU;AAC/E,eAAW,UAAU,OAAO,YAAY;AACtC,wBAAkB,UAAU,MAAM;AAAA,IACpC;AAAA,EACF;AAEA,SAAO;AACT;AAEA,SAAS,mBAAmB,QAA4C;AACtE,MAAI,OAAO,gBAAgB;AACzB,UAAMC,gBAAe,UAAU,aAAa;AAAA,MAC1C,IAAI,IAAI,OAAO,cAAc;AAAA,MAC7B,OAAO;AAAA,IACT;AACA,QAAI,OAAO,gBAAiB,CAAAA,cAAa,qBAAqB,OAAO;AACrE,IAAAA,cAAa,4BAA4B,cAAc,MAAM;AAC7D,WAAOA;AAAA,EACT;AACA,MAAI,OAAO,YAAY;AACrB,UAAMA,gBAAe,UAAU,aAAa;AAAA,MAC1C,IAAI,IAAI,OAAO,UAAU;AAAA,MACzB,OAAO;AAAA,IACT;AACA,QAAI,OAAO,gBAAiB,CAAAA,cAAa,qBAAqB,OAAO;AACrE,IAAAA,cAAa,4BAA4B,cAAc,MAAM;AAC7D,WAAOA;AAAA,EACT;AACA,MAAI,OAAO,iBAAiB;AAC1B,UAAMA,gBAAe,UAAU,aAAa;AAAA,MAC1C,OAAO;AAAA,MACP,OAAO;AAAA,IACT;AACA,IAAAA,cAAa,4BAA4B,cAAc,MAAM;AAC7D,WAAOA;AAAA,EACT;AACA,QAAM,eAAe,UAAU,aAAa;AAAA,IAC1C,OAAO;AAAA,IACP,OAAO;AAAA,EACT;AACA,eAAa,4BAA4B,cAAc,MAAM;AAC7D,SAAO;AACT;AAEA,SAAS,cAAc,QAA4B;AACjD,SAAO,OAAO,SAAS,CAAC,KAAK;AAC/B;AAEA,SAAS,8BACP,cACA,aACA,WAC4B;AAC5B,eAAa;AAAA,IACX,UAAU,WAAW;AAAA,IACrB;AAAA,EACF;AACA,QAAM,iCACJ,UAAU,+BAA+B,cAAc,SAAS;AAClE,SAAO,UAAU,iBAAiB;AAAA,IAChC;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAEA,SAAS,cAAc,OAAgC;AACrD,QAAM,OAAO,IAAI,WAAW,MAAM,KAAK,QAAQ,MAAM,KAAK,YAAY,MAAM,KAAK,UAAU;AAC3F,QAAM,SAAS,IAAI,YAAY,KAAK,UAAU;AAC9C,MAAI,WAAW,MAAM,EAAE,IAAI,IAAI;AAC/B,SAAO;AACT;AAEA,eAAe,YAAY,SAAwB,WAAkC;AACnF,MAAI;AACJ,MAAI;AACF,UAAM,QAAQ,KAAK;AAAA,MACjB;AAAA,MACA,IAAI,QAAc,CAAC,GAAG,WAAW;AAC/B,kBAAU;AAAA,UACR,MAAM,OAAO,IAAI,mBAAmB,EAAE,SAAS,qBAAqB,CAAC,CAAC;AAAA,UACtE;AAAA,QACF;AAAA,MACF,CAAC;AAAA,IACH,CAAC;AAAA,EACH,UAAE;AACA,QAAI,QAAS,cAAa,OAAO;AAAA,EACnC;AACF;","names":["stt","speechConfig"]}