@livekit/agents-plugin-deepgram 1.0.19 → 1.0.21

package/dist/stt.cjs

@@ -50,6 +50,7 @@ class STT extends import_agents.stt.STT {
   #opts;
   #logger = (0, import_agents.log)();
   label = "deepgram.STT";
+  abortController = new AbortController();
   constructor(opts = defaultSTTOptions) {
     super({
       streaming: true,
@@ -89,20 +90,16 @@ class STT extends import_agents.stt.STT {
     this.#opts = { ...this.#opts, ...opts };
   }
   stream() {
-    return new SpeechStream(this, this.#opts);
+    return new SpeechStream(this, this.#opts, this.abortController);
+  }
+  async close() {
+    this.abortController.abort();
   }
 }
 class SpeechStream extends import_agents.stt.SpeechStream {
-  #opts;
-  #audioEnergyFilter;
-  #logger = (0, import_agents.log)();
-  #speaking = false;
-  #resetWS = new import_agents.Future();
-  #requestId = "";
-  #audioDurationCollector;
-  label = "deepgram.SpeechStream";
-  constructor(stt2, opts) {
+  constructor(stt2, opts, abortController) {
     super(stt2, opts.sampleRate);
+    this.abortController = abortController;
     this.#opts = opts;
     this.closed = false;
     this.#audioEnergyFilter = new import_agents.AudioEnergyFilter();
@@ -111,11 +108,19 @@ class SpeechStream extends import_agents.stt.SpeechStream {
       { duration: 5 }
     );
   }
+  #opts;
+  #audioEnergyFilter;
+  #logger = (0, import_agents.log)();
+  #speaking = false;
+  #resetWS = new import_agents.Future();
+  #requestId = "";
+  #audioDurationCollector;
+  label = "deepgram.SpeechStream";
   async run() {
     const maxRetry = 32;
     let retries = 0;
     let ws;
-    while (!this.input.closed) {
+    while (!this.input.closed && !this.closed) {
       const streamURL = new URL(API_BASE_URL_V1);
       const params = {
         model: this.#opts.model,
@@ -157,15 +162,21 @@ class SpeechStream extends import_agents.stt.SpeechStream {
         });
         await this.#runWS(ws);
       } catch (e) {
-        if (retries >= maxRetry) {
-          throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
+        if (!this.closed && !this.input.closed) {
+          if (retries >= maxRetry) {
+            throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
+          }
+          const delay = Math.min(retries * 5, 10);
+          retries++;
+          this.#logger.warn(
+            `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`
+          );
+          await new Promise((resolve) => setTimeout(resolve, delay * 1e3));
+        } else {
+          this.#logger.warn(
+            `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`
+          );
         }
-        const delay = Math.min(retries * 5, 10);
-        retries++;
-        this.#logger.warn(
-          `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`
-        );
-        await new Promise((resolve) => setTimeout(resolve, delay * 1e3));
       }
     }
     this.closed = true;
@@ -185,6 +196,17 @@ class SpeechStream extends import_agents.stt.SpeechStream {
         return;
       }
     }, 5e3);
+    const wsMonitor = import_agents.Task.from(async (controller) => {
+      const closed = new Promise(async (_, reject) => {
+        ws.once("close", (code, reason) => {
+          if (!closing) {
+            this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
+            reject(new Error("WebSocket closed"));
+          }
+        });
+      });
+      await Promise.race([closed, (0, import_agents.waitForAbort)(controller.signal)]);
+    });
     const sendTask = async () => {
       const samples100Ms = Math.floor(this.#opts.sampleRate / 10);
       const stream = new import_agents.AudioByteStream(
@@ -192,41 +214,44 @@ class SpeechStream extends import_agents.stt.SpeechStream {
         this.#opts.numChannels,
         samples100Ms
       );
-      for await (const data of this.input) {
-        let frames;
-        if (data === SpeechStream.FLUSH_SENTINEL) {
-          frames = stream.flush();
-          this.#audioDurationCollector.flush();
-        } else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
-          frames = stream.write(data.data.buffer);
-        } else {
-          throw new Error(`sample rate or channel count of frame does not match`);
-        }
-        for await (const frame of frames) {
-          if (this.#audioEnergyFilter.pushFrame(frame)) {
-            const frameDuration = frame.samplesPerChannel / frame.sampleRate;
-            this.#audioDurationCollector.push(frameDuration);
-            ws.send(frame.data.buffer);
+      try {
+        while (!this.closed) {
+          const result = await Promise.race([
+            this.input.next(),
+            (0, import_agents.waitForAbort)(this.abortController.signal)
+          ]);
+          if (result === void 0) return;
+          if (result.done) {
+            break;
+          }
+          const data = result.value;
+          let frames;
+          if (data === SpeechStream.FLUSH_SENTINEL) {
+            frames = stream.flush();
+            this.#audioDurationCollector.flush();
+          } else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
+            frames = stream.write(data.data.buffer);
+          } else {
+            throw new Error(`sample rate or channel count of frame does not match`);
+          }
+          for await (const frame of frames) {
+            if (this.#audioEnergyFilter.pushFrame(frame)) {
+              const frameDuration = frame.samplesPerChannel / frame.sampleRate;
+              this.#audioDurationCollector.push(frameDuration);
+              ws.send(frame.data.buffer);
+            }
           }
         }
+      } finally {
+        closing = true;
+        ws.send(JSON.stringify({ type: "CloseStream" }));
+        wsMonitor.cancel();
       }
-      closing = true;
-      ws.send(JSON.stringify({ type: "CloseStream" }));
     };
-    const wsMonitor = new Promise(
-      (_, reject) => ws.once("close", (code, reason) => {
-        if (!closing) {
-          this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
-          reject(new Error("WebSocket closed"));
-        }
-      })
-    );
-    const listenTask = async () => {
-      while (!this.closed && !closing) {
-        try {
-          await new Promise((resolve) => {
-            ws.once("message", (data) => resolve(data));
-          }).then((msg) => {
+    const listenTask = import_agents.Task.from(async (controller) => {
+      const listenMessage = new Promise((resolve, reject) => {
+        ws.on("message", (msg) => {
+          try {
             const json = JSON.parse(msg.toString());
             switch (json["type"]) {
               case "SpeechStarted": {
@@ -248,7 +273,9 @@ class SpeechStream extends import_agents.stt.SpeechStream {
                 if (alternatives[0] && alternatives[0].text) {
                   if (!this.#speaking) {
                     this.#speaking = true;
-                    this.queue.put({ type: import_agents.stt.SpeechEventType.START_OF_SPEECH });
+                    this.queue.put({
+                      type: import_agents.stt.SpeechEventType.START_OF_SPEECH
+                    });
                   }
                   if (isFinal) {
                     this.queue.put({
@@ -276,14 +303,21 @@ class SpeechStream extends import_agents.stt.SpeechStream {
                 break;
               }
             }
-          });
-        } catch (error) {
-          this.#logger.child({ error }).warn("unrecoverable error, exiting");
-          break;
-        }
-      }
-    };
-    await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);
+            if (this.closed || closing) {
+              resolve();
+            }
+          } catch (err) {
+            this.#logger.error(`STT: Error processing message: ${msg}`);
+            reject(err);
+          }
+        });
+      });
+      await Promise.race([listenMessage, (0, import_agents.waitForAbort)(controller.signal)]);
+    }, this.abortController);
+    await Promise.race([
+      this.#resetWS.await,
+      Promise.all([sendTask(), listenTask.result, wsMonitor])
+    ]);
     closing = true;
     ws.close();
     clearInterval(keepalive);

package/dist/stt.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n  type AudioBuffer,\n  AudioByteStream,\n  AudioEnergyFilter,\n  Future,\n  log,\n  stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n  apiKey?: string;\n  language?: STTLanguages | string;\n  detectLanguage: boolean;\n  interimResults: boolean;\n  punctuate: boolean;\n  model: STTModels;\n  smartFormat: boolean;\n  noDelay: boolean;\n  endpointing: number;\n  fillerWords: boolean;\n  sampleRate: number;\n  numChannels: number;\n  keywords: [string, number][];\n  keyterm: string[];\n  profanityFilter: boolean;\n  dictation: boolean;\n  diarize: boolean;\n  numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n  apiKey: process.env.DEEPGRAM_API_KEY,\n  language: 'en-US',\n  detectLanguage: false,\n  interimResults: true,\n  punctuate: true,\n  model: 'nova-3',\n  smartFormat: true,\n  noDelay: true,\n  endpointing: 25,\n  fillerWords: false,\n  sampleRate: 16000,\n  numChannels: 1,\n  keywords: [],\n  keyterm: [],\n  profanityFilter: false,\n  dictation: false,\n  diarize: false,\n  numerals: false,\n};\n\nexport class STT extends stt.STT {\n  #opts: STTOptions;\n  #logger = log();\n  label = 'deepgram.STT';\n\n  constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n    super({\n      streaming: true,\n      interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n    });\n    if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n      throw new Error(\n        'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n      );\n    }\n\n    this.#opts = { ...defaultSTTOptions, ...opts };\n\n    if (this.#opts.detectLanguage) {\n      this.#opts.language = undefined;\n    } else if (\n      this.#opts.language &&\n      !['en-US', 'en'].includes(this.#opts.language) &&\n      [\n        'nova-2-meeting',\n        'nova-2-phonecall',\n        'nova-2-finance',\n        'nova-2-conversationalai',\n        'nova-2-voicemail',\n        'nova-2-video',\n        'nova-2-medical',\n        'nova-2-drivethru',\n        'nova-2-automotive',\n        'nova-3-general',\n      ].includes(this.#opts.model)\n    ) {\n      this.#logger.warn(\n        `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n      );\n      this.#opts.model = 'nova-2-general';\n    }\n  }\n\n  // eslint-disable-next-line @typescript-eslint/no-unused-vars\n  async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n    throw new Error('Recognize is not supported on Deepgram STT');\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n  }\n\n  stream(): SpeechStream {\n    return new SpeechStream(this, this.#opts);\n  }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n  #opts: STTOptions;\n  #audioEnergyFilter: AudioEnergyFilter;\n  #logger = log();\n  #speaking = false;\n  #resetWS = new Future();\n  #requestId = '';\n  #audioDurationCollector: PeriodicCollector<number>;\n  label = 'deepgram.SpeechStream';\n\n  constructor(stt: STT, opts: STTOptions) {\n    super(stt, opts.sampleRate);\n    this.#opts = opts;\n    this.closed = false;\n    this.#audioEnergyFilter = new AudioEnergyFilter();\n    this.#audioDurationCollector = new PeriodicCollector(\n      (duration) => this.onAudioDurationReport(duration),\n      { duration: 5.0 },\n    );\n  }\n\n  protected async run() {\n    const maxRetry = 32;\n    let retries = 0;\n    let ws: WebSocket;\n    while (!this.input.closed) {\n      const streamURL = new URL(API_BASE_URL_V1);\n      const params = {\n        model: this.#opts.model,\n        punctuate: this.#opts.punctuate,\n        smart_format: this.#opts.smartFormat,\n        dictation: this.#opts.dictation,\n        diarize: this.#opts.diarize,\n        numerals: this.#opts.numerals,\n        no_delay: this.#opts.noDelay,\n        interim_results: this.#opts.interimResults,\n        encoding: 'linear16',\n        vad_events: true,\n        sample_rate: this.#opts.sampleRate,\n        channels: this.#opts.numChannels,\n        endpointing: this.#opts.endpointing || false,\n        filler_words: this.#opts.fillerWords,\n        keywords: this.#opts.keywords.map((x) => x.join(':')),\n        keyterm: this.#opts.keyterm,\n        profanity_filter: this.#opts.profanityFilter,\n        language: this.#opts.language,\n      };\n      Object.entries(params).forEach(([k, v]) => {\n        if (v !== undefined) {\n          if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n            streamURL.searchParams.append(k, encodeURIComponent(v));\n          } else {\n            v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n          }\n        }\n      });\n\n      ws = new WebSocket(streamURL, {\n        headers: { Authorization: `Token ${this.#opts.apiKey}` },\n      });\n\n      try {\n        await new Promise((resolve, reject) => {\n          ws.on('open', resolve);\n          ws.on('error', (error) => reject(error));\n          ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n        });\n\n        await this.#runWS(ws);\n      } catch (e) {\n        if (retries >= maxRetry) {\n          throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n        }\n\n        const delay = Math.min(retries * 5, 10);\n        retries++;\n\n        this.#logger.warn(\n          `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n        );\n        await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n      }\n    }\n\n    this.closed = true;\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n    this.#resetWS.resolve();\n  }\n\n  async #runWS(ws: WebSocket) {\n    this.#resetWS = new Future();\n    let closing = false;\n\n    const keepalive = setInterval(() => {\n      try {\n        ws.send(JSON.stringify({ type: 'KeepAlive' }));\n      } catch {\n        clearInterval(keepalive);\n        return;\n      }\n    }, 5000);\n\n    const sendTask = async () => {\n      const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n      const stream = new AudioByteStream(\n        this.#opts.sampleRate,\n        this.#opts.numChannels,\n        samples100Ms,\n      );\n\n      for await (const data of this.input) {\n        let frames: AudioFrame[];\n        if (data === SpeechStream.FLUSH_SENTINEL) {\n          frames = stream.flush();\n          this.#audioDurationCollector.flush();\n        } else if (\n          data.sampleRate === this.#opts.sampleRate ||\n          data.channels === this.#opts.numChannels\n        ) {\n          frames = stream.write(data.data.buffer);\n        } else {\n          throw new Error(`sample rate or channel count of frame does not match`);\n        }\n\n        for await (const frame of frames) {\n          if (this.#audioEnergyFilter.pushFrame(frame)) {\n            const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n            this.#audioDurationCollector.push(frameDuration);\n            ws.send(frame.data.buffer);\n          }\n        }\n      }\n\n      closing = true;\n      ws.send(JSON.stringify({ type: 'CloseStream' }));\n    };\n\n    const wsMonitor = new Promise<void>((_, reject) =>\n      ws.once('close', (code, reason) => {\n        if (!closing) {\n          this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n          reject(new Error('WebSocket closed'));\n        }\n      }),\n    );\n\n    const listenTask = async () => {\n      while (!this.closed && !closing) {\n        try {\n          await new Promise<RawData>((resolve) => {\n            ws.once('message', (data) => resolve(data));\n          }).then((msg) => {\n            const json = JSON.parse(msg.toString());\n            switch (json['type']) {\n              case 'SpeechStarted': {\n                // This is a normal case. Deepgram's SpeechStarted events\n                // are not correlated with speech_final or utterance end.\n                // It's possible that we receive two in a row without an endpoint\n                // It's also possible we receive a transcript without a SpeechStarted event.\n                if (this.#speaking) return;\n                this.#speaking = true;\n                this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n                break;\n              }\n              // see this page:\n              // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n              // for more information about the different types of events\n              case 'Results': {\n                const metadata = json['metadata'];\n                const requestId = metadata['request_id'];\n                const isFinal = json['is_final'];\n                const isEndpoint = json['speech_final'];\n                this.#requestId = requestId;\n\n                const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n                // If, for some reason, we didn't get a SpeechStarted event but we got\n                // a transcript with text, we should start speaking. It's rare but has\n                // been observed.\n                if (alternatives[0] && alternatives[0].text) {\n                  if (!this.#speaking) {\n                    this.#speaking = true;\n                    this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n                  }\n\n                  if (isFinal) {\n                    this.queue.put({\n                      type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  } else {\n                    this.queue.put({\n                      type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  }\n                }\n\n                // if we receive an endpoint, only end the speech if\n                // we either had a SpeechStarted event or we have a seen\n                // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n                if (isEndpoint && this.#speaking) {\n                  this.#speaking = false;\n                  this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n                }\n\n                break;\n              }\n              case 'Metadata': {\n                break;\n              }\n              default: {\n                this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n                break;\n              }\n            }\n          });\n        } catch (error) {\n          this.#logger.child({ error }).warn('unrecoverable error, exiting');\n          break;\n        }\n      }\n    };\n\n    await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n    closing = true;\n    ws.close();\n    clearInterval(keepalive);\n  }\n\n  private onAudioDurationReport(duration: number) {\n    const usageEvent: stt.SpeechEvent = {\n      type: stt.SpeechEventType.RECOGNITION_USAGE,\n      requestId: this.#requestId,\n      recognitionUsage: {\n        audioDuration: duration,\n      },\n    };\n    this.queue.put(usageEvent);\n  }\n}\n\nconst liveTranscriptionToSpeechData = (\n  language: STTLanguages | string,\n  data: { [id: string]: any },\n): stt.SpeechData[] => {\n  const alts: any[] = data['channel']['alternatives'];\n\n  return alts.map((alt) => ({\n    language,\n    startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n    endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n    confidence: alt['confidence'],\n    text: alt['transcript'],\n  }));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBAOO;AAEP,gBAAwC;AACxC,mBAAkC;AAGlC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,qBAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,MAAK,KAAK,UAAU;AAC1B,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,gCAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,qBAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AACtB,eAAK,wBAAwB,MAAM;AAAA,QACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,kBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,iBAAK,wBAAwB,KAAK,aAAa;AAC/C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,QACtC;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;AACf,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,kBAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
+{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n  type AudioBuffer,\n  AudioByteStream,\n  AudioEnergyFilter,\n  Future,\n  Task,\n  log,\n  stt,\n  waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n  apiKey?: string;\n  language?: STTLanguages | string;\n  detectLanguage: boolean;\n  interimResults: boolean;\n  punctuate: boolean;\n  model: STTModels;\n  smartFormat: boolean;\n  noDelay: boolean;\n  endpointing: number;\n  fillerWords: boolean;\n  sampleRate: number;\n  numChannels: number;\n  keywords: [string, number][];\n  keyterm: string[];\n  profanityFilter: boolean;\n  dictation: boolean;\n  diarize: boolean;\n  numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n  apiKey: process.env.DEEPGRAM_API_KEY,\n  language: 'en-US',\n  detectLanguage: false,\n  interimResults: true,\n  punctuate: true,\n  model: 'nova-3',\n  smartFormat: true,\n  noDelay: true,\n  endpointing: 25,\n  fillerWords: false,\n  sampleRate: 16000,\n  numChannels: 1,\n  keywords: [],\n  keyterm: [],\n  profanityFilter: false,\n  dictation: false,\n  diarize: false,\n  numerals: false,\n};\n\nexport class STT extends stt.STT {\n  #opts: STTOptions;\n  #logger = log();\n  label = 'deepgram.STT';\n  private abortController = new AbortController();\n\n  constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n    super({\n      streaming: true,\n      interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n    });\n    if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n      throw new Error(\n        'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n      );\n    }\n\n    this.#opts = { ...defaultSTTOptions, ...opts };\n\n    if (this.#opts.detectLanguage) {\n      this.#opts.language = undefined;\n    } else if (\n      this.#opts.language &&\n      !['en-US', 'en'].includes(this.#opts.language) &&\n      [\n        'nova-2-meeting',\n        'nova-2-phonecall',\n        'nova-2-finance',\n        'nova-2-conversationalai',\n        'nova-2-voicemail',\n        'nova-2-video',\n        'nova-2-medical',\n        'nova-2-drivethru',\n        'nova-2-automotive',\n        'nova-3-general',\n      ].includes(this.#opts.model)\n    ) {\n      this.#logger.warn(\n        `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n      );\n      this.#opts.model = 'nova-2-general';\n    }\n  }\n\n  // eslint-disable-next-line @typescript-eslint/no-unused-vars\n  async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n    throw new Error('Recognize is not supported on Deepgram STT');\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n  }\n\n  stream(): SpeechStream {\n    return new SpeechStream(this, this.#opts, this.abortController);\n  }\n\n  async close() {\n    this.abortController.abort();\n  }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n  #opts: STTOptions;\n  #audioEnergyFilter: AudioEnergyFilter;\n  #logger = log();\n  #speaking = false;\n  #resetWS = new Future();\n  #requestId = '';\n  #audioDurationCollector: PeriodicCollector<number>;\n  label = 'deepgram.SpeechStream';\n\n  constructor(\n    stt: STT,\n    opts: STTOptions,\n    private abortController: AbortController,\n  ) {\n    super(stt, opts.sampleRate);\n    this.#opts = opts;\n    this.closed = false;\n    this.#audioEnergyFilter = new AudioEnergyFilter();\n    this.#audioDurationCollector = new PeriodicCollector(\n      (duration) => this.onAudioDurationReport(duration),\n      { duration: 5.0 },\n    );\n  }\n\n  protected async run() {\n    const maxRetry = 32;\n    let retries = 0;\n    let ws: WebSocket;\n\n    while (!this.input.closed && !this.closed) {\n      const streamURL = new URL(API_BASE_URL_V1);\n      const params = {\n        model: this.#opts.model,\n        punctuate: this.#opts.punctuate,\n        smart_format: this.#opts.smartFormat,\n        dictation: this.#opts.dictation,\n        diarize: this.#opts.diarize,\n        numerals: this.#opts.numerals,\n        no_delay: this.#opts.noDelay,\n        interim_results: this.#opts.interimResults,\n        encoding: 'linear16',\n        vad_events: true,\n        sample_rate: this.#opts.sampleRate,\n        channels: this.#opts.numChannels,\n        endpointing: this.#opts.endpointing || false,\n        filler_words: this.#opts.fillerWords,\n        keywords: this.#opts.keywords.map((x) => x.join(':')),\n        keyterm: this.#opts.keyterm,\n        profanity_filter: this.#opts.profanityFilter,\n        language: this.#opts.language,\n      };\n      Object.entries(params).forEach(([k, v]) => {\n        if (v !== undefined) {\n          if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n            streamURL.searchParams.append(k, encodeURIComponent(v));\n          } else {\n            v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n          }\n        }\n      });\n\n      ws = new WebSocket(streamURL, {\n        headers: { Authorization: `Token ${this.#opts.apiKey}` },\n      });\n\n      try {\n        await new Promise((resolve, reject) => {\n          ws.on('open', resolve);\n          ws.on('error', (error) => reject(error));\n          ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n        });\n\n        await this.#runWS(ws);\n      } catch (e) {\n        if (!this.closed && !this.input.closed) {\n          if (retries >= maxRetry) {\n            throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n          }\n\n          const delay = Math.min(retries * 5, 10);\n          retries++;\n\n          this.#logger.warn(\n            `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n          );\n          await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n        } else {\n          this.#logger.warn(\n            `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n          );\n        }\n      }\n    }\n\n    this.closed = true;\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n    this.#resetWS.resolve();\n  }\n\n  async #runWS(ws: WebSocket) {\n    this.#resetWS = new Future();\n    let closing = false;\n\n    const keepalive = setInterval(() => {\n      try {\n        ws.send(JSON.stringify({ type: 'KeepAlive' }));\n      } catch {\n        clearInterval(keepalive);\n        return;\n      }\n    }, 5000);\n\n    // gets cancelled also when sendTask is complete\n    const wsMonitor = Task.from(async (controller) => {\n      const closed = new Promise<void>(async (_, reject) => {\n        ws.once('close', (code, reason) => {\n          if (!closing) {\n            this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n            reject(new Error('WebSocket closed'));\n          }\n        });\n      });\n\n      await Promise.race([closed, waitForAbort(controller.signal)]);\n    });\n\n    const sendTask = async () => {\n      const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n      const stream = new AudioByteStream(\n        this.#opts.sampleRate,\n        this.#opts.numChannels,\n        samples100Ms,\n      );\n\n      try {\n        while (!this.closed) {\n          const result = await Promise.race([\n            this.input.next(),\n            waitForAbort(this.abortController.signal),\n          ]);\n\n          if (result === undefined) return; // aborted\n          if (result.done) {\n            break;\n          }\n\n          const data = result.value;\n\n          let frames: AudioFrame[];\n          if (data === SpeechStream.FLUSH_SENTINEL) {\n            frames = stream.flush();\n            this.#audioDurationCollector.flush();\n          } else if (\n            data.sampleRate === this.#opts.sampleRate ||\n            data.channels === this.#opts.numChannels\n          ) {\n            frames = stream.write(data.data.buffer as ArrayBuffer);\n          } else {\n            throw new Error(`sample rate or channel count of frame does not match`);\n          }\n\n          for await (const frame of frames) {\n            if (this.#audioEnergyFilter.pushFrame(frame)) {\n              const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n              this.#audioDurationCollector.push(frameDuration);\n              ws.send(frame.data.buffer);\n            }\n          }\n        }\n      } finally {\n        closing = true;\n        ws.send(JSON.stringify({ type: 'CloseStream' }));\n        wsMonitor.cancel();\n      }\n    };\n\n    const listenTask = Task.from(async (controller) => {\n      const listenMessage = new Promise<void>((resolve, reject) => {\n        ws.on('message', (msg) => {\n          try {\n            const json = JSON.parse(msg.toString());\n            switch (json['type']) {\n              case 'SpeechStarted': {\n                // This is a normal case. Deepgram's SpeechStarted events\n                // are not correlated with speech_final or utterance end.\n                // It's possible that we receive two in a row without an endpoint\n                // It's also possible we receive a transcript without a SpeechStarted event.\n                if (this.#speaking) return;\n                this.#speaking = true;\n                this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n                break;\n              }\n              // see this page:\n              // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n              // for more information about the different types of events\n              case 'Results': {\n                const metadata = json['metadata'];\n                const requestId = metadata['request_id'];\n                const isFinal = json['is_final'];\n                const isEndpoint = json['speech_final'];\n                this.#requestId = requestId;\n\n                const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n                // If, for some reason, we didn't get a SpeechStarted event but we got\n                // a transcript with text, we should start speaking. It's rare but has\n                // been observed.\n                if (alternatives[0] && alternatives[0].text) {\n                  if (!this.#speaking) {\n                    this.#speaking = true;\n                    this.queue.put({\n                      type: stt.SpeechEventType.START_OF_SPEECH,\n                    });\n                  }\n\n                  if (isFinal) {\n                    this.queue.put({\n                      type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  } else {\n                    this.queue.put({\n                      type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  }\n                }\n\n                // if we receive an endpoint, only end the speech if\n                // we either had a SpeechStarted event or we have a seen\n                // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n                if (isEndpoint && this.#speaking) {\n                  this.#speaking = false;\n                  this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n                }\n\n                break;\n              }\n              case 'Metadata': {\n                break;\n              }\n              default: {\n                this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n                break;\n              }\n            }\n\n            if (this.closed || closing) {\n              resolve();\n            }\n          } catch (err) {\n            this.#logger.error(`STT: Error processing message: ${msg}`);\n            reject(err);\n          }\n        });\n      });\n\n      await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n    }, this.abortController);\n\n    await Promise.race([\n      this.#resetWS.await,\n      Promise.all([sendTask(), listenTask.result, wsMonitor]),\n    ]);\n    closing = true;\n    ws.close();\n    clearInterval(keepalive);\n  }\n\n  private onAudioDurationReport(duration: number) {\n    const usageEvent: stt.SpeechEvent = {\n      type: stt.SpeechEventType.RECOGNITION_USAGE,\n      requestId: this.#requestId,\n      recognitionUsage: {\n        audioDuration: duration,\n      },\n    };\n    this.queue.put(usageEvent);\n  }\n}\n\nconst liveTranscriptionToSpeechData = (\n  language: STTLanguages | string,\n  data: { [id: string]: any },\n): stt.SpeechData[] => {\n  const alts: any[] = data['channel']['alternatives'];\n\n  return alts.map((alt) => ({\n    language,\n    startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n    endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n    confidence: alt['confidence'],\n    text: alt['transcript'],\n  }));\n};\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,oBASO;AAEP,gBAA0B;AAC1B,mBAAkC;AAGlC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,kBAAI,IAAI;AAAA,EAC/B;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,KAAK,eAAe;AAAA,EAChE;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAEO,MAAM,qBAAqB,kBAAI,aAAa;AAAA,EAUjD,YACEA,MACA,MACQ,iBACR;AACA,UAAMA,MAAK,KAAK,UAAU;AAFlB;AAGR,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,gCAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAtBA;AAAA,EACA;AAAA,EACA,cAAU,mBAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,qBAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAiBR,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AAEJ,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,oBAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,UACjF;AAEA,gBAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,eAAK,QAAQ;AAAA,YACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UAC3F;AACA,gBAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,QAClE,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,gDAAgD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAChH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,qBAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAGP,UAAM,YAAY,mBAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,OAAO,GAAG,WAAW;AACpD,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,YAAQ,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,YAChC,KAAK,MAAM,KAAK;AAAA,gBAChB,4BAAa,KAAK,gBAAgB,MAAM;AAAA,UAC1C,CAAC;AAED,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,MAAM;AACf;AAAA,UACF;AAEA,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,OAAO,MAAM;AACtB,iBAAK,wBAAwB,MAAM;AAAA,UACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,qBAAS,OAAO,MAAM,KAAK,KAAK,MAAqB;AAAA,UACvD,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,oBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,mBAAK,wBAAwB,KAAK,aAAa;AAC/C,iBAAG,KAAK,MAAM,KAAK,MAAM;AAAA,YAC3B;AAAA,UACF;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAC/C,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,UAAM,aAAa,mBAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,WAAG,GAAG,WAAW,CAAC,QAAQ;AACxB,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,oBAC5B,CAAC;AAAA,kBACH;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,kBAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,kBAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,mBAAe,4BAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,GAAG,KAAK,eAAe;AAEvB,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,SAAS;AAAA,MACd,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,SAAS,CAAC;AAAA,IACxD,CAAC;AACD,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,kBAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}

package/dist/stt.d.cts

@@ -23,15 +23,18 @@ export interface STTOptions {
 export declare class STT extends stt.STT {
     #private;
     label: string;
+    private abortController;
     constructor(opts?: Partial<STTOptions>);
     _recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
     updateOptions(opts: Partial<STTOptions>): void;
     stream(): SpeechStream;
+    close(): Promise<void>;
 }
 export declare class SpeechStream extends stt.SpeechStream {
     #private;
+    private abortController;
     label: string;
-    constructor(stt: STT, opts: STTOptions);
+    constructor(stt: STT, opts: STTOptions, abortController: AbortController);
     protected run(): Promise<void>;
     updateOptions(opts: Partial<STTOptions>): void;
     private onAudioDurationReport;

package/dist/stt.d.ts

@@ -23,15 +23,18 @@ export interface STTOptions {
 export declare class STT extends stt.STT {
     #private;
     label: string;
+    private abortController;
     constructor(opts?: Partial<STTOptions>);
     _recognize(_: AudioBuffer): Promise<stt.SpeechEvent>;
     updateOptions(opts: Partial<STTOptions>): void;
     stream(): SpeechStream;
+    close(): Promise<void>;
 }
 export declare class SpeechStream extends stt.SpeechStream {
     #private;
+    private abortController;
     label: string;
-    constructor(stt: STT, opts: STTOptions);
+    constructor(stt: STT, opts: STTOptions, abortController: AbortController);
     protected run(): Promise<void>;
     updateOptions(opts: Partial<STTOptions>): void;
     private onAudioDurationReport;

package/dist/stt.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,WAAW,EAKhB,GAAG,EACJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAI3D,MAAM,WAAW,UAAU;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;IACjC,cAAc,EAAE,OAAO,CAAC;IACxB,cAAc,EAAE,OAAO,CAAC;IACxB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,WAAW,EAAE,OAAO,CAAC;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;IAC7B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,eAAe,EAAE,OAAO,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;CACnB;AAuBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAkB;gBAEX,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAuCnD,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI1D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAIvC,MAAM,IAAI,YAAY;CAGvB;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAQhD,KAAK,SAA2B;gBAEpB,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,UAAU;cAWtB,GAAG;IAkEnB,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAkJvC,OAAO,CAAC,qBAAqB;CAU9B"}
+{"version":3,"file":"stt.d.ts","sourceRoot":"","sources":["../src/stt.ts"],"names":[],"mappings":"AAGA,OAAO,EACL,KAAK,WAAW,EAMhB,GAAG,EAEJ,MAAM,iBAAiB,CAAC;AAIzB,OAAO,KAAK,EAAE,YAAY,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAI3D,MAAM,WAAW,UAAU;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,QAAQ,CAAC,EAAE,YAAY,GAAG,MAAM,CAAC;IACjC,cAAc,EAAE,OAAO,CAAC;IACxB,cAAc,EAAE,OAAO,CAAC;IACxB,SAAS,EAAE,OAAO,CAAC;IACnB,KAAK,EAAE,SAAS,CAAC;IACjB,WAAW,EAAE,OAAO,CAAC;IACrB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;IACrB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,QAAQ,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,EAAE,CAAC;IAC7B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,eAAe,EAAE,OAAO,CAAC;IACzB,SAAS,EAAE,OAAO,CAAC;IACnB,OAAO,EAAE,OAAO,CAAC;IACjB,QAAQ,EAAE,OAAO,CAAC;CACnB;AAuBD,qBAAa,GAAI,SAAQ,GAAG,CAAC,GAAG;;IAG9B,KAAK,SAAkB;IACvB,OAAO,CAAC,eAAe,CAAyB;gBAEpC,IAAI,GAAE,OAAO,CAAC,UAAU,CAAqB;IAuCnD,UAAU,CAAC,CAAC,EAAE,WAAW,GAAG,OAAO,CAAC,GAAG,CAAC,WAAW,CAAC;IAI1D,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IAIvC,MAAM,IAAI,YAAY;IAIhB,KAAK;CAGZ;AAED,qBAAa,YAAa,SAAQ,GAAG,CAAC,YAAY;;IAa9C,OAAO,CAAC,eAAe;IALzB,KAAK,SAA2B;gBAG9B,GAAG,EAAE,GAAG,EACR,IAAI,EAAE,UAAU,EACR,eAAe,EAAE,eAAe;cAY1B,GAAG;IAyEnB,aAAa,CAAC,IAAI,EAAE,OAAO,CAAC,UAAU,CAAC;IA+KvC,OAAO,CAAC,qBAAqB;CAU9B"}

package/dist/stt.js

@@ -2,8 +2,10 @@ import {
   AudioByteStream,
   AudioEnergyFilter,
   Future,
+  Task,
   log,
-  stt
+  stt,
+  waitForAbort
 } from "@livekit/agents";
 import { WebSocket } from "ws";
 import { PeriodicCollector } from "./_utils.js";
@@ -32,6 +34,7 @@ class STT extends stt.STT {
   #opts;
   #logger = log();
   label = "deepgram.STT";
+  abortController = new AbortController();
   constructor(opts = defaultSTTOptions) {
     super({
       streaming: true,
@@ -71,20 +74,16 @@ class STT extends stt.STT {
     this.#opts = { ...this.#opts, ...opts };
   }
   stream() {
-    return new SpeechStream(this, this.#opts);
+    return new SpeechStream(this, this.#opts, this.abortController);
+  }
+  async close() {
+    this.abortController.abort();
   }
 }
 class SpeechStream extends stt.SpeechStream {
-  #opts;
-  #audioEnergyFilter;
-  #logger = log();
-  #speaking = false;
-  #resetWS = new Future();
-  #requestId = "";
-  #audioDurationCollector;
-  label = "deepgram.SpeechStream";
-  constructor(stt2, opts) {
+  constructor(stt2, opts, abortController) {
     super(stt2, opts.sampleRate);
+    this.abortController = abortController;
     this.#opts = opts;
     this.closed = false;
     this.#audioEnergyFilter = new AudioEnergyFilter();
@@ -93,11 +92,19 @@ class SpeechStream extends stt.SpeechStream {
       { duration: 5 }
     );
   }
+  #opts;
+  #audioEnergyFilter;
+  #logger = log();
+  #speaking = false;
+  #resetWS = new Future();
+  #requestId = "";
+  #audioDurationCollector;
+  label = "deepgram.SpeechStream";
   async run() {
     const maxRetry = 32;
     let retries = 0;
     let ws;
-    while (!this.input.closed) {
+    while (!this.input.closed && !this.closed) {
       const streamURL = new URL(API_BASE_URL_V1);
       const params = {
         model: this.#opts.model,
@@ -139,15 +146,21 @@ class SpeechStream extends stt.SpeechStream {
         });
         await this.#runWS(ws);
       } catch (e) {
-        if (retries >= maxRetry) {
-          throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
+        if (!this.closed && !this.input.closed) {
+          if (retries >= maxRetry) {
+            throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
+          }
+          const delay = Math.min(retries * 5, 10);
+          retries++;
+          this.#logger.warn(
+            `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`
+          );
+          await new Promise((resolve) => setTimeout(resolve, delay * 1e3));
+        } else {
+          this.#logger.warn(
+            `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`
+          );
         }
-        const delay = Math.min(retries * 5, 10);
-        retries++;
-        this.#logger.warn(
-          `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`
-        );
-        await new Promise((resolve) => setTimeout(resolve, delay * 1e3));
       }
     }
     this.closed = true;
@@ -167,6 +180,17 @@ class SpeechStream extends stt.SpeechStream {
         return;
       }
     }, 5e3);
+    const wsMonitor = Task.from(async (controller) => {
+      const closed = new Promise(async (_, reject) => {
+        ws.once("close", (code, reason) => {
+          if (!closing) {
+            this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
+            reject(new Error("WebSocket closed"));
+          }
+        });
+      });
+      await Promise.race([closed, waitForAbort(controller.signal)]);
+    });
     const sendTask = async () => {
       const samples100Ms = Math.floor(this.#opts.sampleRate / 10);
       const stream = new AudioByteStream(
@@ -174,41 +198,44 @@ class SpeechStream extends stt.SpeechStream {
         this.#opts.numChannels,
         samples100Ms
       );
-      for await (const data of this.input) {
-        let frames;
-        if (data === SpeechStream.FLUSH_SENTINEL) {
-          frames = stream.flush();
-          this.#audioDurationCollector.flush();
-        } else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
-          frames = stream.write(data.data.buffer);
-        } else {
-          throw new Error(`sample rate or channel count of frame does not match`);
-        }
-        for await (const frame of frames) {
-          if (this.#audioEnergyFilter.pushFrame(frame)) {
-            const frameDuration = frame.samplesPerChannel / frame.sampleRate;
-            this.#audioDurationCollector.push(frameDuration);
-            ws.send(frame.data.buffer);
+      try {
+        while (!this.closed) {
+          const result = await Promise.race([
+            this.input.next(),
+            waitForAbort(this.abortController.signal)
+          ]);
+          if (result === void 0) return;
+          if (result.done) {
+            break;
+          }
+          const data = result.value;
+          let frames;
+          if (data === SpeechStream.FLUSH_SENTINEL) {
+            frames = stream.flush();
+            this.#audioDurationCollector.flush();
+          } else if (data.sampleRate === this.#opts.sampleRate || data.channels === this.#opts.numChannels) {
+            frames = stream.write(data.data.buffer);
+          } else {
+            throw new Error(`sample rate or channel count of frame does not match`);
+          }
+          for await (const frame of frames) {
+            if (this.#audioEnergyFilter.pushFrame(frame)) {
+              const frameDuration = frame.samplesPerChannel / frame.sampleRate;
+              this.#audioDurationCollector.push(frameDuration);
+              ws.send(frame.data.buffer);
+            }
           }
         }
+      } finally {
+        closing = true;
+        ws.send(JSON.stringify({ type: "CloseStream" }));
+        wsMonitor.cancel();
       }
-      closing = true;
-      ws.send(JSON.stringify({ type: "CloseStream" }));
     };
-    const wsMonitor = new Promise(
-      (_, reject) => ws.once("close", (code, reason) => {
-        if (!closing) {
-          this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
-          reject(new Error("WebSocket closed"));
-        }
-      })
-    );
-    const listenTask = async () => {
-      while (!this.closed && !closing) {
-        try {
-          await new Promise((resolve) => {
-            ws.once("message", (data) => resolve(data));
-          }).then((msg) => {
+    const listenTask = Task.from(async (controller) => {
+      const listenMessage = new Promise((resolve, reject) => {
+        ws.on("message", (msg) => {
+          try {
             const json = JSON.parse(msg.toString());
             switch (json["type"]) {
               case "SpeechStarted": {
@@ -230,7 +257,9 @@ class SpeechStream extends stt.SpeechStream {
                 if (alternatives[0] && alternatives[0].text) {
                   if (!this.#speaking) {
                     this.#speaking = true;
-                    this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });
+                    this.queue.put({
+                      type: stt.SpeechEventType.START_OF_SPEECH
+                    });
                   }
                   if (isFinal) {
                     this.queue.put({
@@ -258,14 +287,21 @@ class SpeechStream extends stt.SpeechStream {
                 break;
               }
             }
-          });
-        } catch (error) {
-          this.#logger.child({ error }).warn("unrecoverable error, exiting");
-          break;
-        }
-      }
-    };
-    await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);
+            if (this.closed || closing) {
+              resolve();
+            }
+          } catch (err) {
+            this.#logger.error(`STT: Error processing message: ${msg}`);
+            reject(err);
+          }
+        });
+      });
+      await Promise.race([listenMessage, waitForAbort(controller.signal)]);
+    }, this.abortController);
+    await Promise.race([
+      this.#resetWS.await,
+      Promise.all([sendTask(), listenTask.result, wsMonitor])
+    ]);
     closing = true;
     ws.close();
     clearInterval(keepalive);

package/dist/stt.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n  type AudioBuffer,\n  AudioByteStream,\n  AudioEnergyFilter,\n  Future,\n  log,\n  stt,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { type RawData, WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n  apiKey?: string;\n  language?: STTLanguages | string;\n  detectLanguage: boolean;\n  interimResults: boolean;\n  punctuate: boolean;\n  model: STTModels;\n  smartFormat: boolean;\n  noDelay: boolean;\n  endpointing: number;\n  fillerWords: boolean;\n  sampleRate: number;\n  numChannels: number;\n  keywords: [string, number][];\n  keyterm: string[];\n  profanityFilter: boolean;\n  dictation: boolean;\n  diarize: boolean;\n  numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n  apiKey: process.env.DEEPGRAM_API_KEY,\n  language: 'en-US',\n  detectLanguage: false,\n  interimResults: true,\n  punctuate: true,\n  model: 'nova-3',\n  smartFormat: true,\n  noDelay: true,\n  endpointing: 25,\n  fillerWords: false,\n  sampleRate: 16000,\n  numChannels: 1,\n  keywords: [],\n  keyterm: [],\n  profanityFilter: false,\n  dictation: false,\n  diarize: false,\n  numerals: false,\n};\n\nexport class STT extends stt.STT {\n  #opts: STTOptions;\n  #logger = log();\n  label = 'deepgram.STT';\n\n  constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n    super({\n      streaming: true,\n      interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n    });\n    if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n      throw new Error(\n        'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n      );\n    }\n\n    this.#opts = { ...defaultSTTOptions, ...opts };\n\n    if (this.#opts.detectLanguage) {\n      this.#opts.language = undefined;\n    } else if (\n      this.#opts.language &&\n      !['en-US', 'en'].includes(this.#opts.language) &&\n      [\n        'nova-2-meeting',\n        'nova-2-phonecall',\n        'nova-2-finance',\n        'nova-2-conversationalai',\n        'nova-2-voicemail',\n        'nova-2-video',\n        'nova-2-medical',\n        'nova-2-drivethru',\n        'nova-2-automotive',\n        'nova-3-general',\n      ].includes(this.#opts.model)\n    ) {\n      this.#logger.warn(\n        `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n      );\n      this.#opts.model = 'nova-2-general';\n    }\n  }\n\n  // eslint-disable-next-line @typescript-eslint/no-unused-vars\n  async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n    throw new Error('Recognize is not supported on Deepgram STT');\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n  }\n\n  stream(): SpeechStream {\n    return new SpeechStream(this, this.#opts);\n  }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n  #opts: STTOptions;\n  #audioEnergyFilter: AudioEnergyFilter;\n  #logger = log();\n  #speaking = false;\n  #resetWS = new Future();\n  #requestId = '';\n  #audioDurationCollector: PeriodicCollector<number>;\n  label = 'deepgram.SpeechStream';\n\n  constructor(stt: STT, opts: STTOptions) {\n    super(stt, opts.sampleRate);\n    this.#opts = opts;\n    this.closed = false;\n    this.#audioEnergyFilter = new AudioEnergyFilter();\n    this.#audioDurationCollector = new PeriodicCollector(\n      (duration) => this.onAudioDurationReport(duration),\n      { duration: 5.0 },\n    );\n  }\n\n  protected async run() {\n    const maxRetry = 32;\n    let retries = 0;\n    let ws: WebSocket;\n    while (!this.input.closed) {\n      const streamURL = new URL(API_BASE_URL_V1);\n      const params = {\n        model: this.#opts.model,\n        punctuate: this.#opts.punctuate,\n        smart_format: this.#opts.smartFormat,\n        dictation: this.#opts.dictation,\n        diarize: this.#opts.diarize,\n        numerals: this.#opts.numerals,\n        no_delay: this.#opts.noDelay,\n        interim_results: this.#opts.interimResults,\n        encoding: 'linear16',\n        vad_events: true,\n        sample_rate: this.#opts.sampleRate,\n        channels: this.#opts.numChannels,\n        endpointing: this.#opts.endpointing || false,\n        filler_words: this.#opts.fillerWords,\n        keywords: this.#opts.keywords.map((x) => x.join(':')),\n        keyterm: this.#opts.keyterm,\n        profanity_filter: this.#opts.profanityFilter,\n        language: this.#opts.language,\n      };\n      Object.entries(params).forEach(([k, v]) => {\n        if (v !== undefined) {\n          if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n            streamURL.searchParams.append(k, encodeURIComponent(v));\n          } else {\n            v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n          }\n        }\n      });\n\n      ws = new WebSocket(streamURL, {\n        headers: { Authorization: `Token ${this.#opts.apiKey}` },\n      });\n\n      try {\n        await new Promise((resolve, reject) => {\n          ws.on('open', resolve);\n          ws.on('error', (error) => reject(error));\n          ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n        });\n\n        await this.#runWS(ws);\n      } catch (e) {\n        if (retries >= maxRetry) {\n          throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n        }\n\n        const delay = Math.min(retries * 5, 10);\n        retries++;\n\n        this.#logger.warn(\n          `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n        );\n        await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n      }\n    }\n\n    this.closed = true;\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n    this.#resetWS.resolve();\n  }\n\n  async #runWS(ws: WebSocket) {\n    this.#resetWS = new Future();\n    let closing = false;\n\n    const keepalive = setInterval(() => {\n      try {\n        ws.send(JSON.stringify({ type: 'KeepAlive' }));\n      } catch {\n        clearInterval(keepalive);\n        return;\n      }\n    }, 5000);\n\n    const sendTask = async () => {\n      const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n      const stream = new AudioByteStream(\n        this.#opts.sampleRate,\n        this.#opts.numChannels,\n        samples100Ms,\n      );\n\n      for await (const data of this.input) {\n        let frames: AudioFrame[];\n        if (data === SpeechStream.FLUSH_SENTINEL) {\n          frames = stream.flush();\n          this.#audioDurationCollector.flush();\n        } else if (\n          data.sampleRate === this.#opts.sampleRate ||\n          data.channels === this.#opts.numChannels\n        ) {\n          frames = stream.write(data.data.buffer);\n        } else {\n          throw new Error(`sample rate or channel count of frame does not match`);\n        }\n\n        for await (const frame of frames) {\n          if (this.#audioEnergyFilter.pushFrame(frame)) {\n            const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n            this.#audioDurationCollector.push(frameDuration);\n            ws.send(frame.data.buffer);\n          }\n        }\n      }\n\n      closing = true;\n      ws.send(JSON.stringify({ type: 'CloseStream' }));\n    };\n\n    const wsMonitor = new Promise<void>((_, reject) =>\n      ws.once('close', (code, reason) => {\n        if (!closing) {\n          this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n          reject(new Error('WebSocket closed'));\n        }\n      }),\n    );\n\n    const listenTask = async () => {\n      while (!this.closed && !closing) {\n        try {\n          await new Promise<RawData>((resolve) => {\n            ws.once('message', (data) => resolve(data));\n          }).then((msg) => {\n            const json = JSON.parse(msg.toString());\n            switch (json['type']) {\n              case 'SpeechStarted': {\n                // This is a normal case. Deepgram's SpeechStarted events\n                // are not correlated with speech_final or utterance end.\n                // It's possible that we receive two in a row without an endpoint\n                // It's also possible we receive a transcript without a SpeechStarted event.\n                if (this.#speaking) return;\n                this.#speaking = true;\n                this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n                break;\n              }\n              // see this page:\n              // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n              // for more information about the different types of events\n              case 'Results': {\n                const metadata = json['metadata'];\n                const requestId = metadata['request_id'];\n                const isFinal = json['is_final'];\n                const isEndpoint = json['speech_final'];\n                this.#requestId = requestId;\n\n                const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n                // If, for some reason, we didn't get a SpeechStarted event but we got\n                // a transcript with text, we should start speaking. It's rare but has\n                // been observed.\n                if (alternatives[0] && alternatives[0].text) {\n                  if (!this.#speaking) {\n                    this.#speaking = true;\n                    this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n                  }\n\n                  if (isFinal) {\n                    this.queue.put({\n                      type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  } else {\n                    this.queue.put({\n                      type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  }\n                }\n\n                // if we receive an endpoint, only end the speech if\n                // we either had a SpeechStarted event or we have a seen\n                // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n                if (isEndpoint && this.#speaking) {\n                  this.#speaking = false;\n                  this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n                }\n\n                break;\n              }\n              case 'Metadata': {\n                break;\n              }\n              default: {\n                this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n                break;\n              }\n            }\n          });\n        } catch (error) {\n          this.#logger.child({ error }).warn('unrecoverable error, exiting');\n          break;\n        }\n      }\n    };\n\n    await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);\n    closing = true;\n    ws.close();\n    clearInterval(keepalive);\n  }\n\n  private onAudioDurationReport(duration: number) {\n    const usageEvent: stt.SpeechEvent = {\n      type: stt.SpeechEventType.RECOGNITION_USAGE,\n      requestId: this.#requestId,\n      recognitionUsage: {\n        audioDuration: duration,\n      },\n    };\n    this.queue.put(usageEvent);\n  }\n}\n\nconst liveTranscriptionToSpeechData = (\n  language: STTLanguages | string,\n  data: { [id: string]: any },\n): stt.SpeechData[] => {\n  const alts: any[] = data['channel']['alternatives'];\n\n  return alts.map((alt) => ({\n    language,\n    startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n    endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n    confidence: alt['confidence'],\n    text: alt['transcript'],\n  }));\n};\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAuB,iBAAiB;AACxC,SAAS,yBAAyB;AAGlC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EAER,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,KAAK;AAAA,EAC1C;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EACjD;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,OAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAER,YAAYA,MAAU,MAAkB;AACtC,UAAMA,MAAK,KAAK,UAAU;AAC1B,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAEA,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AACJ,WAAO,CAAC,KAAK,MAAM,QAAQ;AACzB,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,WAAW,UAAU;AACvB,gBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,QACjF;AAEA,cAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,aAAK,QAAQ;AAAA,UACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,QAC3F;AACA,cAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,MAClE;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,OAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAEP,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,uBAAiB,QAAQ,KAAK,OAAO;AACnC,YAAI;AACJ,YAAI,SAAS,aAAa,gBAAgB;AACxC,mBAAS,OAAO,MAAM;AACtB,eAAK,wBAAwB,MAAM;AAAA,QACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,mBAAS,OAAO,MAAM,KAAK,KAAK,MAAM;AAAA,QACxC,OAAO;AACL,gBAAM,IAAI,MAAM,sDAAsD;AAAA,QACxE;AAEA,yBAAiB,SAAS,QAAQ;AAChC,cAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,kBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,iBAAK,wBAAwB,KAAK,aAAa;AAC/C,eAAG,KAAK,MAAM,KAAK,MAAM;AAAA,UAC3B;AAAA,QACF;AAAA,MACF;AAEA,gBAAU;AACV,SAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAAA,IACjD;AAEA,UAAM,YAAY,IAAI;AAAA,MAAc,CAAC,GAAG,WACtC,GAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,YAAI,CAAC,SAAS;AACZ,eAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,iBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,QACtC;AAAA,MACF,CAAC;AAAA,IACH;AAEA,UAAM,aAAa,YAAY;AAC7B,aAAO,CAAC,KAAK,UAAU,CAAC,SAAS;AAC/B,YAAI;AACF,gBAAM,IAAI,QAAiB,CAAC,YAAY;AACtC,eAAG,KAAK,WAAW,CAAC,SAAS,QAAQ,IAAI,CAAC;AAAA,UAC5C,CAAC,EAAE,KAAK,CAAC,QAAQ;AACf,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAAA,kBAC9D;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAAA,UACF,CAAC;AAAA,QACH,SAAS,OAAO;AACd,eAAK,QAAQ,MAAM,EAAE,MAAM,CAAC,EAAE,KAAK,8BAA8B;AACjE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,UAAM,QAAQ,KAAK,CAAC,KAAK,SAAS,OAAO,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC,CAAC,CAAC;AAC5F,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}
+{"version":3,"sources":["../src/stt.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n  type AudioBuffer,\n  AudioByteStream,\n  AudioEnergyFilter,\n  Future,\n  Task,\n  log,\n  stt,\n  waitForAbort,\n} from '@livekit/agents';\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { WebSocket } from 'ws';\nimport { PeriodicCollector } from './_utils.js';\nimport type { STTLanguages, STTModels } from './models.js';\n\nconst API_BASE_URL_V1 = 'wss://api.deepgram.com/v1/listen';\n\nexport interface STTOptions {\n  apiKey?: string;\n  language?: STTLanguages | string;\n  detectLanguage: boolean;\n  interimResults: boolean;\n  punctuate: boolean;\n  model: STTModels;\n  smartFormat: boolean;\n  noDelay: boolean;\n  endpointing: number;\n  fillerWords: boolean;\n  sampleRate: number;\n  numChannels: number;\n  keywords: [string, number][];\n  keyterm: string[];\n  profanityFilter: boolean;\n  dictation: boolean;\n  diarize: boolean;\n  numerals: boolean;\n}\n\nconst defaultSTTOptions: STTOptions = {\n  apiKey: process.env.DEEPGRAM_API_KEY,\n  language: 'en-US',\n  detectLanguage: false,\n  interimResults: true,\n  punctuate: true,\n  model: 'nova-3',\n  smartFormat: true,\n  noDelay: true,\n  endpointing: 25,\n  fillerWords: false,\n  sampleRate: 16000,\n  numChannels: 1,\n  keywords: [],\n  keyterm: [],\n  profanityFilter: false,\n  dictation: false,\n  diarize: false,\n  numerals: false,\n};\n\nexport class STT extends stt.STT {\n  #opts: STTOptions;\n  #logger = log();\n  label = 'deepgram.STT';\n  private abortController = new AbortController();\n\n  constructor(opts: Partial<STTOptions> = defaultSTTOptions) {\n    super({\n      streaming: true,\n      interimResults: opts.interimResults ?? defaultSTTOptions.interimResults,\n    });\n    if (opts.apiKey === undefined && defaultSTTOptions.apiKey === undefined) {\n      throw new Error(\n        'Deepgram API key is required, whether as an argument or as $DEEPGRAM_API_KEY',\n      );\n    }\n\n    this.#opts = { ...defaultSTTOptions, ...opts };\n\n    if (this.#opts.detectLanguage) {\n      this.#opts.language = undefined;\n    } else if (\n      this.#opts.language &&\n      !['en-US', 'en'].includes(this.#opts.language) &&\n      [\n        'nova-2-meeting',\n        'nova-2-phonecall',\n        'nova-2-finance',\n        'nova-2-conversationalai',\n        'nova-2-voicemail',\n        'nova-2-video',\n        'nova-2-medical',\n        'nova-2-drivethru',\n        'nova-2-automotive',\n        'nova-3-general',\n      ].includes(this.#opts.model)\n    ) {\n      this.#logger.warn(\n        `${this.#opts.model} does not support language ${this.#opts.language}, falling back to nova-2-general`,\n      );\n      this.#opts.model = 'nova-2-general';\n    }\n  }\n\n  // eslint-disable-next-line @typescript-eslint/no-unused-vars\n  async _recognize(_: AudioBuffer): Promise<stt.SpeechEvent> {\n    throw new Error('Recognize is not supported on Deepgram STT');\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n  }\n\n  stream(): SpeechStream {\n    return new SpeechStream(this, this.#opts, this.abortController);\n  }\n\n  async close() {\n    this.abortController.abort();\n  }\n}\n\nexport class SpeechStream extends stt.SpeechStream {\n  #opts: STTOptions;\n  #audioEnergyFilter: AudioEnergyFilter;\n  #logger = log();\n  #speaking = false;\n  #resetWS = new Future();\n  #requestId = '';\n  #audioDurationCollector: PeriodicCollector<number>;\n  label = 'deepgram.SpeechStream';\n\n  constructor(\n    stt: STT,\n    opts: STTOptions,\n    private abortController: AbortController,\n  ) {\n    super(stt, opts.sampleRate);\n    this.#opts = opts;\n    this.closed = false;\n    this.#audioEnergyFilter = new AudioEnergyFilter();\n    this.#audioDurationCollector = new PeriodicCollector(\n      (duration) => this.onAudioDurationReport(duration),\n      { duration: 5.0 },\n    );\n  }\n\n  protected async run() {\n    const maxRetry = 32;\n    let retries = 0;\n    let ws: WebSocket;\n\n    while (!this.input.closed && !this.closed) {\n      const streamURL = new URL(API_BASE_URL_V1);\n      const params = {\n        model: this.#opts.model,\n        punctuate: this.#opts.punctuate,\n        smart_format: this.#opts.smartFormat,\n        dictation: this.#opts.dictation,\n        diarize: this.#opts.diarize,\n        numerals: this.#opts.numerals,\n        no_delay: this.#opts.noDelay,\n        interim_results: this.#opts.interimResults,\n        encoding: 'linear16',\n        vad_events: true,\n        sample_rate: this.#opts.sampleRate,\n        channels: this.#opts.numChannels,\n        endpointing: this.#opts.endpointing || false,\n        filler_words: this.#opts.fillerWords,\n        keywords: this.#opts.keywords.map((x) => x.join(':')),\n        keyterm: this.#opts.keyterm,\n        profanity_filter: this.#opts.profanityFilter,\n        language: this.#opts.language,\n      };\n      Object.entries(params).forEach(([k, v]) => {\n        if (v !== undefined) {\n          if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {\n            streamURL.searchParams.append(k, encodeURIComponent(v));\n          } else {\n            v.forEach((x) => streamURL.searchParams.append(k, encodeURIComponent(x)));\n          }\n        }\n      });\n\n      ws = new WebSocket(streamURL, {\n        headers: { Authorization: `Token ${this.#opts.apiKey}` },\n      });\n\n      try {\n        await new Promise((resolve, reject) => {\n          ws.on('open', resolve);\n          ws.on('error', (error) => reject(error));\n          ws.on('close', (code) => reject(`WebSocket returned ${code}`));\n        });\n\n        await this.#runWS(ws);\n      } catch (e) {\n        if (!this.closed && !this.input.closed) {\n          if (retries >= maxRetry) {\n            throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);\n          }\n\n          const delay = Math.min(retries * 5, 10);\n          retries++;\n\n          this.#logger.warn(\n            `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,\n          );\n          await new Promise((resolve) => setTimeout(resolve, delay * 1000));\n        } else {\n          this.#logger.warn(\n            `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,\n          );\n        }\n      }\n    }\n\n    this.closed = true;\n  }\n\n  updateOptions(opts: Partial<STTOptions>) {\n    this.#opts = { ...this.#opts, ...opts };\n    this.#resetWS.resolve();\n  }\n\n  async #runWS(ws: WebSocket) {\n    this.#resetWS = new Future();\n    let closing = false;\n\n    const keepalive = setInterval(() => {\n      try {\n        ws.send(JSON.stringify({ type: 'KeepAlive' }));\n      } catch {\n        clearInterval(keepalive);\n        return;\n      }\n    }, 5000);\n\n    // gets cancelled also when sendTask is complete\n    const wsMonitor = Task.from(async (controller) => {\n      const closed = new Promise<void>(async (_, reject) => {\n        ws.once('close', (code, reason) => {\n          if (!closing) {\n            this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);\n            reject(new Error('WebSocket closed'));\n          }\n        });\n      });\n\n      await Promise.race([closed, waitForAbort(controller.signal)]);\n    });\n\n    const sendTask = async () => {\n      const samples100Ms = Math.floor(this.#opts.sampleRate / 10);\n      const stream = new AudioByteStream(\n        this.#opts.sampleRate,\n        this.#opts.numChannels,\n        samples100Ms,\n      );\n\n      try {\n        while (!this.closed) {\n          const result = await Promise.race([\n            this.input.next(),\n            waitForAbort(this.abortController.signal),\n          ]);\n\n          if (result === undefined) return; // aborted\n          if (result.done) {\n            break;\n          }\n\n          const data = result.value;\n\n          let frames: AudioFrame[];\n          if (data === SpeechStream.FLUSH_SENTINEL) {\n            frames = stream.flush();\n            this.#audioDurationCollector.flush();\n          } else if (\n            data.sampleRate === this.#opts.sampleRate ||\n            data.channels === this.#opts.numChannels\n          ) {\n            frames = stream.write(data.data.buffer as ArrayBuffer);\n          } else {\n            throw new Error(`sample rate or channel count of frame does not match`);\n          }\n\n          for await (const frame of frames) {\n            if (this.#audioEnergyFilter.pushFrame(frame)) {\n              const frameDuration = frame.samplesPerChannel / frame.sampleRate;\n              this.#audioDurationCollector.push(frameDuration);\n              ws.send(frame.data.buffer);\n            }\n          }\n        }\n      } finally {\n        closing = true;\n        ws.send(JSON.stringify({ type: 'CloseStream' }));\n        wsMonitor.cancel();\n      }\n    };\n\n    const listenTask = Task.from(async (controller) => {\n      const listenMessage = new Promise<void>((resolve, reject) => {\n        ws.on('message', (msg) => {\n          try {\n            const json = JSON.parse(msg.toString());\n            switch (json['type']) {\n              case 'SpeechStarted': {\n                // This is a normal case. Deepgram's SpeechStarted events\n                // are not correlated with speech_final or utterance end.\n                // It's possible that we receive two in a row without an endpoint\n                // It's also possible we receive a transcript without a SpeechStarted event.\n                if (this.#speaking) return;\n                this.#speaking = true;\n                this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });\n                break;\n              }\n              // see this page:\n              // https://developers.deepgram.com/docs/understand-endpointing-interim-results#using-endpointing-speech_final\n              // for more information about the different types of events\n              case 'Results': {\n                const metadata = json['metadata'];\n                const requestId = metadata['request_id'];\n                const isFinal = json['is_final'];\n                const isEndpoint = json['speech_final'];\n                this.#requestId = requestId;\n\n                const alternatives = liveTranscriptionToSpeechData(this.#opts.language!, json);\n\n                // If, for some reason, we didn't get a SpeechStarted event but we got\n                // a transcript with text, we should start speaking. It's rare but has\n                // been observed.\n                if (alternatives[0] && alternatives[0].text) {\n                  if (!this.#speaking) {\n                    this.#speaking = true;\n                    this.queue.put({\n                      type: stt.SpeechEventType.START_OF_SPEECH,\n                    });\n                  }\n\n                  if (isFinal) {\n                    this.queue.put({\n                      type: stt.SpeechEventType.FINAL_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  } else {\n                    this.queue.put({\n                      type: stt.SpeechEventType.INTERIM_TRANSCRIPT,\n                      alternatives: [alternatives[0], ...alternatives.slice(1)],\n                    });\n                  }\n                }\n\n                // if we receive an endpoint, only end the speech if\n                // we either had a SpeechStarted event or we have a seen\n                // a non-empty transcript (deepgram doesn't have a SpeechEnded event)\n                if (isEndpoint && this.#speaking) {\n                  this.#speaking = false;\n                  this.queue.put({ type: stt.SpeechEventType.END_OF_SPEECH });\n                }\n\n                break;\n              }\n              case 'Metadata': {\n                break;\n              }\n              default: {\n                this.#logger.child({ msg: json }).warn('received unexpected message from Deepgram');\n                break;\n              }\n            }\n\n            if (this.closed || closing) {\n              resolve();\n            }\n          } catch (err) {\n            this.#logger.error(`STT: Error processing message: ${msg}`);\n            reject(err);\n          }\n        });\n      });\n\n      await Promise.race([listenMessage, waitForAbort(controller.signal)]);\n    }, this.abortController);\n\n    await Promise.race([\n      this.#resetWS.await,\n      Promise.all([sendTask(), listenTask.result, wsMonitor]),\n    ]);\n    closing = true;\n    ws.close();\n    clearInterval(keepalive);\n  }\n\n  private onAudioDurationReport(duration: number) {\n    const usageEvent: stt.SpeechEvent = {\n      type: stt.SpeechEventType.RECOGNITION_USAGE,\n      requestId: this.#requestId,\n      recognitionUsage: {\n        audioDuration: duration,\n      },\n    };\n    this.queue.put(usageEvent);\n  }\n}\n\nconst liveTranscriptionToSpeechData = (\n  language: STTLanguages | string,\n  data: { [id: string]: any },\n): stt.SpeechData[] => {\n  const alts: any[] = data['channel']['alternatives'];\n\n  return alts.map((alt) => ({\n    language,\n    startTime: alt['words'].length ? alt['words'][0]['start'] : 0,\n    endTime: alt['words'].length ? alt['words'][alt['words'].length - 1]['end'] : 0,\n    confidence: alt['confidence'],\n    text: alt['transcript'],\n  }));\n};\n"],"mappings":"AAGA;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAC1B,SAAS,yBAAyB;AAGlC,MAAM,kBAAkB;AAuBxB,MAAM,oBAAgC;AAAA,EACpC,QAAQ,QAAQ,IAAI;AAAA,EACpB,UAAU;AAAA,EACV,gBAAgB;AAAA,EAChB,gBAAgB;AAAA,EAChB,WAAW;AAAA,EACX,OAAO;AAAA,EACP,aAAa;AAAA,EACb,SAAS;AAAA,EACT,aAAa;AAAA,EACb,aAAa;AAAA,EACb,YAAY;AAAA,EACZ,aAAa;AAAA,EACb,UAAU,CAAC;AAAA,EACX,SAAS,CAAC;AAAA,EACV,iBAAiB;AAAA,EACjB,WAAW;AAAA,EACX,SAAS;AAAA,EACT,UAAU;AACZ;AAEO,MAAM,YAAY,IAAI,IAAI;AAAA,EAC/B;AAAA,EACA,UAAU,IAAI;AAAA,EACd,QAAQ;AAAA,EACA,kBAAkB,IAAI,gBAAgB;AAAA,EAE9C,YAAY,OAA4B,mBAAmB;AACzD,UAAM;AAAA,MACJ,WAAW;AAAA,MACX,gBAAgB,KAAK,kBAAkB,kBAAkB;AAAA,IAC3D,CAAC;AACD,QAAI,KAAK,WAAW,UAAa,kBAAkB,WAAW,QAAW;AACvE,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,SAAK,QAAQ,EAAE,GAAG,mBAAmB,GAAG,KAAK;AAE7C,QAAI,KAAK,MAAM,gBAAgB;AAC7B,WAAK,MAAM,WAAW;AAAA,IACxB,WACE,KAAK,MAAM,YACX,CAAC,CAAC,SAAS,IAAI,EAAE,SAAS,KAAK,MAAM,QAAQ,KAC7C;AAAA,MACE;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF,EAAE,SAAS,KAAK,MAAM,KAAK,GAC3B;AACA,WAAK,QAAQ;AAAA,QACX,GAAG,KAAK,MAAM,KAAK,8BAA8B,KAAK,MAAM,QAAQ;AAAA,MACtE;AACA,WAAK,MAAM,QAAQ;AAAA,IACrB;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,GAA0C;AACzD,UAAM,IAAI,MAAM,4CAA4C;AAAA,EAC9D;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AAAA,EACxC;AAAA,EAEA,SAAuB;AACrB,WAAO,IAAI,aAAa,MAAM,KAAK,OAAO,KAAK,eAAe;AAAA,EAChE;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AACF;AAEO,MAAM,qBAAqB,IAAI,aAAa;AAAA,EAUjD,YACEA,MACA,MACQ,iBACR;AACA,UAAMA,MAAK,KAAK,UAAU;AAFlB;AAGR,SAAK,QAAQ;AACb,SAAK,SAAS;AACd,SAAK,qBAAqB,IAAI,kBAAkB;AAChD,SAAK,0BAA0B,IAAI;AAAA,MACjC,CAAC,aAAa,KAAK,sBAAsB,QAAQ;AAAA,MACjD,EAAE,UAAU,EAAI;AAAA,IAClB;AAAA,EACF;AAAA,EAtBA;AAAA,EACA;AAAA,EACA,UAAU,IAAI;AAAA,EACd,YAAY;AAAA,EACZ,WAAW,IAAI,OAAO;AAAA,EACtB,aAAa;AAAA,EACb;AAAA,EACA,QAAQ;AAAA,EAiBR,MAAgB,MAAM;AACpB,UAAM,WAAW;AACjB,QAAI,UAAU;AACd,QAAI;AAEJ,WAAO,CAAC,KAAK,MAAM,UAAU,CAAC,KAAK,QAAQ;AACzC,YAAM,YAAY,IAAI,IAAI,eAAe;AACzC,YAAM,SAAS;AAAA,QACb,OAAO,KAAK,MAAM;AAAA,QAClB,WAAW,KAAK,MAAM;AAAA,QACtB,cAAc,KAAK,MAAM;AAAA,QACzB,WAAW,KAAK,MAAM;AAAA,QACtB,SAAS,KAAK,MAAM;AAAA,QACpB,UAAU,KAAK,MAAM;AAAA,QACrB,UAAU,KAAK,MAAM;AAAA,QACrB,iBAAiB,KAAK,MAAM;AAAA,QAC5B,UAAU;AAAA,QACV,YAAY;AAAA,QACZ,aAAa,KAAK,MAAM;AAAA,QACxB,UAAU,KAAK,MAAM;AAAA,QACrB,aAAa,KAAK,MAAM,eAAe;AAAA,QACvC,cAAc,KAAK,MAAM;AAAA,QACzB,UAAU,KAAK,MAAM,SAAS,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,CAAC;AAAA,QACpD,SAAS,KAAK,MAAM;AAAA,QACpB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,UAAU,KAAK,MAAM;AAAA,MACvB;AACA,aAAO,QAAQ,MAAM,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,MAAM;AACzC,YAAI,MAAM,QAAW;AACnB,cAAI,OAAO,MAAM,YAAY,OAAO,MAAM,YAAY,OAAO,MAAM,WAAW;AAC5E,sBAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC;AAAA,UACxD,OAAO;AACL,cAAE,QAAQ,CAAC,MAAM,UAAU,aAAa,OAAO,GAAG,mBAAmB,CAAC,CAAC,CAAC;AAAA,UAC1E;AAAA,QACF;AAAA,MACF,CAAC;AAED,WAAK,IAAI,UAAU,WAAW;AAAA,QAC5B,SAAS,EAAE,eAAe,SAAS,KAAK,MAAM,MAAM,GAAG;AAAA,MACzD,CAAC;AAED,UAAI;AACF,cAAM,IAAI,QAAQ,CAAC,SAAS,WAAW;AACrC,aAAG,GAAG,QAAQ,OAAO;AACrB,aAAG,GAAG,SAAS,CAAC,UAAU,OAAO,KAAK,CAAC;AACvC,aAAG,GAAG,SAAS,CAAC,SAAS,OAAO,sBAAsB,IAAI,EAAE,CAAC;AAAA,QAC/D,CAAC;AAED,cAAM,KAAK,OAAO,EAAE;AAAA,MACtB,SAAS,GAAG;AACV,YAAI,CAAC,KAAK,UAAU,CAAC,KAAK,MAAM,QAAQ;AACtC,cAAI,WAAW,UAAU;AACvB,kBAAM,IAAI,MAAM,uCAAuC,OAAO,cAAc,CAAC,EAAE;AAAA,UACjF;AAEA,gBAAM,QAAQ,KAAK,IAAI,UAAU,GAAG,EAAE;AACtC;AAEA,eAAK,QAAQ;AAAA,YACX,8CAA8C,KAAK,aAAa,CAAC,KAAK,OAAO,IAAI,QAAQ;AAAA,UAC3F;AACA,gBAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,QAAQ,GAAI,CAAC;AAAA,QAClE,OAAO;AACL,eAAK,QAAQ;AAAA,YACX,gDAAgD,CAAC,kBAAkB,KAAK,MAAM,MAAM,eAAe,KAAK,MAAM;AAAA,UAChH;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,cAAc,MAA2B;AACvC,SAAK,QAAQ,EAAE,GAAG,KAAK,OAAO,GAAG,KAAK;AACtC,SAAK,SAAS,QAAQ;AAAA,EACxB;AAAA,EAEA,MAAM,OAAO,IAAe;AAC1B,SAAK,WAAW,IAAI,OAAO;AAC3B,QAAI,UAAU;AAEd,UAAM,YAAY,YAAY,MAAM;AAClC,UAAI;AACF,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,YAAY,CAAC,CAAC;AAAA,MAC/C,QAAQ;AACN,sBAAc,SAAS;AACvB;AAAA,MACF;AAAA,IACF,GAAG,GAAI;AAGP,UAAM,YAAY,KAAK,KAAK,OAAO,eAAe;AAChD,YAAM,SAAS,IAAI,QAAc,OAAO,GAAG,WAAW;AACpD,WAAG,KAAK,SAAS,CAAC,MAAM,WAAW;AACjC,cAAI,CAAC,SAAS;AACZ,iBAAK,QAAQ,MAAM,8BAA8B,IAAI,KAAK,MAAM,EAAE;AAClE,mBAAO,IAAI,MAAM,kBAAkB,CAAC;AAAA,UACtC;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,QAAQ,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IAC9D,CAAC;AAED,UAAM,WAAW,YAAY;AAC3B,YAAM,eAAe,KAAK,MAAM,KAAK,MAAM,aAAa,EAAE;AAC1D,YAAM,SAAS,IAAI;AAAA,QACjB,KAAK,MAAM;AAAA,QACX,KAAK,MAAM;AAAA,QACX;AAAA,MACF;AAEA,UAAI;AACF,eAAO,CAAC,KAAK,QAAQ;AACnB,gBAAM,SAAS,MAAM,QAAQ,KAAK;AAAA,YAChC,KAAK,MAAM,KAAK;AAAA,YAChB,aAAa,KAAK,gBAAgB,MAAM;AAAA,UAC1C,CAAC;AAED,cAAI,WAAW,OAAW;AAC1B,cAAI,OAAO,MAAM;AACf;AAAA,UACF;AAEA,gBAAM,OAAO,OAAO;AAEpB,cAAI;AACJ,cAAI,SAAS,aAAa,gBAAgB;AACxC,qBAAS,OAAO,MAAM;AACtB,iBAAK,wBAAwB,MAAM;AAAA,UACrC,WACE,KAAK,eAAe,KAAK,MAAM,cAC/B,KAAK,aAAa,KAAK,MAAM,aAC7B;AACA,qBAAS,OAAO,MAAM,KAAK,KAAK,MAAqB;AAAA,UACvD,OAAO;AACL,kBAAM,IAAI,MAAM,sDAAsD;AAAA,UACxE;AAEA,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,KAAK,mBAAmB,UAAU,KAAK,GAAG;AAC5C,oBAAM,gBAAgB,MAAM,oBAAoB,MAAM;AACtD,mBAAK,wBAAwB,KAAK,aAAa;AAC/C,iBAAG,KAAK,MAAM,KAAK,MAAM;AAAA,YAC3B;AAAA,UACF;AAAA,QACF;AAAA,MACF,UAAE;AACA,kBAAU;AACV,WAAG,KAAK,KAAK,UAAU,EAAE,MAAM,cAAc,CAAC,CAAC;AAC/C,kBAAU,OAAO;AAAA,MACnB;AAAA,IACF;AAEA,UAAM,aAAa,KAAK,KAAK,OAAO,eAAe;AACjD,YAAM,gBAAgB,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3D,WAAG,GAAG,WAAW,CAAC,QAAQ;AACxB,cAAI;AACF,kBAAM,OAAO,KAAK,MAAM,IAAI,SAAS,CAAC;AACtC,oBAAQ,KAAK,MAAM,GAAG;AAAA,cACpB,KAAK,iBAAiB;AAKpB,oBAAI,KAAK,UAAW;AACpB,qBAAK,YAAY;AACjB,qBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,gBAAgB,CAAC;AAC5D;AAAA,cACF;AAAA;AAAA;AAAA;AAAA,cAIA,KAAK,WAAW;AACd,sBAAM,WAAW,KAAK,UAAU;AAChC,sBAAM,YAAY,SAAS,YAAY;AACvC,sBAAM,UAAU,KAAK,UAAU;AAC/B,sBAAM,aAAa,KAAK,cAAc;AACtC,qBAAK,aAAa;AAElB,sBAAM,eAAe,8BAA8B,KAAK,MAAM,UAAW,IAAI;AAK7E,oBAAI,aAAa,CAAC,KAAK,aAAa,CAAC,EAAE,MAAM;AAC3C,sBAAI,CAAC,KAAK,WAAW;AACnB,yBAAK,YAAY;AACjB,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,oBAC5B,CAAC;AAAA,kBACH;AAEA,sBAAI,SAAS;AACX,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH,OAAO;AACL,yBAAK,MAAM,IAAI;AAAA,sBACb,MAAM,IAAI,gBAAgB;AAAA,sBAC1B,cAAc,CAAC,aAAa,CAAC,GAAG,GAAG,aAAa,MAAM,CAAC,CAAC;AAAA,oBAC1D,CAAC;AAAA,kBACH;AAAA,gBACF;AAKA,oBAAI,cAAc,KAAK,WAAW;AAChC,uBAAK,YAAY;AACjB,uBAAK,MAAM,IAAI,EAAE,MAAM,IAAI,gBAAgB,cAAc,CAAC;AAAA,gBAC5D;AAEA;AAAA,cACF;AAAA,cACA,KAAK,YAAY;AACf;AAAA,cACF;AAAA,cACA,SAAS;AACP,qBAAK,QAAQ,MAAM,EAAE,KAAK,KAAK,CAAC,EAAE,KAAK,2CAA2C;AAClF;AAAA,cACF;AAAA,YACF;AAEA,gBAAI,KAAK,UAAU,SAAS;AAC1B,sBAAQ;AAAA,YACV;AAAA,UACF,SAAS,KAAK;AACZ,iBAAK,QAAQ,MAAM,kCAAkC,GAAG,EAAE;AAC1D,mBAAO,GAAG;AAAA,UACZ;AAAA,QACF,CAAC;AAAA,MACH,CAAC;AAED,YAAM,QAAQ,KAAK,CAAC,eAAe,aAAa,WAAW,MAAM,CAAC,CAAC;AAAA,IACrE,GAAG,KAAK,eAAe;AAEvB,UAAM,QAAQ,KAAK;AAAA,MACjB,KAAK,SAAS;AAAA,MACd,QAAQ,IAAI,CAAC,SAAS,GAAG,WAAW,QAAQ,SAAS,CAAC;AAAA,IACxD,CAAC;AACD,cAAU;AACV,OAAG,MAAM;AACT,kBAAc,SAAS;AAAA,EACzB;AAAA,EAEQ,sBAAsB,UAAkB;AAC9C,UAAM,aAA8B;AAAA,MAClC,MAAM,IAAI,gBAAgB;AAAA,MAC1B,WAAW,KAAK;AAAA,MAChB,kBAAkB;AAAA,QAChB,eAAe;AAAA,MACjB;AAAA,IACF;AACA,SAAK,MAAM,IAAI,UAAU;AAAA,EAC3B;AACF;AAEA,MAAM,gCAAgC,CACpC,UACA,SACqB;AACrB,QAAM,OAAc,KAAK,SAAS,EAAE,cAAc;AAElD,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB;AAAA,IACA,WAAW,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,CAAC,EAAE,OAAO,IAAI;AAAA,IAC5D,SAAS,IAAI,OAAO,EAAE,SAAS,IAAI,OAAO,EAAE,IAAI,OAAO,EAAE,SAAS,CAAC,EAAE,KAAK,IAAI;AAAA,IAC9E,YAAY,IAAI,YAAY;AAAA,IAC5B,MAAM,IAAI,YAAY;AAAA,EACxB,EAAE;AACJ;","names":["stt"]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@livekit/agents-plugin-deepgram",
-  "version": "1.0.19",
+  "version": "1.0.21",
   "description": "Deepgram plugin for LiveKit Agents for Node.js",
   "main": "dist/index.js",
   "require": "dist/index.cjs",
@@ -30,16 +30,16 @@
     "@types/ws": "^8.5.10",
     "tsup": "^8.3.5",
     "typescript": "^5.0.0",
-    "@livekit/agents": "1.0.19",
-    "@livekit/agents-plugin-silero": "1.0.19",
-    "@livekit/agents-plugins-test": "1.0.19"
+    "@livekit/agents": "1.0.21",
+    "@livekit/agents-plugin-silero": "1.0.21",
+    "@livekit/agents-plugins-test": "1.0.21"
   },
   "dependencies": {
     "ws": "^8.16.0"
   },
   "peerDependencies": {
     "@livekit/rtc-node": "^0.13.12",
-    "@livekit/agents": "1.0.19"
+    "@livekit/agents": "1.0.21"
   },
   "scripts": {
     "build": "tsup --onSuccess \"pnpm build:types\"",

package/src/stt.ts

@@ -6,11 +6,13 @@ import {
   AudioByteStream,
   AudioEnergyFilter,
   Future,
+  Task,
   log,
   stt,
+  waitForAbort,
 } from '@livekit/agents';
 import type { AudioFrame } from '@livekit/rtc-node';
-import { type RawData, WebSocket } from 'ws';
+import { WebSocket } from 'ws';
 import { PeriodicCollector } from './_utils.js';
 import type { STTLanguages, STTModels } from './models.js';
 
@@ -62,6 +64,7 @@ export class STT extends stt.STT {
   #opts: STTOptions;
   #logger = log();
   label = 'deepgram.STT';
+  private abortController = new AbortController();
 
   constructor(opts: Partial<STTOptions> = defaultSTTOptions) {
     super({
@@ -111,7 +114,11 @@ export class STT extends stt.STT {
   }
 
   stream(): SpeechStream {
-    return new SpeechStream(this, this.#opts);
+    return new SpeechStream(this, this.#opts, this.abortController);
+  }
+
+  async close() {
+    this.abortController.abort();
   }
 }
 
@@ -125,7 +132,11 @@ export class SpeechStream extends stt.SpeechStream {
   #audioDurationCollector: PeriodicCollector<number>;
   label = 'deepgram.SpeechStream';
 
-  constructor(stt: STT, opts: STTOptions) {
+  constructor(
+    stt: STT,
+    opts: STTOptions,
+    private abortController: AbortController,
+  ) {
     super(stt, opts.sampleRate);
     this.#opts = opts;
     this.closed = false;
@@ -140,7 +151,8 @@ export class SpeechStream extends stt.SpeechStream {
     const maxRetry = 32;
     let retries = 0;
     let ws: WebSocket;
-    while (!this.input.closed) {
+
+    while (!this.input.closed && !this.closed) {
       const streamURL = new URL(API_BASE_URL_V1);
       const params = {
         model: this.#opts.model,
@@ -185,17 +197,23 @@ export class SpeechStream extends stt.SpeechStream {
 
         await this.#runWS(ws);
       } catch (e) {
-        if (retries >= maxRetry) {
-          throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
-        }
+        if (!this.closed && !this.input.closed) {
+          if (retries >= maxRetry) {
+            throw new Error(`failed to connect to Deepgram after ${retries} attempts: ${e}`);
+          }
 
-        const delay = Math.min(retries * 5, 10);
-        retries++;
+          const delay = Math.min(retries * 5, 10);
+          retries++;
 
-        this.#logger.warn(
-          `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,
-        );
-        await new Promise((resolve) => setTimeout(resolve, delay * 1000));
+          this.#logger.warn(
+            `failed to connect to Deepgram, retrying in ${delay} seconds: ${e} (${retries}/${maxRetry})`,
+          );
+          await new Promise((resolve) => setTimeout(resolve, delay * 1000));
+        } else {
+          this.#logger.warn(
+            `Deepgram disconnected, connection is closed: ${e} (inputClosed: ${this.input.closed}, isClosed: ${this.closed})`,
+          );
+        }
       }
     }
 
@@ -220,6 +238,20 @@ export class SpeechStream extends stt.SpeechStream {
       }
     }, 5000);
 
+    // gets cancelled also when sendTask is complete
+    const wsMonitor = Task.from(async (controller) => {
+      const closed = new Promise<void>(async (_, reject) => {
+        ws.once('close', (code, reason) => {
+          if (!closing) {
+            this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
+            reject(new Error('WebSocket closed'));
+          }
+        });
+      });
+
+      await Promise.race([closed, waitForAbort(controller.signal)]);
+    });
+
     const sendTask = async () => {
       const samples100Ms = Math.floor(this.#opts.sampleRate / 10);
       const stream = new AudioByteStream(
@@ -228,48 +260,52 @@ export class SpeechStream extends stt.SpeechStream {
         samples100Ms,
       );
 
-      for await (const data of this.input) {
-        let frames: AudioFrame[];
-        if (data === SpeechStream.FLUSH_SENTINEL) {
-          frames = stream.flush();
-          this.#audioDurationCollector.flush();
-        } else if (
-          data.sampleRate === this.#opts.sampleRate ||
-          data.channels === this.#opts.numChannels
-        ) {
-          frames = stream.write(data.data.buffer);
-        } else {
-          throw new Error(`sample rate or channel count of frame does not match`);
-        }
+      try {
+        while (!this.closed) {
+          const result = await Promise.race([
+            this.input.next(),
+            waitForAbort(this.abortController.signal),
+          ]);
+
+          if (result === undefined) return; // aborted
+          if (result.done) {
+            break;
+          }
+
+          const data = result.value;
+
+          let frames: AudioFrame[];
+          if (data === SpeechStream.FLUSH_SENTINEL) {
+            frames = stream.flush();
+            this.#audioDurationCollector.flush();
+          } else if (
+            data.sampleRate === this.#opts.sampleRate ||
+            data.channels === this.#opts.numChannels
+          ) {
+            frames = stream.write(data.data.buffer as ArrayBuffer);
+          } else {
+            throw new Error(`sample rate or channel count of frame does not match`);
+          }
 
-        for await (const frame of frames) {
-          if (this.#audioEnergyFilter.pushFrame(frame)) {
-            const frameDuration = frame.samplesPerChannel / frame.sampleRate;
-            this.#audioDurationCollector.push(frameDuration);
-            ws.send(frame.data.buffer);
+          for await (const frame of frames) {
+            if (this.#audioEnergyFilter.pushFrame(frame)) {
+              const frameDuration = frame.samplesPerChannel / frame.sampleRate;
+              this.#audioDurationCollector.push(frameDuration);
+              ws.send(frame.data.buffer);
+            }
           }
         }
+      } finally {
+        closing = true;
+        ws.send(JSON.stringify({ type: 'CloseStream' }));
+        wsMonitor.cancel();
       }
-
-      closing = true;
-      ws.send(JSON.stringify({ type: 'CloseStream' }));
     };
 
-    const wsMonitor = new Promise<void>((_, reject) =>
-      ws.once('close', (code, reason) => {
-        if (!closing) {
-          this.#logger.error(`WebSocket closed with code ${code}: ${reason}`);
-          reject(new Error('WebSocket closed'));
-        }
-      }),
-    );
-
-    const listenTask = async () => {
-      while (!this.closed && !closing) {
-        try {
-          await new Promise<RawData>((resolve) => {
-            ws.once('message', (data) => resolve(data));
-          }).then((msg) => {
+    const listenTask = Task.from(async (controller) => {
+      const listenMessage = new Promise<void>((resolve, reject) => {
+        ws.on('message', (msg) => {
+          try {
             const json = JSON.parse(msg.toString());
             switch (json['type']) {
               case 'SpeechStarted': {
@@ -300,7 +336,9 @@ export class SpeechStream extends stt.SpeechStream {
                 if (alternatives[0] && alternatives[0].text) {
                   if (!this.#speaking) {
                     this.#speaking = true;
-                    this.queue.put({ type: stt.SpeechEventType.START_OF_SPEECH });
+                    this.queue.put({
+                      type: stt.SpeechEventType.START_OF_SPEECH,
+                    });
                   }
 
                   if (isFinal) {
@@ -334,15 +372,24 @@ export class SpeechStream extends stt.SpeechStream {
                 break;
               }
             }
-          });
-        } catch (error) {
-          this.#logger.child({ error }).warn('unrecoverable error, exiting');
-          break;
-        }
-      }
-    };
 
-    await Promise.race([this.#resetWS.await, Promise.all([sendTask(), listenTask(), wsMonitor])]);
+            if (this.closed || closing) {
+              resolve();
+            }
+          } catch (err) {
+            this.#logger.error(`STT: Error processing message: ${msg}`);
+            reject(err);
+          }
+        });
+      });
+
+      await Promise.race([listenMessage, waitForAbort(controller.signal)]);
+    }, this.abortController);
+
+    await Promise.race([
+      this.#resetWS.await,
+      Promise.all([sendTask(), listenTask.result, wsMonitor]),
+    ]);
     closing = true;
     ws.close();
     clearInterval(keepalive);