fixed undefine issue when setting tts streaming channel vars (#1456)

* SoundHound support audio endpoint from speech credential

* add requestInfo and sampleRate to houndify channel variable

* add requestInfo and sampleRate to houndify channel variable

* wip

* wip

* wip

* wip

* wip

* wip

* wip

lib/config.js

@@ -119,7 +119,7 @@ const ENCRYPTION_SECRET = process.env.ENCRYPTION_SECRET;
 const HTTP_POOL = process.env.HTTP_POOL && parseInt(process.env.HTTP_POOL);
 const HTTP_POOLSIZE = parseInt(process.env.HTTP_POOLSIZE, 10) || 10;
 const HTTP_PIPELINING = parseInt(process.env.HTTP_PIPELINING, 10) || 1;
-const HTTP_TIMEOUT = 10000;
+const HTTP_TIMEOUT = parseInt(process.env.JAMBONES_HTTP_TIMEOUT, 10) || 10000;
 const HTTP_PROXY_IP = process.env.JAMBONES_HTTP_PROXY_IP;
 const HTTP_PROXY_PORT = process.env.JAMBONES_HTTP_PROXY_PORT;
 const HTTP_PROXY_PROTOCOL = process.env.JAMBONES_HTTP_PROXY_PROTOCOL || 'http';

lib/session/call-session.js

@@ -1195,7 +1195,8 @@ class CallSession extends Emitter {
             speech_credential_sid: credential.speech_credential_sid,
             client_id: credential.client_id,
             client_key: credential.client_key,
-            user_id: credential.user_id
+            user_id: credential.user_id,
+            houndify_server_uri: credential.houndify_server_uri
           };
         }
         else if ('deepgramflux' === vendor) {

lib/tasks/dial.js

@@ -576,7 +576,7 @@ class TaskDial extends Task {
       proxy: `sip:${sbcAddress}`,
       callingNumber: this.callerId || fromUri.user,
       ...(this.callerName && {callingName: this.callerName}),
-      opusFirst: isOpusFirst(this.cs.ep.remote.sdp),
+      opusFirst: isOpusFirst(this.cs.ep.local.sdp),
       isVideoCall: this.cs.ep.remote.sdp.includes('m=video')
     };
 

lib/tasks/gather.js

@@ -258,7 +258,7 @@ class TaskGather extends SttTask {
             startDtmfListener();
           }
           this._stopVad();
-          if (!this.killed) {
+          if (!this.killed && !this.resolved) {
             startListening(cs, ep);
             if (this.input.includes('speech') && this.vendor === 'nuance' && this.listenDuringPrompt) {
               this.logger.debug('Gather:exec - starting transcription timers after say completes');
@@ -296,7 +296,7 @@ class TaskGather extends SttTask {
             startDtmfListener();
           }
           this._stopVad();
-          if (!this.killed) {
+          if (!this.killed && !this.resolved) {
             startListening(cs, ep);
             if (this.input.includes('speech') && this.vendor === 'nuance' && this.listenDuringPrompt) {
               this.logger.debug('Gather:exec - starting transcription timers after play completes');
@@ -1161,7 +1161,7 @@ class TaskGather extends SttTask {
   }
 
   async _startFallback(cs, ep, evt) {
-    if (this.canFallback) {
+    if (this.canFallback()) {
       this._stopTranscribing(ep);
       try {
         this.logger.debug('gather:_startFallback');

lib/tasks/stt-task.js

@@ -171,7 +171,7 @@ class SttTask extends Task {
       try {
         this.sttCredentials = await this._initSpeechCredentials(this.cs, this.vendor, this.label);
       } catch (error) {
-        if (this.canFallback) {
+        if (this.canFallback()) {
           this.notifyError(
             {
               msg: 'ASR error', details:`Invalid vendor ${this.vendor}, Error: ${error}`,
@@ -260,8 +260,19 @@ class SttTask extends Task {
     ep.addCustomEventListener(event, handler);
   }
 
-  removeCustomEventListeners() {
-    this.eventHandlers.forEach((h) => h.ep.removeCustomEventListener(h.event, h.handler));
+  removeCustomEventListeners(ep) {
+    if (ep) {
+      // for specific endpoint
+      this.eventHandlers.filter((h) => h.ep === ep).forEach((h) => {
+        h.ep.removeCustomEventListener(h.event, h.handler);
+      });
+      this.eventHandlers = this.eventHandlers.filter((h) => h.ep !== ep);
+      return;
+    } else {
+      // for all endpoints
+      this.eventHandlers.forEach((h) => h.ep.removeCustomEventListener(h.event, h.handler));
+      this.eventHandlers = [];
+    }
   }
 
   async _initSpeechCredentials(cs, vendor, label) {
@@ -329,11 +340,13 @@ class SttTask extends Task {
     return credentials;
   }
 
-  get canFallback() {
+  canFallback() {
     return this.fallbackVendor && this.isHandledByPrimaryProvider && !this.cs.hasFallbackAsr;
   }
 
-  async _initFallback() {
+  // ep is optional for gather or any verb that have single ep,
+  // but transcribe does need as it might has 2 eps
+  async _initFallback(ep) {
     assert(this.fallbackVendor, 'fallback failed without fallbackVendor configuration');
     this.logger.info(`Failed to use primary STT provider, fallback to ${this.fallbackVendor}`);
     this.isHandledByPrimaryProvider = false;
@@ -346,7 +359,7 @@ class SttTask extends Task {
     this.data.recognizer.label = this.label;
     this.sttCredentials = await this._initSpeechCredentials(this.cs, this.vendor, this.label);
     // cleanup previous listener from previous vendor
-    this.removeCustomEventListeners();
+    this.removeCustomEventListeners(ep);
   }
 
   async compileHintsForCobalt(ep, hostport, model, token, hints) {

lib/tasks/transcribe.js

@@ -70,6 +70,9 @@ class TaskTranscribe extends SttTask {
     this._bufferedTranscripts = [ [], [] ];  // for channel 1 and 2
     this.bugname_prefix = 'transcribe_';
     this.paused = false;
+    // fallback flags
+    this.isHandledByPrimaryProviderForEp1 = true;
+    this.isHandledByPrimaryProviderForEp2 = true;
   }
 
   get name() { return TaskName.Transcribe; }
@@ -776,7 +779,7 @@ class TaskTranscribe extends SttTask {
   }
 
   async _startFallback(cs, _ep, evt) {
-    if (this.canFallback) {
+    if (this.canFallback(_ep)) {
       _ep.stopTranscription({
         vendor: this.vendor,
         bugname: this.bugname,
@@ -786,7 +789,7 @@ class TaskTranscribe extends SttTask {
       try {
         this.notifyError({ msg: 'ASR error',
           details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'in progress'});
-        await this._initFallback();
+        await this._initFallback(_ep);
         let channel = 1;
         if (this.ep !== _ep) {
           channel = 2;
@@ -895,6 +898,41 @@ class TaskTranscribe extends SttTask {
     if (this._asrTimer) clearTimeout(this._asrTimer);
     this._asrTimer = null;
   }
+
+  // We need to keep track the fallback is happened for each endpoint
+  // override the canFallback and _initFallback methods to make sure that
+  // we only fallback once per endpoint
+  // we want to keep track this on task level instead of endpoint level
+  // because the endpoint instance is used across multiple tasks.
+  canFallback(ep) {
+    let isHandledByPrimaryProvider = this.isHandledByPrimaryProvider;
+    if (ep === this.ep) {
+      isHandledByPrimaryProvider = this.isHandledByPrimaryProviderForEp1;
+    } else if (ep === this.ep2) {
+      isHandledByPrimaryProvider = this.isHandledByPrimaryProviderForEp2;
+    }
+
+    const isOneOfEndpointAlreadyFallenBack = !!this.ep && !!this.ep2 &&
+      this.isHandledByPrimaryProviderForEp1 !== this.isHandledByPrimaryProviderForEp2;
+
+    // fallback is configured
+    return this.fallbackVendor &&
+      // has this endpoint already fallen back
+      isHandledByPrimaryProvider &&
+      // in global level, is there any fallback is already happened
+      // one fallen endpoint will mark cs.hasFallbackAsr to true,
+      // so if one endpoint was fallen, the other endpoint would be able to fallback.
+      (isOneOfEndpointAlreadyFallenBack || !this.cs.hasFallbackAsr);
+  }
+
+  _initFallback(ep) {
+    if (ep === this.ep) {
+      this.isHandledByPrimaryProviderForEp1 = false;
+    } else if (ep === this.ep2) {
+      this.isHandledByPrimaryProviderForEp2 = false;
+    }
+    return super._initFallback(ep);
+  }
 }
 
 module.exports = TaskTranscribe;

lib/tasks/tts-task.js

@@ -89,8 +89,9 @@ class TtsTask extends Task {
     // api_key, model_id, api_uri, custom_tts_streaming_url, and auth_token are encoded in the credentials
     // allow them to be overriden via config, using options
     // give preference to options passed in via config
-    const local_options = {...JSON.parse(options), ...this.options};
-    const local_voice_settings = {...JSON.parse(options).voice_settings, ...this.options.voice_settings};
+    const parsed_options = options ? JSON.parse(options) : {};
+    const local_options = {...parsed_options, ...this.options};
+    const local_voice_settings = {...(parsed_options.voice_settings || {}), ...(this.options.voice_settings || {})};
     const local_api_key =  local_options.api_key ?? api_key;
     const local_model_id = local_options.model_id ?? model_id;
     const local_api_uri = local_options.api_uri ?? api_uri;

lib/utils/db-utils.js

@@ -152,6 +152,7 @@ const speechMapper = (cred) => {
       obj.client_id = o.client_id;
       obj.client_key = o.client_key;
       obj.user_id = o.user_id;
+      obj.houndify_server_uri = o.houndify_server_uri;
     }
     else if ('voxist' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));

lib/utils/http-requestor.js

@@ -191,7 +191,7 @@ class HttpRequestor extends BaseRequestor {
         method,
         headers: hdrs,
         ...('POST' === method && {body: JSON.stringify(payload)}),
-        timeout: HTTP_TIMEOUT,
+        headersTimeout: HTTP_TIMEOUT,
         followRedirects: false
       };
 

lib/utils/transcription-utils.js

@@ -920,7 +920,7 @@ module.exports = (logger) => {
         ...(rOpts.initialSpeechTimeoutMs > 0 &&
           {AZURE_INITIAL_SPEECH_TIMEOUT_MS: rOpts.initialSpeechTimeoutMs}),
         ...(rOpts.requestSnr && {AZURE_REQUEST_SNR: 1}),
-        ...(rOpts.audioLogging && {AZURE_AUDIO_LOGGING: 1}),
+        ...(azureOptions.audioLogging && {AZURE_AUDIO_LOGGING: 1}),
         ...{AZURE_USE_OUTPUT_FORMAT_DETAILED: 1},
         ...(azureOptions.speechSegmentationSilenceTimeoutMs &&
           {AZURE_SPEECH_SEGMENTATION_SILENCE_TIMEOUT_MS: azureOptions.speechSegmentationSilenceTimeoutMs}),
@@ -1226,8 +1226,10 @@ module.exports = (logger) => {
         audioFormat, enableNoiseReduction, enableProfanityFilter, enablePunctuation,
         enableCapitalization, confidenceThreshold, enableDisfluencyFilter,
         maxResults, enableWordTimestamps, maxAlternatives, partialTranscriptInterval,
-        sessionTimeout, connectionTimeout, customVocabulary, languageModel
+        sessionTimeout, connectionTimeout, customVocabulary, languageModel,
+        requestInfo, sampleRate
       } = rOpts.houndifyOptions || {};
+      const audioEndpointUri = audioEndpoint || sttCredentials.houndify_server_uri;
 
       opts = {
         ...opts,
@@ -1263,10 +1265,12 @@ module.exports = (logger) => {
         ...(country && {HOUNDIFY_COUNTRY: country}),
         ...(timeZone && {HOUNDIFY_TIMEZONE: timeZone}),
         ...(domain && {HOUNDIFY_DOMAIN: domain}),
-        ...(audioEndpoint && {HOUNDIFY_AUDIO_ENDPOINT: audioEndpoint}),
+        ...(audioEndpointUri && {HOUNDIFY_AUDIO_ENDPOINT: audioEndpointUri}),
         ...(customVocabulary && {HOUNDIFY_CUSTOM_VOCABULARY:
           Array.isArray(customVocabulary) ? customVocabulary.join(',') : customVocabulary}),
         ...(languageModel && {HOUNDIFY_LANGUAGE_MODEL: languageModel}),
+        ...(requestInfo && {HOUNDIFY_REQUEST_INFO: JSON.stringify(requestInfo)}),
+        ...(sampleRate && {HOUNDIFY_SAMPLING_RATE: sampleRate}),
       };
     }
     else if ('voxist' === vendor) {

package-lock.json

@@ -18,7 +18,7 @@
         "@jambonz/speech-utils": "^0.2.26",
         "@jambonz/stats-collector": "^0.1.10",
         "@jambonz/time-series": "^0.2.14",
-        "@jambonz/verb-specifications": "^0.0.119",
+        "@jambonz/verb-specifications": "^0.0.122",
         "@modelcontextprotocol/sdk": "^1.9.0",
         "@opentelemetry/api": "^1.8.0",
         "@opentelemetry/exporter-jaeger": "^1.23.0",
@@ -1533,9 +1533,9 @@
       }
     },
     "node_modules/@jambonz/verb-specifications": {
-      "version": "0.0.119",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.119.tgz",
-      "integrity": "sha512-AQsZ7EY2bBOjdhufKtZpdrhFobo6LBMLkuDJY058Q3qxGBtlZr3Wx9zrlVjJuNOGCEJSmMP7Gr/EtjtFFX2iTw==",
+      "version": "0.0.122",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.122.tgz",
+      "integrity": "sha512-7xqaULhKFywJ2ZuyiYt77iiJwJ+8b98Zt1X4+OqZ7Cdjhfo7S6KnR66XRVJHnekXbmfVv58kB0KWUux5TG//Sw==",
       "license": "MIT",
       "dependencies": {
         "debug": "^4.3.4",
@@ -6543,6 +6543,8 @@
     },
     "node_modules/microsoft-cognitiveservices-speech-sdk/node_modules/utf-8-validate": {
       "version": "5.0.10",
+      "resolved": "https://registry.npmjs.org/utf-8-validate/-/utf-8-validate-5.0.10.tgz",
+      "integrity": "sha512-Z6czzLq4u8fPOyx7TU6X3dvUZVvoJmxSQ+IcrlmagKhilxlhZgxPK6C5Jqbkw1IDUmFTM+cz9QDnnLTwDz/2gQ==",
       "hasInstallScript": true,
       "license": "MIT",
       "optional": true,

package.json

@@ -34,7 +34,7 @@
     "@jambonz/speech-utils": "^0.2.26",
     "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.14",
-    "@jambonz/verb-specifications": "^0.0.119",
+    "@jambonz/verb-specifications": "^0.0.122",
     "@modelcontextprotocol/sdk": "^1.9.0",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",