fix onholdHOok (#540)

* fix onholdHOok

* wip

* wip

* wip

* wip

* adding more debug log

* wip

* wip

* wip

lib/config.js

@@ -120,6 +120,7 @@ const HTTP_TIMEOUT = 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';
+const HTTP_USER_AGENT_HEADER = process.env.JAMBONES_HTTP_USER_AGENT_HEADER || 'jambonz';
 
 const OPTIONS_PING_INTERVAL = parseInt(process.env.OPTIONS_PING_INTERVAL, 10) || 30000;
 
@@ -129,6 +130,8 @@ const JAMBONZ_RECORD_WS_PASSWORD = process.env.JAMBONZ_RECORD_WS_PASSWORD || pro
 const JAMBONZ_DISABLE_DIAL_PAI_HEADER = process.env.JAMBONZ_DISABLE_DIAL_PAI_HEADER || false;
 const JAMBONES_DISABLE_DIRECT_P2P_CALL = process.env.JAMBONES_DISABLE_DIRECT_P2P_CALL || false;
 
+const JAMBONES_EAGERLY_PRE_CACHE_AUDIO = process.env.JAMBONES_EAGERLY_PRE_CACHE_AUDIO;
+
 module.exports = {
   JAMBONES_MYSQL_HOST,
   JAMBONES_MYSQL_USER,
@@ -151,6 +154,7 @@ module.exports = {
   JAMBONES_API_BASE_URL,
   JAMBONES_TIME_SERIES_HOST,
   JAMBONES_INJECT_CONTENT,
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   JAMBONES_ESL_LISTEN_ADDRESS,
   JAMBONES_SBCS,
   JAMBONES_OTEL_ENABLED,
@@ -193,6 +197,7 @@ module.exports = {
   HTTP_PROXY_IP,
   HTTP_PROXY_PORT,
   HTTP_PROXY_PROTOCOL,
+  HTTP_USER_AGENT_HEADER,
   OPTIONS_PING_INTERVAL,
   RESPONSE_TIMEOUT_MS,
   JAMBONES_WS_HANDSHAKE_TIMEOUT_MS,

lib/middleware.js

@@ -382,7 +382,7 @@ module.exports = function(srf, logger) {
           },
           recognizer: {
             vendor: app.speech_recognizer_vendor,
-            ...(app.speech_synthesis_label && {label: app.speech_synthesis_label}),
+            ...(app.speech_recognizer_label && {label: app.speech_recognizer_label}),
             language: app.speech_recognizer_language,
             ...(app.fallback_speech_recognizer_vendor && {fallback_vendor: app.fallback_speech_recognizer_vendor}),
             ...(app.fallback_speech_recognizer_label && {fallback_label: app.fallback_speech_recognizer_label}),

lib/session/call-session.js

@@ -19,6 +19,7 @@ const HttpRequestor = require('../utils/http-requestor');
 const WsRequestor = require('../utils/ws-requestor');
 const {
   JAMBONES_INJECT_CONTENT,
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   AWS_REGION,
 } = require('../config');
 const BackgroundTaskManager = require('../utils/background-task-manager');
@@ -1330,6 +1331,35 @@ Duration=${duration} `
     this.taskIdx = 0;
   }
 
+  _preCacheAudio(newTasks) {
+    for (const task of newTasks) {
+      if (task.name === TaskName.Config && task.hasSynthesizer) {
+        /* if they change synthesizer settings don't try to precache */
+        break;
+      }
+      if (task.name === TaskName.Say) {
+        /* identify vendor language, voice, and label */
+        const vendor = task.synthesizer.vendor && task.synthesizer.vendor !== 'default' ?
+          task.synthesizer.vendor :
+          this.speechSynthesisVendor;
+        const language = task.synthesizer.language && task.synthesizer.language !== 'default' ?
+          task.synthesizer.language :
+          this.speechSynthesisLanguage ;
+        const voice =  task.synthesizer.voice && task.synthesizer.voice !== 'default' ?
+          task.synthesizer.voice :
+          this.speechSynthesisVoice;
+        const label = task.synthesizer.label && task.synthesizer.label !== 'default' ?
+          task.synthesizer.label :
+          this.speechSynthesisLabel;
+
+        this.logger.info({vendor, language, voice, label},
+          'CallSession:_preCacheAudio - precaching audio for future prompt');
+        task._synthesizeWithSpecificVendor(this, this.ep, {vendor, language, voice, label, preCache: true})
+          .catch((err) => this.logger.error(err, 'CallSession:_preCacheAudio - error precaching audio'));
+      }
+    }
+  }
+
   /**
    * Append tasks to the current execution stack UNLESS there is a gather in the stack.
    * in that case, insert the tasks before the gather AND if the tasks include
@@ -1387,10 +1417,12 @@ Duration=${duration} `
             this.replaceApplication(t);
           }
           else if (JAMBONES_INJECT_CONTENT) {
+            if (JAMBONES_EAGERLY_PRE_CACHE_AUDIO) this._preCacheAudio(t);
             this._injectTasks(t);
             this.logger.info({tasks: listTaskNames(this.tasks)}, 'CallSession:_onCommand - updated task list');
           }
           else {
+            if (JAMBONES_EAGERLY_PRE_CACHE_AUDIO) this._preCacheAudio(t);
             this.tasks.push(...t);
             this.logger.info({tasks: listTaskNames(this.tasks)}, 'CallSession:_onCommand - updated task list');
           }
@@ -1712,7 +1744,7 @@ Duration=${duration} `
           res.send(200, {body: this.ep.local.sdp});
         }
         else {
-          if (this.currentTask.name === TaskName.Dial && this.currentTask.isOnHold) {
+          if (this.currentTask.name === TaskName.Dial && this.currentTask.isOnHoldEnabled) {
             this.logger.info('onholdMusic reINVITE after media has been released');
             await this.currentTask.handleReinviteAfterMediaReleased(req, res);
           } else {

lib/tasks/dial.js

@@ -138,13 +138,14 @@ class TaskDial extends Task {
 
   get name() { return TaskName.Dial; }
 
-  get isOnHold() {
-    return this.isIncomingLegHold || this.isOutgoingLegHold;
+  get isOnHoldEnabled() {
+    return !!this.data.onHoldHook;
   }
 
   get canReleaseMedia() {
     const keepAnchor = this.data.anchorMedia ||
     this.cs.isBackGroundListen ||
+    this.cs.onHoldMusic ||
     ANCHOR_MEDIA_ALWAYS ||
     this.listenTask ||
     this.transcribeTask ||
@@ -570,7 +571,8 @@ class TaskDial extends Task {
           accountInfo: cs.accountInfo,
           rootSpan: cs.rootSpan,
           startSpan: this.startSpan.bind(this),
-          dialTask: this
+          dialTask: this,
+          onHoldMusic: this.cs.onHoldMusic
         });
         this.dials.set(sd.callSid, sd);
 
@@ -679,22 +681,43 @@ class TaskDial extends Task {
   async _onReinvite(req, res) {
     try {
       let isHandled = false;
-      if (this.cs.onHoldMusic) {
-        if (isOnhold(req.body) && !this.epOther && !this.ep) {
-          await this.cs.handleReinviteAfterMediaReleased(req, res);
-          // Onhold but media is already released
-          // reconnect A Leg and Response B leg
-          await this.reAnchorMedia(this.cs, this.sd);
-          this.isOutgoingLegHold = true;
+      if (this.isOnHoldEnabled) {
+        if (isOnhold(req.body)) {
+          this.logger.debug('Dial: _onReinvite receive hold Request');
+          if (!this.epOther && !this.ep) {
+            this.logger.debug(`Dial: _onReinvite receive hold Request,
+ media already released, reconnect media server`);
+            // update caller leg for new SDP from callee.
+            await this.cs.handleReinviteAfterMediaReleased(req, res);
+            // Freeswitch media is released, reconnect
+            await this.reAnchorMedia(this.cs, this.sd);
+            this.isOutgoingLegHold = true;
+          } else {
+            this.logger.debug('Dial: _onReinvite receive hold Request, update SDP');
+            const newSdp = await this.ep.modify(req.body);
+            res.send(200, {body: newSdp});
+          }
           isHandled = true;
-          this._onHoldHook();
-        } else if (!isOnhold(req.body) && this.epOther && this.ep && this.isOutgoingLegHold && this.canReleaseMedia) {
-          // Offhold, time to release media
-          const newSdp = await this.ep.modify(req.body);
-          await res.send(200, {body: newSdp});
-          await this._releaseMedia(this.cs, this.sd);
+          // Media already connected, ask for onHoldHook
+          this._onHoldHook(req);
+        } else if (!isOnhold(req.body)) {
+          this.logger.debug('Dial: _onReinvite receive unhold Request');
+          if (this.epOther && this.ep && this.isOutgoingLegHold && this.canReleaseMedia) {
+            this.logger.debug('Dial: _onReinvite receive unhold Request, release media');
+            // Offhold, time to release media
+            const newSdp = await this.ep.modify(req.body);
+            await res.send(200, {body: newSdp});
+            await this._releaseMedia(this.cs, this.sd);
+            this.isOutgoingLegHold = false;
+          } else {
+            this.logger.debug('Dial: _onReinvite receive unhold Request, update media server');
+            const newSdp = await this.ep.modify(req.body);
+            res.send(200, {body: newSdp});
+          }
+          if (this._onHoldSession) {
+            this._onHoldSession.kill();
+          }
           isHandled = true;
-          this.isOutgoingLegHold = false;
         }
       }
       if (!isHandled) {
@@ -811,23 +834,34 @@ class TaskDial extends Task {
     this.epOther = cs.ep;
   }
 
+  // Handle RE-INVITE hold from caller leg.
   async handleReinviteAfterMediaReleased(req, res) {
     let isHandled = false;
-    if (isOnhold(req.body) && !this.epOther && !this.ep) {
-      const sdp = await this.dlg.modify(req.body);
-      res.send(200, {body: sdp});
-      // Onhold but media is already released
-      await this.reAnchorMedia(this.cs, this.sd);
-      isHandled = true;
-      this.isIncomingLegHold = true;
-      this._onHoldHook();
-    } else if (!isOnhold(req.body) && this.epOther && this.ep && this.isIncomingLegHold && this.canReleaseMedia) {
-      // Offhold, time to release media
-      const newSdp = await this.epOther.modify(req.body);
-      await res.send(200, {body: newSdp});
-      await this._releaseMedia(this.cs, this.sd);
-      isHandled = true;
-      this.isIncomingLegHold = false;
+    if (this.isOnHoldEnabled) {
+      if (isOnhold(req.body)) {
+        if (!this.epOther && !this.ep) {
+          // update callee leg for new SDP from caller.
+          const sdp = await this.dlg.modify(req.body);
+          res.send(200, {body: sdp});
+          // Onhold but media is already released, reconnect
+          await this.reAnchorMedia(this.cs, this.sd);
+          isHandled = true;
+          this.isIncomingLegHold = true;
+        }
+        this._onHoldHook(req);
+      } else if (!isOnhold(req.body)) {
+        if (this.epOther && this.ep && this.isIncomingLegHold && this.canReleaseMedia) {
+          // Offhold, time to release media
+          const newSdp = await this.epOther.modify(req.body);
+          await res.send(200, {body: newSdp});
+          await this._releaseMedia(this.cs, this.sd);
+          isHandled = true;
+        }
+        this.isIncomingLegHold = false;
+        if (this._onHoldSession) {
+          this._onHoldSession.kill();
+        }
+      }
     }
 
     if (!isHandled) {
@@ -846,7 +880,7 @@ class TaskDial extends Task {
       });
   }
 
-  async _onHoldHook(allowed = [TaskName.Play, TaskName.Say, TaskName.Pause]) {
+  async _onHoldHook(req, allowed = [TaskName.Play, TaskName.Say, TaskName.Pause]) {
     if (this.data.onHoldHook) {
       // send silence for keep Voice quality
       await this.epOther.play('silence_stream://500');
@@ -856,7 +890,13 @@ class TaskDial extends Task {
           const b3 = this.getTracingPropagation();
           const httpHeaders = b3 && {b3};
           const json = await this.cs.application.requestor.
-            request('verb:hook', this.data.onHoldHook, this.cs.callInfo.toJSON(), httpHeaders);
+            request('verb:hook', this.data.onHoldHook, {
+              ...this.cs.callInfo.toJSON(),
+              hold_detail: {
+                from: req.get('From'),
+                to: req.get('To')
+              }
+            }, httpHeaders);
           const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
           allowedTasks = tasks.filter((t) => allowed.includes(t.name));
           if (tasks.length !== allowedTasks.length) {
@@ -865,7 +905,7 @@ class TaskDial extends Task {
           }
           this.logger.debug(`DialTask:_onHoldHook: executing ${tasks.length} tasks`);
           if (tasks.length) {
-            this._playSession = new ConfirmCallSession({
+            this._onHoldSession = new ConfirmCallSession({
               logger: this.logger,
               application: this.cs.application,
               dlg: this.isIncomingLegHold ? this.dlg : this.cs.dlg,
@@ -875,12 +915,12 @@ class TaskDial extends Task {
               tasks,
               rootSpan: this.cs.rootSpan
             });
-            await this._playSession.exec();
-            this._playSession = null;
+            await this._onHoldSession.exec();
+            this._onHoldSession = null;
           }
         } catch (error) {
           this.logger.info(error, 'DialTask:_onHoldHook: failed retrieving waitHook');
-          this._playSession = null;
+          this._onHoldSession = null;
           break;
         }
       } while (allowedTasks && allowedTasks.length > 0 && !this.killed && this.isOnHold);

lib/tasks/gather.js

@@ -301,7 +301,7 @@ class TaskGather extends SttTask {
       if (this.data.recognizer?.deepgramOptions?.shortUtterance) this.shortUtterance = true;
     }
 
-    const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
+    const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.language, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
         this.bugname = `${this.bugname_prefix}google_transcribe`;

lib/tasks/say.js

@@ -59,7 +59,7 @@ class TaskSay extends Task {
     }
   }
 
-  async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label}) {
+  async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
     const {srf} = cs;
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
     const {writeAlerts, AlertType, stats} = srf.locals;
@@ -97,7 +97,7 @@ class TaskSay extends Task {
       voice = this.options.voice_id || voice;
     }
 
-    this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
+    if (!preCache) this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
     try {
       if (!credentials) {
         writeAlerts({
@@ -120,11 +120,15 @@ class TaskSay extends Task {
         if (text.startsWith('silence_stream://')) return text;
 
         /* otel: trace time for tts */
-        const {span} = this.startChildSpan('tts-generation', {
-          'tts.vendor': vendor,
-          'tts.language': language,
-          'tts.voice': voice
-        });
+        let otelSpan;
+        if (!preCache)  {
+          const {span} = this.startChildSpan('tts-generation', {
+            'tts.vendor': vendor,
+            'tts.language': language,
+            'tts.voice': voice
+          });
+          otelSpan = span;
+        }
         try {
           const {filePath, servedFromCache, rtt} = await synthAudio(stats, {
             account_sid: cs.accountSid,
@@ -146,9 +150,9 @@ class TaskSay extends Task {
             updateSpeechCredentialLastUsed(credentials.speech_credential_sid)
               .catch(() => {/*already logged error */});
           }
-          span.setAttributes({'tts.cached': servedFromCache});
-          span.end();
-          if (!servedFromCache && rtt) {
+          if (otelSpan) otelSpan.setAttributes({'tts.cached': servedFromCache});
+          if (otelSpan) otelSpan.end();
+          if (!servedFromCache && rtt && !preCache) {
             this.notifyStatus({
               event: 'synthesized-audio',
               vendor,
@@ -160,7 +164,7 @@ class TaskSay extends Task {
           return filePath;
         } catch (err) {
           this.logger.info({err}, 'Error synthesizing tts');
-          span.end();
+          if (otelSpan) otelSpan.end();
           writeAlerts({
             account_sid: cs.accountSid,
             alert_type: AlertType.TTS_FAILURE,

lib/tasks/transcribe.js

@@ -89,13 +89,13 @@ class TaskTranscribe extends SttTask {
       stopTranscription = true;
       this.ep.stopTranscription({
         vendor: this.vendor,
-        bugname: this.ep.transcribe_bugname
+        bugname: this.bugname
       })
         .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
     }
     if (this.separateRecognitionPerChannel && this.ep2 && this.ep2.connected) {
       stopTranscription = true;
-      this.ep2.stopTranscription({vendor: this.vendor, bugname: this.ep2.transcribe_bugname})
+      this.ep2.stopTranscription({vendor: this.vendor})
         .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
     }
 
@@ -132,17 +132,16 @@ class TaskTranscribe extends SttTask {
   async _setSpeechHandlers(cs, ep, channel) {
     if (this[`_speechHandlersSet_${channel}`]) return;
     this[`_speechHandlersSet_${channel}`] = true;
-    let bugname;
 
     /* some special deepgram logic */
     if (this.vendor === 'deepgram') {
       if (this.isContinuousAsr) this._doContinuousAsrWithDeepgram(this.asrTimeout);
     }
 
-    const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
+    const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.language, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
-        bugname = `${this.bugname_prefix}google_transcribe`;
+        this.bugname = `${this.bugname_prefix}google_transcribe`;
         this.addCustomEventListener(ep, GoogleTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         this.addCustomEventListener(ep, GoogleTranscriptionEvents.NoAudioDetected,
@@ -153,7 +152,7 @@ class TaskTranscribe extends SttTask {
 
       case 'aws':
       case 'polly':
-        bugname = `${this.bugname_prefix}aws_transcribe`;
+        this.bugname = `${this.bugname_prefix}aws_transcribe`;
         this.addCustomEventListener(ep, AwsTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         this.addCustomEventListener(ep, AwsTranscriptionEvents.NoAudioDetected,
@@ -162,19 +161,19 @@ class TaskTranscribe extends SttTask {
           this._onMaxDurationExceeded.bind(this, cs, ep, channel));
         break;
       case 'microsoft':
-        bugname = `${this.bugname_prefix}azure_transcribe`;
+        this.bugname = `${this.bugname_prefix}azure_transcribe`;
         this.addCustomEventListener(ep, AzureTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         this.addCustomEventListener(ep, AzureTranscriptionEvents.NoSpeechDetected,
           this._onNoAudio.bind(this, cs, ep, channel));
         break;
       case 'nuance':
-        bugname = `${this.bugname_prefix}nuance_transcribe`;
+        this.bugname = `${this.bugname_prefix}nuance_transcribe`;
         this.addCustomEventListener(ep, NuanceTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         break;
       case 'deepgram':
-        bugname = `${this.bugname_prefix}deepgram_transcribe`;
+        this.bugname = `${this.bugname_prefix}deepgram_transcribe`;
         this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Connect,
@@ -187,12 +186,12 @@ class TaskTranscribe extends SttTask {
 
         break;
       case 'soniox':
-        bugname = `${this.bugname_prefix}soniox_transcribe`;
+        this.bugname = `${this.bugname_prefix}soniox_transcribe`;
         this.addCustomEventListener(ep, SonioxTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         break;
       case 'cobalt':
-        bugname = `${this.bugname_prefix}cobalt_transcribe`;
+        this.bugname = `${this.bugname_prefix}cobalt_transcribe`;
         this.addCustomEventListener(ep, CobaltTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
 
@@ -222,7 +221,7 @@ class TaskTranscribe extends SttTask {
         break;
 
       case 'ibm':
-        bugname = `${this.bugname_prefix}ibm_transcribe`;
+        this.bugname = `${this.bugname_prefix}ibm_transcribe`;
         this.addCustomEventListener(ep, IbmTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         this.addCustomEventListener(ep, IbmTranscriptionEvents.Connect,
@@ -232,13 +231,13 @@ class TaskTranscribe extends SttTask {
         break;
 
       case 'nvidia':
-        bugname = `${this.bugname_prefix}nvidia_transcribe`;
+        this.bugname = `${this.bugname_prefix}nvidia_transcribe`;
         this.addCustomEventListener(ep, NvidiaTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         break;
 
       case 'assemblyai':
-        bugname = `${this.bugname_prefix}assemblyai_transcribe`;
+        this.bugname = `${this.bugname_prefix}assemblyai_transcribe`;
         this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         this.addCustomEventListener(ep,
@@ -250,7 +249,7 @@ class TaskTranscribe extends SttTask {
 
       default:
         if (this.vendor.startsWith('custom:')) {
-          bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
+          this.bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
           this.addCustomEventListener(ep, JambonzTranscriptionEvents.Transcription,
             this._onTranscription.bind(this, cs, ep, channel));
           this.addCustomEventListener(ep, JambonzTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
@@ -264,8 +263,7 @@ class TaskTranscribe extends SttTask {
           throw new Error(`Invalid vendor ${this.vendor}`);
         }
     }
-    // save dedicated bugname for each endpoint
-    ep.transcribe_bugname = `${bugname}_${Date.now()}`;
+
     /* common handler for all stt engine errors */
     this.addCustomEventListener(ep, JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
     await ep.set(opts)
@@ -283,13 +281,13 @@ class TaskTranscribe extends SttTask {
 
   async _transcribe(ep) {
     this.logger.debug(
-      `TaskTranscribe:_transcribe - starting transcription vendor ${this.vendor} bugname ${ep.transcribe_bugname}`);
+      `TaskTranscribe:_transcribe - starting transcription vendor ${this.vendor} bugname ${this.bugname}`);
     await ep.startTranscription({
       vendor: this.vendor,
       interim: this.interim ? true : false,
       locale: this.language,
       channels: /*this.separateRecognitionPerChannel ? 2 : */ 1,
-      bugname: ep.transcribe_bugname,
+      bugname: this.bugname,
       hostport: this.hostport
     });
   }
@@ -297,7 +295,7 @@ class TaskTranscribe extends SttTask {
   async _onTranscription(cs, ep, channel, evt, fsEvent) {
     // make sure this is not a transcript from answering machine detection
     const bugname = fsEvent.getHeader('media-bugname');
-    if (bugname && ep.transcribe_bugname !== bugname) return;
+    if (bugname && this.bugname !== bugname) return;
 
     if (this.vendor === 'ibm' && evt?.state === 'listening') return;
 
@@ -445,7 +443,7 @@ class TaskTranscribe extends SttTask {
     if (this.isHandledByPrimaryProvider && this.fallbackVendor) {
       _ep.stopTranscription({
         vendor: this.vendor,
-        bugname: _ep.transcribe_bugname
+        bugname: this.bugname
       })
         .catch((err) => this.logger.error({err}, `Error stopping transcription for primary vendor ${this.vendor}`));
       const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);

lib/utils/amd-utils.js

@@ -266,7 +266,7 @@ module.exports = (logger) => {
 
     /* set stt options */
     logger.info(`starting amd for vendor ${vendor} and language ${language}`);
-    const sttOpts = amd.setChannelVarsForStt({name: TaskName.Gather}, sttCredentials, {
+    const sttOpts = amd.setChannelVarsForStt({name: TaskName.Gather}, sttCredentials, language, {
       vendor,
       hints,
       enhancedModel: true,

lib/utils/http-requestor.js

@@ -14,6 +14,7 @@ const {
   HTTP_PROXY_PORT,
   HTTP_PROXY_PROTOCOL,
   NODE_ENV,
+  HTTP_USER_AGENT_HEADER,
 } = require('../config');
 
 const toBase64 = (str) => Buffer.from(str || '', 'utf8').toString('base64');
@@ -116,6 +117,10 @@ class HttpRequestor extends BaseRequestor {
     const url = hook.url || hook;
     const method = hook.method || 'POST';
     let buf = '';
+    httpHeaders = {
+      ...httpHeaders,
+      ...(HTTP_USER_AGENT_HEADER && {'user-agent' : HTTP_USER_AGENT_HEADER})
+    };
 
     assert.ok(url, 'HttpRequestor:request url was not provided');
     assert.ok, (['GET', 'POST'].includes(method), `HttpRequestor:request method must be 'GET' or 'POST' not ${method}`);

lib/utils/place-outdial.js

@@ -18,7 +18,8 @@ const WsRequestor = require('./ws-requestor');
 const {makeOpusFirst} = require('./sdp-utils');
 
 class SingleDialer extends Emitter {
-  constructor({logger, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask}) {
+  constructor({logger, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask,
+    onHoldMusic}) {
     super();
     assert(target.type);
 
@@ -41,6 +42,7 @@ class SingleDialer extends Emitter {
 
     this.callSid = uuidv4();
     this.dialTask = dialTask;
+    this.onHoldMusic = onHoldMusic;
 
     this.on('callStatusChange', this._notifyCallStatusChange.bind(this));
   }
@@ -131,6 +133,7 @@ class SingleDialer extends Emitter {
       this.serviceUrl = srf.locals.serviceUrl;
 
       this.ep = await ms.createEndpoint();
+      this._configMsEndpoint();
       this.logger.debug(`SingleDialer:exec - created endpoint ${this.ep.uuid}`);
 
       /**
@@ -253,7 +256,7 @@ class SingleDialer extends Emitter {
         .on('modify', async(req, res) => {
           try {
             if (this.ep) {
-              if (this.dialTask && this.dialTask.isOnHold) {
+              if (this.dialTask && this.dialTask.isOnHoldEnabled) {
                 this.logger.info('dial is onhold, emit event');
                 this.emit('reinvite', req, res);
               } else {
@@ -320,6 +323,12 @@ class SingleDialer extends Emitter {
     }
   }
 
+  _configMsEndpoint() {
+    if (this.onHoldMusic) {
+      this.ep.set({hold_music: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`});
+    }
+  }
+
   /**
    * Run an application on the call after answer, e.g. call screening.
    * Once the application completes in some fashion, emit an 'accepted' event
@@ -436,6 +445,7 @@ class SingleDialer extends Emitter {
   async reAnchorMedia() {
     assert(this.dlg && this.dlg.connected && !this.ep);
     this.ep = await this.ms.createEndpoint({remoteSdp: this.dlg.remote.sdp});
+    this._configMsEndpoint();
     await this.dlg.modify(this.ep.local.sdp, {
       headers: {
         'X-Reason': 'anchor-media'
@@ -466,11 +476,12 @@ class SingleDialer extends Emitter {
 }
 
 function placeOutdial({
-  logger, srf, ms, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask
+  logger, srf, ms, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask,
+  onHoldMusic
 }) {
   const myOpts = deepcopy(opts);
   const sd = new SingleDialer({
-    logger, sbcAddress, target, myOpts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask
+    logger, sbcAddress, target, myOpts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask, onHoldMusic
   });
   sd.exec(srf, ms, myOpts);
   return sd;

lib/utils/transcription-utils.js

@@ -102,6 +102,50 @@ const stickyVars = {
   ]
 };
 
+const optimalDeepramModels = {
+  zh: ['base', 'base'],
+  'zh-CN':['base', 'base'],
+  'zh-TW': ['base', 'base'],
+  da: ['enhanced', 'enhanced'],
+  en: ['nova-2-conversationalai', 'nova-2'],
+  'en-US': ['nova-2-conversationalai', 'nova-2'],
+  'en-AU': ['nova-2-conversationalai', 'nova-2'],
+  'en-GB': ['nova-2-conversationalai', 'nova-2'],
+  'en-IN': ['nova-2-conversationalai', 'nova-2'],
+  'en-NZ': ['nova-2-conversationalai', 'nova-2'],
+  nl: ['nova-2-conversationalai', 'nova-2'],
+  fr: ['nova-2-conversationalai', 'nova-2'],
+  'fr-CA': ['nova-2-conversationalai', 'nova-2'],
+  de: ['nova-2-conversationalai', 'nova-2'],
+  hi: ['nova-2-conversationalai', 'nova-2'],
+  'hi-Latn': ['nova-2-conversationalai', 'nova-2'],
+  id: ['base', 'base'],
+  it: ['enhanced', 'enhanced'],
+  ja: ['enhanced', 'enhanced'],
+  ko: ['enhanced', 'enhanced'],
+  no: ['enhanced', 'enhanced'],
+  pl: ['enhanced', 'enhanced'],
+  pt: ['nova-2-conversationalai', 'nova-2'],
+  'pt-BR': ['nova-2-conversationalai', 'nova-2'],
+  'pt-PT': ['base', 'base'],
+  ru: ['base', 'base'],
+  es: ['nova-2-conversationalai', 'nova-2'],
+  'es-419': ['nova-2-conversationalai', 'nova-2'],
+  'es-LATAM': ['enhanced', 'enhanced'],
+  sv: ['enhanced', 'enhanced'],
+  ta: ['enhanced', 'enhanced'],
+  taq: ['enhanced', 'enhanced'],
+  tr: ['base', 'base'],
+  uk: ['base', 'base']
+};
+
+const selectDefaultDeepgramModel = (task, language) => {
+  if (language in optimalDeepramModels) {
+    const [gather, transcribe] = optimalDeepramModels[language];
+    return task.name === TaskName.Gather ? gather : transcribe;
+  }
+};
+
 const consolidateTranscripts = (bufferedTranscripts, channel, language) => {
   if (bufferedTranscripts.length === 1) return bufferedTranscripts[0];
   let totalConfidence = 0;
@@ -424,7 +468,7 @@ module.exports = (logger) => {
     }
   };
 
-  const setChannelVarsForStt = (task, sttCredentials, rOpts = {}) => {
+  const setChannelVarsForStt = (task, sttCredentials, language, rOpts = {}) => {
     let opts = {};
     const {enable, voiceMs = 0, mode = -1} = rOpts.vad || {};
     const vad = {enable, voiceMs, mode};
@@ -568,6 +612,9 @@ module.exports = (logger) => {
     }
     else if ('deepgram' === vendor) {
       const {deepgramOptions = {}} = rOpts;
+      if (!deepgramOptions.model) {
+        deepgramOptions.model = selectDefaultDeepgramModel(task, language);
+      }
       opts = {
         ...opts,
         ...(sttCredentials.api_key) &&

lib/utils/ws-requestor.js

@@ -9,7 +9,8 @@ const {
   JAMBONES_WS_PING_INTERVAL_MS,
   MAX_RECONNECTS,
   JAMBONES_WS_HANDSHAKE_TIMEOUT_MS,
-  JAMBONES_WS_MAX_PAYLOAD
+  JAMBONES_WS_MAX_PAYLOAD,
+  HTTP_USER_AGENT_HEADER
 } = require('../config');
 
 class WsRequestor extends BaseRequestor {
@@ -228,6 +229,9 @@ class WsRequestor extends BaseRequestor {
         maxRedirects: 2,
         handshakeTimeout,
         maxPayload: JAMBONES_WS_MAX_PAYLOAD ? parseInt(JAMBONES_WS_MAX_PAYLOAD) : 24 * 1024,
+        headers: {
+          ...(HTTP_USER_AGENT_HEADER && {'user-agent' : HTTP_USER_AGENT_HEADER})
+        }
       };
       if (this.username && this.password) opts = {...opts, auth: `${this.username}:${this.password}`};
 

package-lock.json

@@ -7530,9 +7530,9 @@
       "dev": true
     },
     "node_modules/follow-redirects": {
-      "version": "1.15.2",
-      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.2.tgz",
-      "integrity": "sha512-VQLG33o04KaQ8uYi2tVNbdrWp1QWxNNea+nmIB4EVM28v0hmP17z7aG1+wAkNzVq4KeXTq3221ye5qTJP91JwA==",
+      "version": "1.15.4",
+      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.4.tgz",
+      "integrity": "sha512-Cr4D/5wlrb0z9dgERpUL3LrmPKVDsETIJhaCMeDfuFYcqa5bldGV6wBsAN6X/vxlXQtFBMrXdXxdL8CbDTGniw==",
       "funding": [
         {
           "type": "individual",
@@ -18056,9 +18056,9 @@
       "dev": true
     },
     "follow-redirects": {
-      "version": "1.15.2",
-      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.2.tgz",
-      "integrity": "sha512-VQLG33o04KaQ8uYi2tVNbdrWp1QWxNNea+nmIB4EVM28v0hmP17z7aG1+wAkNzVq4KeXTq3221ye5qTJP91JwA=="
+      "version": "1.15.4",
+      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.4.tgz",
+      "integrity": "sha512-Cr4D/5wlrb0z9dgERpUL3LrmPKVDsETIJhaCMeDfuFYcqa5bldGV6wBsAN6X/vxlXQtFBMrXdXxdL8CbDTGniw=="
     },
     "for-each": {
       "version": "0.3.3",

test/create-call-test.js

@@ -99,6 +99,8 @@ test('test create-call call-hook basic authentication', async(t) => {
     let obj = await getJSON(`http:127.0.0.1:3100/lastRequest/${from}`)
     t.ok(obj.headers.Authorization = 'Basic dXNlcm5hbWU6cGFzc3dvcmQ=',
       'create-call: call-hook contains basic authentication header');
+    t.ok(obj.headers['user-agent'] = 'jambonz',
+    'create-call: call-hook contains user-agent header');
     disconnect();
   } catch (err) {
     console.log(`error received: ${err}`);