standardize on passing .query args as array (#356)

Add support for sending 'amd' property in createCall REST API and also added support for using any of the speech vendors for STT
---------

Co-authored-by: Dave Horton <daveh@beachdognet.com>

.github/workflows/docker-publish.yml

@@ -18,7 +18,7 @@ jobs:
       - name: prepare tag 
         id: prepare_tag
         run: |
-            IMAGE_ID=$GITHUB_REPOSITORY
+            IMAGE_ID=feature-server
 
             # Strip git ref prefix from version
             VERSION=$(echo "${{ github.ref }}" | sed -e 's,.*/\(.*\),\1,')

Dockerfile

@@ -1,4 +1,4 @@
-FROM --platform=linux/amd64 node:18.14.1-alpine3.16 as base
+FROM --platform=linux/amd64 node:18.15-alpine3.16 as base
 
 RUN apk --update --no-cache add --virtual .builds-deps build-base python3
 

data/example-voicemail-greetings.json

@@ -5,14 +5,11 @@
     "at the tone",
     "leave a message",
     "leave me a message",
-    "not available right now",
-    "not available to take your call",
+    "not available",
     "can't take your call",
-    "I will get back to you",
+    "will get back to you",
     "I'll get back to you",
-    "we will get back to you",
-    "we are unable",
-    "we are not available"
+    "we are unable"
   ],
   "es-ES": [
     "le pasamos la llamada",
@@ -48,5 +45,18 @@
     "ens posarem en contacto",
     "ara no estem disponibles",
     "no hi som"
+  ],
+  "de-DE": [
+    "nicht erreichbar",
+    "nnruf wurde weitergeleitet",
+    "beim piepsen",
+    "am ton",
+    "eine nachricht hinterlassen",
+    "hinterlasse mir eine Nachricht",
+    "nicht verfügbar",
+    "kann ihren anruf nicht entgegennehmen",
+    "wird sich bei Ihnen melden",
+    "ich melde mich bei dir",
+    "wir können nicht"
   ]
 }

lib/http-routes/api/create-call.js

@@ -27,6 +27,7 @@ router.post('/', async(req, res) => {
     const target = restDial.to;
     const opts = {
       callingNumber: restDial.from,
+      ...(restDial.callerName && {callingName: restDial.callerName}),
       headers: req.body.headers || {}
     };
 

lib/middleware.js

@@ -248,8 +248,9 @@ module.exports = function(srf, logger) {
       const app2 = JAMBONES_MYSQL_REFRESH_TTL ? JSON.parse(JSON.stringify(app)) : app;
       if ('WS' === app.call_hook?.method ||
         app.call_hook?.url.startsWith('ws://') || app.call_hook?.url.startsWith('wss://')) {
-        app2.requestor = new WsRequestor(logger, account_sid, app.call_hook, accountInfo.account.webhook_secret) ;
-        app2.notifier = app.requestor;
+        const requestor = new WsRequestor(logger, account_sid, app.call_hook, accountInfo.account.webhook_secret) ;
+        app2.requestor = requestor;
+        app2.notifier = requestor;
         app2.call_hook.method = 'WS';
       }
       else {

lib/session/call-session.js

@@ -64,6 +64,16 @@ class CallSession extends Emitter {
     this.notifiedComplete = false;
     this.rootSpan = rootSpan;
 
+    this._origRecognizerSettings = {
+      vendor: this.application?.speech_recognizer_vendor,
+      language: this.application?.speech_recognizer_language,
+    };
+    this._origSynthesizerSettings = {
+      vendor: this.application?.speech_synthesis_vendor,
+      language: this.application?.speech_synthesis_language,
+      voice: this.application?.speech_synthesis_voice,
+    };
+
     assert(rootSpan);
 
     this._recordState = RecordState.RecordingOff;
@@ -324,6 +334,22 @@ class CallSession extends Emitter {
     return this._globalSttPunctuation !== undefined;
   }
 
+  resetRecognizer() {
+    this._globalSttHints = undefined;
+    this._globalSttPunctuation = undefined;
+    this._globalAltLanguages = undefined;
+    this.isContinuousAsr = false;
+    this.asrDtmfTerminationDigits = undefined;
+    this.speechRecognizerLanguage = this._origRecognizerSettings.language;
+    this.speechRecognizerVendor = this._origRecognizerSettings.vendor;
+  }
+
+  resetSynthesizer() {
+    this.speechSynthesisLanguage = this._origSynthesizerSettings.language;
+    this.speechSynthesisVendor = this._origSynthesizerSettings.vendor;
+    this.speechSynthesisVoice = this._origSynthesizerSettings.voice;
+  }
+
   async notifyRecordOptions(opts) {
     const {action} = opts;
     this.logger.debug({opts}, 'CallSession:notifyRecordOptions');
@@ -529,7 +555,7 @@ class CallSession extends Emitter {
           this.logger.info({currInput, newInput},
             'CallSession:enableBotMode - restarting background gather to apply new input type');
           this.backgroundGatherTask.sticky = false;
-          this.disableBotMode();
+          await this.disableBotMode();
         }
       }
       this.backgroundGatherTask = task;
@@ -568,12 +594,12 @@ class CallSession extends Emitter {
       this.logger.info({err, gather}, 'CallSession:enableBotMode - Error creating gather task');
     }
   }
-  disableBotMode() {
+  async disableBotMode() {
     this._bargeInEnabled = false;
     if (this.backgroundGatherTask) {
       try {
         this.backgroundGatherTask.removeAllListeners();
-        this.backgroundGatherTask.kill().catch((err) => {});
+        await this.backgroundGatherTask.kill();
       } catch (err) {}
       this.backgroundGatherTask = null;
     }
@@ -680,6 +706,12 @@ class CallSession extends Emitter {
             stt_region: credential.stt_region
           };
         }
+        else if ('nvidia' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            riva_server_uri: credential.riva_server_uri
+          };
+        }
         else if (vendor.startsWith('custom:')) {
           return {
             speech_credential_sid: credential.speech_credential_sid,
@@ -724,11 +756,12 @@ class CallSession extends Emitter {
           }
           else {
             this.logger.info('CallSession:exec disabling bot mode to start gather with new options');
-            this.disableBotMode();
+            await this.disableBotMode();
           }
         }
         if (!skip) {
           const {span, ctx} = this.rootSpan.startChildSpan(`verb:${task.summary}`);
+          span.setAttributes({'verb.summary': task.summary});
           task.span = span;
           task.ctx = ctx;
           await task.exec(this, resources);
@@ -750,19 +783,22 @@ class CallSession extends Emitter {
       }
 
       if (0 === this.tasks.length && this.requestor instanceof WsRequestor && !this.callGone) {
-        let span;
+        //let span;
         try {
-          const {span} = this.rootSpan.startChildSpan('waiting for commands');
-          const {reason, queue, command} = await this._awaitCommandsOrHangup();
+          //const {span} = this.rootSpan.startChildSpan('waiting for commands');
+          //const {reason, queue, command} = await this._awaitCommandsOrHangup();
+          /*
           span.setAttributes({
             'completion.reason': reason,
             'async.request.queue': queue,
             'async.request.command': command
           });
           span.end();
+          */
+          await this._awaitCommandsOrHangup();
           if (this.callGone) break;
         } catch (err) {
-          span.end();
+          //span.end();
           this.logger.info(err, 'CallSession:exec - error waiting for new commands');
           break;
         }
@@ -1192,7 +1228,7 @@ class CallSession extends Emitter {
   }
 
   _onCommand({msgid, command, call_sid, queueCommand, data}) {
-    this.logger.info({msgid, command, queueCommand}, 'CallSession:_onCommand - received command');
+    this.logger.info({msgid, command, queueCommand, data}, 'CallSession:_onCommand - received command');
     const resolution = {reason: 'received command', queue: queueCommand, command};
     switch (command) {
       case 'redirect':
@@ -1255,15 +1291,17 @@ class CallSession extends Emitter {
       this.wakeupResolver(resolution);
       this.wakeupResolver = null;
     }
+    /*
     else {
-      const {span} = this.rootSpan.startChildSpan('async command');
       const {queue, command} = resolution;
+      const {span} = this.rootSpan.startChildSpan(`recv cmd: ${command}`);
       span.setAttributes({
         'async.request.queue': queue,
         'async.request.command': command
       });
       span.end();
     }
+    */
   }
 
   _onWsConnectionDropped() {
@@ -1549,7 +1587,7 @@ class CallSession extends Emitter {
       const pp = this._pool.promise();
       try {
         this.logger.info({accountSid: this.accountSid}, 'performQueueWebhook: looking up account');
-        const [r] = await pp.query(sqlRetrieveQueueEventHook, this.accountSid);
+        const [r] = await pp.query(sqlRetrieveQueueEventHook, [this.accountSid]);
         if (0 === r.length) {
           this.logger.info({accountSid: this.accountSid}, 'performQueueWebhook: no webhook provisioned');
           this.queueEventHookRequestor = null;

lib/session/rest-call-session.js

@@ -21,6 +21,10 @@ class RestCallSession extends CallSession {
     });
     this.req = req;
     this.ep = ep;
+    // keep restDialTask reference for closing AMD
+    if (tasks.length) {
+      this.restDialTask = tasks[0];
+    }
 
     this.on('callStatusChange', this._notifyCallStatusChange.bind(this));
     this._notifyCallStatusChange({
@@ -44,6 +48,10 @@ class RestCallSession extends CallSession {
    * This is invoked when the called party hangs up, in order to calculate the call duration.
    */
   _callerHungup() {
+    if (this.restDialTask) {
+      this.logger.info('RestCallSession: releasing AMD');
+      this.restDialTask.turnOffAmd();
+    }
     this.callInfo.callTerminationBy = 'caller';
     const duration = moment().diff(this.dlg.connectTime, 'seconds');
     this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});

lib/tasks/config.js

@@ -30,6 +30,12 @@ class TaskConfig extends Task {
         if (this.bargeIn[k]) this.gatherOpts[k] = this.bargeIn[k];
       });
     }
+
+    if (this.data.reset) {
+      if (typeof this.data.reset === 'string') this.data.reset = [this.data.reset];
+    }
+    else this.data.reset = [];
+
     if (this.bargeIn.sticky) this.autoEnable = true;
     this.preconditions = (this.bargeIn.enable || this.record?.action || this.listen?.url || this.data.amd) ?
       TaskPreconditions.Endpoint :
@@ -45,6 +51,10 @@ class TaskConfig extends Task {
 
   get summary() {
     const phrase = [];
+
+    /* reset recognizer and/or synthesizer to default values? */
+    if (this.data.reset.length) phrase.push(`reset ${this.data.reset.join(',')}`);
+
     if (this.bargeIn.enable) phrase.push('enable barge-in');
     if (this.hasSynthesizer) {
       const {vendor:v, language:l, voice} = this.synthesizer;
@@ -62,7 +72,7 @@ class TaskConfig extends Task {
     }
     if (this.data.amd) phrase.push('enable amd');
     if (this.notifyEvents) phrase.push(`event notification ${this.notifyEvents ? 'on' : 'off'}`);
-    return `${this.name}{${phrase.join(',')}`;
+    return `${this.name}{${phrase.join(',')}}`;
   }
 
   async exec(cs, {ep} = {}) {
@@ -86,6 +96,11 @@ class TaskConfig extends Task {
       }
     }
 
+    this.data.reset.forEach((k) => {
+      if (k === 'synthesizer') cs.resetSynthesizer();
+      else if (k === 'recognizer') cs.resetRecognizer();
+    });
+
     if (this.hasSynthesizer) {
       cs.speechSynthesisVendor = this.synthesizer.vendor !== 'default'
         ? this.synthesizer.vendor

lib/tasks/dial.js

@@ -85,6 +85,7 @@ class TaskDial extends Task {
 
     this.earlyMedia = this.data.answerOnBridge === true;
     this.callerId = this.data.callerId;
+    this.callerName = this.data.callerName;
     this.dialMusic = this.data.dialMusic;
     this.headers = this.data.headers || {};
     this.method = this.data.method || 'POST';
@@ -415,7 +416,8 @@ class TaskDial extends Task {
     const opts = {
       headers: this.headers,
       proxy: `sip:${sbcAddress}`,
-      callingNumber: this.callerId || req.callingNumber
+      callingNumber: this.callerId || req.callingNumber,
+      ...(this.callerName && {callingName: this.callerName})
     };
 
     const t = this.target.find((t) => t.type === 'teams');

lib/tasks/rest_dial.js

@@ -11,6 +11,7 @@ class TaskRestDial extends Task {
     super(logger, opts);
 
     this.from = this.data.from;
+    this.callerName = this.data.callerName;
     this.fromHost = this.data.fromHost;
     this.to = this.data.to;
     this.call_hook = this.data.call_hook;
@@ -27,18 +28,29 @@ class TaskRestDial extends Task {
   */
   async exec(cs) {
     await super.exec(cs);
+    this.cs = cs;
     this.canCancel = true;
 
+    if (this.data.amd) {
+      this.startAmd = cs.startAmd;
+      this.stopAmd = cs.stopAmd;
+      this.on('amd', this._onAmdEvent.bind(this, cs));
+    }
+
     this._setCallTimer();
     await this.awaitTaskDone();
   }
 
+  turnOffAmd() {
+    if (this.callSession.ep && this.callSession.ep.amd) this.stopAmd(this.callSession.ep, this);
+  }
+
   kill(cs) {
     super.kill(cs);
     this._clearCallTimer();
-    if (this.canCancel && cs?.req) {
+    if (this.canCancel) {
       this.canCancel = false;
-      cs.req.cancel();
+      cs?.req?.cancel();
     }
     this.notifyTaskDone();
   }
@@ -65,6 +77,13 @@ class TaskRestDial extends Task {
           }
         }
       };
+      if (this.startAmd) {
+        try {
+          this.startAmd(this.callSession, this.callSession.ep, this, this.data.amd);
+        } catch (err) {
+          this.logger.info({err}, 'Rest:dial:Call established - Error calling startAmd');
+        }
+      }
       const tasks = await cs.requestor.request('session:new', this.call_hook, params, httpHeaders);
       if (tasks && Array.isArray(tasks)) {
         this.logger.debug({tasks: tasks}, `TaskRestDial: replacing application with ${tasks.length} tasks`);
@@ -97,7 +116,16 @@ class TaskRestDial extends Task {
   _onCallTimeout() {
     this.logger.debug('TaskRestDial: timeout expired without answer, killing task');
     this.timer = null;
-    this.kill();
+    this.kill(this.cs);
+  }
+
+  _onAmdEvent(cs, evt) {
+    this.logger.info({evt}, 'Rest:dial:_onAmdEvent');
+    const {actionHook} = this.data.amd;
+    this.performHook(cs, actionHook, evt)
+      .catch((err) => {
+        this.logger.error({err}, 'Rest:dial:_onAmdEvent - error calling actionHook');
+      });
   }
 }
 

lib/tasks/say.js

@@ -36,6 +36,7 @@ class TaskSay extends Task {
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
     this.synthesizer = this.data.synthesizer || {};
     this.disableTtsCache = this.data.disableTtsCache;
+    this.options = this.synthesizer.options || {};
   }
 
   get name() { return TaskName.Say; }
@@ -66,7 +67,7 @@ class TaskSay extends Task {
       cs.speechSynthesisVoice;
     const engine = this.synthesizer.engine || 'standard';
     const salt = cs.callSid;
-    const credentials = cs.getSpeechCredentials(vendor, 'tts');
+    let credentials = cs.getSpeechCredentials(vendor, 'tts');
 
     /* parse Nuance voices into name and model */
     let model;
@@ -78,6 +79,16 @@ class TaskSay extends Task {
       }
     }
 
+    /* allow for microsoft custom region voice and api_key to be specified as an override */
+    if (vendor === 'microsoft' && this.options.deploymentId) {
+      credentials = credentials || {};
+      credentials.use_custom_tts = true;
+      credentials.custom_tts_endpoint = this.options.deploymentId;
+      credentials.api_key = this.options.apiKey || credentials.apiKey;
+      credentials.region = this.options.region || credentials.region;
+      voice = this.options.voice || voice;
+    }
+
     this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
     this.ep = ep;
     try {

lib/tasks/task.js

@@ -155,7 +155,7 @@ class Task extends Emitter {
     if (this.actionHook) {
       const type = this.name === TaskName.Redirect ? 'session:redirect' : 'verb:hook';
       const params = results ? Object.assign(this.cs.callInfo.toJSON(), results) : this.cs.callInfo.toJSON();
-      const span = this.startSpan(type, {'hook.url': this.actionHook});
+      const span = this.startSpan(`${type} (${this.actionHook})`);
       const b3 = this.getTracingPropagation('b3', span);
       const httpHeaders = b3 && {b3};
       span.setAttributes({'http.body': JSON.stringify(params)});

lib/utils/amd-utils.js

@@ -1,9 +1,16 @@
 const Emitter = require('events');
 const {readFile} = require('fs');
 const {
+  TaskName,
   GoogleTranscriptionEvents,
   AwsTranscriptionEvents,
   AzureTranscriptionEvents,
+  NuanceTranscriptionEvents,
+  NvidiaTranscriptionEvents,
+  IbmTranscriptionEvents,
+  SonioxTranscriptionEvents,
+  DeepgramTranscriptionEvents,
+  JambonzTranscriptionEvents,
   AmdEvents,
   AvmdEvents
 } = require('./constants');
@@ -54,6 +61,11 @@ class Amd extends Emitter {
     this.thresholdWordCount = opts.thresholdWordCount || 9;
     const {normalizeTranscription} = require('./transcription-utils')(logger);
     this.normalizeTranscription = normalizeTranscription;
+    const {getNuanceAccessToken, getIbmAccessToken} = cs.srf.locals.dbHelpers;
+    this.getNuanceAccessToken = getNuanceAccessToken;
+    this.getIbmAccessToken = getIbmAccessToken;
+    const {setChannelVarsForStt} = require('./transcription-utils')(logger);
+    this.setChannelVarsForStt = setChannelVarsForStt;
 
     const {
       noSpeechTimeoutMs = 5000,
@@ -184,7 +196,7 @@ module.exports = (logger) => {
     const {vendor, language} = ep.amd;
     ep.startTranscription({
       vendor,
-      language,
+      locale: language,
       interim: true,
       bugname
     }).catch((err) => {
@@ -229,51 +241,92 @@ module.exports = (logger) => {
 
   const startAmd = async(cs, ep, task, opts) => {
     const amd = ep.amd = new Amd(logger, cs, opts);
-    const {vendor, language, sttCredentials} = amd;
-    const sttOpts = {};
+    const {vendor, language} = amd;
+    let sttCredentials = amd.sttCredentials;
     const hints = voicemailHints[language] || [];
 
+    if (vendor === 'nuance' && sttCredentials.client_id) {
+      /* get nuance access token */
+      const {getNuanceAccessToken} = amd;
+      const {client_id, secret} = sttCredentials;
+      const {access_token, servedFromCache} = await getNuanceAccessToken(client_id, secret, 'asr tts');
+      logger.debug({client_id}, `Gather:exec - got nuance access token ${servedFromCache ? 'from cache' : ''}`);
+      sttCredentials = {...sttCredentials, access_token};
+    }
+    else if (vendor == 'ibm' && sttCredentials.stt_api_key) {
+      /* get ibm access token */
+      const {getIbmAccessToken} = amd;
+      const {stt_api_key, stt_region} = sttCredentials;
+      const {access_token, servedFromCache} = await getIbmAccessToken(stt_api_key);
+      logger.debug({stt_api_key}, `Gather:exec - got ibm access token ${servedFromCache ? 'from cache' : ''}`);
+      sttCredentials = {...sttCredentials, access_token, stt_region};
+    }
+
     /* set stt options */
     logger.info(`starting amd for vendor ${vendor} and language ${language}`);
-    if ('google' === vendor) {
-      sttOpts.GOOGLE_APPLICATION_CREDENTIALS = JSON.stringify(sttCredentials.credentials);
-      sttOpts.GOOGLE_SPEECH_USE_ENHANCED = true;
-      sttOpts.GOOGLE_SPEECH_HINTS = hints.join(',');
-      if (opts.recognizer?.altLanguages) {
-        sttOpts.GOOGLE_SPEECH_ALTERNATIVE_LANGUAGE_CODES = opts.recognizer.altLanguages.join(',');
-      }
-      ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription, onTranscription.bind(null, cs, ep, task));
-      ep.addCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance, onEndOfUtterance.bind(null, cs, ep, task));
-    }
-    else if (['aws', 'polly'].includes(vendor)) {
-      Object.assign(sttOpts, {
-        AWS_ACCESS_KEY_ID: sttCredentials.accessKeyId,
-        AWS_SECRET_ACCESS_KEY: sttCredentials.secretAccessKey,
-        AWS_REGION: sttCredentials.region
-      });
-      ep.addCustomEventListener(AwsTranscriptionEvents.Transcription, onTranscription.bind(null, cs, ep, task));
-    }
-    else if ('microsoft' === vendor) {
-      Object.assign(sttOpts, {
-        'AZURE_SUBSCRIPTION_KEY': sttCredentials.api_key,
-        'AZURE_REGION': sttCredentials.region
-      });
-      sttOpts.AZURE_SPEECH_HINTS = hints.join(',');
-      if (opts.recognizer?.altLanguages) {
-        sttOpts.AZURE_SPEECH_ALTERNATIVE_LANGUAGE_CODES = opts.recognizer.altLanguages.join(',');
-      }
-      sttOpts.AZURE_INITIAL_SPEECH_TIMEOUT_MS = opts.resolveTimeoutMs || 20000;
+    const sttOpts = amd.setChannelVarsForStt({name: TaskName.Gather}, sttCredentials, {
+      vendor,
+      hints,
+      enhancedModel: true,
+      altLanguages: opts.recognizer?.altLanguages || [],
+      initialSpeechTimeoutMs: opts.resolveTimeoutMs,
+    });
 
-      ep.addCustomEventListener(AzureTranscriptionEvents.Transcription, onTranscription.bind(null, cs, ep, task));
-      ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected, onNoSpeechDetected.bind(null, cs, ep, task));
-    }
     await ep.set(sttOpts).catch((err) => logger.info(err, 'Error setting channel variables'));
 
+    amd.transcriptionHandler = onTranscription.bind(null, cs, ep, task);
+    amd.EndOfUtteranceHandler = onEndOfUtterance.bind(null, cs, ep, task);
+    amd.noSpeechHandler = onNoSpeechDetected.bind(null, cs, ep, task);
+
+    switch (vendor) {
+      case 'google':
+        ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        ep.addCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance, amd.EndOfUtteranceHandler);
+        break;
+
+      case 'aws':
+      case 'polly':
+        ep.addCustomEventListener(AwsTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        break;
+      case 'microsoft':
+        ep.addCustomEventListener(AzureTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected, amd.noSpeechHandler);
+        break;
+      case 'nuance':
+        ep.addCustomEventListener(NuanceTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        break;
+
+      case 'deepgram':
+        ep.addCustomEventListener(DeepgramTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        break;
+
+      case 'soniox':
+        amd.bugname = 'soniox_amd_transcribe';
+        ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        break;
+
+      case 'ibm':
+        ep.addCustomEventListener(IbmTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        break;
+
+      case 'nvidia':
+        ep.addCustomEventListener(NvidiaTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        break;
+
+      default:
+        if (vendor.startsWith('custom:')) {
+          ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription, amd.transcriptionHandler);
+          break;
+        }
+        else {
+          throw new Error(`Invalid vendor ${this.vendor}`);
+        }
+    }
     amd
       .on(AmdEvents.NoSpeechDetected, (evt) => {
         task.emit('amd', {type: AmdEvents.NoSpeechDetected, ...evt});
         try {
-          ep.connected && ep.stopTranscription({vendor, bugname});
+          stopAmd(ep, task);
         } catch (err) {
           logger.info({err}, 'Error stopping transcription');
         }
@@ -281,7 +334,7 @@ module.exports = (logger) => {
       .on(AmdEvents.HumanDetected, (evt) => {
         task.emit('amd', {type: AmdEvents.HumanDetected, ...evt});
         try {
-          ep.connected && ep.stopTranscription({vendor, bugname});
+          stopAmd(ep, task);
         } catch (err) {
           logger.info({err}, 'Error stopping transcription');
         }
@@ -292,7 +345,7 @@ module.exports = (logger) => {
       .on(AmdEvents.DecisionTimeout, (evt) => {
         task.emit('amd', {type: AmdEvents.DecisionTimeout, ...evt});
         try {
-          ep.connected && ep.stopTranscription({vendor, bugname});
+          stopAmd(ep, task);
         } catch (err) {
           logger.info({err}, 'Error stopping transcription');
         }
@@ -300,7 +353,7 @@ module.exports = (logger) => {
       .on(AmdEvents.ToneTimeout, (evt) => {
         //task.emit('amd', {type: AmdEvents.ToneTimeout, ...evt});
         try {
-          ep.connected && ep.execute('avmd_stop').catch((err) => logger.info(err, 'Error stopping avmd'));
+          stopAmd(ep, task);
         } catch (err) {
           logger.info({err}, 'Error stopping avmd');
         }
@@ -308,7 +361,7 @@ module.exports = (logger) => {
       .on(AmdEvents.MachineStoppedSpeaking, () => {
         task.emit('amd', {type: AmdEvents.MachineStoppedSpeaking});
         try {
-          ep.connected && ep.stopTranscription({vendor, bugname});
+          stopAmd(ep, task);
         } catch (err) {
           logger.info({err}, 'Error stopping transcription');
         }
@@ -327,6 +380,19 @@ module.exports = (logger) => {
     if (ep.amd) {
       vendor = ep.amd.vendor;
       ep.amd.stopAllTimers();
+
+      ep.removeListener(GoogleTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(GoogleTranscriptionEvents.EndOfUtterance, ep.amd.EndOfUtteranceHandler);
+      ep.removeListener(AwsTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(AzureTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(AzureTranscriptionEvents.NoSpeechDetected, ep.amd.noSpeechHandler);
+      ep.removeListener(NuanceTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(DeepgramTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(SonioxTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(IbmTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(NvidiaTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+      ep.removeListener(JambonzTranscriptionEvents.Transcription, ep.amd.transcriptionHandler);
+
       ep.amd = null;
     }
 

lib/utils/db-utils.js

@@ -78,6 +78,10 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
     }
+    else if ('nvidia' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.riva_server_uri = o.riva_server_uri;
+    }
     else if (obj.vendor.startsWith('custom:')) {
       const o = JSON.parse(decrypt(credential));
       obj.auth_token = o.auth_token;
@@ -96,13 +100,13 @@ module.exports = (logger, srf) => {
 
   const lookupAccountDetails = async(account_sid) => {
 
-    const [r] = await pp.query({sql: sqlAccountDetails, nestTables: true}, account_sid);
+    const [r] = await pp.query({sql: sqlAccountDetails, nestTables: true}, [account_sid]);
     if (0 === r.length) throw new Error(`invalid accountSid: ${account_sid}`);
-    const [r2] = await pp.query(sqlSpeechCredentials, account_sid);
+    const [r2] = await pp.query(sqlSpeechCredentials, [account_sid]);
     const speech = r2.map(speechMapper);
 
     /* add service provider creds unless we have that vendor at the account level */
-    const [r3] = await pp.query(sqlSpeechCredentialsForSP, account_sid);
+    const [r3] = await pp.query(sqlSpeechCredentialsForSP, [account_sid]);
     r3.forEach((s) => {
       if (!speech.find((s2) => s2.vendor === s.vendor)) {
         speech.push(speechMapper(s));

lib/utils/transcription-utils.js

@@ -549,6 +549,7 @@ module.exports = (logger) => {
     }
     else if ('nvidia' === vendor) {
       const {nvidiaOptions = {}} = rOpts;
+      const rivaUri = nvidiaOptions.rivaUri || sttCredentials.riva_server_uri;
       opts = {
         ...opts,
         ...((nvidiaOptions.profanityFilter || rOpts.profanityFilter) && {NVIDIA_PROFANITY_FILTER: 1}),
@@ -560,7 +561,7 @@ module.exports = (logger) => {
         ...(nvidiaOptions.maxAlternatives && {NVIDIA_MAX_ALTERNATIVES: nvidiaOptions.maxAlternatives}),
         ...(!nvidiaOptions.maxAlternatives && {NVIDIA_MAX_ALTERNATIVES: 1}),
         ...(rOpts.model && {NVIDIA_MODEL: rOpts.model}),
-        ...(nvidiaOptions.rivaUri && {NVIDIA_RIVA_URI: nvidiaOptions.rivaUri}),
+        ...(rivaUri && {NVIDIA_RIVA_URI: rivaUri}),
         ...(nvidiaOptions.verbatimTranscripts && {NVIDIA_VERBATIM_TRANSCRIPTS: 1}),
         ...(rOpts.diarization && {NVIDIA_SPEAKER_DIARIZATION: 1}),
         ...(rOpts.diarization && rOpts.diarizationMaxSpeakers > 0 &&

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jambonz-feature-server",
-  "version": "v0.8.2",
+  "version": "0.8.3",
   "main": "app.js",
   "engines": {
     "node": ">= 10.16.0"
@@ -25,13 +25,13 @@
     "jslint:fix": "eslint app.js tracer.js lib --fix"
   },
   "dependencies": {
-    "@jambonz/db-helpers": "^0.7.4",
+    "@jambonz/db-helpers": "^0.8.1",
     "@jambonz/http-health-check": "^0.0.1",
-    "@jambonz/realtimedb-helpers": "^0.7.0",
-    "@jambonz/speech-utils": "^0.0.12",
+    "@jambonz/realtimedb-helpers": "^0.7.2",
+    "@jambonz/speech-utils": "^0.0.13",
     "@jambonz/stats-collector": "^0.1.8",
     "@jambonz/time-series": "^0.2.5",
-    "@jambonz/verb-specifications": "^0.0.12",
+    "@jambonz/verb-specifications": "^0.0.22",
     "@opentelemetry/api": "^1.4.0",
     "@opentelemetry/exporter-jaeger": "^1.9.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
@@ -45,7 +45,7 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.20",
+    "drachtio-fsmrf": "^3.0.21",
     "drachtio-srf": "^4.5.23",
     "express": "^4.18.2",
     "ip": "^1.1.8",

test/create-call-test.js

@@ -106,3 +106,61 @@ test('test create-call call-hook basic authentication', async(t) => {
     t.error(err);
   }
 });
+
+test('test create-call amd', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+
+    // GIVEN
+    let from = 'create-call-amd';
+    let account_sid = 'bb845d4b-83a9-4cde-a6e9-50f3743bab3f';
+
+    // Give UAS app time to come up
+    const p = sippUac('uas.xml', '172.38.0.10', from);
+    await waitFor(1000);
+
+    const post = bent('http://127.0.0.1:3000/', 'POST', 'json', 201);
+    post('v1/createCall', {
+      'account_sid':account_sid,
+      "call_hook": {
+        "url": "http://127.0.0.1:3100/",
+        "method": "POST",
+        "username": "username",
+        "password": "password"
+      },
+      "from": from,
+      "to": {
+        "type": "phone",
+        "number": "15583084809"
+      },
+      "amd": {
+        "actionHook": "/actionHook"
+      },
+      "speech_recognizer_vendor": "google",
+      "speech_recognizer_language": "en"
+    });
+
+    let verbs = [
+      {
+        "verb": "pause",
+        "length": 7
+      }
+    ];
+    provisionCallHook(from, verbs);
+    //THEN
+    await p;
+
+    let obj = await getJSON(`http:127.0.0.1:3100/lastRequest/${from}_actionHook`)
+    t.ok(obj.body.type = 'amd_no_speech_detected',
+      'create-call: AMD detected');
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});

test/dial-tests.js

@@ -31,6 +31,7 @@ test('\'dial-phone\'', async(t) => {
       {
         "verb": "dial",
         "callerId": from,
+        "callerName": "test_callerName",
         "actionHook": "/actionHook",
         "timeLimit": 5,
         "target": [
@@ -56,6 +57,7 @@ test('\'dial-phone\'', async(t) => {
         "method": "POST",
       },
       "from": from,
+      "callerName": "Tom",
       "to": {
         "type": "phone",
         "number": "15583084808"

test/gather-tests.js

@@ -102,6 +102,53 @@ test('\'gather\' test - default (google)', async(t) => {
   }
 });
 
+test('\'config\' test - reset to app defaults', async(t) => {
+  if (!GCP_JSON_KEY) {
+    t.pass('skipping config tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "config",
+        "recognizer": {
+          "vendor": "google",
+          "language": "fr-FR"
+        },
+      },
+      {
+        "verb": "config",
+        "reset": ['recognizer'],
+      },
+      {
+        "verb": "gather",
+        "input": ["speech"],
+        "timeout": 10,
+        "actionHook": "/actionHook"
+      }
+    ];
+    let from = "gather_success";
+    provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase() === 'i\'d like to speak to customer support',
+      'config: resets recognizer to app defaults');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
 test('\'gather\' test - microsoft', async(t) => {
   if (!MICROSOFT_REGION || !MICROSOFT_API_KEY) {
     t.pass('skipping microsoft tests');
@@ -214,7 +261,7 @@ test('\'gather\' test - deepgram', async(t) => {
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
     //console.log(JSON.stringify(obj));
-    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
       'gather: succeeds when using deepgram credentials');
     disconnect();
   } catch (err) {

test/say-tests.js

@@ -43,3 +43,84 @@ test('\'say\' tests', async(t) => {
     t.error(err);
   }
 });
+
+test('\'config\' reset synthesizer tests', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const verbs = [
+      {
+        "verb": "config",
+        "synthesizer": {
+          "vendor": "microsft",
+          "voice": "foobar"
+        },
+      },
+      {
+        "verb": "config",
+        "reset": 'synthesizer',
+      },
+      {
+        verb: 'say',
+        text: 'hello'
+      }
+    ];
+
+    const from = 'say_test_success';
+    provisionCallHook(from, verbs)
+
+    // THEN
+    await sippUac('uac-success-received-bye.xml', '172.38.0.10', from);
+    t.pass('say: succeeds when using using account credentials');
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+const {MICROSOFT_CUSTOM_API_KEY, MICROSOFT_DEPLOYMENT_ID, MICROSOFT_CUSTOM_REGION, MICROSOFT_CUSTOM_VOICE} = process.env;
+if (MICROSOFT_CUSTOM_API_KEY && MICROSOFT_DEPLOYMENT_ID && MICROSOFT_CUSTOM_REGION && MICROSOFT_CUSTOM_VOICE) {
+  test('\'say\' tests - microsoft custom voice', async(t) => {
+    clearModule.all();
+    const {srf, disconnect} = require('../app');
+
+    try {
+      await connect(srf);
+
+      // GIVEN
+      const verbs = [
+        {
+          verb: 'say',
+          text: 'hello',
+          synthesizer: {
+            vendor: 'microsoft',
+            voice: MICROSOFT_CUSTOM_VOICE,
+            options: {
+              deploymentId: MICROSOFT_DEPLOYMENT_ID,
+              apiKey: MICROSOFT_CUSTOM_API_KEY,
+              region: MICROSOFT_CUSTOM_REGION,
+            }
+          }
+        }
+      ];
+
+      const from = 'say_test_success';
+      provisionCallHook(from, verbs)
+
+      // THEN
+      await sippUac('uac-success-received-bye.xml', '172.38.0.10', from);
+      t.pass('say: succeeds when using microsoft custom voice');
+      disconnect();
+    } catch (err) {
+      console.log(`error received: ${err}`);
+      disconnect();
+      t.error(err);
+    }
+  });
+}

test/transcribe-tests.js

@@ -166,7 +166,7 @@ test('\'transcribe\' test - deepgram', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
       'transcribe: succeeds when using deepgram credentials');
 
     disconnect();
@@ -206,7 +206,7 @@ test('\'transcribe\' test - soniox', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    console.log(JSON.stringify(obj));
+    //console.log(JSON.stringify(obj));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using soniox credentials');
 

test/webhook/app.js

@@ -1,8 +1,7 @@
 const express = require('express');
 const app = express();
 const Websocket = require('ws');
-const {PORT} = require('../../lib/config');
-const listenPort = PORT;
+const listenPort = process.env.HTTP_PORT || 3000;
 let json_mapping = new Map();
 let hook_mapping = new Map();
 let ws_packet_count = new Map();

tracer.js

@@ -12,6 +12,7 @@ const {
   OTEL_EXPORTER_JAEGER_AGENT_HOST,
   OTEL_EXPORTER_JAEGER_ENDPOINT,
   OTEL_EXPORTER_ZIPKIN_URL,
+  OTEL_EXPORTER_COLLECTOR_URL
 } = require('./lib/config');
 
 module.exports = (serviceName) => {
@@ -33,7 +34,7 @@ module.exports = (serviceName) => {
     }
     else {
       exporter = new OTLPTraceExporter({
-        url: process.OTEL_EXPORTER_COLLECTOR_URL
+        url: OTEL_EXPORTER_COLLECTOR_URL
       });
     }