Feature/azure recognition (#46)

* add support for microsoft speech recognition

* update to drachtio-fsmrf that support microsoft stt

* gather and transcribe now support microsoft

Dockerfile

@@ -1,16 +1,10 @@
-FROM node:alpine as builder
-RUN apk update && apk add --no-cache python make g++
+FROM node:16
 WORKDIR /opt/app/
 COPY package.json ./
 RUN npm install
 RUN npm prune
-
-FROM node:alpine as app
-WORKDIR /opt/app
 COPY . /opt/app
-COPY --from=builder /opt/app/node_modules ./node_modules
-
 ARG NODE_ENV
 ENV NODE_ENV $NODE_ENV
 
-CMD [ "npm", "start" ]
+CMD [ "npm", "start" ]

lib/middleware.js

@@ -137,7 +137,9 @@ module.exports = function(srf, logger) {
       const obj = Object.assign({}, app);
       delete obj.requestor;
       delete obj.notifier;
-      logger.info({app: obj}, `retrieved application for incoming call to ${req.locals.calledNumber}`);
+      // eslint-disable-next-line no-unused-vars
+      const {call_hook, call_status_hook, ...appInfo} = obj;  // mask sensitive data like user/pass on webhook
+      logger.info({app: appInfo}, `retrieved application for incoming call to ${req.locals.calledNumber}`);
       req.locals.callInfo = new CallInfo({req, app, direction: CallDirection.Inbound});
       next();
     } catch (err) {

lib/session/call-session.js

@@ -257,6 +257,13 @@ class CallSession extends Emitter {
             region: process.env.AWS_REGION || credential.aws_region
           };
         }
+        else if ('microsoft' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            api_key: credential.api_key,
+            region: credential.region
+          };
+        }
       }
       else {
         writeAlerts({

lib/tasks/conference.js

@@ -10,6 +10,7 @@ const WAIT = 'wait';
 const JOIN = 'join';
 const START = 'start';
 
+
 function confNoMatch(str) {
   return str.match(/^No active conferences/) || str.match(/Conference.*not found/);
 }
@@ -49,7 +50,7 @@ class Conference extends Task {
       'beep', 'startConferenceOnEnter', 'endConferenceOnExit', 'joinMuted',
       'maxParticipants', 'waitHook', 'statusHook', 'endHook', 'enterHook'
     ].forEach((attr) => this[attr] = this.data[attr]);
-
+    this.record = this.data.record || {};
     this.statusEvents = [];
     if (this.statusHook) {
       ['start', 'end', 'join', 'leave', 'start-talking', 'stop-talking'].forEach((e) => {
@@ -68,6 +69,9 @@ class Conference extends Task {
 
   get name() { return TaskName.Conference; }
 
+  get shouldRecord() { return this.record.path; }
+  get isRecording() { return this.recordingInProgress; }
+
   async exec(cs, ep) {
     await super.exec(cs);
     this.ep = ep;
@@ -346,6 +350,16 @@ class Conference extends Task {
       if (response.body && /\d+/.test(response.body)) this.participantCount = parseInt(response.body);
       this._notifyConferenceEvent(cs, 'join');
 
+      // start recording if requested and we just started the conference
+      if (startConf && this.shouldRecord) {
+        this.logger.info(`recording conference to ${this.record.path}`);
+        try {
+          await this.ep.api(`conference ${this.confName} record ${this.record.path}`);
+        } catch (err) {
+          this.logger.info({err}, 'Conference:_joinConference - failed to start recording');
+        }
+      }
+
       // listen for conference events
       this.ep.filter('Conference-Unique-ID', this.confUuid);
       this.ep.conn.on('esl::event::CUSTOM::*', this.__onConferenceEvent.bind(this, cs)) ;

lib/tasks/dial.js

@@ -161,6 +161,10 @@ class TaskDial extends Task {
   async kill(cs, reason) {
     super.kill(cs);
     this.killReason = reason || KillReason.Hangup;
+    if (this.timerMaxCallDuration) {
+      clearTimeout(this.timerMaxCallDuration);
+      this.timerMaxCallDuration = null;
+    }
     this._removeDtmfDetection(cs.dlg);
     this._removeDtmfDetection(this.dlg);
     this._killOutdials();
@@ -171,7 +175,6 @@ class TaskDial extends Task {
     if (this.callSid) sessionTracker.remove(this.callSid);
     if (this.listenTask) await this.listenTask.kill(cs);
     if (this.transcribeTask) await this.transcribeTask.kill(cs);
-    if (this.timerMaxCallDuration) clearTimeout(this.timerMaxCallDuration);
     this.notifyTaskDone();
   }
 
@@ -458,7 +461,7 @@ class TaskDial extends Task {
     if (this.timeLimit) {
       this.timerMaxCallDuration = setTimeout(() => {
         this.logger.info(`Dial:_selectSingleDial tearing down call as it has reached ${this.timeLimit}s`);
-        this.ep.unbridge();
+        this.ep && this.ep.unbridge();
         this.kill(cs);
       }, this.timeLimit * 1000);
     }
@@ -468,7 +471,10 @@ class TaskDial extends Task {
       if (this.dlg) {
         this.logger.debug('Dial:_selectSingleDial called party hungup, ending dial operation');
         sessionTracker.remove(this.callSid);
-        if (this.timerMaxCallDuration) clearTimeout(this.timerMaxCallDuration);
+        if (this.timerMaxCallDuration) {
+          clearTimeout(this.timerMaxCallDuration);
+          this.timerMaxCallDuration = null;
+        }
         this.ep && this.ep.unbridge();
         this.kill(cs);
       }

lib/tasks/gather.js

@@ -3,7 +3,8 @@ const {
   TaskName,
   TaskPreconditions,
   GoogleTranscriptionEvents,
-  AwsTranscriptionEvents
+  AwsTranscriptionEvents,
+  AzureTranscriptionEvents
 } = require('../utils/constants');
 
 const makeTask = require('./make_task');
@@ -33,6 +34,12 @@ class TaskGather extends Task {
       this.vocabularyName = recognizer.vocabularyName;
       this.vocabularyFilterName = recognizer.vocabularyFilterName;
       this.filterMethod = recognizer.filterMethod;
+
+      /* microsoft options */
+      this.outputFormat = recognizer.outputFormat || 'simple';
+      this.profanityOption = recognizer.profanityOption || 'raw';
+      this.requestSnr = recognizer.requestSnr || false;
+      this.initialSpeechTimeoutMs = recognizer.initialSpeechTimeoutMs || 0;
     }
 
     this.digitBuffer = '';
@@ -46,6 +53,8 @@ class TaskGather extends Task {
 
   get name() { return TaskName.Gather; }
 
+  get needsStt() { return this.input.includes('speech'); }
+
   get earlyMedia() {
     return (this.sayTask && this.sayTask.earlyMedia) ||
       (this.playTask && this.playTask.earlyMedia);
@@ -59,9 +68,9 @@ class TaskGather extends Task {
     if ('default' === this.vendor || !this.vendor) this.vendor = cs.speechRecognizerVendor;
     if ('default' === this.language || !this.language) this.language = cs.speechRecognizerLanguage;
     this.sttCredentials = cs.getSpeechCredentials(this.vendor, 'stt');
-    if (!this.sttCredentials) {
+    if (this.needsStt && !this.sttCredentials) {
       const {writeAlerts, AlertType} = cs.srf.locals;
-      this.logger.info(`TaskGather:exec - ERROR stt using ${this.vendor} requested but not creds supplied`);
+      this.logger.info(`TaskGather:exec - ERROR stt using ${this.vendor} requested but creds not supplied`);
       writeAlerts({
         account_sid: cs.accountSid,
         alert_type: AlertType.STT_NOT_PROVISIONED,
@@ -104,6 +113,8 @@ class TaskGather extends Task {
     ep.removeCustomEventListener(GoogleTranscriptionEvents.Transcription);
     ep.removeCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance);
     ep.removeCustomEventListener(AwsTranscriptionEvents.Transcription);
+    ep.removeCustomEventListener(AzureTranscriptionEvents.Transcription);
+    ep.removeCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected);
   }
 
   kill(cs) {
@@ -133,7 +144,9 @@ class TaskGather extends Task {
         GOOGLE_SPEECH_SINGLE_UTTERANCE: true,
         GOOGLE_SPEECH_MODEL: 'command_and_search'
       });
-      if (this.hints && this.hints.length > 1) opts.GOOGLE_SPEECH_HINTS = this.hints.join(',');
+      if (this.hints && this.hints.length > 1) {
+        opts.GOOGLE_SPEECH_HINTS = this.hints.map((h) => h.trim()).join(',');
+      }
       if (this.altLanguages && this.altLanguages.length > 1) {
         opts.GOOGLE_SPEECH_ALTERNATIVE_LANGUAGE_CODES = this.altLanguages.join(',');
       }
@@ -143,22 +156,41 @@ class TaskGather extends Task {
       ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
       ep.addCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance, this._onEndOfUtterance.bind(this, cs, ep));
     }
-    else {
+    else if (['aws', 'polly'].includes(this.vendor)) {
       if (this.vocabularyName) opts.AWS_VOCABULARY_NAME = this.vocabularyName;
       if (this.vocabularyFilterName) {
         opts.AWS_VOCABULARY_NAME = this.vocabularyFilterName;
         opts.AWS_VOCABULARY_FILTER_METHOD = this.filterMethod || 'mask';
       }
-      Object.assign(opts, {
-        AWS_ACCESS_KEY_ID: this.sttCredentials.accessKeyId,
-        AWS_SECRET_ACCESS_KEY: this.sttCredentials.secretAccessKey,
-        AWS_REGION: this.sttCredentials.region
-      });
+      if (this.sttCredentials) {
+        Object.assign(opts, {
+          AWS_ACCESS_KEY_ID: this.sttCredentials.accessKeyId,
+          AWS_SECRET_ACCESS_KEY: this.sttCredentials.secretAccessKey,
+          AWS_REGION: this.sttCredentials.region
+        });
+      }
       ep.addCustomEventListener(AwsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
     }
+    else if ('microsoft' === this.vendor) {
+      if (this.sttCredentials) {
+        Object.assign(opts, {
+          'AZURE_SUBSCRIPTION_KEY': this.sttCredentials.api_key,
+          'AZURE_REGION': this.sttCredentials.region
+        });
+      }
+      if (this.hints && this.hints.length > 1) {
+        opts.AZURE_SPEECH_HINTS = this.hints.map((h) => h.trim()).join(',');
+      }
+      //if (this.requestSnr) opts.AZURE_REQUEST_SNR = 1;
+      //if (this.profanityOption !== 'raw') opts.AZURE_PROFANITY_OPTION = this.profanityOption;
+      if (this.initialSpeechTimeoutMs > 0) opts.AZURE_INITIAL_SPEECH_TIMEOUT_MS = this.initialSpeechTimeoutMs;
+      opts.AZURE_USE_OUTPUT_FORMAT_DETAILED = 1;
+
+      ep.addCustomEventListener(AzureTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+      ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected, this._onNoSpeechDetected.bind(this, cs, ep));
+    }
     await ep.set(opts)
       .catch((err) => this.logger.info(err, 'Error setting channel variables'));
-
   }
 
   _startTranscribing(ep) {
@@ -206,11 +238,21 @@ class TaskGather extends Task {
 
   _onTranscription(cs, ep, evt) {
     if ('aws' === this.vendor && Array.isArray(evt) && evt.length > 0) evt = evt[0];
-    this.logger.debug(evt, 'TaskGather:_onTranscription');
-    const final = evt.is_final;
-    if (final) {
-      this._resolve('speech', evt);
+    if ('microsoft' === this.vendor) {
+      const nbest = evt.NBest;
+      const newEvent = {
+        is_final: evt.RecognitionStatus === 'Success',
+        alternatives: [
+          {
+            confidence: nbest[0].Confidence,
+            transcript: nbest[0].Display
+          }
+        ]
+      };
+      evt = newEvent;
     }
+    this.logger.debug(evt, 'TaskGather:_onTranscription');
+    if (evt.is_final) this._resolve('speech', evt);
     else if (this.partialResultHook) {
       this.cs.requestor.request(this.partialResultHook,  Object.assign({speech: evt}, this.cs.callInfo))
         .catch((err) => this.logger.info(err, 'GatherTask:_onTranscription error'));
@@ -223,6 +265,10 @@ class TaskGather extends Task {
     }
   }
 
+  _onNoSpeechDetected(cs, ep) {
+    this._resolve('timeout');
+  }
+
   async _resolve(reason, evt) {
     if (this.resolved) return;
     this.resolved = true;

lib/tasks/make_task.js

@@ -17,8 +17,10 @@ function makeTask(logger, obj, parent) {
     case TaskName.SipDecline:
       const TaskSipDecline = require('./sip_decline');
       return new TaskSipDecline(logger, data, parent);
+    case TaskName.SipRefer:
+      const TaskSipRefer = require('./sip_refer');
+      return new TaskSipRefer(logger, data, parent);
     case TaskName.Conference:
-      logger.debug({data}, 'Conference verb');
       const TaskConference = require('./conference');
       return new TaskConference(logger, data, parent);
     case TaskName.Dial:

lib/tasks/say.js

@@ -21,19 +21,15 @@ class TaskSay extends Task {
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
     const {writeAlerts, AlertType, stats} = srf.locals;
     const {synthAudio} = srf.locals.dbHelpers;
-    const vendor = ('default' === this.synthesizer.vendor  || !this.synthesizer.vendor) ?
-      cs.speechSynthesisVendor :
-      this.synthesizer.vendor;
-    const language = ('default' === this.synthesizer.language  || !this.synthesizer.language) ?
-      cs.speechSynthesisLanguage :
-      this.synthesizer.language;
-    const voice = ('default' === this.synthesizer.voice  || !this.synthesizer.voice) ?
-      cs.speechSynthesisVoice :
-      this.synthesizer.voice;
+    const hasVerbLevelTts = this.synthesizer.vendor && this.synthesizer.vendor !== 'default';
+    const vendor = hasVerbLevelTts ? this.synthesizer.vendor : cs.speechSynthesisVendor ;
+    const language = hasVerbLevelTts ? this.synthesizer.language : cs.speechSynthesisLanguage ;
+    const voice = hasVerbLevelTts ? this.synthesizer.voice : cs.speechSynthesisVoice ;
+
     const salt = cs.callSid;
     const credentials = cs.getSpeechCredentials(vendor, 'tts');
 
-    this.logger.info({vendor, credentials}, 'Task:say - using vendor');
+    this.logger.info({language, voice}, `Task:say - using vendor: ${vendor}`);
     this.ep = ep;
     try {
       if (!credentials) {

lib/tasks/sip_refer.js

@@ -0,0 +1,101 @@
+const Task = require('./task');
+const {TaskName, TaskPreconditions} = require('../utils/constants');
+const {parseUri} = require('drachtio-srf');
+
+/**
+ * sends a sip REFER to transfer the existing call
+ */
+class TaskSipRefer extends Task {
+  constructor(logger, opts) {
+    super(logger, opts);
+    this.preconditions = TaskPreconditions.StableCall;
+
+    this.referTo = this.data.referTo;
+    this.referredBy = this.data.referredBy;
+    this.headers = this.data.headers || {};
+    this.eventHook = this.data.eventHook;
+  }
+
+  get name() { return TaskName.SipRefer; }
+
+  async exec(cs) {
+    super.exec(cs);
+    const {dlg} = cs;
+    const {referTo, referredBy} = this._normalizeReferHeaders(cs, dlg);
+
+    try {
+      this.notifyHandler = this._handleNotify.bind(this, cs, dlg);
+      dlg.on('notify', this.notifyHandler);
+      const response = await dlg.request({
+        method: 'REFER',
+        headers: {
+          ...this.headers,
+          'Refer-To': referTo,
+          'Referred-By': referredBy
+        }
+      });
+      this.referStatus = response.status;
+      this.logger.info(`TaskSipRefer:exec - received ${this.referStatus} to REFER`);
+
+      /* if we fail, fall through to next verb.  If success, we should get BYE from far end */
+      if (this.referStatus === 202) {
+        await this.awaitTaskDone();
+      }
+      else await this.performAction({refer_status: this.referStatus});
+    } catch (err) {
+      this.logger.info({err}, 'TaskSipRefer:exec - error sending REFER');
+    }
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+    const {dlg} = cs;
+    dlg.off('notify', this.notifyHandler);
+  }
+
+  async _handleNotify(cs, dlg, req, res) {
+    res.send(200);
+
+    const contentType = req.get('Content-Type');
+    this.logger.debug({body: req.body}, `TaskSipRefer:_handleNotify got ${contentType}`);
+
+    if (contentType === 'message/sipfrag') {
+      const arr = /SIP\/2\.0\s+(\d+)/.exec(req.body);
+      if (arr) {
+        const status = arr[1];
+        this.logger.debug(`TaskSipRefer:_handleNotify: call got status ${status}`);
+        if (this.eventHook) {
+          await cs.requestor.request(this.eventHook, {event: 'transfer-status', call_status: status});
+        }
+        if (status >= 200) {
+          await this.performAction({refer_status: 202, final_referred_call_status: status});
+          this.notifyTaskDone();
+        }
+      }
+    }
+  }
+
+  _normalizeReferHeaders(cs, dlg) {
+    let {referTo, referredBy} = this;
+
+    /* get IP address of the SBC to use as hostname if needed */
+    const {host} = parseUri(dlg.remote.uri);
+
+    if (!referTo.startsWith('<') && !referTo.startsWith('sip') && !referTo.startsWith('"')) {
+      /* they may have only provided a phone number/user */
+      referTo = `sip:${referTo}@${host}`;
+    }
+    if (!referredBy) {
+      /* default */
+      referredBy = cs.req?.callingNumber || dlg.local.uri;
+      this.logger.info({referredBy}, 'setting referredby');
+    }
+    if (!referredBy.startsWith('<') && !referredBy.startsWith('sip') && !referredBy.startsWith('"')) {
+      /* they may have only provided a phone number/user */
+      referredBy = `sip:${referredBy}@${host}`;
+    }
+    return {referTo, referredBy};
+  }
+}
+
+module.exports = TaskSipRefer;

lib/tasks/specs.json

@@ -9,6 +9,18 @@
       "status"
     ]
   },
+  "sip:refer": {
+    "properties": {
+      "referTo": "string",
+      "referredBy": "string",
+      "headers": "object",
+      "actionHook": "object|string",
+      "eventHook": "object|string"
+    },
+    "required": [
+      "referTo"
+    ]
+  },
   "dequeue": {
     "properties": {
       "name": "string",
@@ -92,7 +104,8 @@
       "waitHook": "object|string",
       "statusEvents": "array",
       "statusHook": "object|string",
-      "enterHook": "object|string"
+      "enterHook": "object|string",
+      "record": "#record"
     },
     "required": [
       "name"
@@ -241,6 +254,14 @@
       "url"
     ]
   },
+  "record": {
+    "properties": {
+      "path": "string"
+    },
+    "required": [
+      "path"
+    ]
+  },
   "redirect": {
     "properties": {
       "actionHook": "object|string"
@@ -327,7 +348,7 @@
     "properties": {
       "vendor": {
         "type": "string",
-        "enum": ["google", "aws", "polly", "default"]
+        "enum": ["google", "aws", "polly", "microsoft", "default"]
       },
       "language": "string",
       "voice": "string",
@@ -344,7 +365,7 @@
     "properties": {
       "vendor": {
         "type": "string",
-        "enum": ["google", "aws", "default"]
+        "enum": ["google", "aws", "microsoft", "default"]
       },
       "language": "string",
       "hints": "array",
@@ -384,7 +405,24 @@
           "mask",
           "tag"
         ]
-      }
+      },
+      "outputFormat": {
+        "type": "string",
+        "enum": [
+          "simple",
+          "detailed"
+        ]
+      },
+      "profanityOption": {
+        "type": "string",
+        "enum": [
+          "masked",
+          "removed",
+          "raw"
+        ]
+      },
+      "requestSnr": "boolean",
+      "initialSpeechTimeoutMs": "number"
     },
     "required": [
       "vendor"

lib/tasks/transcribe.js

@@ -3,6 +3,7 @@ const {
   TaskName,
   TaskPreconditions,
   GoogleTranscriptionEvents,
+  AzureTranscriptionEvents,
   AwsTranscriptionEvents
 } = require('../utils/constants');
 
@@ -38,6 +39,12 @@ class TaskTranscribe extends Task {
     this.vocabularyName = recognizer.vocabularyName;
     this.vocabularyFilterName = recognizer.vocabularyFilterName;
     this.filterMethod = recognizer.filterMethod;
+
+    /* microsoft options */
+    this.outputFormat = recognizer.outputFormat || 'simple';
+    this.profanityOption = recognizer.profanityOption || 'raw';
+    this.requestSnr = recognizer.requestSnr || false;
+    this.initialSpeechTimeoutMs = recognizer.initialSpeechTimeoutMs || 0;
   }
 
   get name() { return TaskName.Transcribe; }
@@ -53,7 +60,13 @@ class TaskTranscribe extends Task {
 
     try {
       if (!this.sttCredentials) {
-        // TODO: generate alert (actually should be done by cs.getSpeechCredentials)
+        const {writeAlerts, AlertType} = cs.srf.locals;
+        this.logger.info(`TaskTranscribe:exec - ERROR stt using ${this.vendor} requested but creds not supplied`);
+        writeAlerts({
+          account_sid: cs.accountSid,
+          alert_type: AlertType.STT_NOT_PROVISIONED,
+          vendor: this.vendor
+        }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
         throw new Error('no provisioned speech credentials for TTS');
       }
       await this._startTranscribing(cs, ep);
@@ -70,6 +83,8 @@ class TaskTranscribe extends Task {
     ep.removeCustomEventListener(AwsTranscriptionEvents.Transcription);
     ep.removeCustomEventListener(AwsTranscriptionEvents.NoAudioDetected);
     ep.removeCustomEventListener(AwsTranscriptionEvents.MaxDurationExceeded);
+    ep.removeCustomEventListener(AzureTranscriptionEvents.Transcription);
+    ep.removeCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected);
   }
 
   async kill(cs) {
@@ -96,6 +111,8 @@ class TaskTranscribe extends Task {
     ep.addCustomEventListener(AwsTranscriptionEvents.NoAudioDetected, this._onNoAudio.bind(this, cs, ep));
     ep.addCustomEventListener(AwsTranscriptionEvents.MaxDurationExceeded,
       this._onMaxDurationExceeded.bind(this, cs, ep));
+    ep.addCustomEventListener(AzureTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+    ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected, this._onNoAudio.bind(this, cs, ep));
 
     if (this.vendor === 'google') {
       if (this.sttCredentials) opts.GOOGLE_APPLICATION_CREDENTIALS = JSON.stringify(this.sttCredentials.credentials);
@@ -164,6 +181,22 @@ class TaskTranscribe extends Task {
       await ep.set(opts)
         .catch((err) => this.logger.info(err, 'TaskTranscribe:_startTranscribing with aws'));
     }
+    else if (this.vendor === 'microsoft') {
+      Object.assign(opts, {
+        'AZURE_SUBSCRIPTION_KEY': this.sttCredentials.api_key,
+        'AZURE_REGION': this.sttCredentials.region
+      });
+      if (this.hints && this.hints.length > 1) {
+        opts.AZURE_SPEECH_HINTS = this.hints.map((h) => h.trim()).join(',');
+      }
+      if (this.requestSnr) opts.AZURE_REQUEST_SNR = 1;
+      if (this.profanityOption !== 'raw') opts.AZURE_PROFANITY_OPTION = this.profanityOption;
+      if (this.initialSpeechTimeoutMs > 0) opts.AZURE_INITIAL_SPEECH_TIMEOUT_MS = this.initialSpeechTimeoutMs;
+      if (this.outputFormat !== 'simple') opts.AZURE_USE_OUTPUT_FORMAT_DETAILED = 1;
+
+      await ep.set(opts)
+        .catch((err) => this.logger.info(err, 'TaskTranscribe:_startTranscribing with azure'));
+    }
     await this._transcribe(ep);
   }
 
@@ -178,6 +211,20 @@ class TaskTranscribe extends Task {
 
   _onTranscription(cs, ep, evt) {
     if ('aws' === this.vendor && Array.isArray(evt) && evt.length > 0) evt = evt[0];
+    if ('microsoft' === this.vendor) {
+      const nbest = evt.NBest;
+      const alternatives = nbest.map((n) => {
+        return {
+          confidence: n.Confidence,
+          transcript: n.Display
+        };
+      });
+      const newEvent = {
+        is_final: evt.RecognitionStatus === 'Success',
+        alternatives
+      };
+      evt = newEvent;
+    }
     this.logger.debug(evt, 'TaskTranscribe:_onTranscription');
 
     this.cs.requestor.request(this.transcriptionHook, Object.assign({speech: evt}, this.cs.callInfo))

lib/utils/constants.json

@@ -18,6 +18,7 @@
     "Redirect": "redirect",
     "RestDial": "rest:dial",
     "SipDecline": "sip:decline",
+    "SipRefer": "sip:refer",
     "SipNotify": "sip:notify",
     "SipRedirect": "sip:redirect",
     "Say": "say",
@@ -64,6 +65,12 @@
     "NoAudioDetected": "aws_transcribe::no_audio_detected",
     "MaxDurationExceeded": "aws_transcribe::max_duration_exceeded"
   },
+  "AzureTranscriptionEvents": {
+    "Transcription": "azure_transcribe::transcription",
+    "StartOfUtterance": "azure_transcribe::start_of_utterance",
+    "EndOfUtterance": "azure_transcribe::end_of_utterance",
+    "NoSpeechDetected": "azure_transcribe::no_speech_detected"
+  },
   "ListenEvents": {
     "Connect": "mod_audio_fork::connect",
     "ConnectFailure": "mod_audio_fork::connect_failed",

lib/utils/db-utils.js

@@ -21,6 +21,11 @@ const speechMapper = (cred) => {
     obj.access_key_id = o.access_key_id;
     obj.secret_access_key = o.secret_access_key;
   }
+  else if ('microsoft' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = o.api_key;
+    obj.region = o.region;
+  }
   return obj;
 };
 
@@ -37,7 +42,8 @@ module.exports = (logger, srf) => {
     /* search at the service provider level if we don't find it at the account level */
     const haveGoogle = speech.find((s) => s.vendor === 'google');
     const haveAws = speech.find((s) => s.vendor === 'aws');
-    if (!haveGoogle || !haveAws) {
+    const haveMicrosoft = speech.find((s) => s.vendor === 'microsoft');
+    if (!haveGoogle || !haveAws || !haveMicrosoft) {
       const [r3] = await pp.query(sqlSpeechCredentialsForSP, account_sid);
       if (r3.length) {
         if (!haveGoogle) {
@@ -48,6 +54,10 @@ module.exports = (logger, srf) => {
           const aws = r3.find((s) => s.vendor === 'aws');
           if (aws) speech.push(speechMapper(aws));
         }
+        if (!haveMicrosoft) {
+          const ms = r3.find((s) => s.vendor === 'microsoft');
+          if (ms) speech.push(speechMapper(ms));
+        }
       }
     }
 

lib/utils/encrypt-decrypt.js

@@ -1,5 +1,5 @@
 const crypto = require('crypto');
-const algorithm = 'aes-256-ctr';
+const algorithm = 'aes-256-cbc';
 const iv = crypto.randomBytes(16);
 const secretKey = crypto.createHash('sha256')
   .update(String(process.env.JWT_SECRET))

lib/utils/requestor.js

@@ -100,15 +100,15 @@ class Requestor {
 
     assert.ok(url, 'Requestor:request url was not provided');
     assert.ok, (['GET', 'POST'].includes(method), `Requestor:request method must be 'GET' or 'POST' not ${method}`);
-
-    this.logger.debug({hook, payload}, `Requestor:request ${method} ${url}`);
+    const {url: urlInfo = hook, method: methodInfo = 'POST'} = hook;  // mask user/pass
+    this.logger.debug({url: urlInfo, method: methodInfo, payload}, `Requestor:request ${method} ${url}`);
     const startAt = process.hrtime();
 
     let buf;
     try {
       const sigHeader = generateSigHeader(payload, this.secret);
       const headers = {...sigHeader, ...this.authHeader};
-      this.logger.info({url, headers}, 'send webhook');
+      //this.logger.info({url, headers}, 'send webhook');
       buf = isRelativeUrl(url) ?
         await this.post(url, payload, headers) :
         await bent(method, 'buffer', 200, 201, 202)(url, payload, headers);

package.json

@@ -26,9 +26,9 @@
     "jslint": "eslint app.js lib"
   },
   "dependencies": {
-    "@jambonz/db-helpers": "^0.6.13",
+    "@jambonz/db-helpers": "^0.6.14",
     "@jambonz/mw-registrar": "^0.2.1",
-    "@jambonz/realtimedb-helpers": "^0.4.8",
+    "@jambonz/realtimedb-helpers": "^0.4.9",
     "@jambonz/stats-collector": "^0.1.5",
     "@jambonz/time-series": "^0.1.5",
     "aws-sdk": "^2.846.0",
@@ -36,7 +36,7 @@
     "cidr-matcher": "^2.1.1",
     "debug": "^4.3.1",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^2.0.7",
+    "drachtio-fsmrf": "^2.0.11",
     "drachtio-srf": "^4.4.55",
     "express": "^4.17.1",
     "ip": "^1.1.5",

test/db/create-and-populate-schema.sql

@@ -587,7 +587,8 @@ DROP TABLE IF EXISTS `speech_credentials`;
 /*!50503 SET character_set_client = utf8mb4 */;
 CREATE TABLE `speech_credentials` (
   `speech_credential_sid` char(36) NOT NULL,
-  `account_sid` char(36) NOT NULL,
+ `service_provider_sid` CHAR(36),
+ `account_sid` char(36) NOT NULL,
   `vendor` varchar(255) NOT NULL,
   `credential` VARCHAR(8192) NOT NULL,
   `use_for_tts` tinyint(1) DEFAULT '1',
@@ -611,7 +612,7 @@ CREATE TABLE `speech_credentials` (
 
 LOCK TABLES `speech_credentials` WRITE;
 /*!40000 ALTER TABLE `speech_credentials` DISABLE KEYS */;
-INSERT INTO `speech_credentials` VALUES ('2add163c-34f2-45c6-a016-f955d218ffb6','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','google','credential-goes-here',1,1,NULL,'2021-04-03 15:42:10',1,1),('84154212-5c99-4c94-8993-bc2a46288daa','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','aws','credential-goes-here',0,0,NULL,NULL,NULL,NULL);
+INSERT INTO `speech_credentials` VALUES ('2add163c-34f2-45c6-a016-f955d218ffb6',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','google','credential-goes-here',1,1,NULL,'2021-04-03 15:42:10',1,1),('84154212-5c99-4c94-8993-bc2a46288daa',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','aws','credential-goes-here',0,0,NULL,NULL,NULL,NULL);
 /*!40000 ALTER TABLE `speech_credentials` ENABLE KEYS */;
 UNLOCK TABLES;
 

test/db/jambones-sql.sql

@@ -20,12 +20,16 @@ DROP TABLE IF EXISTS lcr_routes;
 
 DROP TABLE IF EXISTS predefined_sip_gateways;
 
+DROP TABLE IF EXISTS predefined_smpp_gateways;
+
 DROP TABLE IF EXISTS predefined_carriers;
 
 DROP TABLE IF EXISTS account_offers;
 
 DROP TABLE IF EXISTS products;
 
+DROP TABLE IF EXISTS schema_version;
+
 DROP TABLE IF EXISTS api_keys;
 
 DROP TABLE IF EXISTS sbc_addresses;
@@ -148,6 +152,20 @@ predefined_carrier_sid CHAR(36) NOT NULL,
 PRIMARY KEY (predefined_sip_gateway_sid)
 );
 
+CREATE TABLE predefined_smpp_gateways
+(
+predefined_smpp_gateway_sid CHAR(36) NOT NULL UNIQUE ,
+ipv4 VARCHAR(128) NOT NULL COMMENT 'ip address or DNS name of the gateway. ',
+port INTEGER NOT NULL DEFAULT 2775 COMMENT 'smpp signaling port',
+inbound BOOLEAN NOT NULL COMMENT 'if true, whitelist this IP to allow inbound SMS from the gateway',
+outbound BOOLEAN NOT NULL COMMENT 'i',
+netmask INTEGER NOT NULL DEFAULT 32,
+is_primary BOOLEAN NOT NULL DEFAULT 1,
+use_tls BOOLEAN DEFAULT 0,
+predefined_carrier_sid CHAR(36) NOT NULL,
+PRIMARY KEY (predefined_smpp_gateway_sid)
+);
+
 CREATE TABLE products
 (
 product_sid CHAR(36) NOT NULL UNIQUE ,
@@ -174,6 +192,11 @@ stripe_product_id VARCHAR(56) NOT NULL,
 PRIMARY KEY (account_offer_sid)
 );
 
+CREATE TABLE schema_version
+(
+version VARCHAR(16)
+);
+
 CREATE TABLE api_keys
 (
 api_key_sid CHAR(36) NOT NULL UNIQUE ,
@@ -420,6 +443,10 @@ CREATE INDEX predefined_sip_gateway_sid_idx ON predefined_sip_gateways (predefin
 CREATE INDEX predefined_carrier_sid_idx ON predefined_sip_gateways (predefined_carrier_sid);
 ALTER TABLE predefined_sip_gateways ADD FOREIGN KEY predefined_carrier_sid_idxfk (predefined_carrier_sid) REFERENCES predefined_carriers (predefined_carrier_sid);
 
+CREATE INDEX predefined_smpp_gateway_sid_idx ON predefined_smpp_gateways (predefined_smpp_gateway_sid);
+CREATE INDEX predefined_carrier_sid_idx ON predefined_smpp_gateways (predefined_carrier_sid);
+ALTER TABLE predefined_smpp_gateways ADD FOREIGN KEY predefined_carrier_sid_idxfk_1 (predefined_carrier_sid) REFERENCES predefined_carriers (predefined_carrier_sid);
+
 CREATE INDEX product_sid_idx ON products (product_sid);
 CREATE INDEX account_product_sid_idx ON account_products (account_product_sid);
 CREATE INDEX account_subscription_sid_idx ON account_products (account_subscription_sid);
@@ -545,4 +572,4 @@ ALTER TABLE accounts ADD FOREIGN KEY queue_event_hook_sid_idxfk (queue_event_hoo
 
 ALTER TABLE accounts ADD FOREIGN KEY device_calling_application_sid_idxfk (device_calling_application_sid) REFERENCES applications (application_sid);
 
-SET FOREIGN_KEY_CHECKS=1;
+SET FOREIGN_KEY_CHECKS=0;