support verbio speech (#757)

* support verbio speech

* wip

* wip

* wip

* update speech utils

* update verb specification

.github/workflows/build.yml

@@ -6,10 +6,10 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
-      - uses: actions/setup-node@v3
+      - uses: actions/setup-node@v4
         with:
-          node-version: 18
+          node-version: 20
       - run: npm ci
       - run: npm run jslint
       - run: docker pull drachtio/sipp

.gitignore

@@ -42,3 +42,4 @@ ecosystem.config.js
 test/credentials/*.json
 run-tests.sh
 run-coverage.sh
+.vscode

.vscode/launch.json

@@ -1,17 +0,0 @@
-{
-  // Use IntelliSense to learn about possible attributes.
-  // Hover to view descriptions of existing attributes.
-  // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
-  "version": "0.2.0",
-  "configurations": [
-    {
-      "type": "node",
-      "request": "launch",
-      "name": "Launch Program",
-      "program": "${workspaceFolder}/test/index.js",
-      "env": {
-        "NODE_ENV": "test"
-      }
-    }
-  ]
-}

README.md

@@ -40,6 +40,8 @@ Configuration is provided via environment variables:
 |JAMBONZ_RECORD_WS_BASE_URL| recording websocket URL to send the recording audio|no|
 |JAMBONZ_RECORD_WS_USERNAME| recording websocket username|no|
 |JAMBONZ_RECORD_WS_PASSWORD| recording websocket password|no|
+|ANCHOR_MEDIA_ALWAYS| keep media on media server|no|
+|JAMBONZ_DISABLE_DIAL_PAI_HEADER| control P-Asserted-Identity header on B-Leg|no|
 
 ### running under pm2
 Typically, this application runs under [pm2](https://pm2.io) using an [ecosystem.config.js](https://pm2.keymetrics.io/docs/usage/application-declaration/) file similar to this:

app.js

@@ -109,7 +109,7 @@ const disconnect = () => {
     httpServer?.on('close', resolve);
     httpServer?.close();
     srf.disconnect();
-    srf.locals.mediaservers.forEach((ms) => ms.disconnect());
+    srf.locals.mediaservers?.forEach((ms) => ms.disconnect());
   });
 };
 

data/example-voicemail-greetings.json

@@ -9,7 +9,112 @@
     "can't take your call",
     "will get back to you",
     "I'll get back to you",
-    "we are unable"
+    "we are unable",
+    "Unable to take your call now",
+    "I'll reply soon",
+    "I'll call back",
+    "I'll reach out to you as soon as possible",
+    "Leave a message",
+    "Away from phone",
+    "Not available now",
+    "I'll return call",
+    "On another call",
+    "Currently on another call",
+    "I will return call later",
+    "Busy please leave message",
+    "Message will be returned promptly",
+    "Currently unavailable to answer",
+    "Planning to return your call soon",
+    "Apologies for missing your call",
+    "Not by the phone at the moment",
+    "Expecting to return your call",
+    "Currently not accessible",
+    "Intend to call back",
+    "Appreciate your patience!",
+    "Engaged in another conversation",
+    "I Will respond promptly",
+    "Kindly leave a message",
+    "Currently occupied leave a message",
+    "Unfortunately unable to answer right now",
+    "Occupied at the moment",
+    "Not present leave a message",
+    "Regrettably unavailable kindly leave a message",
+    "Will ensure a prompt response to your message",
+    "Currently engaged",
+    "Will return your call at the earliest opportunity",
+    "Your message will receive my prompt attention",
+    "I'll respond as soon as I can",
+    "Your message is important please leave it after the beep",
+    "Away from the phone at the moment",
+    "Unable to answer right now",
+    "Engaged in another task",
+    "Not by the phone presently",
+    "I'll respond at my earliest convenience",
+    "Away from the phone momentarily",
+    "I'll return your call shortly",
+    "Currently not able to answer",
+    "Your message is important please leave it after the tone",
+    "I'm unable to take your call right now",
+    "Please leave your message for me",
+    "I'll get back to you soon",
+    "Your call has been missed",
+    "Please leave a detailed message for me to respond to",
+    "Leave a message I'll make sure to respond",
+    "Feel free to leave a message",
+    "Your call is important to me",
+    "I'll get back to you shortly",
+    "Your message will be attended to promptly",
+    "Not available at the moment",
+    "I'll be sure to get back to you",
+    "I'll call you back soon",
+    "I'll ensure a prompt response",
+    "Sorry for the inconvenience",
+    "I'll return your call",
+    "I'll make sure to get back to you",
+    "I'll call you back shortly",
+    "I'll return your call as soon as possible",
+    "Apologies for the inconvenience leave your message",
+    "Your call is appreciated",
+    "I'm unavailable to answer",
+    "I'm currently away",
+    "I'll return your call as soon as I can",
+    "I'm away from the phone",
+    "I'm currently unavailable to take your call",
+    "Sorry for missing your call",
+    "I'll ensure it receives my immediate attention",
+    "I'm away from the phone momentarily",
+    "I'll reach out to you shortly",
+    "Apologies for the inconvenience",
+    "Currently occupied",
+    "Unable to answer your call at the moment",
+    "I'll make sure to follow up with you",
+    "Sorry for not being available",
+    "I'll reach out to you as soon as I can",
+    "I'm currently engaged",
+    "I'm currently busy",
+    "I'm currently unavailable",
+    "I'll respond to you at my earliest convenience",
+    "Your message is appreciated",
+    "I'll get back to you promptly",
+    "I'll get back to you without delay",
+    "Currently away from the phone",
+    "I'll return your call at my earliest opportunity",
+    "Sorry for the missed call",
+    "I'll make sure to address your concerns",
+    "Please provide your details for a callback",
+    "I'll make every effort to respond promptly",
+    "I'll ensure it's attended to promptly",
+    "Away from the phone temporarily",
+    "I'll get back to you as soon as I return",
+    "Currently not in a position to answer your call",
+    "Your call cannot be answered at the moment",
+    "I'll ensure to respond as soon as I'm able",
+    "Your call is important please leave a message",
+    "Unable to answer right now please leave your message",
+    "Currently not accessible intending to return your call",
+    "I'll respond promptly to your message",
+    "leave a memo",
+    "please leave a memo"
   ],
   "es-ES": [
     "le pasamos la llamada",

docs/contributing.md

@@ -45,7 +45,7 @@ The GCP credential is the JSON service key in stringified format.
 
 #### Install Docker
 
-The test suite ralso equires [Docker](https://www.docker.com/) and docker-compose to be installed on your laptop.  Docker is used to set up a network with all of the elements required to test the jambonz-feature-server in a black-box type of fashion.
+The test suite also requires [Docker](https://www.docker.com/) and docker-compose to be installed on your laptop.  Docker is used to set up a network with all of the elements required to test the jambonz-feature-server in a black-box type of fashion.
 
 Once you have docker installed, you can optionally make sure everything Docker-wise is working properly by running this command from the project folder:
 

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 +130,10 @@ 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 = parseInt(process.env.JAMBONES_EAGERLY_PRE_CACHE_AUDIO, 10) || 0;
+
+const JAMBONES_USE_FREESWITCH_TIMER_FD = process.env.JAMBONES_USE_FREESWITCH_TIMER_FD;
+
 module.exports = {
   JAMBONES_MYSQL_HOST,
   JAMBONES_MYSQL_USER,
@@ -151,6 +156,7 @@ 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 +199,7 @@ 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,
@@ -208,5 +215,6 @@ module.exports = {
   JAMBONZ_RECORD_WS_USERNAME,
   JAMBONZ_RECORD_WS_PASSWORD,
   JAMBONZ_DISABLE_DIAL_PAI_HEADER,
-  JAMBONES_DISABLE_DIRECT_P2P_CALL
+  JAMBONES_DISABLE_DIRECT_P2P_CALL,
+  JAMBONES_USE_FREESWITCH_TIMER_FD
 };

lib/dynamic-apps.js

@@ -30,6 +30,20 @@ const appsMap = {
         }
       ]
     }]
+  },
+  conference: {
+    // Dummy hook to follow later feature server logic.
+    call_hook: {
+      url: 'https://jambonz.org',
+      method: 'GET'
+    },
+    account_sid: '',
+    app_json: [{
+      verb: 'conference',
+      name: '',
+      beep: false,
+      startConferenceOnEnter: true
+    }]
   }
 };
 
@@ -38,6 +52,7 @@ const createJambonzApp = (type, {account_sid, name, caller_id}) => {
   app.account_sid = account_sid;
   switch (type) {
     case 'queue':
+    case 'conference':
       app.app_json[0].name = name;
       break;
     case 'user':

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

@@ -5,7 +5,7 @@ const CallInfo = require('../../session/call-info');
 const {CallDirection, CallStatus} = require('../../utils/constants');
 const uuidv4 = require('uuid-random');
 const SipError = require('drachtio-srf').SipError;
-const { validationResult } = require('express-validator');
+const { validationResult, body } = require('express-validator');
 const { validate } = require('@jambonz/verb-specifications');
 const sysError = require('./error');
 const HttpRequestor = require('../../utils/http-requestor');
@@ -13,7 +13,7 @@ const WsRequestor = require('../../utils/ws-requestor');
 const RootSpan = require('../../utils/call-tracer');
 const dbUtils = require('../../utils/db-utils');
 const { mergeSdpMedia, extractSdpMedia } = require('../../utils/sdp-utils');
-const { createCallSchema } = require('../schemas/create-call');
+const { createCallSchema, customSanitizeFunction } = require('../schemas/create-call');
 
 const removeNullProperties = (obj) => (Object.keys(obj).forEach((key) => obj[key] === null && delete obj[key]), obj);
 const removeNulls = (req, res, next) => {
@@ -24,6 +24,12 @@ const removeNulls = (req, res, next) => {
 router.post('/',
   removeNulls,
   createCallSchema,
+  body('tag').custom((value) => {
+    if (value) {
+      customSanitizeFunction(value);
+    }
+    return true;
+  }),
   async(req, res) => {
     const {logger} = req.app.locals;
     const errors = validationResult(req);

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

@@ -45,7 +45,7 @@ function retrieveCallSession(callSid, opts) {
 router.post('/:callSid', async(req, res) => {
   const logger = req.app.locals.logger;
   const callSid = req.params.callSid;
-  logger.debug({body: req.body}, 'got upateCall request');
+  logger.debug({body: req.body}, 'got updateCall request');
   try {
     const cs = retrieveCallSession(callSid, req.body);
     if (!cs) {

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

@@ -46,11 +46,6 @@ const createCallSchema = checkSchema({
     optional: true,
     errorMessage: 'Invalid tag',
   },
-  'tag.*': {
-    trim: true,
-    escape: true,
-    stripLow: true,
-  },
   app_json: {
     isString: true,
     optional: true,
@@ -109,6 +104,31 @@ const createCallSchema = checkSchema({
   }
 }, ['body']);
 
-module.exports = {
+const customSanitizeFunction = (value) => {
-  createCallSchema
+  try {
+    if (Array.isArray(value)) {
+      value = value.map((item) => customSanitizeFunction(item));
+    } else if (typeof value === 'object') {
+      Object.keys(value).forEach((key) => {
+        value[key] = customSanitizeFunction(value[key]);
+      });
+    } else if (typeof value === 'string') {
+      /* trims characters at the beginning and at the end of a string */
+      value = value.trim();
+
+      /* Verify strings including 'http' via new URL */
+      if (value.includes('http')) {
+        value = new URL(value).toString();
+      }
+    }
+  } catch (error) {
+    value = `Error: ${error.message}`;
+  }
+
+  return value;
+};
+
+module.exports = {
+  createCallSchema,
+  customSanitizeFunction
 };

lib/middleware.js

@@ -52,14 +52,20 @@ module.exports = function(srf, logger) {
     let clientDb = null;
     if (req.has('X-Authenticated-User')) {
       req.locals.originatingUser = req.get('X-Authenticated-User');
+      let clientSettings;
       const arr = /^(.*)@(.*)/.exec(req.locals.originatingUser);
       if (arr) {
-        [clientDb] = await lookupClientByAccountAndUsername(account_sid, arr[1]);
+        [clientSettings] = await lookupClientByAccountAndUsername(account_sid, arr[1]);
       }
+      clientDb = await registrar.query(req.locals.originatingUser);
+      clientDb = {
+        ...clientDb,
+        ...clientSettings,
+      };
     }
 
     // check for call to application
-    if (uri.user?.startsWith('app-') && req.locals.originatingUser && clientDb.allow_direct_app_calling) {
+    if (uri.user?.startsWith('app-') && req.locals.originatingUser && clientDb?.allow_direct_app_calling) {
       const application_sid = uri.user.match(/app-(.*)/)[1];
       logger.debug(`got application from Request URI header: ${application_sid}`);
       req.locals.application_sid = application_sid;
@@ -69,13 +75,19 @@ module.exports = function(srf, logger) {
       req.locals.application_sid = application_sid;
     }
     // check for call to queue
-    if (uri.user?.startsWith('queue-') && req.locals.originatingUser && clientDb.allow_direct_queue_calling) {
+    else if (uri.user?.startsWith('queue-') && req.locals.originatingUser && clientDb?.allow_direct_queue_calling) {
       const queue_name = uri.user.match(/queue-(.*)/)[1];
       logger.debug(`got Queue from Request URI header: ${queue_name}`);
       req.locals.queue_name = queue_name;
     }
+    // check for call to conference
+    else if (uri.user?.startsWith('conference-') && req.locals.originatingUser && clientDb?.allow_direct_app_calling) {
+      const conference_id = uri.user.match(/conference-(.*)/)[1];
+      logger.debug(`got Conference from Request URI header: ${conference_id}`);
+      req.locals.conference_id = conference_id;
+    }
     // check for call to registered user
-    if (!JAMBONES_DISABLE_DIRECT_P2P_CALL && req.locals.originatingUser && clientDb.allow_direct_user_calling) {
+    else if (!JAMBONES_DISABLE_DIRECT_P2P_CALL && req.locals.originatingUser && clientDb?.allow_direct_user_calling) {
       const arr = /^(.*)@(.*)/.exec(req.locals.originatingUser);
       if (arr) {
         const sipRealm = arr[2];
@@ -91,7 +103,7 @@ module.exports = function(srf, logger) {
     if (req.has('X-MS-Teams-Tenant-FQDN')) req.locals.msTeamsTenant = req.get('X-MS-Teams-Tenant-FQDN');
     if (req.has('X-Cisco-Recording-Participant')) {
       const ciscoParticipants = req.get('X-Cisco-Recording-Participant');
-      const regex = /sip:[\d]+@[\d]+\.[\d]+\.[\d]+\.[\d]+/g;
+      const regex = /sip:[a-zA-Z0-9]+@[a-zA-Z0-9.-_]+/g;
       const sipURIs = ciscoParticipants.match(regex);
       logger.info(`X-Cisco-Recording-Participant : ${sipURIs} `);
       if (sipURIs && sipURIs.length > 0) {
@@ -160,7 +172,7 @@ module.exports = function(srf, logger) {
         };
         logger.info({callId, metadata, sdp}, 'successfully parsed SIPREC payload');
       } catch (err) {
-        logger.info({callId}, 'Error parsing multipart payload');
+        logger.info({err, callId}, 'Error parsing multipart payload');
         return res.send(503);
       }
     }
@@ -231,6 +243,9 @@ module.exports = function(srf, logger) {
         logger.debug(`calling to registered user ${req.locals.called_user}, generating dial app`);
         app = createJambonzApp('user',
           {account_sid, name: req.locals.called_user, caller_id: req.locals.callingNumber});
+      } else if (req.locals.conference_id) {
+        logger.debug(`calling to conference ${req.locals.conference_id}, generating conference app`);
+        app = createJambonzApp('conference', {account_sid, name: req.locals.conference_id});
       } else if (req.locals.application_sid) {
         app = await lookupAppBySid(req.locals.application_sid);
       } else if (req.locals.originatingUser) {
@@ -376,7 +391,7 @@ 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/adulting-call-session.js

@@ -1,4 +1,6 @@
 const CallSession = require('./call-session');
+const {CallStatus} = require('../utils/constants');
+const moment = require('moment');
 
 /**
  * @classdesc Subclass of CallSession.  Represents a CallSession
@@ -19,12 +21,14 @@ class AdultingCallSession extends CallSession {
       rootSpan
     });
     this.sd = singleDialer;
+    this.req = callInfo.req;
 
     this.sd.dlg.on('destroy', () => {
       this.logger.info('AdultingCallSession: called party hung up');
       this._callReleased();
     });
     this.sd.emit('adulting');
+    this.on('callStatusChange', this._notifyCallStatusChange.bind(this));
   }
 
   get dlg() {
@@ -49,6 +53,26 @@ class AdultingCallSession extends CallSession {
   }
 
   _callerHungup() {
+    this._hangup('caller');
+  }
+
+  _jambonzHangup() {
+    this._hangup();
+  }
+
+  _hangup(terminatedBy = 'jambonz') {
+    if (this.dlg.connectTime) {
+      const duration = moment().diff(this.dlg.connectTime, 'seconds');
+      this.rootSpan.setAttributes({'call.termination': `hangup by ${terminatedBy}`});
+      this.callInfo.callTerminationBy = terminatedBy;
+      this.emit('callStatusChange', {
+        callStatus: CallStatus.Completed,
+        duration
+      });
+    }
+    this.logger.info(`InboundCallSession: ${terminatedBy} hung up`);
+    this._callReleased();
+    this.req.removeAllListeners('cancel');
   }
 }
 

lib/session/confirm-call-session.js

@@ -34,6 +34,9 @@ class ConfirmCallSession extends CallSession {
   _callerHungup() {
   }
 
+  _jambonzHangup() {
+  }
+
 
 }
 

lib/session/inbound-call-session.js

@@ -67,15 +67,27 @@ class InboundCallSession extends CallSession {
    * This is invoked when the caller hangs up, in order to calculate the call duration.
    */
   _callerHungup() {
+    this._hangup('caller');
+  }
+
+  _jambonzHangup() {
+    this._hangup();
+  }
+
+  _hangup(terminatedBy = 'jambonz') {
+    if (this.dlg === null) {
+      this.logger.info('InboundCallSession:_hangup - race condition, dlg cleared by app hangup');
+      return;
+    }
+    this.logger.info(`InboundCallSession: ${terminatedBy} hung up`);
     assert(this.dlg.connectTime);
     const duration = moment().diff(this.dlg.connectTime, 'seconds');
-    this.rootSpan.setAttributes({'call.termination': 'hangup by caller'});
+    this.rootSpan.setAttributes({'call.termination': `hangup by ${terminatedBy}`});
-    this.callInfo.callTerminationBy = 'caller';
+    this.callInfo.callTerminationBy = terminatedBy;
     this.emit('callStatusChange', {
       callStatus: CallStatus.Completed,
       duration
     });
-    this.logger.info('InboundCallSession: caller hung up');
     this._callReleased();
     this.req.removeAllListeners('cancel');
   }

lib/session/rest-call-session.js

@@ -49,13 +49,21 @@ class RestCallSession extends CallSession {
    * This is invoked when the called party hangs up, in order to calculate the call duration.
    */
   _callerHungup() {
+    this._hangup('caller');
+  }
+
+  _jambonzHangup() {
+    this._hangup();
+  }
+
+  _hangup(terminatedBy = 'jamboz') {
     if (this.restDialTask) {
       this.restDialTask.turnOffAmd();
     }
-    this.callInfo.callTerminationBy = 'caller';
+    this.callInfo.callTerminationBy = terminatedBy;
     const duration = moment().diff(this.dlg.connectTime, 'seconds');
     this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});
-    this.logger.debug('RestCallSession: called party hung up');
+    this.logger.debug(`RestCallSession: called party hung up by ${terminatedBy}`);
     this._callReleased();
   }
 

lib/tasks/answer.js

@@ -0,0 +1,22 @@
+const Task = require('./task');
+const {TaskName, TaskPreconditions} = require('../utils/constants');
+
+/**
+ * Answer the call.
+ * Note: This is rarely used, as the call is typically answered automatically when required by the app,
+ * but it can be useful to force an answer before a pause in some cases
+ */
+class TaskAnswer extends Task {
+  constructor(logger, opts) {
+    super(logger, opts);
+    this.preconditions = TaskPreconditions.Endpoint;
+  }
+
+  get name() { return TaskName.Answer; }
+
+  async exec(cs) {
+    super.exec(cs);
+  }
+}
+
+module.exports = TaskAnswer;

lib/tasks/conference.js

@@ -60,6 +60,8 @@ class Conference extends Task {
 
     this.emitter = new Emitter();
     this.results = {};
+    this.coaching = [];
+    this.speakOnlyTo = this.data.speakOnlyTo;
 
     // transferred from another server in order to bridge to a local caller?
     if (this.data._ && this.data._.connectTime) {
@@ -348,16 +350,27 @@ class Conference extends Task {
       Object.assign(opts, {flags: {
         ...(this.endConferenceOnExit && {endconf: true}),
         ...(this.startConferenceOnEnter && {moderator: true}),
-        ...(this.joinMuted && {joinMuted: true}),
+        ...((this.joinMuted || this.speakOnlyTo) && {joinMuted: true}),
       }});
+
+      /**
+       * Note on the above: if we are joining in "coaching" mode (ie only going to heard by a subset of participants)
+       * then we join muted temporarily, and then unmute ourselves once we have identified the subset of participants
+       * to whom we will be speaking.
+       */
     }
 
     try {
       const {memberId, confUuid} = await this.ep.join(this.confName, opts);
       this.logger.debug({memberId, confUuid}, `Conference:_joinConference: successfully joined ${this.confName}`);
-      this.memberId = memberId;
+      this.memberId = parseInt(memberId, 10);
       this.confUuid = confUuid;
 
+      // set a tag for this member, if provided
+      if (this.data.memberTag) {
+        this.setMemberTag(this.data.memberTag);
+      }
+
       cs.setConferenceDetails(memberId, this.confName, confUuid);
       const response = await this.ep.api('conference', [this.confName, 'get', 'count']);
       if (response.body && /\d+/.test(response.body)) this.participantCount = parseInt(response.body);
@@ -384,6 +397,9 @@ class Conference extends Task {
           .catch((err) => {});
       }
 
+      if (this.speakOnlyTo) {
+        this.setCoachMode(this.speakOnlyTo);
+      }
     } catch (err) {
       this.logger.error(err, `Failed to join conference ${this.confName}`);
       throw err;
@@ -428,7 +444,15 @@ class Conference extends Task {
     }
   }
 
-  async doConferenceHold(cs, opts) {
+  doConferenceMute(cs, opts) {
+    assert (cs.isInConference);
+
+    const mute = opts.conf_mute_status === 'mute';
+    this.ep.api(`conference ${this.confName} ${mute ? 'mute' : 'unmute'} ${this.memberId}`)
+      .catch((err) => this.logger.info({err}, 'Error muting or unmuting participant'));
+  }
+
+  doConferenceHold(cs, opts) {
     assert (cs.isInConference);
 
     const {conf_hold_status, wait_hook} = opts;
@@ -465,6 +489,40 @@ class Conference extends Task {
     }
   }
 
+  async doConferenceParticipantAction(cs, opts) {
+    const {action, tag} = opts;
+
+    switch (action) {
+      case 'tag':
+        await this.setMemberTag(tag);
+        break;
+      case 'untag':
+        await this.clearMemberTag();
+        break;
+      case 'coach':
+        await this.setCoachMode(tag);
+        break;
+      case 'uncoach':
+        await this.clearCoachMode();
+        break;
+      case 'hold':
+        this.doConferenceHold(cs, {conf_hold_status: 'hold'});
+        break;
+      case 'unhold':
+        this.doConferenceHold(cs, {conf_hold_status: 'unhold'});
+        break;
+      case 'mute':
+        this.doConferenceMute(cs, {conf_mute_status: 'mute'});
+        break;
+      case 'unmute':
+        this.doConferenceMute(cs, {conf_mute_status: 'unmute'});
+        break;
+      default:
+        this.logger.info(`Conference:doConferenceParticipantAction - unhandled action ${action}`);
+        break;
+    }
+  }
+
   async _doWaitHookWhileOnHold(cs, dlg, wait_hook) {
     do {
       try {
@@ -530,7 +588,7 @@ class Conference extends Task {
       const response = await this.ep.api('conference', [this.confName, 'get', 'count']);
       if (response.body && confNoMatch(response.body)) this.participantCount = 0;
       else if (response.body && /^\d+$/.test(response.body)) this.participantCount = parseInt(response.body) - 1;
-      this.logger.debug({response}, `Conference:_doFinalMemberCheck conference count ${this.participantCount}`);
+      this.logger.debug(`Conference:_doFinalMemberCheck conference count ${this.participantCount}`);
     } catch (err) {
       this.logger.info({err}, 'Conference:_doFinalMemberCheck error retrieving count (we were probably kicked');
     }
@@ -642,11 +700,19 @@ class Conference extends Task {
   }
 
   // conference event handlers
+  _onAddMember(logger, cs, evt) {
+    const memberId = parseInt(evt.getHeader('Member-ID')) ;
+    if (this.speakOnlyTo) {
+      logger.debug(`Conference:_onAddMember - member ${memberId} added to ${this.confName}, updating coaching mode`);
+      this.setCoachMode(this.speakOnlyTo).catch(() => {});
+    }
+    else logger.debug(`Conference:_onAddMember - member ${memberId} added to conference ${this.confName}`);
+  }
   _onDelMember(logger, cs, evt) {
     const memberId = parseInt(evt.getHeader('Member-ID')) ;
     this.participantCount = parseInt(evt.getHeader('Conference-Size'));
     if (memberId === this.memberId) {
-      this.logger.info(`Conference:_onDelMember - I was dropped from conference ${this.confName}, task is complete`);
+      logger.info(`Conference:_onDelMember - I was dropped from conference ${this.confName}, task is complete`);
       this.replaceEndpointAndEnd(cs);
     }
   }
@@ -675,6 +741,89 @@ class Conference extends Task {
     }
   }
 
+  _onTag(logger, cs, evt) {
+    const memberId = parseInt(evt.getHeader('Member-ID')) ;
+    const tag = evt.getHeader('Tag') || '';
+    if (memberId !== this.memberId && this.speakOnlyTo) {
+      logger.info(`Conference:_onTag - member ${memberId} set tag to '${tag }'; updating coach mode accordingly`);
+      this.setCoachMode(this.speakOnlyTo).catch(() => {});
+    }
+  }
+
+  /**
+   * Set the conference to "coaching" mode, where the audio of the participant is only heard
+   * by a subset of the participants in the conference.
+   * We do this by first getting all of the members who do *not* have this tag, and then
+   * we configure this members audio to not be sent to them.
+   * @param {string} speakOnlyTo - tag of the members who should receive our audio
+   *
+   * N.B.: this feature requires jambonz patches to freeswitch mod_conference
+   */
+  async setCoachMode(speakOnlyTo) {
+    this.speakOnlyTo = speakOnlyTo;
+    if (!this.memberId) {
+      this.logger.info('Conference:_setCoachMode: no member id yet');
+      return;
+    }
+    try {
+      const members = (await this.ep.getNonMatchingConfParticipants(this.confName, speakOnlyTo))
+        .filter((m) => m !== this.memberId);
+      if (members.length === 0) {
+        this.logger.info({members}, 'Conference:_setCoachMode: all participants have the tag, so all will hear me');
+        if (this.coaching.length) {
+          await this.ep.api('conference', [this.confName, 'relate', this.memberId, this.coaching.join(','), 'clear']);
+          this.coaching = [];
+        }
+      }
+      else {
+        const memberList = members.join(',');
+        this.logger.info(`Conference:_setCoachMode: my audio will NOT be sent to ${memberList}`);
+        await this.ep.api('conference', [this.confName, 'relate', this.memberId, memberList, 'nospeak']);
+        this.coaching = members;
+      }
+    } catch (err) {
+      this.logger.error({err, speakOnlyTo}, '_setCoachMode: Error');
+    }
+  }
+
+  async clearCoachMode() {
+    if (!this.memberId) return;
+    try {
+      if (this.coaching.length === 0) {
+        this.logger.info('Conference:_clearCoachMode: no coaching mode to clear');
+      }
+      else {
+        const memberList = this.coaching.join(',');
+        this.logger.info(`Conference:_clearCoachMode: now sending my audio to all, including ${memberList}`);
+        await this.ep.api('conference', [this.confName, 'relate', this.memberId, memberList, 'clear']);
+      }
+      this.speakOnlyTo = null;
+      this.coaching = [];
+    } catch (err) {
+      this.logger.error({err}, '_clearCoachMode: Error');
+    }
+  }
+
+  async setMemberTag(tag) {
+    try {
+      await this.ep.api('conference', [this.confName, 'tag', this.memberId, tag]);
+      this.logger.info(`Conference:setMemberTag: set tag for ${this.memberId} to ${tag}`);
+      this.memberTag = tag;
+    } catch (err) {
+      this.logger.error({err}, `Error setting tag for ${this.memberId} to ${tag}`);
+    }
+  }
+
+  async clearMemberTag() {
+    try {
+      await this.ep.api('conference', [this.confName, 'tag', this.memberId]);
+      this.logger.info(`Conference:setMemberTag: clearing tag for ${this.memberId}`);
+      this.memberTag = null;
+    } catch (err) {
+      this.logger.error({err}, `Error clearing tag for ${this.memberId}`);
+    }
+  }
+
 }
 
 module.exports = Conference;

lib/tasks/config.js

@@ -1,15 +1,22 @@
 const Task = require('./task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
+const parseDecibels = require('../utils/parse-decibels');
 
 class TaskConfig extends Task {
   constructor(logger, opts) {
     super(logger, opts);
+
     [
       'synthesizer',
       'recognizer',
       'bargeIn',
       'record',
-      'listen'
+      'listen',
+      'transcribe',
+      'fillerNoise',
+      'actionHookDelayAction',
+      'boostAudioSignal',
+      'vad'
     ].forEach((k) => this[k] = this.data[k] || {});
 
     if ('notifyEvents' in this.data) {
@@ -30,6 +37,13 @@ class TaskConfig extends Task {
         if (this.bargeIn[k]) this.gatherOpts[k] = this.bargeIn[k];
       });
     }
+    if (this.transcribe?.enable) {
+      this.transcribeOpts = {
+        verb: 'transcribe',
+        ...this.transcribe
+      };
+      delete this.transcribeOpts.enable;
+    }
 
     if (this.data.reset) {
       if (typeof this.data.reset === 'string') this.data.reset = [this.data.reset];
@@ -37,7 +51,12 @@ class TaskConfig extends Task {
     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) ?
+    this.preconditions = (this.bargeIn.enable ||
+      this.record?.action ||
+      this.listen?.url ||
+      this.data.amd ||
+      'boostAudioSignal' in this.data ||
+      this.transcribe?.enable) ?
       TaskPreconditions.Endpoint :
       TaskPreconditions.None;
 
@@ -50,6 +69,10 @@ class TaskConfig extends Task {
   get hasRecognizer() { return Object.keys(this.recognizer).length; }
   get hasRecording() { return Object.keys(this.record).length; }
   get hasListen() { return Object.keys(this.listen).length; }
+  get hasTranscribe() { return Object.keys(this.transcribe).length; }
+  get hasDub() { return Object.keys(this.dub).length; }
+  get hasVad() { return Object.keys(this.vad).length; }
+  get hasFillerNoise() { return Object.keys(this.fillerNoise).length; }
 
   get summary() {
     const phrase = [];
@@ -72,9 +95,14 @@ class TaskConfig extends Task {
     if (this.hasListen) {
       phrase.push(this.listen.enable ? `listen ${this.listen.url}` : 'stop listen');
     }
+    if (this.hasTranscribe) {
+      phrase.push(this.transcribe.enable ? `transcribe ${this.transcribe.transcriptionHook}` : 'stop transcribe');
+    }
+    if (this.hasFillerNoise) phrase.push(`fillerNoise ${this.fillerNoise.enable ? 'on' : 'off'}`);
     if (this.data.amd) phrase.push('enable amd');
     if (this.notifyEvents) phrase.push(`event notification ${this.notifyEvents ? 'on' : 'off'}`);
     if (this.onHoldMusic) phrase.push(`onHoldMusic: ${this.onHoldMusic}`);
+    if ('boostAudioSignal' in this.data) phrase.push(`setGain ${this.data.boostAudioSignal}`);
     return `${this.name}{${phrase.join(',')}}`;
   }
 
@@ -109,6 +137,7 @@ class TaskConfig extends Task {
     });
 
     if (this.hasSynthesizer) {
+      cs.synthesizer = this.synthesizer;
       cs.speechSynthesisVendor = this.synthesizer.vendor !== 'default'
         ? this.synthesizer.vendor
         : cs.speechSynthesisVendor;
@@ -138,6 +167,7 @@ class TaskConfig extends Task {
       this.logger.info({synthesizer: this.synthesizer}, 'Config: updated synthesizer');
     }
     if (this.hasRecognizer) {
+      cs.recognizer = this.recognizer;
       cs.speechRecognizerVendor = this.recognizer.vendor !== 'default'
         ? this.recognizer.vendor
         : cs.speechRecognizerVendor;
@@ -210,15 +240,65 @@ class TaskConfig extends Task {
       const {enable, ...opts} = this.listen;
       if (enable) {
         this.logger.debug({opts}, 'Config: enabling listen');
-        cs.startBackgroundListen({verb: 'listen', ...opts});
+        cs.startBackgroundTask('listen', {verb: 'listen', ...opts});
       } else {
         this.logger.info('Config: disabling listen');
-        cs.stopBackgroundListen();
+        cs.stopBackgroundTask('listen');
       }
     }
+    if (this.hasTranscribe) {
+      if (this.transcribe.enable) {
+        this.transcribeOpts.recognizer = this.hasRecognizer ?
+          this.recognizer :
+          {
+            vendor: cs.speechRecognizerVendor,
+            language: cs.speechRecognizerLanguage
+          };
+        this.logger.debug(this.transcribeOpts, 'Config: enabling transcribe');
+        cs.startBackgroundTask('transcribe', this.transcribeOpts);
+      } else {
+        this.logger.info('Config: disabling transcribe');
+        cs.stopBackgroundTask('transcribe');
+      }
+    }
+    if (Object.keys(this.actionHookDelayAction).length !== 0) {
+      cs.actionHookDelayEnabled = this.actionHookDelayAction.enabled || false;
+      cs.actionHookNoResponseTimeout = this.actionHookDelayAction.noResponseTimeout || 0;
+      cs.actionHookNoResponseGiveUpTimeout = this.actionHookDelayAction.noResponseGiveUpTimeout || 0;
+      cs.actionHookDelayRetries = this.actionHookDelayAction.retries || 1;
+      cs.actionHookDelayActions = this.actionHookDelayAction.actions || [];
+    }
     if (this.data.sipRequestWithinDialogHook) {
       cs.sipRequestWithinDialogHook = this.data.sipRequestWithinDialogHook;
     }
+
+    if ('boostAudioSignal' in this.data) {
+      const db = parseDecibels(this.data.boostAudioSignal);
+      this.logger.info(`Config: boosting audio signal by ${db} dB`);
+      const args = [ep.uuid, 'setGain', db];
+      ep.api('uuid_dub', args).catch((err) => {
+        this.logger.error(err, 'Error boosting audio signal');
+      });
+    }
+
+    if (this.hasFillerNoise) {
+      const {enable, ...opts} = this.fillerNoise;
+      this.logger.info({fillerNoise: this.fillerNoise}, 'Config: fillerNoise');
+      if (!enable) cs.disableFillerNoise();
+      else {
+        cs.enableFillerNoise(opts);
+      }
+    }
+
+    if (this.hasVad) {
+      cs.vad = {
+        enable: this.vad.enable || false,
+        voiceMs: this.vad.voiceMs || 250,
+        silenceMs: this.vad.silenceMs || 150,
+        strategy: this.vad.strategy || 'one-shot',
+        mode: this.vad.mod || 2
+      };
+    }
   }
 
   async kill(cs) {

lib/tasks/dequeue.js

@@ -73,7 +73,8 @@ class TaskDequeue extends Task {
         try {
           let url;
           if (this.callSid) {
-            url = await retrieveByPatternSortedSet(this.queueName, `*${this.callSid}`);
+            const r = await retrieveByPatternSortedSet(this.queueName, `*${this.callSid}`);
+            url = r[0];
           } else {
             url = await retrieveFromSortedSet(this.queueName);
           }

lib/tasks/dial.js

@@ -14,10 +14,11 @@ const sessionTracker = require('../session/session-tracker');
 const DtmfCollector = require('../utils/dtmf-collector');
 const ConfirmCallSession = require('../session/confirm-call-session');
 const dbUtils = require('../utils/db-utils');
+const parseDecibels = require('../utils/parse-decibels');
 const debug = require('debug')('jambonz:feature-server');
 const {parseUri} = require('drachtio-srf');
 const {ANCHOR_MEDIA_ALWAYS, JAMBONZ_DISABLE_DIAL_PAI_HEADER} = require('../config');
-const { isOnhold } = require('../utils/sdp-utils');
+const { isOnhold, isOpusFirst } = require('../utils/sdp-utils');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 
 function parseDtmfOptions(logger, dtmfCapture) {
@@ -100,6 +101,8 @@ class TaskDial extends Task {
     this.referHook = this.data.referHook;
     this.dtmfHook = this.data.dtmfHook;
     this.proxy = this.data.proxy;
+    this.tag = this.data.tag;
+    this.boostAudioSignal = this.data.boostAudioSignal;
 
     if (this.dtmfHook) {
       const {parentDtmfCollector, childDtmfCollector} = parseDtmfOptions(logger, this.data.dtmfCapture || {});
@@ -117,6 +120,9 @@ class TaskDial extends Task {
     if (this.data.transcribe) {
       this.transcribeTask = makeTask(logger, {'transcribe' : this.data.transcribe}, this);
     }
+    if (this.data.dub && Array.isArray(this.data.dub) && this.data.dub.length > 0) {
+      this.dubTasks = this.data.dub.map((d) => makeTask(logger, {'dub': d}, this));
+    }
 
     this.results = {};
     this.bridged = false;
@@ -138,15 +144,17 @@ class TaskDial extends Task {
 
   get name() { return TaskName.Dial; }
 
-  get isOnHold() {
+  get isOnHoldEnabled() {
-    return this.isIncomingLegHold || this.isOutgoingLegHold;
+    return !!this.data.onHoldHook;
   }
 
   get canReleaseMedia() {
     const keepAnchor = this.data.anchorMedia ||
     this.cs.isBackGroundListen ||
+    this.cs.onHoldMusic ||
     ANCHOR_MEDIA_ALWAYS ||
     this.listenTask ||
+    this.dubTasks ||
     this.transcribeTask ||
     this.startAmd;
 
@@ -323,7 +331,7 @@ class TaskDial extends Task {
         const by = parseUri(req.getParsedHeader('Referred-By').uri);
         this.logger.info({to}, 'refer to parsed');
         const json = await cs.requestor.request('verb:hook', this.referHook, {
-          ...callInfo,
+          ...(callInfo.toJSON()),
           refer_details: {
             sip_refer_to: req.get('Refer-To'),
             sip_referred_by: req.get('Referred-By'),
@@ -488,7 +496,8 @@ class TaskDial extends Task {
       headers: this.headers,
       proxy: `sip:${sbcAddress}`,
       callingNumber: this.callerId || req.callingNumber,
-      ...(this.callerName && {callingName: this.callerName})
+      ...(this.callerName && {callingName: this.callerName}),
+      opusFirst: isOpusFirst(this.cs.ep.remote.sdp)
     };
 
     const t = this.target.find((t) => t.type === 'teams');
@@ -548,9 +557,9 @@ class TaskDial extends Task {
           const str = this.callerId || req.callingNumber || '';
           const callingNumber = str.startsWith('+') ? str.substring(1) : str;
           const voip_carrier_sid = await lookupCarrierByPhoneNumber(cs.accountSid, callingNumber);
-          this.logger.info(
-            `Dial:_attemptCalls: selected ${voip_carrier_sid} for requested phone number: ${callingNumber}`);
           if (voip_carrier_sid) {
+            this.logger.info(
+              `Dial:_attemptCalls: selected voip_carrier_sid ${voip_carrier_sid} for callingNumber: ${callingNumber}`);
             opts.headers['X-Requested-Carrier-Sid'] = voip_carrier_sid;
           }
         }
@@ -569,7 +578,8 @@ 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);
 
@@ -626,6 +636,8 @@ class TaskDial extends Task {
               await this._connectSingleDial(cs, sd);
             } catch (err) {
               this.logger.info({err}, 'Dial:_attemptCalls - Error calling _connectSingleDial ');
+              sd.removeAllListeners();
+              this.kill(cs);
             }
           })
           .on('decline', () => {
@@ -678,22 +690,43 @@ class TaskDial extends Task {
   async _onReinvite(req, res) {
     try {
       let isHandled = false;
-      if (this.cs.onHoldMusic) {
+      if (this.isOnHoldEnabled) {
-        if (isOnhold(req.body) && !this.epOther && !this.ep) {
+        if (isOnhold(req.body)) {
-          await this.cs.handleReinviteAfterMediaReleased(req, res);
+          this.logger.debug('Dial: _onReinvite receive hold Request');
-          // Onhold but media is already released
+          if (!this.epOther && !this.ep) {
-          // reconnect A Leg and Response B leg
+            this.logger.debug(`Dial: _onReinvite receive hold Request,
-          await this.reAnchorMedia(this.cs, this.sd);
+ media already released, reconnect media server`);
-          this.isOutgoingLegHold = true;
+            // 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();
+          // Media already connected, ask for onHoldHook
-        } else if (!isOnhold(req.body) && this.epOther && this.ep && this.isOutgoingLegHold && this.canReleaseMedia) {
+          this._onHoldHook(req);
-          // Offhold, time to release media
+        } else if (!isOnhold(req.body)) {
-          const newSdp = await this.ep.modify(req.body);
+          this.logger.debug('Dial: _onReinvite receive unhold Request');
-          await res.send(200, {body: newSdp});
+          if (this.epOther && this.ep && this.isOutgoingLegHold && this.canReleaseMedia) {
-          await this._releaseMedia(this.cs, this.sd);
+            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) {
@@ -754,6 +787,17 @@ class TaskDial extends Task {
       dialCallSid: sd.callSid,
     });
 
+    if (this.dubTasks) {
+      for (const dub of this.dubTasks) {
+        try {
+          await dub.exec(cs, {ep: sd.ep});
+        }
+        catch (err) {
+          this.logger.error({err}, 'Dial:_selectSingleDial - error executing dubTask');
+        }
+      }
+    }
+
     if (this.parentDtmfCollector) this._installDtmfDetection(cs, cs.dlg);
     if (this.childDtmfCollector) this._installDtmfDetection(cs, this.dlg);
     if (cs.sipRequestWithinDialogHook) this._initSipIndialogRequestListener(cs, this.dlg);
@@ -768,6 +812,18 @@ class TaskDial extends Task {
       }
     }
 
+    /* boost audio signal if requested */
+    if (this.boostAudioSignal) {
+      try {
+        const db = parseDecibels(this.boostAudioSignal);
+        this.logger.info(`Dial: boosting audio signal by ${db} dB`);
+        const args = [this.ep.uuid, 'setGain', db];
+        await this.ep.api('uuid_dub', args);
+      } catch (err) {
+        this.logger.info({err}, 'Dial:_selectSingleDial - Error boosting audio signal');
+      }
+    }
+
     /* if we can release the media back to the SBC, do so now */
     if (this.canReleaseMedia) setTimeout(this._releaseMedia.bind(this, cs, sd), 200);
   }
@@ -790,9 +846,11 @@ class TaskDial extends Task {
     assert(cs.ep && sd.ep);
 
     try {
+      // Wait until we got new SDP from B leg to ofter to A Leg
       const aLegSdp = cs.ep.remote.sdp;
+      await sd.releaseMediaToSBC(aLegSdp, cs.ep.local.sdp);
       const bLegSdp = sd.dlg.remote.sdp;
-      await Promise.all[sd.releaseMediaToSBC(aLegSdp, cs.ep.local.sdp), cs.releaseMediaToSBC(bLegSdp)];
+      await cs.releaseMediaToSBC(bLegSdp);
       this.epOther = null;
       this.logger.info('Dial:_releaseMedia - successfully released media from freewitch');
     } catch (err) {
@@ -808,23 +866,34 @@ 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) {
+    if (this.isOnHoldEnabled) {
-      const sdp = await this.dlg.modify(req.body);
+      if (isOnhold(req.body)) {
-      res.send(200, {body: sdp});
+        if (!this.epOther && !this.ep) {
-      // Onhold but media is already released
+          // update callee leg for new SDP from caller.
-      await this.reAnchorMedia(this.cs, this.sd);
+          const sdp = await this.dlg.modify(req.body);
-      isHandled = true;
+          res.send(200, {body: sdp});
-      this.isIncomingLegHold = true;
+          // Onhold but media is already released, reconnect
-      this._onHoldHook();
+          await this.reAnchorMedia(this.cs, this.sd);
-    } else if (!isOnhold(req.body) && this.epOther && this.ep && this.isIncomingLegHold && this.canReleaseMedia) {
+          isHandled = true;
-      // Offhold, time to release media
+          this.isIncomingLegHold = true;
-      const newSdp = await this.epOther.modify(req.body);
+        }
-      await res.send(200, {body: newSdp});
+        this._onHoldHook(req);
-      await this._releaseMedia(this.cs, this.sd);
+      } else if (!isOnhold(req.body)) {
-      isHandled = true;
+        if (this.epOther && this.ep && this.isIncomingLegHold && this.canReleaseMedia) {
-      this.isIncomingLegHold = false;
+          // 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) {
@@ -843,7 +912,7 @@ 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');
@@ -853,7 +922,13 @@ 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) {
@@ -862,7 +937,7 @@ 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,
@@ -872,12 +947,12 @@ class TaskDial extends Task {
               tasks,
               rootSpan: this.cs.rootSpan
             });
-            await this._playSession.exec();
+            await this._onHoldSession.exec();
-            this._playSession = null;
+            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/dub.js

@@ -0,0 +1,144 @@
+const {TaskName} = require('../utils/constants');
+const TtsTask = require('./tts-task');
+const assert = require('assert');
+const parseDecibels = require('../utils/parse-decibels');
+
+/**
+ * Dub task: add or remove additional audio tracks into the call
+ */
+class TaskDub extends TtsTask {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+
+    this.logger.debug({opts: this.data}, 'TaskDub constructor');
+    ['action', 'track', 'play', 'say', 'loop'].forEach((prop) => {
+      this[prop] = this.data[prop];
+    });
+    this.gain = parseDecibels(this.data.gain);
+
+    assert.ok(this.action, 'TaskDub: action is required');
+    assert.ok(this.track, 'TaskDub: track is required');
+  }
+
+  get name() { return TaskName.Dub; }
+
+  async exec(cs, {ep}) {
+    super.exec(cs);
+
+    try {
+      switch (this.action) {
+        case 'addTrack':
+          await this._addTrack(cs, ep);
+          break;
+        case 'removeTrack':
+          await this._removeTrack(cs, ep);
+          break;
+        case 'silenceTrack':
+          await this._silenceTrack(cs, ep);
+          break;
+        case 'playOnTrack':
+          await this._playOnTrack(cs, ep);
+          break;
+        case 'sayOnTrack':
+          await this._sayOnTrack(cs, ep);
+          break;
+        default:
+          throw new Error(`TaskDub: unsupported action ${this.action}`);
+      }
+    } catch (err) {
+      this.logger.error(err, 'Error executing dub task');
+    }
+  }
+
+  async _addTrack(cs, ep) {
+    this.logger.info(`adding track: ${this.track}`);
+    await ep.dub({
+      action: 'addTrack',
+      track: this.track
+    });
+
+    if (this.play) await this._playOnTrack(cs, ep);
+    else if (this.say) await this._sayOnTrack(cs, ep);
+  }
+
+  async _removeTrack(_cs, ep) {
+    this.logger.info(`removing track: ${this.track}`);
+    await ep.dub({
+      action: 'removeTrack',
+      track: this.track
+    });
+  }
+
+  async _silenceTrack(_cs, ep) {
+    this.logger.info(`silencing track: ${this.track}`);
+    await ep.dub({
+      action: 'silenceTrack',
+      track: this.track
+    });
+  }
+
+  async _playOnTrack(_cs, ep) {
+    this.logger.info(`playing on track: ${this.track}`);
+    await ep.dub({
+      action: 'playOnTrack',
+      track: this.track,
+      play: this.play,
+      loop: this.loop ? 'loop' : 'once',
+      gain: this.gain
+    });
+  }
+
+  async _sayOnTrack(cs, ep) {
+    const text = this.say.text || this.say;
+    this.synthesizer = this.say.synthesizer || {};
+
+    if (Object.keys(this.synthesizer).length) {
+      this.logger.info({synthesizer: this.synthesizer},
+        `saying on track ${this.track}: ${text} with synthesizer options`);
+    }
+    else {
+      this.logger.info(`saying on track ${this.track}: ${text}`);
+    }
+    this.synthesizer = this.synthesizer || {};
+
+    this.text = [text];
+
+    const vendor = this.synthesizer.vendor && this.synthesizer.vendor !== 'default' ?
+      this.synthesizer.vendor :
+      cs.speechSynthesisVendor;
+    const language = this.synthesizer.language && this.synthesizer.language !== 'default' ?
+      this.synthesizer.language :
+      cs.speechSynthesisLanguage ;
+    const voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
+      this.synthesizer.voice :
+      cs.speechSynthesisVoice;
+    const label = this.synthesizer.label && this.synthesizer.label !== 'default' ?
+      this.synthesizer.label :
+      cs.speechSynthesisLabel;
+
+    const disableTtsStreaming = false;
+    const filepath = await this._synthesizeWithSpecificVendor(cs, ep, {
+      vendor, language, voice, label, disableTtsStreaming
+    });
+    assert.ok(filepath.length === 1, 'TaskDub: no filepath returned from synthesizer');
+
+    const path = filepath[0];
+    if (!path.startsWith('say:{')) {
+      /* we have a local file of mp3 or r8 of synthesized speech audio to play */
+      this.logger.info(`playing synthesized speech from file on track ${this.track}: ${path}`);
+      this.play = path;
+      await this._playOnTrack(cs, ep);
+    }
+    else {
+      this.logger.info(`doing actual text to speech file on track ${this.track}: ${path}`);
+      await ep.dub({
+        action: 'sayOnTrack',
+        track: this.track,
+        say: path,
+        gain: this.gain
+      });
+    }
+  }
+}
+
+module.exports = TaskDub;

lib/tasks/enqueue.js

@@ -338,6 +338,7 @@ class TaskEnqueue extends Task {
       this.logger.error({err}, `TaskEnqueue:_playHook error retrieving list info for queue ${this.queueName}`);
     }
     const json = await cs.application.requestor.request('verb:hook', hook, params, httpHeaders);
+    this.logger.debug({json}, 'TaskEnqueue:_playHook: received response from waitHook');
     const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
 
     const allowedTasks = tasks.filter((t) => allowed.includes(t.name));

lib/tasks/gather.js

@@ -10,7 +10,9 @@ const {
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
-  AssemblyAiTranscriptionEvents
+  AssemblyAiTranscriptionEvents,
+  VadDetection,
+  VerbioTranscriptionEvents
 } = require('../utils/constants.json');
 const {
   JAMBONES_GATHER_EARLY_HINTS_MATCH,
@@ -27,9 +29,13 @@ class TaskGather extends SttTask {
     [
       'finishOnKey', 'input', 'numDigits', 'minDigits', 'maxDigits',
       'interDigitTimeout', 'partialResultHook', 'bargein', 'dtmfBargein',
-      'speechTimeout', 'timeout', 'say', 'play'
+      'speechTimeout', 'timeout', 'say', 'play', 'actionHookDelayAction', 'fillerNoise', 'vad'
     ].forEach((k) => this[k] = this.data[k]);
 
+    // gather default input is digits
+    if (!this.input) {
+      this.input = ['digits'];
+    }
     /* when collecting dtmf, bargein on dtmf is true unless explicitly set to false */
     if (this.dtmfBargein !== false  && this.input.includes('digits')) this.dtmfBargein = true;
 
@@ -37,7 +43,8 @@ class TaskGather extends SttTask {
     this.timeout = this.timeout === 0 ? 0 : (this.timeout || 15) * 1000;
     this.interim = !!this.partialResultHook || this.bargein || (this.timeout > 0);
     this.listenDuringPrompt = this.data.listenDuringPrompt === false ? false : true;
-    this.minBargeinWordCount = this.data.minBargeinWordCount || 1;
+    this.minBargeinWordCount = this.data.minBargeinWordCount !== undefined ? this.data.minBargeinWordCount : 1;
+    this._vadEnabled = this.minBargeinWordCount === 0;
     if (this.data.recognizer) {
       /* continuous ASR (i.e. compile transcripts until a special timeout or dtmf key) */
       this.asrTimeout = typeof this.data.recognizer.asrTimeout === 'number' ?
@@ -71,6 +78,7 @@ class TaskGather extends SttTask {
     /* buffer speech for continuous asr */
     this._bufferedTranscripts = [];
     this.partialTranscriptsCount = 0;
+    this.bugname_prefix = 'gather_';
   }
 
   get name() { return TaskName.Gather; }
@@ -86,6 +94,18 @@ class TaskGather extends SttTask {
       (this.playTask && this.playTask.earlyMedia);
   }
 
+  get hasFillerNoise() {
+    return Object.keys(this.fillerNoise).length > 0 && this.fillerNoise.enabled !== false;
+  }
+
+  get fillerNoiseUrl() {
+    return this.fillerNoise.url;
+  }
+
+  get fillerNoiseStartDelaySecs() {
+    return this.fillerNoise.startDelaySecs;
+  }
+
   get summary() {
     let s = `${this.name}{`;
     if (this.input.length === 2) s += 'inputs=[speech,digits],';
@@ -103,9 +123,19 @@ class TaskGather extends SttTask {
 
   async exec(cs, {ep}) {
     this.logger.debug({options: this.data}, 'Gather:exec');
-    await super.exec(cs);
+    await super.exec(cs, {ep});
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
 
+    this.fillerNoise = {
+      ...(cs.fillerNoise || {}),
+      ...(this.fillerNoise || {})
+    };
+
+    this.vad = {
+      ...(cs.vad || {}),
+      ...(this.vad || {})
+    };
+
     if (cs.hasGlobalSttHints && !this.maskGlobalSttHints) {
       const {hints, hintsBoost} = cs.globalSttHints;
       const setOfHints = new Set((this.data.recognizer.hints || [])
@@ -116,14 +146,6 @@ class TaskGather extends SttTask {
       this.logger.debug({hints: this.data.recognizer.hints, hintsBoost: this.data.recognizer.hintsBoost},
         'Gather:exec - applying global sttHints');
     }
-    if (cs.hasAltLanguages) {
-      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages.concat(cs.altLanguages);
-      this.logger.debug({altLanguages: this.data.recognizer?.altLanguages},
-        'Gather:exec - applying altLanguages');
-    }
-    if (cs.hasGlobalSttPunctuation && !this.data.recognizer.punctuation) {
-      this.data.recognizer.punctuation = cs.globalSttPunctuation;
-    }
     if (!this.isContinuousAsr && cs.isContinuousAsr) {
       this.isContinuousAsr = true;
       this.asrTimeout = cs.asrTimeout * 1000;
@@ -141,58 +163,30 @@ class TaskGather extends SttTask {
       this.interim = true;
       this.logger.debug('Gather:exec - early hints match enabled');
     }
+    // actionHook delay
+    this._hookDelayEn = cs.actionHookDelayEnabled || !!this.actionHookDelayAction;
 
-    this.ep = ep;
+    this._hookDelayActions = this.actionHookDelayAction?.actions || cs.actionHookDelayActions || [];
-    if ('default' === this.vendor || !this.vendor) {
-      this.vendor = cs.speechRecognizerVendor;
-      if (this.data.recognizer) this.data.recognizer.vendor = this.vendor;
-    }
-    if ('default' === this.language || !this.language) {
-      this.language = cs.speechRecognizerLanguage;
-      if (this.data.recognizer) this.data.recognizer.language = this.language;
-    }
-    if ('default' === this.label || !this.label) {
-      this.label = cs.speechRecognizerLabel;
-      if (this.data.recognizer) this.data.recognizer.label = this.label;
-    }
-    // Fallback options
-    if ('default' === this.fallbackVendor || !this.fallbackVendor) {
-      this.fallbackVendor = cs.fallbackSpeechRecognizerVendor;
-      if (this.data.recognizer) this.data.recognizer.fallbackVendor = this.fallbackVendor;
-    }
-    if ('default' === this.fallbackLanguage || !this.fallbackLanguage) {
-      this.fallbackLanguage = cs.fallbackSpeechRecognizerLanguage;
-      if (this.data.recognizer) this.data.recognizer.fallbackLanguage = this.fallbackLanguage;
-    }
-    if ('default' === this.fallbackLabel || !this.fallbackLabel) {
-      this.fallbackLabel = cs.fallbackSpeechRecognizerLabel;
-      if (this.data.recognizer) this.data.recognizer.fallbackLabel = this.fallbackLabel;
-    }
-    if (!this.data.recognizer.vendor) {
-      this.data.recognizer.vendor = this.vendor;
-    }
-    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
-      // By default, application saves cobalt model in language
-      this.data.recognizer.model = cs.speechRecognizerLanguage;
-    }
 
-    if (this.needsStt && !this.sttCredentials) {
+    // Only enable NoResponseTimeout if there is _hookDelayActions
-      try {
+    this._hookNoResponseTimeout = (this._hookDelayActions?.length ?
-        this.sttCredentials = await this._initSpeechCredentials(cs, this.vendor, this.label);
+      (this.actionHookDelayAction?.noResponseTimeout || cs.actionHookNoResponseTimeout || 0)
-      } catch (error) {
+      : 0) * 1000;
-        if (this.fallbackVendor && this.isHandledByPrimaryProvider) {
-          await this._fallback();
-        } else {
-          throw error;
-        }
-      }
-    }
 
-    /* when using cobalt model is required */
+    this._hookNoResponseGiveUpTimeout = (this.actionHookDelayAction?.noResponseGiveUpTimeout ||
-    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
+      cs.actionHookNoResponseGiveUpTimeout || 0) * 1000;
-      this.notifyError({ msg: 'ASR error', details:'Cobalt requires a model to be specified'});
+
-      throw new Error('Cobalt requires a model to be specified');
+    this._hookDelayRetries = this.actionHookDelayAction?.retries || cs.actionHookDelayRetries || 1;
-    }
+    this._hookDelayRetryCount = 0;
+    this.hookDelayActionOpts = {
+      enabled: this._hookDelayEn,
+      actions: this._hookDelayActions,
+      noResponseTimeoutMs: this._hookNoResponseTimeout,
+      noResponseGiveUpTimeoutMs: this._hookNoResponseGiveUpTimeout,
+      retries: this._hookDelayRetries
+    };
+
+    this._startVad();
 
     const startListening = async(cs, ep) => {
       this._startTimer();
@@ -207,12 +201,7 @@ class TaskGather extends SttTask {
           this._startTranscribing(ep);
           return updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
         } catch (e) {
-          if (this.fallbackVendor && this.isHandledByPrimaryProvider) {
+          await this._startFallback(cs, ep, {error: e});
-            await this._fallback();
-            startListening(cs, ep);
-          } else {
-            this.logger.error({error: e}, 'error in initSpeech');
-          }
         }
       }
     };
@@ -220,13 +209,9 @@ class TaskGather extends SttTask {
     try {
       if (this.sayTask) {
         const {span, ctx} = this.startChildSpan(`nested:${this.sayTask.summary}`);
-        this.sayTask.span = span;
+        const process = () => {
-        this.sayTask.ctx = ctx;
-        this.sayTask.exec(cs, {ep});  // kicked off, _not_ waiting for it to complete
-        this.sayTask.on('playDone', (err) => {
-          span.end();
-          if (err) this.logger.error({err}, 'Gather:exec Error playing tts');
           this.logger.debug('Gather: nested say task completed');
+          this._stopVad();
           if (!this.killed) {
             startListening(cs, ep);
             if (this.input.includes('speech') && this.vendor === 'nuance' && this.listenDuringPrompt) {
@@ -236,17 +221,24 @@ class TaskGather extends SttTask {
               });
             }
           }
+        };
+        this.sayTask.span = span;
+        this.sayTask.ctx = ctx;
+        this.sayTask.exec(cs, {ep})  // kicked off, _not_ waiting for it to complete
+          .catch((err) => {
+            process();
+          });
+        this.sayTask.on('playDone', (err) => {
+          span.end();
+          if (err) this.logger.error({err}, 'Gather:exec Error playing tts');
+          process();
         });
       }
       else if (this.playTask) {
         const {span, ctx} = this.startChildSpan(`nested:${this.playTask.summary}`);
-        this.playTask.span = span;
+        const process = () => {
-        this.playTask.ctx = ctx;
-        this.playTask.exec(cs, {ep});  // kicked off, _not_ waiting for it to complete
-        this.playTask.on('playDone', (err) => {
-          span.end();
-          if (err) this.logger.error({err}, 'Gather:exec Error playing url');
           this.logger.debug('Gather: nested play task completed');
+          this._stopVad();
           if (!this.killed) {
             startListening(cs, ep);
             if (this.input.includes('speech') && this.vendor === 'nuance' && this.listenDuringPrompt) {
@@ -256,6 +248,17 @@ class TaskGather extends SttTask {
               });
             }
           }
+        };
+        this.playTask.span = span;
+        this.playTask.ctx = ctx;
+        this.playTask.exec(cs, {ep})  // kicked off, _not_ waiting for it to complete
+          .catch((err) => {
+            process();
+          });
+        this.playTask.on('playDone', (err) => {
+          span.end();
+          if (err) this.logger.error({err}, 'Gather:exec Error playing url');
+          process();
         });
       }
       else {
@@ -268,9 +271,14 @@ class TaskGather extends SttTask {
 
       if (this.input.includes('speech') && this.listenDuringPrompt) {
         await this._setSpeechHandlers(cs, ep);
-        this._startTranscribing(ep);
+        if (!this.resolved && !this.killed) {
-        updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid)
+          this._startTranscribing(ep);
-          .catch(() => {/*already logged error */});
+          updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid)
+            .catch(() => {/*already logged error */});
+        }
+        else {
+          this.logger.info('Gather:exec - task was killed or resolved quickly, not starting transcription');
+        }
       }
 
       if (this.input.includes('digits') || this.dtmfBargein || this.asrDtmfTerminationDigit) {
@@ -278,20 +286,24 @@ class TaskGather extends SttTask {
       }
 
       await this.awaitTaskDone();
+      this._killAudio(cs);
     } catch (err) {
       this.logger.error(err, 'TaskGather:exec error');
     }
-    this.removeSpeechListeners(ep);
+    this.removeCustomEventListeners();
   }
 
   kill(cs) {
     super.kill(cs);
     this._killAudio(cs);
+    this._killActionHookDelayAction();
+    this._clearFillerNoiseTimer();
     this.ep.removeAllListeners('dtmf');
     clearTimeout(this.interDigitTimer);
     this._clearAsrTimer();
     this.playTask?.span.end();
     this.sayTask?.span.end();
+    this._stopVad();
     this._resolve('killed');
   }
 
@@ -319,6 +331,16 @@ class TaskGather extends SttTask {
       this._resolve('dtmf-terminator-key');
     }
     else if (this.input.includes('digits')) {
+      if (this.digitBuffer.length === 0 && this.needsStt) {
+        // DTMF is higher priority than STT.
+        this.removeCustomEventListeners();
+        ep.stopTranscription({
+          vendor: this.vendor,
+          bugname: this.bugname,
+        })
+          .catch((err) => this.logger.error({err},
+            ` Received DTMF, Error stopping transcription for vendor ${this.vendor}`));
+      }
       this.digitBuffer += evt.dtmf;
       const len = this.digitBuffer.length;
       if (len === this.numDigits || len === this.maxDigits) {
@@ -351,38 +373,36 @@ 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 = 'google_transcribe';
+        this.bugname = `${this.bugname_prefix}google_transcribe`;
-        ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
-        ep.addCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance, this._onEndOfUtterance.bind(this, cs, ep));
+          ep, GoogleTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(GoogleTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, GoogleTranscriptionEvents.EndOfUtterance, this._onEndOfUtterance.bind(this, cs, ep));
         break;
 
       case 'aws':
       case 'polly':
-        this.bugname = 'aws_transcribe';
+        this.bugname = `${this.bugname_prefix}aws_transcribe`;
-        ep.addCustomEventListener(AwsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(AwsTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
         break;
       case 'microsoft':
-        this.bugname = 'azure_transcribe';
+        this.bugname = `${this.bugname_prefix}azure_transcribe`;
-        ep.addCustomEventListener(AzureTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
-        ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected,
+          ep, AzureTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-          this._onNoSpeechDetected.bind(this, cs, ep));
+        //this.addCustomEventListener(ep, AzureTranscriptionEvents.NoSpeechDetected,
-        ep.addCustomEventListener(AzureTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
+        //this._onNoSpeechDetected.bind(this, cs, ep));
         break;
       case 'nuance':
-        this.bugname = 'nuance_transcribe';
+        this.bugname = `${this.bugname_prefix}nuance_transcribe`;
-        ep.addCustomEventListener(NuanceTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.StartOfSpeech,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.StartOfSpeech,
           this._onStartOfSpeech.bind(this, cs, ep));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.TranscriptionComplete,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.TranscriptionComplete,
           this._onTranscriptionComplete.bind(this, cs, ep));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.VadDetected,
-          this._onVadDetected.bind(this, cs, ep));
 
         /* stall timers until prompt finishes playing */
         if ((this.sayTask || this.playTask) && this.listenDuringPrompt) {
@@ -391,24 +411,30 @@ class TaskGather extends SttTask {
         break;
 
       case 'deepgram':
-        this.bugname = 'deepgram_transcribe';
+        this.bugname = `${this.bugname_prefix}deepgram_transcribe`;
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+          ep, DeepgramTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.ConnectFailure,
           this._onVendorConnectFailure.bind(this, cs, ep));
-
-        /* if app sets deepgramOptions.utteranceEndMs they essentially want continuous asr */
-        if (opts.DEEPGRAM_SPEECH_UTTERANCE_END_MS) this.isContinuousAsr = true;
         break;
 
       case 'soniox':
-        this.bugname = 'soniox_transcribe';
+        this.bugname = `${this.bugname_prefix}soniox_transcribe`;
-        ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, SonioxTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        break;
+
+      case 'verbio':
+        this.bugname = `${this.bugname_prefix}verbio_transcribe`;
+        this.addCustomEventListener(
+          ep, VerbioTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
         break;
 
       case 'cobalt':
-        this.bugname = 'cobalt_transcribe';
+        this.bugname = `${this.bugname_prefix}cobalt_transcribe`;
-        ep.addCustomEventListener(CobaltTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, CobaltTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
 
         /* cobalt doesnt have language, it has model, which is required */
         if (!this.data.recognizer.model) {
@@ -437,23 +463,21 @@ class TaskGather extends SttTask {
         break;
 
       case 'ibm':
-        this.bugname = 'ibm_transcribe';
+        this.bugname = `${this.bugname_prefix}ibm_transcribe`;
-        ep.addCustomEventListener(IbmTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(IbmTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
-        ep.addCustomEventListener(IbmTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.ConnectFailure,
           this._onVendorConnectFailure.bind(this, cs, ep));
         break;
 
       case 'nvidia':
-        this.bugname = 'nvidia_transcribe';
+        this.bugname = `${this.bugname_prefix}nvidia_transcribe`;
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.StartOfSpeech,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.StartOfSpeech,
           this._onStartOfSpeech.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.TranscriptionComplete,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.TranscriptionComplete,
           this._onTranscriptionComplete.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.VadDetected,
-          this._onVadDetected.bind(this, cs, ep));
 
         /* I think nvidia has this (??) - stall timers until prompt finishes playing */
         if ((this.sayTask || this.playTask) && this.listenDuringPrompt) {
@@ -462,20 +486,22 @@ class TaskGather extends SttTask {
         break;
 
       case 'assemblyai':
-        this.bugname = 'assemblyai_transcribe';
+        this.bugname = `${this.bugname_prefix}assemblyai_transcribe`;
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+          ep, AssemblyAiTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.ConnectFailure,
           this._onVendorConnectFailure.bind(this, cs, ep));
         break;
       default:
         if (this.vendor.startsWith('custom:')) {
-          this.bugname = `${this.vendor}_transcribe`;
+          this.bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
-          ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+          this.addCustomEventListener(
-          ep.addCustomEventListener(JambonzTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+            ep, JambonzTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-          ep.addCustomEventListener(JambonzTranscriptionEvents.ConnectFailure,
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.ConnectFailure,
             this._onVendorConnectFailure.bind(this, cs, ep));
           break;
         }
@@ -487,7 +513,7 @@ class TaskGather extends SttTask {
     }
 
     /* common handler for all stt engine errors */
-    ep.addCustomEventListener(JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
+    this.addCustomEventListener(ep, JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
     await ep.set(opts)
       .catch((err) => this.logger.info(err, 'Error setting channel variables'));
   }
@@ -528,7 +554,10 @@ class TaskGather extends SttTask {
     this._clearTimer();
     this._timeoutTimer = setTimeout(() => {
       if (this.isContinuousAsr) this._startAsrTimer();
-      else this._resolve(this.digitBuffer.length >= this.minDigits ? 'dtmf-num-digits' : 'timeout');
+      if (this.interDigitTimer) return; // let the inter-digit timer complete
+      else {
+        this._resolve(this.digitBuffer.length >= this.minDigits ? 'dtmf-num-digits' : 'timeout');
+      }
     }, this.timeout);
   }
 
@@ -536,7 +565,9 @@ class TaskGather extends SttTask {
     if (this._timeoutTimer) {
       clearTimeout(this._timeoutTimer);
       this._timeoutTimer = null;
+      return true;
     }
+    return false;
   }
 
   _startAsrTimer() {
@@ -545,7 +576,7 @@ class TaskGather extends SttTask {
     this._clearAsrTimer();
     this._asrTimer = setTimeout(() => {
       this.logger.debug('_startAsrTimer - asr timer went off');
-      const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+      const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
       this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
     }, this.asrTimeout);
     this.logger.debug(`_startAsrTimer: set for ${this.asrTimeout}ms`);
@@ -556,6 +587,103 @@ class TaskGather extends SttTask {
     this._asrTimer = null;
   }
 
+  _hangupCall() {
+    this.logger.debug('_hangupCall');
+    this.cs.hangup();
+  }
+
+  _actionHookDelaySayAction(verb) {
+    delete verb.verb;
+    this.logger.debug(`_actionHookDelaySayAction ${verb}`);
+    this._actionHookDelaySayTask = makeTask(this.logger, {say: verb}, this);
+    const {span, ctx} = this.startChildSpan(`actionHookDelayAction:${this._actionHookDelaySayTask.summary}`);
+    this._actionHookDelaySayTask.span = span;
+    this._actionHookDelaySayTask.ctx = ctx;
+    this._actionHookDelaySayTask.exec(this.cs, {ep: this.ep});
+    this._actionHookDelaySayTask.on('playDone', (err) => {
+      this._actionHookDelaySayTask = null;
+      span.end();
+      if (err) this.logger.error({err}, 'Gather:actionHookDelay Error playing tts');
+    });
+  }
+
+  _killActionHookDelayAction() {
+    if (this._actionHookDelaySayTask && !this._actionHookDelaySayTask.killed) {
+      this._actionHookDelaySayTask.removeAllListeners('playDone');
+      this._actionHookDelaySayTask.kill(this.cs);
+      this._actionHookDelaySayTask.span.end();
+      this._actionHookDelaySayTask = null;
+    }
+
+    if (this._actionHookDelayPlayTask && !this._actionHookDelayPlayTask.killed) {
+      this._actionHookDelayPlayTask.removeAllListeners('playDone');
+      this._actionHookDelayPlayTask.kill(this.cs);
+      this._actionHookDelayPlayTask.span.end();
+      this._actionHookDelayPlayTask = null;
+    }
+  }
+
+  _actionHookDelayPlayAction(verb) {
+    delete verb.verb;
+    this.logger.debug(`_actionHookDelayPlayAction ${verb}`);
+    this._actionHookDelayPlayTask = makeTask(this.logger, {play: verb}, this);
+    const {span, ctx} = this.startChildSpan(`actionHookDelayAction:${this._actionHookDelayPlayTask.summary}`);
+    this._actionHookDelayPlayTask.span = span;
+    this._actionHookDelayPlayTask.ctx = ctx;
+    this._actionHookDelayPlayTask.exec(this.cs, {ep: this.ep});
+    this._actionHookDelayPlayTask.on('playDone', (err) => {
+      this._actionHookDelayPlayTask = null;
+      span.end();
+      if (err) this.logger.error({err}, 'Gather:actionHookDelay Error playing tts');
+    });
+
+  }
+
+  _startActionHookNoResponseTimer() {
+    assert(this._hookNoResponseTimeout > 0);
+    this._clearActionHookNoResponseTimer();
+    this.logger.debug('startActionHookNoResponseTimer');
+    this._actionHookNoResponseTimer = setTimeout(() => {
+      if (this._hookDelayRetryCount >= this._hookDelayRetries) {
+        this._hangupCall();
+        return;
+      }
+      const verb = this._hookDelayActions[this._hookDelayRetryCount % this._hookDelayActions.length];
+      if (verb.verb === 'say') {
+        this._actionHookDelaySayAction(verb);
+      } else if (verb.verb === 'play') {
+        this._actionHookDelayPlayAction(verb);
+      }
+      this._hookDelayRetryCount++;
+      this._startActionHookNoResponseTimer();
+
+    }, this._hookNoResponseTimeout);
+
+  }
+
+  _clearActionHookNoResponseTimer() {
+    if (this._actionHookNoResponseTimer) {
+      clearTimeout(this._actionHookNoResponseTimer);
+    }
+    this._actionHookNoResponseTimer = null;
+  }
+
+  _startActionHookNoResponseGiveUpTimer() {
+    assert(this._hookNoResponseGiveUpTimeout > 0);
+    this._clearActionHookNoResponseGiveUpTimer();
+    this.logger.debug('startActionHookNoResponseGiveUpTimer');
+    this._actionHookNoResponseGiveUpTimer = setTimeout(() => {
+      this._hangupCall();
+    }, this._hookNoResponseGiveUpTimeout);
+  }
+
+  _clearActionHookNoResponseGiveUpTimer() {
+    if (this._actionHookNoResponseGiveUpTimer) {
+      clearTimeout(this._actionHookNoResponseGiveUpTimer);
+    }
+    this._actionHookNoResponseGiveUpTimer = null;
+  }
+
   _startFastRecognitionTimer(evt) {
     assert(this.fastRecognitionTimeout > 0);
     this._clearFastRecognitionTimer();
@@ -576,7 +704,7 @@ class TaskGather extends SttTask {
     this._clearFinalAsrTimer();
     this._finalAsrTimer = setTimeout(() => {
       this.logger.debug('_startFinalAsrTimer - final asr timer went off');
-      const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+      const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
       this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
     }, 1000);
     this.logger.debug('_startFinalAsrTimer: set for 1 second');
@@ -587,11 +715,65 @@ class TaskGather extends SttTask {
     this._finalAsrTimer = null;
   }
 
+
+  _startVad() {
+    if (!this._vadStarted && this._vadEnabled) {
+      this.logger.debug('_startVad');
+      this.addCustomEventListener(this.ep, VadDetection.Detection, this._onVadDetected.bind(this, this.cs, this.ep));
+      this.ep?.startVadDetection(this.vad);
+      this._vadStarted = true;
+    }
+  }
+
+  _stopVad() {
+    if (this._vadStarted) {
+      this.logger.debug('_stopVad');
+      this.ep?.stopVadDetection(this.vad);
+      this.ep?.removeCustomEventListener(VadDetection.Detection, this._onVadDetected);
+      this._vadStarted = false;
+    }
+  }
+
+  _startFillerNoise() {
+    this.logger.debug('Gather:_startFillerNoise - playing filler noise');
+    this.ep?.play(this.fillerNoise.url);
+    this._fillerNoiseOn = true;
+    this.ep.once('playback-start', (evt) => {
+      if (evt.file === this.fillerNoise.url && !this._fillerNoiseOn) {
+        this.logger.info({evt}, 'Gather:_startFillerNoise - race condition - kill filler noise here');
+        this.ep.api('uuid_break', this.ep.uuid)
+          .catch((err) => this.logger.info(err, 'Error killing filler noise'));
+        return;
+      } else this.logger.debug({evt}, 'Gather:_startFillerNoise - playback started');
+    });
+  }
+
+  _startFillerNoiseTimer() {
+    this._clearFillerNoiseTimer();
+    this._fillerNoiseTimer = setTimeout(() => {
+      this.logger.debug('Gather:_startFillerNoiseTimer - playing filler noise');
+      this._startFillerNoise();
+    }, this.fillerNoise.startDelaySecs * 1000);
+  }
+
+  _clearFillerNoiseTimer() {
+    if (this._fillerNoiseTimer) clearTimeout(this._fillerNoiseTimer);
+    this._fillerNoiseTimer = null;
+  }
+
+  _killFillerNoise() {
+    if (this._fillerNoiseTimer) {
+      this.logger.debug('Gather:_killFillerNoise');
+      this.ep?.api('uuid_break', this.ep.uuid);
+    }
+  }
+
   _killAudio(cs) {
-    if (!this.sayTask && !this.playTask && this.bargein) {
+    if (this.hasFillerNoise || (!this.sayTask && !this.playTask && this.bargein)) {
-      if (this.ep?.connected && !this.playComplete) {
+      if (this.ep?.connected && (!this.playComplete || this.hasFillerNoise)) {
         this.logger.debug('Gather:_killAudio: killing playback of any audio');
         this.playComplete = true;
+        this._fillerNoiseOn = false;  // in a race, if we just started audio it may sneak through here
         this.ep.api('uuid_break', this.ep.uuid)
           .catch((err) => this.logger.info(err, 'Error killing audio'));
       }
@@ -613,11 +795,11 @@ class TaskGather extends SttTask {
     // make sure this is not a transcript from answering machine detection
     const bugname = fsEvent.getHeader('media-bugname');
     const finished = fsEvent.getHeader('transcription-session-finished');
-    this.logger.debug({evt, bugname, finished}, `Gather:_onTranscription for vendor ${this.vendor}`);
+    this.logger.debug({evt, bugname, finished, vendor: this.vendor}, 'Gather:_onTranscription raw transcript');
     if (bugname && this.bugname !== bugname) return;
+    if (finished === 'true') return;
 
     if (this.vendor === 'ibm' && evt?.state === 'listening') return;
-
     if (this.vendor === 'deepgram' && evt.type === 'UtteranceEnd') {
       /* we will only get this when we have set utterance_end_ms */
       if (this._bufferedTranscripts.length === 0) {
@@ -625,14 +807,21 @@ class TaskGather extends SttTask {
       }
       else {
         this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram, return buffered transcript');
-        evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+        evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
         this._bufferedTranscripts = [];
         this._resolve('speech', evt);
       }
       return;
     }
+    if (this.vendor === 'deepgram' && evt.type === 'Metadata') {
+      this.logger.debug('Gather:_onTranscription - discarding Metadata event from deepgram');
+      return;
+    }
+
+    evt = this.normalizeTranscription(evt, this.vendor, 1, this.language,
+      this.shortUtterance, this.data.recognizer.punctuation);
+    //this.logger.debug({evt, bugname, finished, vendor: this.vendor}, 'Gather:_onTranscription normalized transcript');
 
-    evt = this.normalizeTranscription(evt, this.vendor, 1, this.language, this.shortUtterance);
     if (evt.alternatives.length === 0) {
       this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, continue listening');
       return;
@@ -658,7 +847,9 @@ class TaskGather extends SttTask {
     if (evt.is_final) {
       if (evt.alternatives[0].transcript === '' && !this.callSession.callGone && !this.killed) {
         emptyTranscript = true;
-        if (finished === 'true' && ['microsoft', 'deepgram'].includes(this.vendor)) {
+        if (finished === 'true' &&
+          ['microsoft', 'deepgram'].includes(this.vendor) &&
+          this._bufferedTranscripts.length === 0) {
           this.logger.debug({evt}, 'TaskGather:_onTranscription - got empty transcript from old gather, disregarding');
           return;
         }
@@ -692,7 +883,7 @@ class TaskGather extends SttTask {
         this._clearTimer();
         if (this._finalAsrTimer) {
           this._clearFinalAsrTimer();
-          const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+          const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
           return this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
         }
         this._startAsrTimer();
@@ -704,8 +895,8 @@ class TaskGather extends SttTask {
         if (this.bargein && (words + bufferedWords) < this.minBargeinWordCount) {
           this.logger.debug({evt, words, bufferedWords},
             'TaskGather:_onTranscription - final transcript but < min barge words');
-          this._bufferedTranscripts.push(evt);
+          if (!emptyTranscript) this._bufferedTranscripts.push(evt);
-          this._startTranscribing(ep);
+          if (!['soniox', 'aws', 'microsoft', 'deepgram'].includes(this.vendor)) this._startTranscribing(ep);
           return;
         }
         else {
@@ -718,14 +909,11 @@ class TaskGather extends SttTask {
           else if (this.vendor === 'deepgram') {
             /* compile transcripts into one */
             if (!emptyTranscript) this._bufferedTranscripts.push(evt);
-            if (this.data.recognizer?.deepgramOptions?.utteranceEndMs) {
+
-              this.logger.debug('TaskGather:_onTranscription - got speech_final waiting for UtteranceEnd event');
+            /* deepgram can send an empty and final transcript; only if we have any buffered should we resolve */
-              return;
+            if (this._bufferedTranscripts.length === 0) return;
-            }
+            evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
-            this.logger.debug({evt}, 'TaskGather:_onTranscription - compiling deepgram transcripts');
-            evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
             this._bufferedTranscripts = [];
-            this.logger.debug({evt}, 'TaskGather:_onTranscription - compiled deepgram transcripts');
           }
 
           /* here is where we return a final transcript */
@@ -734,14 +922,26 @@ class TaskGather extends SttTask {
       }
     }
     else {
-      this._clearTimer();
+      /* deepgram can send a non-final transcript but with words that are final, so we need to buffer */
-      this._startTimer();
+      let emptyTranscript = false;
-      if (this.bargein && (words + bufferedWords) >= this.minBargeinWordCount) {
+      if (this.vendor === 'deepgram') {
-        if (!this.playComplete) {
+        const originalEvent = evt.vendor.evt;
-          this.logger.debug({transcript: evt.alternatives[0].transcript}, 'killing audio due to speech');
+        if (originalEvent.is_final && evt.alternatives[0].transcript !== '') {
-          this.emit('vad');
+          this.logger.debug({evt}, 'Gather:_onTranscription - buffering a completed (partial) deepgram transcript');
+          this._bufferedTranscripts.push(evt);
+        }
+        if (evt.alternatives[0].transcript === '') emptyTranscript = true;
+      }
+
+      if (!emptyTranscript) {
+        if (this._clearTimer()) this._startTimer();
+        if (this.bargein && (words + bufferedWords) >= this.minBargeinWordCount) {
+          if (!this.playComplete) {
+            this.logger.debug({transcript: evt.alternatives[0].transcript}, 'killing audio due to speech');
+            this.emit('vad');
+          }
+          this._killAudio(cs);
         }
-        this._killAudio(cs);
       }
       if (this.fastRecognitionTimeout) {
         this._startFastRecognitionTimer(evt);
@@ -759,14 +959,9 @@ class TaskGather extends SttTask {
           this._sonioxTranscripts.push(evt.vendor.finalWords);
         }
       }
-      /* deepgram can send a non-final transcript but with words that are final, so we need to buffer */
+
-      if (this.vendor === 'deepgram') {
+      /* restart asr timer if we get a partial transcript */
-        const originalEvent = evt.vendor.evt;
+      if (this.isContinuousAsr) this._startAsrTimer();
-        if (originalEvent.is_final && evt.alternatives[0].transcript !== '') {
-          this.logger.debug({evt}, 'Gather:_onTranscription - buffering a completed (partial) deepgram transcript');
-          this._bufferedTranscripts.push(evt);
-        }
-      }
     }
   }
   _onEndOfUtterance(cs, ep) {
@@ -798,24 +993,45 @@ class TaskGather extends SttTask {
   _onTranscriptionComplete(cs, ep) {
     this.logger.debug('TaskGather:_onTranscriptionComplete');
   }
-  async _onJambonzError(cs, ep, evt) {
+
-    this.logger.info({evt}, 'TaskGather:_onJambonzError');
+  async _startFallback(cs, ep, evt) {
-    if (this.isHandledByPrimaryProvider && this.fallbackVendor) {
+    if (this.canFallback) {
-      ep.stopTranscription({vendor: this.vendor})
+      ep.stopTranscription({
+        vendor: this.vendor,
+        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);
       try {
-        await this._fallback();
+        this.logger.debug('gather:_startFallback');
-        await this._initSpeech(cs, ep);
+        this.notifyError({ msg: 'ASR error',
+          details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'in progress'});
+        await this._initFallback();
+        this._speechHandlersSet = false;
+        await this._setSpeechHandlers(cs, ep);
         this._startTranscribing(ep);
         updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
-        return;
+        return true;
       } catch (error) {
         this.logger.info({error}, `There is error while falling back to ${this.fallbackVendor}`);
+        this.notifyError({ msg: 'ASR error',
+          details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'not available'});
       }
+    } else {
+      this.logger.debug('gather:_startFallback no condition for falling back');
+      this.notifyError({ msg: 'ASR error',
+        details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'not available'});
     }
-    const {writeAlerts, AlertType} = cs.srf.locals;
+    return false;
+  }
 
+  async _onJambonzError(cs, ep, evt) {
+    if (this.vendor === 'google' && evt.error_code === 0) {
+      this.logger.info({evt}, 'TaskTranscribe:_onJambonzError - ignoring google error code 0');
+      return;
+    }
+    this.logger.info({evt}, 'TaskGather:_onJambonzError');
+    const {writeAlerts, AlertType} = cs.srf.locals;
     if (this.vendor === 'nuance') {
       const {code, error} = evt;
       if (code === 404 && error === 'No speech') return this._resolve('timeout');
@@ -828,17 +1044,23 @@ class TaskGather extends SttTask {
       message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
       vendor: this.vendor,
     }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
-    this.notifyError({msg: 'ASR error', details:`Custom speech vendor ${this.vendor} error: ${evt.error}`});
+    if (!(await this._startFallback(cs, ep, evt))) {
+      this.notifyTaskDone();
+    }
   }
 
-  _onVendorConnectFailure(cs, _ep, evt) {
+  async _onVendorConnectFailure(cs, _ep, evt) {
     super._onVendorConnectFailure(cs, _ep, evt);
-    this.notifyTaskDone();
+    if (!(await this._startFallback(cs, _ep, evt))) {
+      this.notifyTaskDone();
+    }
   }
 
-  _onVendorError(cs, _ep, evt) {
+  async _onVendorError(cs, _ep, evt) {
     super._onVendorError(cs, _ep, evt);
-    this._resolve('stt-error', evt);
+    if (!(await this._startFallback(cs, _ep, evt))) {
+      this._resolve('stt-error', evt);
+    }
   }
 
   _onVadDetected(cs, ep) {
@@ -847,6 +1069,10 @@ class TaskGather extends SttTask {
       this._killAudio(cs);
       this.emit('vad');
     }
+    if (this.vad?.strategy === 'one-shot') {
+      this.ep?.removeCustomEventListener(VadDetection.Detection, this._onVadDetected);
+      this._vadStarted = false;
+    }
   }
 
   _onNoSpeechDetected(cs, ep, evt, fsEvent) {
@@ -865,13 +1091,24 @@ class TaskGather extends SttTask {
 
   async _resolve(reason, evt) {
     this.logger.debug(`TaskGather:resolve with reason ${reason}`);
+    if (this.needsStt && this.ep && this.ep.connected) {
+      this.ep.stopTranscription({
+        vendor: this.vendor,
+        bugname: this.bugname
+      })
+        .catch((err) => {
+          if (this.resolved) return;
+          this.logger.error({err}, 'Error stopping transcription');
+        });
+    }
     if (this.resolved) return;
 
     this.resolved = true;
-    // Clear dtmf event
+    // If bargin is false and ws application return ack to verb:hook
-    if (this.dtmfBargein) {
+    // the gather should not play any audio
-      this.ep.removeAllListeners('dtmf');
+    this._killAudio(this.cs);
-    }
+    // Clear dtmf events, to avoid any case can leak the listener, just clean it
+    this.ep.removeAllListeners('dtmf');
     clearTimeout(this.interDigitTimer);
     this._clearTimer();
     this._clearFastRecognitionTimer();
@@ -881,10 +1118,6 @@ class TaskGather extends SttTask {
       'stt.resolve': reason,
       'stt.result': JSON.stringify(evt)
     });
-    if (this.needsStt && this.ep && this.ep.connected) {
-      this.ep.stopTranscription({vendor: this.vendor})
-        .catch((err) => this.logger.error({err}, 'Error stopping transcription'));
-    }
 
     if (this.callSession && this.callSession.callGone) {
       this.logger.debug('TaskGather:_resolve - call is gone, not invoking web callback');
@@ -892,6 +1125,25 @@ class TaskGather extends SttTask {
       return;
     }
 
+    // Enabled action Hook delay timer to applied actions
+    if (this._hookNoResponseTimeout > 0) {
+      this._startActionHookNoResponseTimer();
+    }
+
+    if (this._hookNoResponseGiveUpTimeout > 0) {
+      this._startActionHookNoResponseGiveUpTimer();
+    }
+
+    if (this.hasFillerNoise && (reason.startsWith('dtmf') || reason.startsWith('speech'))) {
+      if (this.fillerNoiseStartDelaySecs > 0) {
+        this._startFillerNoiseTimer();
+      }
+      else {
+        this.logger.debug(`TaskGather:_resolve - playing filler noise: ${this.fillerNoiseUrl}`);
+        this._startFillerNoise();
+      }
+    }
+
     try {
       if (reason.startsWith('dtmf')) {
         if (this.parentTask) this.parentTask.emit('dtmf', evt);
@@ -922,6 +1174,12 @@ class TaskGather extends SttTask {
         }
       }
     } catch (err) {  /*already logged error*/ }
+
+    // Gather got response from hook, cancel all delay timers if there is any
+    this._clearActionHookNoResponseTimer();
+    this._clearActionHookNoResponseGiveUpTimer();
+    this._clearFillerNoiseTimer();
+
     this.notifyTaskDone();
   }
 }

lib/tasks/listen.js

@@ -8,6 +8,10 @@ const DTMF_SPAN_NAME = 'dtmf';
 class TaskListen extends Task {
   constructor(logger, opts, parentTask) {
     super(logger, opts);
+    /**
+     * @deprecated
+     * use bidirectionalAudio.enabled
+     */
     this.disableBidirectionalAudio = opts.disableBidirectionalAudio;
     this.preconditions = TaskPreconditions.Endpoint;
 
@@ -25,6 +29,15 @@ class TaskListen extends Task {
     this.results = {};
     this.playAudioQueue = [];
     this.isPlayingAudioFromQueue = false;
+    this.bidirectionalAudio = {
+      enabled: this.disableBidirectionalAudio === true ? false : true,
+      ...(this.data['bidirectionalAudio']),
+    };
+
+    // From drachtio-version 3.0.40, forkAudioStart will send empty bugname, metadata together with
+    // bidirectionalAudio params that cause old version of freeswitch missunderstand between bugname and
+    // bidirectionalAudio params
+    this._bugname = 'audio_fork';
 
     if (this.transcribe) this.transcribeTask = makeTask(logger, {'transcribe': opts.transcribe}, this);
   }
@@ -123,8 +136,6 @@ class TaskListen extends Task {
       ci,
       this.metadata);
     if (this.hook.auth) {
-      this.logger.debug({username: this.hook.auth.username, password: this.hook.auth.password},
-        'TaskListen:_startListening basic auth');
       await this.ep.set({
         'MOD_AUDIO_BASIC_AUTH_USERNAME': this.hook.auth.username,
         'MOD_AUDIO_BASIC_AUTH_PASSWORD': this.hook.auth.password
@@ -135,7 +146,8 @@ class TaskListen extends Task {
       mixType: this.mixType,
       sampling: this.sampleRate,
       ...(this._bugname && {bugname: this._bugname}),
-      metadata
+      metadata,
+      bidirectionalAudio: this.bidirectionalAudio || {}
     });
     this.recordStartTime = moment();
     if (this.maxLength) {
@@ -155,7 +167,7 @@ class TaskListen extends Task {
     }
 
     /* support bi-directional audio */
-    if (!this.disableBidirectionalAudio) {
+    if (this.bidirectionalAudio.enabled) {
       ep.addCustomEventListener(ListenEvents.PlayAudio, this._onPlayAudio.bind(this, ep));
     }
     ep.addCustomEventListener(ListenEvents.KillAudio, this._onKillAudio.bind(this, ep));

lib/tasks/make_task.js

@@ -14,6 +14,9 @@ function makeTask(logger, obj, parent) {
   }
   validateVerb(name, data, logger);
   switch (name) {
+    case TaskName.Answer:
+      const TaskAnswer = require('./answer');
+      return new TaskAnswer(logger, data, parent);
     case TaskName.SipDecline:
       const TaskSipDecline = require('./sip_decline');
       return new TaskSipDecline(logger, data, parent);
@@ -41,6 +44,9 @@ function makeTask(logger, obj, parent) {
     case TaskName.Dtmf:
       const TaskDtmf = require('./dtmf');
       return new TaskDtmf(logger, data, parent);
+    case TaskName.Dub:
+      const TaskDub = require('./dub');
+      return new TaskDub(logger, data, parent);
     case TaskName.Enqueue:
       const TaskEnqueue = require('./enqueue');
       return new TaskEnqueue(logger, data, parent);

lib/tasks/say.js

@@ -1,4 +1,4 @@
-const Task = require('./task');
+const TtsTask = require('./tts-task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
 const pollySSMLSplit = require('polly-ssml-split');
 
@@ -23,9 +23,15 @@ const breakLengthyTextIfNeeded = (logger,  text) => {
   }
 };
 
-class TaskSay extends Task {
+const parseTextFromSayString = (text) => {
+  const closingBraceIndex = text.indexOf('}');
+  if (closingBraceIndex === -1) return text;
+  return text.slice(closingBraceIndex + 1);
+};
+
+class TaskSay extends TtsTask {
   constructor(logger, opts, parentTask) {
-    super(logger, opts);
+    super(logger, opts, parentTask);
     this.preconditions = TaskPreconditions.Endpoint;
 
     this.text = (Array.isArray(this.data.text) ? this.data.text : [this.data.text])
@@ -33,10 +39,6 @@ class TaskSay extends Task {
       .flat();
 
     this.loop = this.data.loop || 1;
-    this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
-    this.synthesizer = this.data.synthesizer || {};
-    this.disableTtsCache = this.data.disableTtsCache;
-    this.options = this.synthesizer.options || {};
     this.isHandledByPrimaryProvider = true;
   }
 
@@ -59,8 +61,8 @@ 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 {srf, accountSid:account_sid} = cs;
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
     const {writeAlerts, AlertType, stats} = srf.locals;
     const {synthAudio} = srf.locals.dbHelpers;
@@ -76,6 +78,8 @@ class TaskSay extends Task {
         voice = arr[1];
         model = arr[2];
       }
+    } else if (vendor === 'deepgram') {
+      model = voice;
     }
 
     /* allow for microsoft custom region voice and api_key to be specified as an override */
@@ -86,20 +90,32 @@ class TaskSay extends Task {
       credentials.api_key = this.options.apiKey || credentials.apiKey;
       credentials.region = this.options.region || credentials.region;
       voice = this.options.voice || voice;
+    } else if (vendor === 'elevenlabs') {
+      credentials = credentials || {};
+      credentials.model_id = this.options.model_id || credentials.model_id;
+      credentials.voice_settings = this.options.voice_settings || {};
+      credentials.optimize_streaming_latency = this.options.optimize_streaming_latency
+      || credentials.optimize_streaming_latency;
+      voice = this.options.voice_id || voice;
     }
 
-    this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
+    ep.set({
+      tts_engine: vendor.startsWith('custom:') ? 'custom' : vendor,
+      tts_voice: voice,
+      cache_speech_handles: !cs.currentTtsVendor || cs.currentTtsVendor === vendor ? 1 : 0,
+    }).catch((err) => this.logger.info({err}, 'Error setting tts_engine on endpoint'));
+    // set the current vendor on the call session
+    // If vendor is changed from the previous one, then reset the cache_speech_handles flag
+    cs.currentTtsVendor = vendor;
+
+    if (!preCache) this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
     try {
       if (!credentials) {
         writeAlerts({
-          account_sid: cs.accountSid,
+          account_sid,
           alert_type: AlertType.TTS_NOT_PROVISIONED,
           vendor
         }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-        this.notifyError({
-          msg: 'TTS error',
-          details:`No speech credentials provisioned for selected vendor ${vendor}`
-        });
         throw new Error('no provisioned speech credentials for TTS');
       }
       // synthesize all of the text elements
@@ -111,14 +127,17 @@ class TaskSay extends Task {
         if (text.startsWith('silence_stream://')) return text;
 
         /* otel: trace time for tts */
-        const {span} = this.startChildSpan('tts-generation', {
+        if (!preCache)  {
-          'tts.vendor': vendor,
+          const {span} = this.startChildSpan('tts-generation', {
-          'tts.language': language,
+            'tts.vendor': vendor,
-          'tts.voice': voice
+            'tts.language': language,
-        });
+            'tts.voice': voice
+          });
+          this.otelSpan = span;
+        }
         try {
           const {filePath, servedFromCache, rtt} = await synthAudio(stats, {
-            account_sid: cs.accountSid,
+            account_sid,
             text,
             vendor,
             language,
@@ -127,37 +146,47 @@ class TaskSay extends Task {
             model,
             salt,
             credentials,
-            disableTtsCache : this.disableTtsCache
+            options: this.options,
+            disableTtsCache : this.disableTtsCache,
+            preCache
           });
-          this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
+          if (!filePath.startsWith('say:')) {
-          if (filePath) cs.trackTmpFile(filePath);
+            this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
-          if (!servedFromCache && !lastUpdated) {
+            if (filePath) cs.trackTmpFile(filePath);
-            lastUpdated = true;
+            if (this.otelSpan) {
-            updateSpeechCredentialLastUsed(credentials.speech_credential_sid)
+              this.otelSpan.setAttributes({'tts.cached': servedFromCache});
-              .catch(() => {/*already logged error */});
+              this.otelSpan.end();
+              this.otelSpan = null;
+            }
+            if (!servedFromCache && !lastUpdated) {
+              lastUpdated = true;
+              updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
+            }
+            if (!servedFromCache && rtt && !preCache) {
+              this.notifyStatus({
+                event: 'synthesized-audio',
+                vendor,
+                language,
+                characters: text.length,
+                elapsedTime: rtt
+              });
+            }
           }
-          span.setAttributes({'tts.cached': servedFromCache});
+          else {
-          span.end();
+            this.logger.debug('a streaming tts api will be used');
-          if (!servedFromCache && rtt) {
+            const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${ep.uuid},`);
-            this.notifyStatus({
+            return modifiedPath;
-              event: 'synthesized-audio',
-              vendor,
-              language,
-              characters: text.length,
-              elapsedTime: rtt
-            });
           }
           return filePath;
         } catch (err) {
           this.logger.info({err}, 'Error synthesizing tts');
-          span.end();
+          if (this.otelSpan) this.otelSpan.end();
           writeAlerts({
             account_sid: cs.accountSid,
             alert_type: AlertType.TTS_FAILURE,
             vendor,
             detail: err.message
           }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
-          this.notifyError({msg: 'TTS error', details: err.message || err});
           throw err;
         }
       };
@@ -172,19 +201,24 @@ class TaskSay extends Task {
   }
 
   async exec(cs, {ep}) {
+    const {srf, accountSid:account_sid} = cs;
+    const {writeAlerts, AlertType} = srf.locals;
+    const {addFileToCache} = srf.locals.dbHelpers;
+    const engine = this.synthesizer.engine || 'standard';
+
     await super.exec(cs);
     this.ep = ep;
 
-    const vendor = this.synthesizer.vendor && this.synthesizer.vendor !== 'default' ?
+    let vendor = this.synthesizer.vendor && this.synthesizer.vendor !== 'default' ?
       this.synthesizer.vendor :
       cs.speechSynthesisVendor;
-    const language = this.synthesizer.language && this.synthesizer.language !== 'default' ?
+    let language = this.synthesizer.language && this.synthesizer.language !== 'default' ?
       this.synthesizer.language :
       cs.speechSynthesisLanguage ;
-    const voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
+    let voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
       this.synthesizer.voice :
       cs.speechSynthesisVoice;
-    const label = this.synthesizer.label && this.synthesizer.label !== 'default' ?
+    let label = this.synthesizer.label && this.synthesizer.label !== 'default' ?
       this.synthesizer.label :
       cs.speechSynthesisLabel;
 
@@ -201,12 +235,19 @@ class TaskSay extends Task {
       this.synthesizer.fallbackLabel :
       cs.fallbackSpeechSynthesisLabel;
 
-    let filepath;
+    if (cs.hasFallbackTts) {
-    try {
+      vendor = fallbackVendor;
-      filepath = await this._synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label});
+      language = fallbackLanguage;
-    } catch (error) {
+      voice = fallbackVoice;
-      if (fallbackVendor && this.isHandledByPrimaryProvider) {
+      label = fallbackLabel;
+    }
+
+    const startFallback = async(error) => {
+      if (fallbackVendor && this.isHandledByPrimaryProvider && !cs.hasFallbackTts) {
+        this.notifyError(
+          { msg: 'TTS error', details:`TTS vendor ${vendor} error: ${error}`, failover: 'in progress'});
         this.isHandledByPrimaryProvider = false;
+        cs.hasFallbackTts = true;
         this.logger.info(`Synthesize error, fallback to ${fallbackVendor}`);
         filepath = await this._synthesizeWithSpecificVendor(cs, ep,
           {
@@ -216,22 +257,94 @@ class TaskSay extends Task {
             label: fallbackLabel
           });
       } else {
+        this.notifyError(
+          { msg: 'TTS error', details:`TTS vendor ${vendor} error: ${error}`, failover: 'not available'});
         throw error;
       }
+    };
+    let filepath;
+    try {
+      filepath = await this._synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label});
+    } catch (error) {
+      await startFallback(error);
     }
     this.notifyStatus({event: 'start-playback'});
 
-    while (!this.killed && (this.loop === 'forever' || this.loop--) && this.ep?.connected) {
+    while (!this.killed && (this.loop === 'forever' || this.loop--) && ep?.connected) {
       let segment = 0;
       while (!this.killed && segment < filepath.length) {
         if (cs.isInConference) {
           const {memberId, confName, confUuid} = cs;
-          await this.playToConfMember(this.ep, memberId, confName, confUuid, filepath[segment]);
+          await this.playToConfMember(ep, memberId, confName, confUuid, filepath[segment]);
         }
         else {
-          this.logger.debug(`Say:exec sending command to play file ${filepath[segment]}`);
+          if (filepath[segment].startsWith('say:{')) {
+            const arr = /^say:\{.*\}\s*(.*)$/.exec(filepath[segment]);
+            if (arr) this.logger.debug(`Say:exec sending streaming tts request: ${arr[1].substring(0, 64)}..`);
+          }
+          else this.logger.debug(`Say:exec sending  ${filepath[segment].substring(0, 64)}`);
+          ep.once('playback-start', (evt) => {
+            this.logger.debug({evt}, 'got playback-start');
+            if (this.otelSpan) {
+              this._addStreamingTtsAttributes(this.otelSpan, evt);
+              this.otelSpan.end();
+              this.otelSpan = null;
+              if (evt.variable_tts_cache_filename) cs.trackTmpFile(evt.variable_tts_cache_filename);
+            }
+          });
+          ep.once('playback-stop', (evt) => {
+            this.logger.debug({evt}, 'got playback-stop');
+            if (evt.variable_tts_error) {
+              writeAlerts({
+                account_sid,
+                alert_type: AlertType.TTS_FAILURE,
+                vendor,
+                detail: evt.variable_tts_error
+              }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
+            }
+            if (evt.variable_tts_cache_filename) {
+              const text = parseTextFromSayString(this.text[segment]);
+              addFileToCache(evt.variable_tts_cache_filename, {
+                account_sid,
+                vendor,
+                language,
+                voice,
+                engine,
+                text
+              }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
+            }
+            if (this._playResolve) {
+              evt.variable_tts_error ? this._playReject(new Error(evt.variable_tts_error)) : this._playResolve();
+            }
+          });
+          // wait for playback-stop event received to confirm if the playback is successful
+          this._playPromise = new Promise((resolve, reject) => {
+            this._playResolve = resolve;
+            this._playReject = reject;
+          });
           await ep.play(filepath[segment]);
-          this.logger.debug(`Say:exec completed play file ${filepath[segment]}`);
+          try {
+            // wait for playback-stop event received to confirm if the playback is successful
+            await this._playPromise;
+          } catch (err) {
+            try {
+              await startFallback(err);
+              continue;
+            } catch (err) {
+              this.logger.info({err}, 'Error waiting for playback-stop event');
+            }
+          } finally {
+            this._playPromise = null;
+            this._playResolve = null;
+            this._playReject = null;
+          }
+          if (filepath[segment].startsWith('say:{')) {
+            const arr = /^say:\{.*\}\s*(.*)$/.exec(filepath[segment]);
+            if (arr) this.logger.debug(`Say:exec complete playing streaming tts request: ${arr[1].substring(0, 64)}..`);
+          } else {
+            // This log will print spech credentials in say command for tts stream mode
+            this.logger.debug(`Say:exec completed play file ${filepath[segment]}`);
+          }
         }
         segment++;
       }
@@ -241,7 +354,7 @@ class TaskSay extends Task {
 
   async kill(cs) {
     super.kill(cs);
-    if (this.ep.connected) {
+    if (this.ep?.connected) {
       this.logger.debug('TaskSay:kill - killing audio');
       if (cs.isInConference) {
         const {memberId, confName} = cs;
@@ -251,8 +364,75 @@ class TaskSay extends Task {
         this.notifyStatus({event: 'kill-playback'});
         this.ep.api('uuid_break', this.ep.uuid);
       }
+      this.ep.removeAllListeners('playback-start');
+      this.ep.removeAllListeners('playback-stop');
+      // if we are waiting for playback-stop event, resolve the promise
+      if (this._playResolve) this._playResolve();
     }
   }
+
+  _addStreamingTtsAttributes(span, evt) {
+    const attrs = {'tts.cached': false};
+    for (const [key, value] of Object.entries(evt)) {
+      if (key.startsWith('variable_tts_')) {
+        let newKey = key.substring('variable_tts_'.length)
+          .replace('whisper_', 'whisper.')
+          .replace('deepgram_', 'deepgram.')
+          .replace('playht_', 'playht.')
+          .replace('rimelabs_', 'rimelabs.')
+          .replace('verbio_', 'verbio.')
+          .replace('elevenlabs_', 'elevenlabs.');
+        if (spanMapping[newKey]) newKey = spanMapping[newKey];
+        attrs[newKey] = value;
+      }
+    }
+    delete attrs['cache_filename']; //no value in adding this to the span
+    span.setAttributes(attrs);
+  }
 }
 
+const spanMapping = {
+  // IMPORTANT!!! JAMBONZ WEBAPP WILL SHOW TEXT PERFECTLY IF THE SPAN NAME IS SMALLER OR EQUAL 25 CHARACTERS.
+  // EX: whisper.ratelim_reqs has length 20 <= 25 which is perfect
+  // Elevenlabs
+  'elevenlabs.reported_latency_ms': 'elevenlabs.latency_ms',
+  'elevenlabs.request_id': 'elevenlabs.req_id',
+  'elevenlabs.history_item_id': 'elevenlabs.item_id',
+  'elevenlabs.optimize_streaming_latency': 'elevenlabs.optimization',
+  'elevenlabs.name_lookup_time_ms': 'name_lookup_ms',
+  'elevenlabs.connect_time_ms': 'connect_ms',
+  'elevenlabs.final_response_time_ms': 'final_response_ms',
+  // Whisper
+  'whisper.reported_latency_ms': 'whisper.latency_ms',
+  'whisper.request_id': 'whisper.req_id',
+  'whisper.reported_organization': 'whisper.organization',
+  'whisper.reported_ratelimit_requests': 'whisper.ratelimit',
+  'whisper.reported_ratelimit_remaining_requests': 'whisper.ratelimit_remain',
+  'whisper.reported_ratelimit_reset_requests': 'whisper.ratelimit_reset',
+  'whisper.name_lookup_time_ms': 'name_lookup_ms',
+  'whisper.connect_time_ms': 'connect_ms',
+  'whisper.final_response_time_ms': 'final_response_ms',
+  // Deepgram
+  'deepgram.request_id': 'deepgram.req_id',
+  'deepgram.reported_model_name': 'deepgram.model_name',
+  'deepgram.reported_model_uuid': 'deepgram.model_uuid',
+  'deepgram.reported_char_count': 'deepgram.char_count',
+  'deepgram.name_lookup_time_ms': 'name_lookup_ms',
+  'deepgram.connect_time_ms': 'connect_ms',
+  'deepgram.final_response_time_ms': 'final_response_ms',
+  // Playht
+  'playht.request_id': 'playht.req_id',
+  'playht.name_lookup_time_ms': 'name_lookup_ms',
+  'playht.connect_time_ms': 'connect_ms',
+  'playht.final_response_time_ms': 'final_response_ms',
+  // Rimelabs
+  'rimelabs.name_lookup_time_ms': 'name_lookup_ms',
+  'rimelabs.connect_time_ms': 'connect_ms',
+  'rimelabs.final_response_time_ms': 'final_response_ms',
+  // verbio
+  'verbio.name_lookup_time_ms': 'name_lookup_ms',
+  'verbio.connect_time_ms': 'connect_ms',
+  'verbio.final_response_time_ms': 'final_response_ms',
+};
+
 module.exports = TaskSay;

lib/tasks/stt-task.js

@@ -14,17 +14,15 @@ class SttTask extends Task {
     const {
       setChannelVarsForStt,
       normalizeTranscription,
-      removeSpeechListeners,
       setSpeechCredentialsAtRuntime,
       compileSonioxTranscripts,
       consolidateTranscripts
     } = require('../utils/transcription-utils')(logger);
     this.setChannelVarsForStt = setChannelVarsForStt;
     this.normalizeTranscription = normalizeTranscription;
-    this.removeSpeechListeners = removeSpeechListeners;
     this.compileSonioxTranscripts = compileSonioxTranscripts;
     this.consolidateTranscripts = consolidateTranscripts;
-
+    this.eventHandlers = [];
     this.isHandledByPrimaryProvider = true;
     if (this.data.recognizer) {
       const recognizer = this.data.recognizer;
@@ -49,11 +47,126 @@ class SttTask extends Task {
 
     /* buffer for soniox transcripts */
     this._sonioxTranscripts = [];
+    /*bug name prefix */
+    this.bugname_prefix = '';
 
   }
 
+  async exec(cs, {ep, ep2}) {
+    super.exec(cs);
+    this.ep = ep;
+    this.ep2 = ep2;
+
+    // use session preferences if we don't have specific verb-level settings.
+    if (cs.recognizer) {
+      for (const k in cs.recognizer) {
+        const newValue = this.data.recognizer && this.data.recognizer[k] !== undefined ?
+          this.data.recognizer[k] :
+          cs.recognizer[k];
+
+        if (Array.isArray(newValue)) {
+          this.data.recognizer[k] = [...(this.data.recognizer[k] || []), ...cs.recognizer[k]];
+        } else if (typeof newValue === 'object' && newValue !== null) {
+          this.data.recognizer[k] = { ...(this.data.recognizer[k] || {}), ...cs.recognizer[k] };
+        } else {
+          this.data.recognizer[k] = newValue;
+        }
+      }
+    }
+    if ('default' === this.vendor || !this.vendor) {
+      this.vendor = cs.speechRecognizerVendor;
+      if (this.data.recognizer) this.data.recognizer.vendor = this.vendor;
+    }
+    if ('default' === this.language || !this.language) {
+      this.language = cs.speechRecognizerLanguage;
+      if (this.data.recognizer) this.data.recognizer.language = this.language;
+    }
+    if ('default' === this.label || !this.label) {
+      this.label = cs.speechRecognizerLabel;
+      if (this.data.recognizer) this.data.recognizer.label = this.label;
+    }
+    // Fallback options
+    if ('default' === this.fallbackVendor || !this.fallbackVendor) {
+      this.fallbackVendor = cs.fallbackSpeechRecognizerVendor;
+      if (this.data.recognizer) this.data.recognizer.fallbackVendor = this.fallbackVendor;
+    }
+    if ('default' === this.fallbackLanguage || !this.fallbackLanguage) {
+      this.fallbackLanguage = cs.fallbackSpeechRecognizerLanguage;
+      if (this.data.recognizer) this.data.recognizer.fallbackLanguage = this.fallbackLanguage;
+    }
+    if ('default' === this.fallbackLabel || !this.fallbackLabel) {
+      this.fallbackLabel = cs.fallbackSpeechRecognizerLabel;
+      if (this.data.recognizer) this.data.recognizer.fallbackLabel = this.fallbackLabel;
+    }
+    // If call is already fallback to 2nd ASR vendor
+    // use that.
+    if (cs.hasFallbackAsr) {
+      this.vendor = this.fallbackVendor;
+      this.language = this.fallbackLanguage;
+      this.label = this.fallbackLabel;
+    }
+    if (!this.data.recognizer.vendor) {
+      this.data.recognizer.vendor = this.vendor;
+    }
+    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
+      // By default, application saves cobalt model in language
+      this.data.recognizer.model = cs.speechRecognizerLanguage;
+    }
+
+    if (
+      // not gather task, such as transcribe
+      (!this.input ||
+      // gather task with speech
+        this.input.includes('speech')) &&
+      !this.sttCredentials) {
+      try {
+        this.sttCredentials = await this._initSpeechCredentials(this.cs, this.vendor, this.label);
+      } catch (error) {
+        if (this.canFallback) {
+          this.notifyError(
+            {
+              msg: 'ASR error', details:`Invalid vendor ${this.vendor}, Error: ${error}`,
+              failover: 'in progress'
+            });
+          await this._initFallback();
+        } else {
+          this.notifyError(
+            {
+              msg: 'ASR error', details:`Invalid vendor ${this.vendor}, Error: ${error}`,
+              failover: 'not available'
+            });
+          throw error;
+        }
+      }
+    }
+
+    /* when using cobalt model is required */
+    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
+      this.notifyError({ msg: 'ASR error', details:'Cobalt requires a model to be specified'});
+      throw new Error('Cobalt requires a model to be specified');
+    }
+
+    if (cs.hasAltLanguages) {
+      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages.concat(cs.altLanguages);
+      this.logger.debug({altLanguages: this.altLanguages},
+        'STT:exec - applying altLanguages');
+    }
+    if (cs.hasGlobalSttPunctuation && !this.data.recognizer.punctuation) {
+      this.data.recognizer.punctuation = cs.globalSttPunctuation;
+    }
+  }
+
+  addCustomEventListener(ep, event, handler) {
+    this.eventHandlers.push({ep, event, handler});
+    ep.addCustomEventListener(event, handler);
+  }
+
+  removeCustomEventListeners() {
+    this.eventHandlers.forEach((h) => h.ep.removeCustomEventListener(h.event, h.handler));
+  }
+
   async _initSpeechCredentials(cs, vendor, label) {
-    const {getNuanceAccessToken, getIbmAccessToken} = this.cs.srf.locals.dbHelpers;
+    const {getNuanceAccessToken, getIbmAccessToken, getAwsAuthToken, getVerbioAccessToken} = cs.srf.locals.dbHelpers;
     let credentials = cs.getSpeechCredentials(vendor, 'stt', label);
 
     if (!credentials) {
@@ -64,11 +177,6 @@ class SttTask extends Task {
         alert_type: AlertType.STT_NOT_PROVISIONED,
         vendor
       }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
-      // Notify application that STT vender is wrong.
-      this.notifyError({
-        msg: 'ASR error',
-        details: `No speech-to-text service credentials for ${vendor} have been configured`
-      });
       this.notifyTaskDone();
       throw new Error(`No speech-to-text service credentials for ${vendor} have been configured`);
     }
@@ -86,13 +194,29 @@ class SttTask extends Task {
       const {access_token, servedFromCache} = await getIbmAccessToken(stt_api_key);
       this.logger.debug({stt_api_key}, `got ibm access token ${servedFromCache ? 'from cache' : ''}`);
       credentials = {...credentials, access_token, stt_region};
+    } else if (['aws', 'polly'].includes(vendor) && credentials.roleArn) {
+      /* get aws access token */
+      const {roleArn} = credentials;
+      const {accessKeyId, secretAccessKey, sessionToken, servedFromCache} = await getAwsAuthToken(roleArn);
+      this.logger.debug({roleArn}, `got aws access token ${servedFromCache ? 'from cache' : ''}`);
+      credentials = {...credentials, accessKeyId, secretAccessKey, sessionToken};
+    } else if (vendor === 'verbio' && credentials.client_id && credentials.client_secret) {
+      const {access_token, servedFromCache} = await getVerbioAccessToken(credentials);
+      this.logger.debug({client_id: credentials.client_id},
+        `got verbio access token ${servedFromCache ? 'from cache' : ''}`);
+      credentials.access_token = access_token;
     }
     return credentials;
   }
 
-  async _fallback() {
+  get canFallback() {
+    return this.fallbackVendor && this.isHandledByPrimaryProvider && !this.cs.hasFallbackAsr;
+  }
+
+  async _initFallback() {
     assert(this.fallbackVendor, 'fallback failed without fallbackVendor configuration');
     this.isHandledByPrimaryProvider = false;
+    this.cs.hasFallbackAsr = true;
     this.logger.info(`Failed to use primary STT provider, fallback to ${this.fallbackVendor}`);
     this.vendor = this.fallbackVendor;
     this.language = this.fallbackLanguage;
@@ -101,6 +225,8 @@ class SttTask extends Task {
     this.data.recognizer.language = this.language;
     this.data.recognizer.label = this.label;
     this.sttCredentials = await this._initSpeechCredentials(this.cs, this.vendor, this.label);
+    // cleanup previous listener from previous vendor
+    this.removeCustomEventListeners();
   }
 
   async compileHintsForCobalt(ep, hostport, model, token, hints) {
@@ -144,6 +270,20 @@ class SttTask extends Task {
   _doContinuousAsrWithDeepgram(asrTimeout) {
     /* deepgram has an utterance_end_ms property that simplifies things */
     assert(this.vendor === 'deepgram');
+    if (asrTimeout < 1000) {
+      this.notifyError({
+        msg: 'ASR error',
+        details:`asrTimeout ${asrTimeout} is too short for deepgram; setting it to 1000ms`
+      });
+      asrTimeout = 1000;
+    }
+    else if (asrTimeout > 5000) {
+      this.notifyError({
+        msg: 'ASR error',
+        details:`asrTimeout ${asrTimeout} is too long for deepgram; setting it to 5000ms`
+      });
+      asrTimeout = 5000;
+    }
     this.logger.debug(`_doContinuousAsrWithDeepgram - setting utterance_end_ms to ${asrTimeout}`);
     const dgOptions = this.data.recognizer.deepgramOptions = this.data.recognizer.deepgramOptions || {};
     dgOptions.utteranceEndMs = dgOptions.utteranceEndMs || asrTimeout;
@@ -163,7 +303,6 @@ class SttTask extends Task {
       detail: evt.error,
       vendor: this.vendor,
     }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
-    this.notifyError({msg: 'ASR error', details:`Failed connecting to speech vendor ${this.vendor}: ${evt.error}`});
   }
 
   _onVendorConnectFailure(cs, _ep, evt) {
@@ -176,7 +315,6 @@ class SttTask extends Task {
       message: `Failed connecting to ${this.vendor} speech recognizer: ${reason}`,
       vendor: this.vendor,
     }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
-    this.notifyError({msg: 'ASR error', details:`Failed connecting to speech vendor ${this.vendor}: ${reason}`});
   }
 }
 

lib/tasks/tag.js

@@ -12,7 +12,7 @@ class TaskTag extends Task {
   async exec(cs) {
     super.exec(cs);
     cs.callInfo.customerData = this.data;
-    //this.logger.debug({callInfo: cs.callInfo.toJSON()}, 'TaskTag:exec set customer data in callInfo');
+    this.logger.debug({customerData: cs.callInfo.customerData}, 'TaskTag:exec set customer data in callInfo');
   }
 }
 

lib/tasks/task.js

@@ -160,14 +160,31 @@ class Task extends Emitter {
       const httpHeaders = b3 && {b3};
       span.setAttributes({'http.body': JSON.stringify(params)});
       try {
+        if (this.id) params.verb_id = this.id;
         const json = await this.cs.requestor.request(type, this.actionHook, params, httpHeaders);
         span.setAttributes({'http.statusCode': 200});
-        span.end();
+        const isWsConnection = this.cs.requestor instanceof WsRequestor;
+        if (!isWsConnection || (expectResponse && json && Array.isArray(json) && json.length)) {
+          span.end();
+        } else {
+          /** we use this span to measure application response latency,
+           * and with websocket connections we generally get the application's response
+           * in a subsequent message from the far end, so we terminate the span when the
+           * first new set of verbs arrive after sending a transcript
+           * */
+          this.emit('VerbHookSpanWaitForEnd', {span});
+
+          // If actionHook delay action is configured, and ws application have not responded yet any verb for actionHook
+          // We have to transfer the task to call-session to await on next ws command verbs, and also run action Hook
+          // delay actions
+          if (this.hookDelayActionOpts) {
+            this.emit('ActionHookDelayActionOptions', this.hookDelayActionOpts);
+          }
+        }
         if (expectResponse && json && Array.isArray(json)) {
           const makeTask = require('./make_task');
           const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
           if (tasks && tasks.length > 0) {
-            this.logger.info({tasks: tasks}, `${this.name} replacing application with ${tasks.length} tasks`);
             this.callSession.replaceApplication(tasks);
           }
         }

lib/tasks/transcribe.js

@@ -31,21 +31,59 @@ class TaskTranscribe extends SttTask {
       this.separateRecognitionPerChannel = this.data.recognizer.separateRecognitionPerChannel;
     }
 
+    /* for nested transcribe in dial, unless the app explicitly says so we want to transcribe both legs */
+    if (this.parentTask?.name === TaskName.Dial) {
+      if (this.data.channel === 1 || this.data.channel === 2) {
+        /* transcribe only the channel specified */
+        this.separateRecognitionPerChannel = false;
+        this.channel = this.data.channel;
+        logger.debug(`TaskTranscribe: transcribing only channel ${this.channel} in the Dial verb`);
+      }
+      else if (this.separateRecognitionPerChannel !== false) {
+        this.separateRecognitionPerChannel = true;
+      }
+      else {
+        this.channel = 1;
+      }
+    }
+    else {
+      this.channel = 1;
+    }
+
     this.childSpan = [null, null];
 
-    // Continuos asr timeout
+    // Continuous asr timeout
     this.asrTimeout = typeof this.data.recognizer.asrTimeout === 'number' ? this.data.recognizer.asrTimeout * 1000 : 0;
     if (this.asrTimeout > 0) {
       this.isContinuousAsr = true;
     }
     /* buffer speech for continuous asr */
-    this._bufferedTranscripts = [];
+    this._bufferedTranscripts = [ [], [] ];  // for channel 1 and 2
+    this.bugname_prefix = 'transcribe_';
+    this.paused = false;
   }
 
   get name() { return TaskName.Transcribe; }
 
+  get transcribing1() {
+    return this.channel === 1 || this.separateRecognitionPerChannel;
+  }
+
+  get transcribing2() {
+    return this.channel === 2 || this.separateRecognitionPerChannel && this.ep2;
+  }
+
   async exec(cs, {ep, ep2}) {
-    super.exec(cs);
+    await super.exec(cs, {ep, ep2});
+
+    if (this.data.recognizer.vendor === 'nuance') {
+      this.data.recognizer.nuanceOptions = {
+        // by default, nuance STT will recognize only 1st utterance.
+        // enable multiple allow nuance detact all utterances
+        utteranceDetectionMode: 'multiple',
+        ...this.data.recognizer.nuanceOptions
+      };
+    }
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
 
     if (cs.hasGlobalSttHints) {
@@ -55,95 +93,42 @@ class TaskTranscribe extends SttTask {
       this.logger.debug({hints: this.data.recognizer.hints, hintsBoost: this.data.recognizer.hintsBoost},
         'Transcribe:exec - applying global sttHints');
     }
-    if (cs.hasAltLanguages) {
-      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages.concat(cs.altLanguages);
-      this.logger.debug({altLanguages: this.altLanguages},
-        'Transcribe:exec - applying altLanguages');
-    }
-    if (cs.hasGlobalSttPunctuation && !this.data.recognizer.punctuation) {
-      this.data.recognizer.punctuation = cs.globalSttPunctuation;
-    }
-
-    this.ep = ep;
-    this.ep2 = ep2;
-    if ('default' === this.vendor || !this.vendor) {
-      this.vendor = cs.speechRecognizerVendor;
-      if (this.data.recognizer) this.data.recognizer.vendor = this.vendor;
-    }
-    if ('default' === this.language || !this.language) {
-      this.language = cs.speechRecognizerLanguage;
-      if (this.data.recognizer) this.data.recognizer.language = this.language;
-    }
-    if ('default' === this.label || !this.label) {
-      this.label = cs.speechRecognizerLabel;
-      if (this.data.recognizer) this.data.recognizer.label = this.label;
-    }
-    // Fallback options
-    if ('default' === this.fallbackVendor || !this.fallbackVendor) {
-      this.fallbackVendor = cs.fallbackSpeechRecognizerVendor;
-      if (this.data.recognizer) this.data.recognizer.fallbackVendor = this.fallbackVendor;
-    }
-    if ('default' === this.fallbackLanguage || !this.fallbackLanguage) {
-      this.fallbackLanguage = cs.fallbackSpeechRecognizerLanguage;
-      if (this.data.recognizer) this.data.recognizer.fallbackLanguage = this.fallbackLanguage;
-    }
-    if ('default' === this.fallbackLabel || !this.fallbackLabel) {
-      this.fallbackLabel = cs.fallbackSpeechRecognizerLabel;
-      if (this.data.recognizer) this.data.recognizer.fallbackLabel = this.fallbackLabel;
-    }
-    if (!this.data.recognizer.vendor) {
-      this.data.recognizer.vendor = this.vendor;
-    }
-    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
-      // By default, application saves cobalt model in language
-      this.data.recognizer.model = cs.speechRecognizerLanguage;
-    }
-
-    if (!this.sttCredentials) {
-      try {
-        this.sttCredentials = await this._initSpeechCredentials(cs, this.vendor, this.label);
-      } catch (error) {
-        if (this.fallbackVendor && this.isHandledByPrimaryProvider) {
-          await this._fallback();
-        } else {
-          throw error;
-        }
-      }
-    }
-
-    /* when using cobalt model is required */
-    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
-      this.notifyError({ msg: 'ASR error', details:'Cobalt requires a model to be specified'});
-      throw new Error('Cobalt requires a model to be specified');
-    }
 
     try {
-      await this._startTranscribing(cs, ep, 1);
+      if (this.transcribing1) {
-      if (this.separateRecognitionPerChannel && ep2) {
+        await this._startTranscribing(cs, ep, 1);
+      }
+      if (this.transcribing2) {
         await this._startTranscribing(cs, ep2, 2);
       }
 
       updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid)
         .catch(() => {/*already logged error */});
-
-      await this.awaitTaskDone();
     } catch (err) {
-      this.logger.info(err, 'TaskTranscribe:exec - error');
+      if (!(await this._startFallback(cs, ep, {error: err}))) {
-      this.parentTask && this.parentTask.emit('error', err);
+        this.logger.info(err, 'TaskTranscribe:exec - error');
+        this.parentTask && this.parentTask.emit('error', err);
+        this.removeCustomEventListeners();
+        return;
+      }
     }
-    this.removeSpeechListeners(ep);
+    await this.awaitTaskDone();
+    this.removeCustomEventListeners();
   }
 
   async _stopTranscription() {
     let stopTranscription = false;
-    if (this.ep?.connected) {
+    if (this.transcribing1 && this.ep?.connected) {
       stopTranscription = true;
-      this.ep.stopTranscription({vendor: this.vendor})
+      this.ep.stopTranscription({
+        vendor: this.vendor,
+        bugname: this.bugname
+      })
         .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
     }
-    if (this.separateRecognitionPerChannel && this.ep2 && this.ep2.connected) {
+    if (this.transcribing2 && this.ep2?.connected) {
       stopTranscription = true;
-      this.ep2.stopTranscription({vendor: this.vendor})
+      this.ep2.stopTranscription({vendor: this.vendor, bugname: this.bugname})
         .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
     }
 
@@ -165,13 +150,13 @@ class TaskTranscribe extends SttTask {
       this.logger.info(`TaskTranscribe:updateTranscribe status ${status}`);
       switch (status) {
         case TranscribeStatus.Pause:
+          this.paused = true;
           await this._stopTranscription();
           break;
         case TranscribeStatus.Resume:
-          await this._startTranscribing(this.cs, this.ep, 1);
+          this.paused = false;
-          if (this.separateRecognitionPerChannel && this.ep2) {
+          if (this.transcribing1) await this._startTranscribing(this.cs, this.ep, 1);
-            await this._startTranscribing(this.cs, this.ep2, 2);
+          if (this.transcribing2) await this._startTranscribing(this.cs, this.ep2, 2);
-          }
           break;
       }
     }
@@ -186,51 +171,47 @@ class TaskTranscribe extends SttTask {
       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':
-        this.bugname = 'google_transcribe';
+        this.bugname = `${this.bugname_prefix}google_transcribe`;
-        ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, GoogleTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(GoogleTranscriptionEvents.NoAudioDetected,
+        this.addCustomEventListener(ep, GoogleTranscriptionEvents.NoAudioDetected,
           this._onNoAudio.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(GoogleTranscriptionEvents.MaxDurationExceeded,
+        this.addCustomEventListener(ep, GoogleTranscriptionEvents.MaxDurationExceeded,
           this._onMaxDurationExceeded.bind(this, cs, ep, channel));
         break;
 
       case 'aws':
       case 'polly':
-        this.bugname = 'aws_transcribe';
+        this.bugname = `${this.bugname_prefix}aws_transcribe`;
-        ep.addCustomEventListener(AwsTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(AwsTranscriptionEvents.NoAudioDetected,
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.NoAudioDetected,
           this._onNoAudio.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(AwsTranscriptionEvents.MaxDurationExceeded,
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.MaxDurationExceeded,
           this._onMaxDurationExceeded.bind(this, cs, ep, channel));
         break;
       case 'microsoft':
-        this.bugname = 'azure_transcribe';
+        this.bugname = `${this.bugname_prefix}azure_transcribe`;
-        ep.addCustomEventListener(AzureTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, AzureTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected,
+        //this.addCustomEventListener(ep, AzureTranscriptionEvents.NoSpeechDetected,
-          this._onNoAudio.bind(this, cs, ep, channel));
+        //  this._onNoAudio.bind(this, cs, ep, channel));
         break;
       case 'nuance':
-        this.bugname = 'nuance_transcribe';
+        this.bugname = `${this.bugname_prefix}nuance_transcribe`;
-        ep.addCustomEventListener(NuanceTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.StartOfSpeech,
-          this._onStartOfSpeech.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.TranscriptionComplete,
-          this._onTranscriptionComplete.bind(this, cs, ep, channel));
         break;
       case 'deepgram':
-        this.bugname = 'deepgram_transcribe';
+        this.bugname = `${this.bugname_prefix}deepgram_transcribe`;
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Connect,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Connect,
           this._onVendorConnect.bind(this, cs, ep));
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.ConnectFailure,
           this._onVendorConnectFailure.bind(this, cs, ep, channel));
 
         /* if app sets deepgramOptions.utteranceEndMs they essentially want continuous asr */
@@ -238,13 +219,13 @@ class TaskTranscribe extends SttTask {
 
         break;
       case 'soniox':
-        this.bugname = 'soniox_transcribe';
+        this.bugname = `${this.bugname_prefix}soniox_transcribe`;
-        ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, SonioxTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         break;
       case 'cobalt':
-        this.bugname = 'cobalt_transcribe';
+        this.bugname = `${this.bugname_prefix}cobalt_transcribe`;
-        ep.addCustomEventListener(CobaltTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, CobaltTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
 
         /* cobalt doesnt have language, it has model, which is required */
@@ -273,45 +254,40 @@ class TaskTranscribe extends SttTask {
         break;
 
       case 'ibm':
-        this.bugname = 'ibm_transcribe';
+        this.bugname = `${this.bugname_prefix}ibm_transcribe`;
-        ep.addCustomEventListener(IbmTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(IbmTranscriptionEvents.Connect,
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Connect,
           this._onVendorConnect.bind(this, cs, ep));
-        ep.addCustomEventListener(IbmTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.ConnectFailure,
           this._onVendorConnectFailure.bind(this, cs, ep, channel));
         break;
 
       case 'nvidia':
-        this.bugname = 'nvidia_transcribe';
+        this.bugname = `${this.bugname_prefix}nvidia_transcribe`;
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.Transcription,
-          this._onTranscription.bind(this, cs, ep));
+          this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.StartOfSpeech,
-          this._onStartOfSpeech.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.TranscriptionComplete,
-          this._onTranscriptionComplete.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.VadDetected,
-          this._onVadDetected.bind(this, cs, ep));
         break;
 
       case 'assemblyai':
-        this.bugname = 'assemblyai_transcribe';
+        this.bugname = `${this.bugname_prefix}assemblyai_transcribe`;
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep,
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+          AssemblyAiTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
-        ep.addCustomEventListener(AssemblyAiTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.ConnectFailure,
           this._onVendorConnectFailure.bind(this, cs, ep, channel));
         break;
 
       default:
         if (this.vendor.startsWith('custom:')) {
-          this.bugname = `${this.vendor}_transcribe`;
+          this.bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
-          ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription,
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.Transcription,
             this._onTranscription.bind(this, cs, ep, channel));
-          ep.addCustomEventListener(JambonzTranscriptionEvents.Connect, this._onJambonzConnect.bind(this, cs, ep));
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
-          ep.addCustomEventListener(JambonzTranscriptionEvents.ConnectFailure,
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.ConnectFailure,
-            this._onJambonzConnectFailure.bind(this, cs, ep));
+            this._onVendorConnectFailure.bind(this, cs, ep));
           break;
         }
         else {
@@ -322,7 +298,7 @@ class TaskTranscribe extends SttTask {
     }
 
     /* common handler for all stt engine errors */
-    ep.addCustomEventListener(JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
+    this.addCustomEventListener(ep, JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
     await ep.set(opts)
       .catch((err) => this.logger.info(err, 'Error setting channel variables'));
   }
@@ -337,11 +313,13 @@ class TaskTranscribe extends SttTask {
   }
 
   async _transcribe(ep) {
+    this.logger.debug(
+      `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,
+      channels: 1,
       bugname: this.bugname,
       hostport: this.hostport
     });
@@ -350,57 +328,122 @@ 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');
+    const finished = fsEvent.getHeader('transcription-session-finished');
+    const bufferedTranscripts = this._bufferedTranscripts[channel - 1];
     if (bugname && this.bugname !== bugname) return;
+    if (this.paused) {
+      this.logger.debug({evt}, 'TaskTranscribe:_onTranscription - paused, ignoring transcript');
+    }
 
     if (this.vendor === 'ibm' && evt?.state === 'listening') return;
 
     if (this.vendor === 'deepgram' && evt.type === 'UtteranceEnd') {
       /* we will only get this when we have set utterance_end_ms */
-      if (this._bufferedTranscripts.length === 0) {
+      if (bufferedTranscripts.length === 0) {
         this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram but no buffered transcripts');
       }
       else {
         this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram, return buffered transcript');
-        evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+        evt = this.consolidateTranscripts(bufferedTranscripts, channel, this.language, this.vendor);
-        this._bufferedTranscripts = [];
+        evt.is_final = true;
-        this._resolve('speech', evt);
+        this._bufferedTranscripts[channel - 1] = [];
+        this._resolve(channel, evt);
       }
       return;
     }
     this.logger.debug({evt}, 'TaskTranscribe:_onTranscription - before normalization');
 
-    evt = this.normalizeTranscription(evt, this.vendor, channel, this.language);
+    evt = this.normalizeTranscription(evt, this.vendor, channel, this.language, undefined,
+      this.data.recognizer.punctuation);
     this.logger.debug({evt}, 'TaskTranscribe:_onTranscription');
     if (evt.alternatives.length === 0) {
       this.logger.info({evt}, 'TaskTranscribe:_onTranscription - got empty transcript, continue listening');
       return;
     }
 
-    if (evt.alternatives[0]?.transcript === '' && !cs.callGone && !this.killed) {
+    let emptyTranscript = false;
-      if (['microsoft', 'deepgram'].includes(this.vendor)) {
+    if (evt.is_final) {
-        this.logger.info({evt}, 'TaskTranscribe:_onTranscription - got empty transcript, continue listening');
+      if (evt.alternatives.length === 0 || evt.alternatives[0].transcript === '' && !cs.callGone && !this.killed) {
+        emptyTranscript = true;
+        if (finished === 'true' &&
+          ['microsoft', 'deepgram'].includes(this.vendor) &&
+          bufferedTranscripts.length === 0) {
+          this.logger.debug({evt}, 'TaskGather:_onTranscription - got empty transcript from old gather, disregarding');
+          return;
+        }
+        else if (this.vendor !== 'deepgram') {
+          this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, continue listening');
+          return;
+        }
+        else if (this.isContinuousAsr) {
+          this.logger.info({evt},
+            'TaskGather:_onTranscription - got empty deepgram transcript during continous asr, continue listening');
+          return;
+        }
+        else if (this.vendor === 'deepgram' && bufferedTranscripts.length > 0) {
+          this.logger.info({evt},
+            'TaskGather:_onTranscription - got empty transcript from deepgram, return the buffered transcripts');
+        }
+      }
+      if (this.isContinuousAsr) {
+        /* append the transcript and start listening again for asrTimeout */
+        const t = evt.alternatives[0].transcript;
+        if (t) {
+          /* remove trailing punctuation */
+          if (/[,;:\.!\?]$/.test(t)) {
+            this.logger.debug('TaskGather:_onTranscription - removing trailing punctuation');
+            evt.alternatives[0].transcript = t.slice(0, -1);
+          }
+        }
+        this.logger.info({evt}, 'TaskGather:_onTranscription - got transcript during continous asr');
+        bufferedTranscripts.push(evt);
+        this._startAsrTimer(channel);
+
+        /* some STT engines will keep listening after a final response, so no need to restart */
+        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google']
+          .includes(this.vendor)) this._startTranscribing(cs, ep, channel);
       }
       else {
-        this.logger.info({evt}, 'TaskTranscribe:_onTranscription - got empty transcript, listen again');
+        if (this.vendor === 'soniox') {
-        this._transcribe(ep);
+          /* compile transcripts into one */
-      }
+          this._sonioxTranscripts.push(evt.vendor.finalWords);
-      return;
+          evt = this.compileSonioxTranscripts(this._sonioxTranscripts, 1, this.language);
-    }
+          this._sonioxTranscripts = [];
+        }
+        else if (this.vendor === 'deepgram') {
+          /* compile transcripts into one */
+          if (!emptyTranscript) bufferedTranscripts.push(evt);
 
-    if (this.vendor === 'soniox') {
+          /* deepgram can send an empty and final transcript; only if we have any buffered should we resolve */
-      /* compile transcripts into one */
+          if (bufferedTranscripts.length === 0) return;
-      this._sonioxTranscripts.push(evt.vendor.finalWords);
+          evt = this.consolidateTranscripts(bufferedTranscripts, channel, this.language);
-      if (evt.is_final) {
+          this._bufferedTranscripts[channel - 1] = [];
-        evt = this.compileSonioxTranscripts(this._sonioxTranscripts, 1, this.language);
+        }
-        this._sonioxTranscripts = [];
+
+        /* here is where we return a final transcript */
+        this.logger.debug({evt}, 'TaskTranscribe:_onTranscription - sending final transcript');
+        this._resolve(channel, evt);
+        /* some STT engines will keep listening after a final response, so no need to restart */
+        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google']
+          .includes(this.vendor)) this._startTranscribing(cs, ep, channel);
       }
     }
+    else {
+      /* interim transcript */
 
-    if (this.isContinuousAsr && evt.is_final) {
+      /* deepgram can send a non-final transcript but with words that are final, so we need to buffer */
-      this._bufferedTranscripts.push(evt);
+      if (this.vendor === 'deepgram') {
-      this._startAsrTimer(channel);
+        const originalEvent = evt.vendor.evt;
-    } else {
+        if (originalEvent.is_final && evt.alternatives[0].transcript !== '') {
-      await this._resolve(channel, evt);
+          this.logger.debug({evt}, 'Gather:_onTranscription - buffering a completed (partial) deepgram transcript');
+          bufferedTranscripts.push(evt);
+        }
+      }
+
+      if (this.interim) {
+        this.logger.debug({evt}, 'TaskTranscribe:_onTranscription - sending interim transcript');
+        this._resolve(channel, evt);
+      }
     }
   }
 
@@ -453,7 +496,8 @@ class TaskTranscribe extends SttTask {
   }
 
   _onNoAudio(cs, ep, channel) {
-    this.logger.debug(`TaskTranscribe:_onNoAudio restarting transcription on channel ${channel}`);
+    this.logger.debug(`TaskTranscribe:_onNoAudio on channel ${channel}`);
+    if (this.paused) return;
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
@@ -469,7 +513,8 @@ class TaskTranscribe extends SttTask {
   }
 
   _onMaxDurationExceeded(cs, ep, channel) {
-    this.logger.debug(`TaskTranscribe:_onMaxDurationExceeded restarting transcription on channel ${channel}`);
+    this.logger.debug(`TaskTranscribe:_onMaxDurationExceeded on channel ${channel}`);
+    if (this.paused) return;
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
@@ -492,44 +537,66 @@ class TaskTranscribe extends SttTask {
     }
   }
 
-  async _onJambonzError(cs, _ep, evt) {
+  async _startFallback(cs, _ep, evt) {
-    this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
+    if (this.canFallback) {
-    if (this.isHandledByPrimaryProvider && this.fallbackVendor) {
+      _ep.stopTranscription({
-      _ep.stopTranscription({vendor: this.vendor})
+        vendor: this.vendor,
+        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);
       try {
-        await this._fallback();
+        this.notifyError({ msg: 'ASR error',
+          details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'in progress'});
+        await this._initFallback();
         let channel = 1;
         if (this.ep !== _ep) {
           channel = 2;
         }
+        this[`_speechHandlersSet_${channel}`] = false;
         this._startTranscribing(cs, _ep, channel);
         updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
-        return;
+        return true;
       } catch (error) {
+        this.notifyError({ msg: 'ASR error',
+          details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'not available'});
         this.logger.info({error}, `There is error while falling back to ${this.fallbackVendor}`);
       }
     } else {
-      const {writeAlerts, AlertType} = cs.srf.locals;
+      this.logger.debug('transcribe:_startFallback no condition for falling back');
+      this.notifyError({ msg: 'ASR error',
+        details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'not available'});
+    }
+    return false;
+  }
 
-      if (this.vendor === 'nuance') {
+  async _onJambonzError(cs, _ep, evt) {
-        const {code, error} = evt;
+    if (this.vendor === 'google' && evt.error_code === 0) {
-        if (code === 404 && error === 'No speech') return this._resolve('timeout');
+      this.logger.info({evt}, 'TaskTranscribe:_onJambonzError - ignoring google error code 0');
-        if (code === 413 && error === 'Too much speech') return this._resolve('timeout');
+      return;
-      }
+    }
-      this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
+    this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
-      writeAlerts({
+    if (this.paused) return;
-        account_sid: cs.accountSid,
+    const {writeAlerts, AlertType} = cs.srf.locals;
-        alert_type: AlertType.STT_FAILURE,
+
-        message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
+    if (this.vendor === 'nuance') {
-        vendor: this.vendor,
+      const {code, error} = evt;
-      }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
+      if (code === 404 && error === 'No speech') return this._resolve('timeout');
-      this.notifyError({msg: 'ASR error', details:`Custom speech vendor ${this.vendor} error: ${evt.error}`});
+      if (code === 413 && error === 'Too much speech') return this._resolve('timeout');
+    }
+    this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
+    writeAlerts({
+      account_sid: cs.accountSid,
+      alert_type: AlertType.STT_FAILURE,
+      message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
+      vendor: this.vendor,
+    }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
+    if (!(await this._startFallback(cs, _ep, evt))) {
+      this.notifyTaskDone();
     }
   }
 
-  _onVendorConnectFailure(cs, _ep, channel, evt) {
+  async _onVendorConnectFailure(cs, _ep, channel, evt) {
     super._onVendorConnectFailure(cs, _ep, evt);
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
@@ -538,7 +605,9 @@ class TaskTranscribe extends SttTask {
       });
       this.childSpan[channel - 1].span.end();
     }
-    this.notifyTaskDone();
+    if (!(await this._startFallback(cs, _ep, evt))) {
+      this.notifyTaskDone();
+    }
   }
 
   _startAsrTimer(channel) {
@@ -547,8 +616,9 @@ class TaskTranscribe extends SttTask {
     this._clearAsrTimer(channel);
     this._asrTimer = setTimeout(() => {
       this.logger.debug(`TaskTranscribe:_startAsrTimer - asr timer went off for channel: ${channel}`);
-      const evt = this.consolidateTranscripts(this._bufferedTranscripts, channel, this.language);
+      const evt = this.consolidateTranscripts(
-      this._bufferedTranscripts = [];
+        this._bufferedTranscripts[channel - 1], channel, this.language, this.vendor);
+      this._bufferedTranscripts[channel - 1] = [];
       this._resolve(channel, evt);
     }, this.asrTimeout);
     this.logger.debug(`TaskTranscribe:_startAsrTimer: set for ${this.asrTimeout}ms for channel ${channel}`);

lib/tasks/tts-task.js

@@ -0,0 +1,180 @@
+const Task = require('./task');
+const { TaskPreconditions } = require('../utils/constants');
+
+class TtsTask extends Task {
+
+  constructor(logger, data, parentTask) {
+    super(logger, data);
+    this.parentTask = parentTask;
+
+    this.preconditions = TaskPreconditions.Endpoint;
+
+    this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
+    this.synthesizer = this.data.synthesizer || {};
+    this.disableTtsCache = this.data.disableTtsCache;
+    this.options = this.synthesizer.options || {};
+  }
+
+  async exec(cs) {
+    super.exec(cs);
+  }
+
+  async _synthesizeWithSpecificVendor(cs, ep, {
+    vendor,
+    language,
+    voice,
+    label,
+    disableTtsStreaming,
+    preCache
+  }) {
+    const {srf, accountSid:account_sid} = cs;
+    const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
+    const {writeAlerts, AlertType, stats} = srf.locals;
+    const {synthAudio} = srf.locals.dbHelpers;
+    const engine = this.synthesizer.engine || 'standard';
+    const salt = cs.callSid;
+
+    let credentials = cs.getSpeechCredentials(vendor, 'tts', label);
+    /* parse Nuance voices into name and model */
+    let model;
+    if (vendor === 'nuance' && voice) {
+      const arr = /([A-Za-z-]*)\s+-\s+(enhanced|standard)/.exec(voice);
+      if (arr) {
+        voice = arr[1];
+        model = arr[2];
+      }
+    } else if (vendor === 'deepgram') {
+      model = voice;
+    }
+
+    /* 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;
+    } else if (vendor === 'elevenlabs') {
+      credentials = credentials || {};
+      credentials.model_id = this.options.model_id || credentials.model_id;
+      credentials.voice_settings = this.options.voice_settings || {};
+      credentials.optimize_streaming_latency = this.options.optimize_streaming_latency
+      || credentials.optimize_streaming_latency;
+      voice = this.options.voice_id || voice;
+    }
+
+    ep.set({
+      tts_engine: vendor,
+      tts_voice: voice,
+      cache_speech_handles: 1,
+    }).catch((err) => this.logger.info({err}, `${this.name}: Error setting tts_engine on endpoint`));
+
+    if (!preCache) this.logger.info({vendor, language, voice, model}, `${this.name}:exec`);
+    try {
+      if (!credentials) {
+        writeAlerts({
+          account_sid,
+          alert_type: AlertType.TTS_NOT_PROVISIONED,
+          vendor
+        }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
+        this.notifyError({
+          msg: 'TTS error',
+          details:`No speech credentials provisioned for selected vendor ${vendor}`
+        });
+        throw new Error('no provisioned speech credentials for TTS');
+      }
+      // synthesize all of the text elements
+      let lastUpdated = false;
+
+      /* produce an audio segment from the provided text */
+      const generateAudio = async(text) => {
+        if (this.killed) return;
+        if (text.startsWith('silence_stream://')) return text;
+
+        /* otel: trace time for tts */
+        if (!preCache && !this.parentTask)  {
+          const {span} = this.startChildSpan('tts-generation', {
+            'tts.vendor': vendor,
+            'tts.language': language,
+            'tts.voice': voice
+          });
+          this.otelSpan = span;
+        }
+        try {
+          const {filePath, servedFromCache, rtt} = await synthAudio(stats, {
+            account_sid,
+            text,
+            vendor,
+            language,
+            voice,
+            engine,
+            model,
+            salt,
+            credentials,
+            options: this.options,
+            disableTtsCache : this.disableTtsCache,
+            disableTtsStreaming,
+            preCache
+          });
+          if (!filePath.startsWith('say:')) {
+            this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
+            if (filePath) cs.trackTmpFile(filePath);
+            if (this.otelSpan) {
+              this.otelSpan.setAttributes({'tts.cached': servedFromCache});
+              this.otelSpan.end();
+              this.otelSpan = null;
+            }
+            if (!servedFromCache && !lastUpdated) {
+              lastUpdated = true;
+              updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
+            }
+            if (!servedFromCache && rtt && !preCache) {
+              this.notifyStatus({
+                event: 'synthesized-audio',
+                vendor,
+                language,
+                characters: text.length,
+                elapsedTime: rtt
+              });
+            }
+          }
+          else {
+            this.logger.debug('a streaming tts api will be used');
+            const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${ep.uuid},`);
+            return modifiedPath;
+          }
+          return filePath;
+        } catch (err) {
+          this.logger.info({err}, 'Error synthesizing tts');
+          if (this.otelSpan) this.otelSpan.end();
+          writeAlerts({
+            account_sid: cs.accountSid,
+            alert_type: AlertType.TTS_FAILURE,
+            vendor,
+            detail: err.message
+          }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
+          this.notifyError({msg: 'TTS error', details: err.message || err});
+          throw err;
+        }
+      };
+
+      const arr = this.text.map((t) => (this._validateURL(t) ? t : generateAudio(t)));
+      return (await Promise.all(arr)).filter((fp) => fp && fp.length);
+    } catch (err) {
+      this.logger.info(err, 'TaskSay:exec error');
+      throw err;
+    }
+  }
+
+  _validateURL(urlString) {
+    try {
+      new URL(urlString);
+      return true;
+    } catch (e) {
+      return false;
+    }
+  }
+}
+
+module.exports = TtsTask;

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/background-task-manager.js

@@ -0,0 +1,195 @@
+const { normalizeJambones } = require('@jambonz/verb-specifications');
+const makeTask = require('../tasks/make_task');
+const { JAMBONZ_RECORD_WS_BASE_URL, JAMBONZ_RECORD_WS_USERNAME, JAMBONZ_RECORD_WS_PASSWORD } = require('../config');
+const Emitter = require('events');
+
+class BackgroundTaskManager extends Emitter {
+  constructor({cs, logger, rootSpan}) {
+    super();
+    this.tasks = new Map();
+    this.cs = cs;
+    this.logger = logger;
+    this.rootSpan = rootSpan;
+  }
+
+  isTaskRunning(type) {
+    return this.tasks.has(type);
+  }
+
+  getTask(type) {
+    if (this.tasks.has(type)) {
+      return this.tasks.get(type);
+    }
+  }
+
+  count() {
+    return this.tasks.size;
+  }
+
+  async newTask(type, opts, sticky = false) {
+    this.logger.info({opts}, `initiating Background task ${type}`);
+    if (this.tasks.has(type)) {
+      this.logger.info(`Background task ${type} is running, skipped`);
+      return;
+    }
+    let task;
+    switch (type) {
+      case 'listen':
+        task = await this._initListen(opts);
+        break;
+      case 'bargeIn':
+        task = await this._initBargeIn(opts);
+        break;
+      case 'record':
+        task = await this._initRecord();
+        break;
+      case 'transcribe':
+        task = await this._initTranscribe(opts);
+        break;
+      default:
+        break;
+    }
+    if (task) {
+      this.tasks.set(type, task);
+    }
+    if (task && sticky) task.sticky = true;
+    return task;
+  }
+
+  stop(type) {
+    const task = this.getTask(type);
+    if (task) {
+      this.logger.info(`stopping background task: ${type}`);
+      task.removeAllListeners();
+      task.span.end();
+      task.kill();
+      // Remove task from managed List
+      this.tasks.delete(type);
+    }
+  }
+
+  stopAll() {
+    this.logger.debug('BackgroundTaskManager:stopAll');
+    for (const key of this.tasks.keys()) {
+      this.stop(key);
+    }
+  }
+
+  // Initiate Listen
+  async _initListen(opts, bugname = 'jambonz-background-listen', ignoreCustomerData = false, type = 'listen') {
+    let task;
+    try {
+      const t = normalizeJambones(this.logger, [opts]);
+      task = makeTask(this.logger, t[0]);
+      task.bugname = bugname;
+      task.ignoreCustomerData = ignoreCustomerData;
+      const resources = await this.cs._evaluatePreconditions(task);
+      const {span, ctx} = this.rootSpan.startChildSpan(`background-${type}:${task.summary}`);
+      task.span = span;
+      task.ctx = ctx;
+      task.exec(this.cs, resources)
+        .then(this._taskCompleted.bind(this, type, task))
+        .catch(this._taskError.bind(this, type, task));
+    } catch (err) {
+      this.logger.info({err, opts}, `BackgroundTaskManager:_initListen - Error creating ${bugname} task`);
+    }
+    return task;
+  }
+
+  // Initiate Gather
+  async _initBargeIn(opts) {
+    let task;
+    try {
+      const t = normalizeJambones(this.logger, [opts]);
+      task = makeTask(this.logger, t[0]);
+      task
+        .once('dtmf', this._bargeInTaskCompleted.bind(this))
+        .once('vad', this._bargeInTaskCompleted.bind(this))
+        .once('transcription', this._bargeInTaskCompleted.bind(this))
+        .once('timeout', this._bargeInTaskCompleted.bind(this));
+      const resources = await this.cs._evaluatePreconditions(task);
+      const {span, ctx} = this.rootSpan.startChildSpan(`background-bargeIn:${task.summary}`);
+      task.span = span;
+      task.ctx = ctx;
+      task.bugname_prefix = 'background_bargeIn_';
+      task.exec(this.cs, resources)
+        .then(() => {
+          this._taskCompleted('bargeIn', task);
+          if (task.sticky && !this.cs.callGone && !this.cs._stopping) {
+            this.logger.info('BackgroundTaskManager:_initBargeIn: restarting background bargeIn');
+            this.newTask('bargeIn', opts, true);
+          }
+          return;
+        })
+        .catch(this._taskError.bind(this, 'bargeIn', task));
+    } catch (err) {
+      this.logger.info(err, 'BackgroundTaskManager:_initGather - Error creating bargeIn task');
+    }
+    return task;
+  }
+
+  // Initiate Record
+  async _initRecord() {
+    if (this.cs.accountInfo.account.record_all_calls || this.cs.application.record_all_calls) {
+      if (!JAMBONZ_RECORD_WS_BASE_URL || !this.cs.accountInfo.account.bucket_credential) {
+        this.logger.error('_initRecord: invalid cfg - missing JAMBONZ_RECORD_WS_BASE_URL or bucket config');
+        return undefined;
+      }
+      const listenOpts = {
+        url: `${JAMBONZ_RECORD_WS_BASE_URL}/record/${this.cs.accountInfo.account.bucket_credential.vendor}`,
+        disableBidirectionalAudio: true,
+        mixType : 'stereo',
+        passDtmf: true
+      };
+      if (JAMBONZ_RECORD_WS_USERNAME && JAMBONZ_RECORD_WS_PASSWORD) {
+        listenOpts.wsAuth = {
+          username: JAMBONZ_RECORD_WS_USERNAME,
+          password: JAMBONZ_RECORD_WS_PASSWORD
+        };
+      }
+      this.logger.debug({listenOpts}, '_initRecord: enabling listen');
+      return await this._initListen({verb: 'listen', ...listenOpts}, 'jambonz-session-record', true, 'record');
+    }
+  }
+
+  // Initiate Transcribe
+  async _initTranscribe(opts) {
+    let task;
+    try {
+      const t = normalizeJambones(this.logger, [opts]);
+      task = makeTask(this.logger, t[0]);
+      const resources = await this.cs._evaluatePreconditions(task);
+      const {span, ctx} = this.rootSpan.startChildSpan(`background-transcribe:${task.summary}`);
+      task.span = span;
+      task.ctx = ctx;
+      task.bugname_prefix = 'background_transcribe_';
+      task.exec(this.cs, resources)
+        .then(this._taskCompleted.bind(this, 'transcribe', task))
+        .catch(this._taskError.bind(this, 'transcribe', task));
+    } catch (err) {
+      this.logger.info(err, 'BackgroundTaskManager:_initTranscribe - Error creating transcribe task');
+    }
+    return task;
+  }
+
+  _taskCompleted(type, task) {
+    this.logger.debug({type, task}, `BackgroundTaskManager:_taskCompleted: task completed, sticky: ${task.sticky}`);
+    task.removeAllListeners();
+    task.span.end();
+    this.tasks.delete(type);
+  }
+  _taskError(type, task, error) {
+    this.logger.info({type, task, error}, 'BackgroundTaskManager:_taskError: task Error');
+    task.removeAllListeners();
+    task.span.end();
+    this.tasks.delete(type);
+  }
+
+  _bargeInTaskCompleted(evt) {
+    this.logger.debug({evt},
+      'BackgroundTaskManager:_bargeInTaskCompleted on event from background bargeIn, emitting bargein-done event');
+    this.emit('bargeIn-done', evt);
+  }
+}
+
+module.exports = BackgroundTaskManager;

lib/utils/constants.json

@@ -1,12 +1,13 @@
 {
   "TaskName": {
-    "Cognigy": "cognigy",
+    "Answer": "answer",
     "Conference": "conference",
     "Config": "config",
     "Dequeue": "dequeue",
     "Dial": "dial",
     "Dialogflow": "dialogflow",
     "Dtmf": "dtmf",
+    "Dub": "dub",
     "Enqueue": "enqueue",
     "Gather": "gather",
     "Hangup": "hangup",
@@ -29,7 +30,8 @@
     "Tag": "tag",
     "Transcribe": "transcribe"
   },
-  "AllowedSipRecVerbs": ["config", "gather", "transcribe", "listen", "tag"],
+  "AllowedSipRecVerbs": ["answer", "config", "gather", "transcribe", "listen", "tag"],
+  "AllowedConfirmSessionVerbs": ["config", "gather", "plays", "say", "tag"],
   "CallStatus": {
     "Trying": "trying",
     "Ringing": "ringing",
@@ -95,6 +97,10 @@
     "Transcription": "soniox_transcribe::transcription",
     "Error": "soniox_transcribe::error"
   },
+  "VerbioTranscriptionEvents": {
+    "Transcription": "verbio_transcribe::transcription",
+    "Error": "verbio_transcribe::error"
+  },
   "CobaltTranscriptionEvents": {
     "Transcription": "cobalt_speech::transcription",
     "CompileContext": "cobalt_speech::compile_context_response",
@@ -132,6 +138,9 @@
     "ConnectFailure": "assemblyai_transcribe::connect_failed",
     "Connect": "assemblyai_transcribe::connect"
   },
+  "VadDetection": {
+    "Detection": "vad_detect:detection"
+  },
   "ListenEvents": {
     "Connect": "mod_audio_fork::connect",
     "ConnectFailure": "mod_audio_fork::connect_failed",
@@ -169,6 +178,7 @@
     "session:new",
     "session:reconnect",
     "session:redirect",
+    "session:adulting",
     "call:status",
     "queue:status",
     "dial:confirm",

lib/utils/db-utils.js

@@ -41,6 +41,7 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.access_key_id = o.access_key_id;
       obj.secret_access_key = o.secret_access_key;
+      obj.role_arn = o.role_arn;
       obj.aws_region = o.aws_region;
     }
     else if ('microsoft' === obj.vendor) {
@@ -75,6 +76,8 @@ const speechMapper = (cred) => {
     else if ('deepgram' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
+      obj.deepgram_stt_uri = o.deepgram_stt_uri;
+      obj.deepgram_stt_use_tls = o.deepgram_stt_use_tls;
     }
     else if ('soniox' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
@@ -91,6 +94,18 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
       obj.model_id = o.model_id;
+      obj.options = o.options;
+    } else if ('playht' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.user_id = o.user_id;
+      obj.voice_engine = o.voice_engine;
+      obj.options = o.options;
+    } else if ('rimelabs' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.model_id = o.model_id;
+      obj.options = o.options;
     } else if ('assemblyai' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
@@ -98,6 +113,11 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
       obj.model_id = o.model_id;
+    }  else if ('verbio' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.client_id = o.client_id;
+      obj.client_secret = o.client_secret;
+      obj.engine_version = o.engine_version;
     } else if (obj.vendor.startsWith('custom:')) {
       const o = JSON.parse(decrypt(credential));
       obj.auth_token = o.auth_token;

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/install-srf-locals.js

@@ -171,14 +171,17 @@ function installSrfLocals(srf, logger) {
     retrieveFromSortedSet,
     retrieveByPatternSortedSet,
     sortedSetLength,
-    sortedSetPositionByPattern
+    sortedSetPositionByPattern,
   } = require('@jambonz/realtimedb-helpers')({}, logger, tracer);
   const registrar = new Registrar(logger, client);
   const {
     synthAudio,
+    addFileToCache,
     getNuanceAccessToken,
     getIbmAccessToken,
-  } = require('@jambonz/speech-utils')({redis_client: client}, logger);
+    getAwsAuthToken,
+    getVerbioAccessToken
+  } = require('@jambonz/speech-utils')({}, logger);
   const {
     writeAlerts,
     AlertType
@@ -215,6 +218,8 @@ function installSrfLocals(srf, logger) {
       listCalls,
       deleteCall,
       synthAudio,
+      getAwsAuthToken,
+      addFileToCache,
       createHash,
       retrieveHash,
       deleteKey,
@@ -235,7 +240,8 @@ function installSrfLocals(srf, logger) {
       retrieveFromSortedSet,
       retrieveByPatternSortedSet,
       sortedSetLength,
-      sortedSetPositionByPattern
+      sortedSetPositionByPattern,
+      getVerbioAccessToken
     },
     parentLogger: logger,
     getSBC,

lib/utils/parse-decibels.js

@@ -0,0 +1,18 @@
+const parseDecibels = (db) => {
+  if (!db) return 0;
+  if (typeof db === 'number') {
+    return db;
+  }
+  else if (typeof db === 'string') {
+    const match = db.match(/([+-]?\d+(\.\d+)?)\s*db/i);
+    if (match) {
+      return Math.trunc(parseFloat(match[1]));
+    } else {
+      return 0;
+    }
+  } else {
+    return 0;
+  }
+};
+
+module.exports = parseDecibels;

lib/utils/place-outdial.js

@@ -13,9 +13,16 @@ const moment = require('moment');
 const stripCodecs = require('./strip-ancillary-codecs');
 const RootSpan = require('./call-tracer');
 const uuidv4 = require('uuid-random');
+const HttpRequestor = require('./http-requestor');
+const WsRequestor = require('./ws-requestor');
+const {makeOpusFirst} = require('./sdp-utils');
+const {
+  JAMBONES_USE_FREESWITCH_TIMER_FD
+} = require('../config');
 
 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);
 
@@ -38,6 +45,7 @@ class SingleDialer extends Emitter {
 
     this.callSid = uuidv4();
     this.dialTask = dialTask;
+    this.onHoldMusic = onHoldMusic;
 
     this.on('callStatusChange', this._notifyCallStatusChange.bind(this));
   }
@@ -128,6 +136,7 @@ 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}`);
 
       /**
@@ -149,7 +158,7 @@ class SingleDialer extends Emitter {
 
       Object.assign(opts, {
         proxy: `sip:${this.sbcAddress}`,
-        localSdp: this.ep.local.sdp
+        localSdp: opts.opusFirst ? makeOpusFirst(this.ep.local.sdp) : this.ep.local.sdp
       });
       if (this.target.auth) opts.auth = this.target.auth;
       inviteSpan = this.startSpan('invite', {
@@ -177,6 +186,7 @@ class SingleDialer extends Emitter {
            *  (a) create a logger for this call
            */
           req.srf = srf;
+          this.req = req;
           this.callInfo = new CallInfo({
             direction: CallDirection.Outbound,
             parentCallInfo: this.parentCallInfo,
@@ -185,6 +195,10 @@ class SingleDialer extends Emitter {
             callSid: this.callSid,
             traceId: this.rootSpan.traceId
           });
+          if (this.dialTask && this.dialTask.tag !== null &&
+            typeof this.dialTask.tag === 'object' && !Array.isArray(this.dialTask.tag)) {
+            this.callInfo.customerData = this.dialTask.tag;
+          }
           this.logger = srf.locals.parentLogger.child({
             callSid: this.callSid,
             parentCallSid: this.parentCallInfo.callSid,
@@ -249,7 +263,7 @@ 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 {
@@ -281,17 +295,17 @@ class SingleDialer extends Emitter {
         if (err.status === 487) status.callStatus = CallStatus.NoAnswer;
         else if ([486, 600].includes(err.status)) status.callStatus = CallStatus.Busy;
         this.logger.info(`SingleDialer:exec outdial failure ${err.status}`);
-        inviteSpan.setAttributes({'invite.status_code': err.status});
+        inviteSpan?.setAttributes({'invite.status_code': err.status});
-        inviteSpan.end();
+        inviteSpan?.end();
       }
       else {
         this.logger.error(err, 'SingleDialer:exec');
         status.sipStatus = 500;
-        inviteSpan.setAttributes({
+        inviteSpan?.setAttributes({
           'invite.status_code': 500,
           'invite.err': err.message
         });
-        inviteSpan.end();
+        inviteSpan?.end();
       }
       this.emit('callStatusChange', status);
       if (this.ep) this.ep.destroy();
@@ -316,6 +330,16 @@ class SingleDialer extends Emitter {
     }
   }
 
+  _configMsEndpoint() {
+    const opts = {
+      ...(this.onHoldMusic && {holdMusic: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`}),
+      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'})
+    };
+    if (Object.keys(opts).length > 0) {
+      this.ep.set(opts);
+    }
+  }
+
   /**
    * Run an application on the call after answer, e.g. call screening.
    * Once the application completes in some fashion, emit an 'accepted' event
@@ -329,6 +353,7 @@ class SingleDialer extends Emitter {
       const json = await this.requestor.request('dial:confirm', confirmHook, this.callInfo.toJSON());
       if (!json || (Array.isArray(json) && json.length === 0)) {
         this.logger.info('SingleDialer:_executeApp: no tasks returned from confirm hook');
+        this.emit('accept');
         return;
       }
       const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
@@ -383,15 +408,45 @@ class SingleDialer extends Emitter {
     this.dlg.linkedSpanId = this.rootSpan.traceId;
     const rootSpan = new RootSpan('outbound-call', this.dlg);
     const newLogger = logger.child({traceId: rootSpan.traceId});
+    //clone application from parent call with new requestor
+    //parrent application will be closed in case the parent hangup
+    const app = {...application};
+    if ('WS' === app.call_hook?.method ||
+      app.call_hook?.url.startsWith('ws://') || app.call_hook?.url.startsWith('wss://')) {
+      if (app.call_hook?.url) app.call_hook.url += '/adulting';
+      const requestor = new WsRequestor(logger, this.accountInfo.account.account_sid,
+        app.call_hook, this.accountInfo.account.webhook_secret);
+      app.requestor = requestor;
+      app.notifier = requestor;
+      app.call_hook.method = 'WS';
+    }
+    else {
+      app.requestor = new HttpRequestor(logger, this.accountInfo.account.account_sid,
+        app.call_hook, this.accountInfo.account.webhook_secret);
+      if (app.call_status_hook) app.notifier = new HttpRequestor(logger,
+        this.accountInfo.account.account_sid, app.call_status_hook,
+        this.accountInfo.account.webhook_secret);
+      else app.notifier = {request: () => {}, close: () => {}};
+    }
+    // Replace old application with new application.
+    this.application = app;
     const cs = new AdultingCallSession({
       logger: newLogger,
       singleDialer: this,
-      application,
+      application: app,
       callInfo: this.callInfo,
       accountInfo: this.accountInfo,
       tasks,
       rootSpan
     });
+    app.requestor.request('session:adulting', '/adulting', {
+      ...cs.callInfo.toJSON(),
+      parentCallInfo: this.parentCallInfo
+    }).catch((err) => {
+      newLogger.error({err}, 'doAdulting: error sending adulting request');
+    });
+
+    cs.req = this.req;
     cs.exec().catch((err) => newLogger.error({err}, 'doAdulting: error executing session'));
     return cs;
   }
@@ -412,6 +467,7 @@ 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'
@@ -442,11 +498,12 @@ 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/sdp-utils.js

@@ -12,6 +12,30 @@ const mergeSdpMedia = (sdp1, sdp2) => {
   return sdpTransform.write(parsedSdp1);
 };
 
+const getCodecPlacement = (parsedSdp, codec) => parsedSdp?.media[0]?.rtp?.findIndex((e) => e.codec === codec);
+
+const isOpusFirst = (sdp) => {
+  return getCodecPlacement(sdpTransform.parse(sdp), 'opus') === 0;
+};
+
+const makeOpusFirst = (sdp) => {
+  const parsedSdp = sdpTransform.parse(sdp);
+  // Find the index of the OPUS codec
+  const opusIndex = getCodecPlacement(parsedSdp, 'opus');
+
+  // Move OPUS codec to the beginning
+  if (opusIndex > 0) {
+    const opusEntry = parsedSdp.media[0].rtp.splice(opusIndex, 1)[0];
+    parsedSdp.media[0].rtp.unshift(opusEntry);
+
+    // Also move the corresponding payload type in the "m" line
+    const opusPayloadType = parsedSdp.media[0].payloads.split(' ')[opusIndex];
+    const otherPayloadTypes = parsedSdp.media[0].payloads.split(' ').filter((pt) => pt != opusPayloadType);
+    parsedSdp.media[0].payloads = [opusPayloadType, ...otherPayloadTypes].join(' ');
+  }
+  return sdpTransform.write(parsedSdp);
+};
+
 const extractSdpMedia = (sdp) => {
   const parsedSdp1 = sdpTransform.parse(sdp);
   if (parsedSdp1.media.length > 1) {
@@ -28,5 +52,7 @@ const extractSdpMedia = (sdp) => {
 module.exports = {
   isOnhold,
   mergeSdpMedia,
-  extractSdpMedia
+  extractSdpMedia,
+  isOpusFirst,
+  makeOpusFirst
 };

lib/utils/siprec-utils.js

@@ -97,8 +97,12 @@ const parseSiprecPayload = (req, logger) => {
             obj[`${prefix}participantstreamassoc`].forEach((ps) => {
               const part = participants[ps.$.participant_id];
               if (part) {
-                part.send = ps[`${prefix}send`][0];
+                if (ps.hasOwnProperty(`${prefix}send`)) {
-                part.recv = ps[`${prefix}recv`][0];
+                  part.send = ps[`${prefix}send`][0];
+                }
+                if (ps.hasOwnProperty(`${prefix}recv`)) {
+                  part.recv = ps[`${prefix}recv`][0];
+                }
               }
             });
           }
@@ -109,9 +113,9 @@ const parseSiprecPayload = (req, logger) => {
           obj[`${prefix}stream`].forEach((s) => {
             const streamId = s.$.stream_id;
             let sender;
-            for (const [k, v] of Object.entries(participants)) {
+            for (const v of Object.values(participants)) {
               if (v.send === streamId) {
-                sender = k;
+                sender = v;
                 break;
               }
             }
@@ -121,9 +125,15 @@ const parseSiprecPayload = (req, logger) => {
 
             sender.label = s[`${prefix}label`][0];
 
-            if (-1 !== ['1', 'a_leg', 'inbound'].indexOf(sender.label)) {
+            if (-1 !== ['1', 'a_leg', 'inbound', '10'].indexOf(sender.label)) {
-              opts.caller.aor = sender.aor ;
+              opts.caller.aor = sender.aor;
               if (sender.name) opts.caller.name = sender.name;
+              // Remap the sdp stream base on sender label
+              if (!opts.sdp1.includes(`a=label:${sender.label}`)) {
+                const tmp = opts.sdp1;
+                opts.sdp1 = opts.sdp2;
+                opts.sdp2 = tmp;
+              }
             }
             else {
               opts.callee.aor = sender.aor ;

lib/utils/transcription-utils.js

@@ -1,15 +1,5 @@
 const {
   TaskName,
-  AzureTranscriptionEvents,
-  GoogleTranscriptionEvents,
-  AwsTranscriptionEvents,
-  NuanceTranscriptionEvents,
-  DeepgramTranscriptionEvents,
-  SonioxTranscriptionEvents,
-  NvidiaTranscriptionEvents,
-  CobaltTranscriptionEvents,
-  JambonzTranscriptionEvents,
-  AssemblyAiTranscriptionEvents
 } = require('./constants.json');
 
 const stickyVars = {
@@ -112,7 +102,55 @@ const stickyVars = {
   ]
 };
 
-const consolidateTranscripts = (bufferedTranscripts, channel, language) => {
+/**
+ * @see https://developers.deepgram.com/docs/models-languages-overview
+ */
+const optimalDeepramModels = {
+  zh: ['base', 'base'],
+  'zh-CN':['base', 'base'],
+  'zh-TW': ['base', 'base'],
+  da: ['enhanced', 'enhanced'],
+  en: ['nova-2-phonecall', 'nova-2'],
+  'en-US': ['nova-2-phonecall', 'nova-2'],
+  'en-AU': ['nova-2', 'nova-2'],
+  'en-GB': ['nova-2', 'nova-2'],
+  'en-IN': ['nova-2', 'nova-2'],
+  'en-NZ': ['nova-2', 'nova-2'],
+  nl: ['nova-2', 'nova-2'],
+  fr: ['nova-2', 'nova-2'],
+  'fr-CA': ['nova-2', 'nova-2'],
+  de: ['nova-2', 'nova-2'],
+  hi: ['nova-2', 'nova-2'],
+  'hi-Latn': ['nova-2', 'nova-2'],
+  id: ['base', 'base'],
+  it: ['nova-2', 'nova-2'],
+  ja: ['enhanced', 'enhanced'],
+  ko: ['nova-2', 'nova-2'],
+  no: ['nova-2', 'nova-2'],
+  pl: ['nova-2', 'nova-2'],
+  pt: ['nova-2', 'nova-2'],
+  'pt-BR': ['nova-2', 'nova-2'],
+  'pt-PT': ['nova-2', 'nova-2'],
+  ru: ['nova-2', 'nova-2'],
+  es: ['nova-2', 'nova-2'],
+  'es-419': ['nova-2', 'nova-2'],
+  'es-LATAM': ['enhanced', 'enhanced'],
+  sv: ['nova-2', 'nova-2'],
+  ta: ['enhanced', 'enhanced'],
+  taq: ['enhanced', 'enhanced'],
+  tr: ['nova-2', 'nova-2'],
+  uk: ['nova-2', 'nova-2']
+};
+
+const selectDefaultDeepgramModel = (task, language) => {
+  if (language in optimalDeepramModels) {
+    const [gather, transcribe] = optimalDeepramModels[language];
+    return task.name === TaskName.Gather ? gather : transcribe;
+  }
+  return 'base';
+};
+
+const consolidateTranscripts = (bufferedTranscripts, channel, language, vendor) => {
   if (bufferedTranscripts.length === 1) return bufferedTranscripts[0];
   let totalConfidence = 0;
   const finalTranscript = bufferedTranscripts.reduce((acc, evt) => {
@@ -153,7 +191,7 @@ const consolidateTranscripts = (bufferedTranscripts, channel, language) => {
     totalConfidence / bufferedTranscripts.length;
   finalTranscript.alternatives[0].transcript = finalTranscript.alternatives[0].transcript.trim();
   finalTranscript.vendor = {
-    name: 'deepgram',
+    name: vendor,
     evt: bufferedTranscripts
   };
   return finalTranscript;
@@ -232,7 +270,7 @@ const normalizeDeepgram = (evt, channel, language, shortUtterance) => {
     language_code: language,
     channel_tag: channel,
     is_final: shortUtterance ? evt.is_final : evt.speech_final,
-    alternatives: [alternatives[0]],
+    alternatives: alternatives.length ? [alternatives[0]] : [],
     vendor: {
       name: 'deepgram',
       evt: copy
@@ -338,19 +376,20 @@ const normalizeNuance = (evt, channel, language) => {
   };
 };
 
-const normalizeMicrosoft = (evt, channel, language) => {
+const normalizeMicrosoft = (evt, channel, language, punctuation = true) => {
   const copy = JSON.parse(JSON.stringify(evt));
   const nbest = evt.NBest;
   const language_code = evt.PrimaryLanguage?.Language || language;
   const alternatives = nbest ? nbest.map((n) => {
     return {
       confidence: n.Confidence,
-      transcript: n.Display
+      // remove all puntuation if needed
+      transcript: punctuation ? n.Display : n.Display.replace(/\p{P}/gu, '')
     };
   }) :
     [
       {
-        transcript: evt.DisplayText || evt.Text
+        transcript: punctuation ? evt.DisplayText || evt.Text : (evt.DisplayText || evt.Text).replace(/\p{P}/gu, '')
       }
     ];
 
@@ -400,14 +439,14 @@ const normalizeAssemblyAi = (evt, channel, language) => {
 };
 
 module.exports = (logger) => {
-  const normalizeTranscription = (evt, vendor, channel, language, shortUtterance) => {
+  const normalizeTranscription = (evt, vendor, channel, language, shortUtterance, punctuation) => {
 
     //logger.debug({ evt, vendor, channel, language }, 'normalizeTranscription');
     switch (vendor) {
       case 'deepgram':
         return normalizeDeepgram(evt, channel, language, shortUtterance);
       case 'microsoft':
-        return normalizeMicrosoft(evt, channel, language);
+        return normalizeMicrosoft(evt, channel, language, punctuation);
       case 'google':
         return normalizeGoogle(evt, channel, language);
       case 'aws':
@@ -433,22 +472,15 @@ 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};
     const vendor = rOpts.vendor;
 
-    /* voice activity detection works across vendors */
-    opts = {
-      ...opts,
-      ...(vad.enable && {START_RECOGNIZING_ON_VAD: 1}),
-      ...(vad.enable && vad.voiceMs && {RECOGNIZER_VAD_VOICE_MS: vad.voiceMs}),
-      ...(vad.enable && typeof vad.mode === 'number' && {RECOGNIZER_VAD_MODE: vad.mode}),
-    };
-
     if ('google' === vendor) {
-      const model = task.name === TaskName.Gather ? 'command_and_search' : 'latest_long';
+      const useV2 = rOpts.googleOptions?.serviceVersion === 'v2';
+      const model = task.name === TaskName.Gather ?
+        (useV2 ? 'telephony_short' : 'command_and_search') :
+        (useV2 ? 'long' : 'latest_long');
       opts = {
         ...opts,
         ...(sttCredentials && {GOOGLE_APPLICATION_CREDENTIALS: JSON.stringify(sttCredentials.credentials)}),
@@ -473,13 +505,34 @@ module.exports = (logger) => {
         ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
           {GOOGLE_SPEECH_HINTS: JSON.stringify(rOpts.hints)}),
         ...(typeof rOpts.hintsBoost === 'number' && {GOOGLE_SPEECH_HINTS_BOOST: rOpts.hintsBoost}),
-        ...(rOpts.altLanguages?.length > 0 &&
+        // When altLanguages is emptylist, we have to send value to freeswitch to clear the previous settings
+        ...(rOpts.altLanguages &&
           {GOOGLE_SPEECH_ALTERNATIVE_LANGUAGE_CODES: [...new Set(rOpts.altLanguages)].join(',')}),
         ...(rOpts.interactionType &&
           {GOOGLE_SPEECH_METADATA_INTERACTION_TYPE: rOpts.interactionType}),
         ...{GOOGLE_SPEECH_MODEL: rOpts.model || model},
         ...(rOpts.naicsCode > 0 && {GOOGLE_SPEECH_METADATA_INDUSTRY_NAICS_CODE: rOpts.naicsCode}),
         GOOGLE_SPEECH_METADATA_RECORDING_DEVICE_TYPE: 'phone_line',
+        ...(useV2 && {
+          GOOGLE_SPEECH_RECOGNIZER_PARENT: `projects/${sttCredentials.credentials.project_id}/locations/global`,
+          GOOGLE_SPEECH_CLOUD_SERVICES_VERSION: 'v2',
+          ...(rOpts.googleOptions?.speechStartTimeoutMs && {
+            GOOGLE_SPEECH_START_TIMEOUT_MS: rOpts.googleOptions.speechStartTimeoutMs
+          }),
+          ...(rOpts.googleOptions?.speechEndTimeoutMs && {
+            GOOGLE_SPEECH_END_TIMEOUT_MS: rOpts.googleOptions.speechEndTimeoutMs
+          }),
+          ...(rOpts.googleOptions?.transcriptNormalization && {
+            GOOGLE_SPEECH_TRANSCRIPTION_NORMALIZATION: JSON.stringify(rOpts.googleOptions.transcriptNormalization)
+          }),
+          ...(rOpts.googleOptions?.enableVoiceActivityEvents && {
+            GOOGLE_SPEECH_ENABLE_VOICE_ACTIVITY_EVENTS: rOpts.googleOptions.enableVoiceActivityEvents
+          }),
+          ...(rOpts.sgoogleOptions?.recognizerId) && {GOOGLE_SPEECH_RECOGNIZER_ID: rOpts.googleOptions.recognizerId},
+          ...(rOpts.googleOptions?.enableVoiceActivityEvents && {
+            GOOGLE_SPEECH_ENABLE_VOICE_ACTIVITY_EVENTS: rOpts.googleOptions.enableVoiceActivityEvents
+          }),
+        }),
       };
     }
     else if (['aws', 'polly'].includes(vendor)) {
@@ -489,9 +542,10 @@ module.exports = (logger) => {
         ...(rOpts.vocabularyFilterName && {AWS_VOCABULARY_FILTER_NAME: rOpts.vocabularyFilterName}),
         ...(rOpts.filterMethod && {AWS_VOCABULARY_FILTER_METHOD: rOpts.filterMethod}),
         ...(sttCredentials && {
-          AWS_ACCESS_KEY_ID: sttCredentials.accessKeyId,
+          ...(sttCredentials.accessKeyId && {AWS_ACCESS_KEY_ID: sttCredentials.accessKeyId}),
-          AWS_SECRET_ACCESS_KEY: sttCredentials.secretAccessKey,
+          ...(sttCredentials.secretAccessKey && {AWS_SECRET_ACCESS_KEY: sttCredentials.secretAccessKey}),
-          AWS_REGION: sttCredentials.region
+          AWS_REGION: sttCredentials.region,
+          ...(sttCredentials.sessionToken && {AWS_SESSION_TOKEN: sttCredentials.sessionToken}),
         }),
       };
     }
@@ -503,7 +557,8 @@ module.exports = (logger) => {
           {AZURE_SPEECH_HINTS: rOpts.hints.map((h) => h.trim()).join(',')}),
         ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
           {AZURE_SPEECH_HINTS: rOpts.hints.map((h) => h.phrase).join(',')}),
-        ...(rOpts.altLanguages && rOpts.altLanguages.length > 0 &&
+        // When altLanguages is emptylist, we have to send value to freeswitch to clear the previous settings
+        ...(rOpts.altLanguages &&
           {AZURE_SPEECH_ALTERNATIVE_LANGUAGE_CODES: [...new Set(rOpts.altLanguages)].join(',')}),
         ...(rOpts.requestSnr && {AZURE_REQUEST_SNR: 1}),
         ...(rOpts.profanityOption && {AZURE_PROFANITY_OPTION: rOpts.profanityOption}),
@@ -517,12 +572,17 @@ module.exports = (logger) => {
         ...{AZURE_USE_OUTPUT_FORMAT_DETAILED: 1},
         ...(azureOptions.speechSegmentationSilenceTimeoutMs &&
           {AZURE_SPEECH_SEGMENTATION_SILENCE_TIMEOUT_MS: azureOptions.speechSegmentationSilenceTimeoutMs}),
+        ...(azureOptions.languageIdMode &&
+          {AZURE_LANGUAGE_ID_MODE: azureOptions.languageIdMode}),
         ...(sttCredentials && {
           ...(sttCredentials.api_key && {AZURE_SUBSCRIPTION_KEY: sttCredentials.api_key}),
           ...(sttCredentials.region && {AZURE_REGION: sttCredentials.region}),
         }),
         ...(sttCredentials.use_custom_stt && sttCredentials.custom_stt_endpoint &&
           {AZURE_SERVICE_ENDPOINT_ID: sttCredentials.custom_stt_endpoint}),
+        //azureSttEndpointId overrides sttCredentials.custom_stt_endpoint
+        ...(rOpts.azureSttEndpointId &&
+          {AZURE_SERVICE_ENDPOINT_ID: rOpts.azureSttEndpointId}),
       };
     }
     else if ('nuance' === vendor) {
@@ -574,15 +634,24 @@ module.exports = (logger) => {
       };
     }
     else if ('deepgram' === vendor) {
+      let {model} = rOpts;
       const {deepgramOptions = {}} = rOpts;
+      const deepgramUri = deepgramOptions.deepgramSttUri || sttCredentials.deepgram_stt_uri;
+      const useTls = deepgramOptions.deepgramSttUseTls || sttCredentials.deepgram_stt_use_tls;
+
+      /* default to a sensible model if not supplied */
+      if (!model) {
+        model = selectDefaultDeepgramModel(task, language);
+      }
       opts = {
         ...opts,
+        DEEPGRAM_SPEECH_MODEL: model,
+        ...(deepgramUri && {DEEPGRAM_URI: deepgramUri}),
+        ...(deepgramUri && useTls && {DEEPGRAM_USE_TLS: 1}),
         ...(sttCredentials.api_key) &&
           {DEEPGRAM_API_KEY: sttCredentials.api_key},
         ...(deepgramOptions.tier) &&
           {DEEPGRAM_SPEECH_TIER: deepgramOptions.tier},
-        ...(deepgramOptions.model) &&
-          {DEEPGRAM_SPEECH_MODEL: deepgramOptions.model},
         ...(deepgramOptions.punctuate) &&
           {DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION: 1},
         ...(deepgramOptions.smartFormatting) &&
@@ -612,7 +681,9 @@ module.exports = (logger) => {
         ...(deepgramOptions.keywords) &&
           {DEEPGRAM_SPEECH_KEYWORDS: deepgramOptions.keywords.join(',')},
         ...('endpointing' in deepgramOptions) &&
-          {DEEPGRAM_SPEECH_ENDPOINTING: deepgramOptions.endpointing},
+          {DEEPGRAM_SPEECH_ENDPOINTING: deepgramOptions.endpointing === false ? 'false' : deepgramOptions.endpointing,
+            // default DEEPGRAM_SPEECH_UTTERANCE_END_MS is 1000, will be override by user settings later if there is.
+            DEEPGRAM_SPEECH_UTTERANCE_END_MS: 1000},
         ...(deepgramOptions.utteranceEndMs) &&
           {DEEPGRAM_SPEECH_UTTERANCE_END_MS: deepgramOptions.utteranceEndMs},
         ...(deepgramOptions.vadTurnoff) &&
@@ -725,8 +796,26 @@ module.exports = (logger) => {
         ...(rOpts.hints?.length > 0 &&
           {ASSEMBLYAI_WORD_BOOST: JSON.stringify(rOpts.hints)})
       };
-    }
+    } else if ('verbio' === vendor) {
-    else if (vendor.startsWith('custom:')) {
+      const {verbioOptions = {}} = rOpts;
+      opts = {
+        ...opts,
+        ...(sttCredentials.access_token && { VERBIO_ACCESS_TOKEN: sttCredentials.access_token}),
+        ...(sttCredentials.engine_version && {VERBIO_ENGINE_VERSION: sttCredentials.engine_version}),
+        ...(language && {VERBIO_LANGUAGE: language}),
+        ...(verbioOptions.enable_formatting && {VERBIO_ENABLE_FORMATTING: verbioOptions.enable_formatting}),
+        ...(verbioOptions.enable_diarization && {VERBIO_ENABLE_DIARIZATION: verbioOptions.enable_diarization}),
+        ...(verbioOptions.topic && {VERBIO_TOPIC: verbioOptions.topic}),
+        ...(verbioOptions.inline_grammar && {VERBIO_INLINE_GRAMMAR: verbioOptions.inline_grammar}),
+        ...(verbioOptions.grammar_uri && {VERBIO_GRAMMAR_URI: verbioOptions.grammar_uri}),
+        ...(verbioOptions.label && {VERBIO_LABEL: verbioOptions.label}),
+        ...(verbioOptions.recognition_timeout && {VERBIO_RECOGNITION_TIMEOUT: verbioOptions.recognition_timeout}),
+        ...(verbioOptions.speech_complete_timeout &&
+          {VERBIO_SPEECH_COMPLETE_TIMEOUT: verbioOptions.speech_complete_timeout}),
+        ...(verbioOptions.speech_incomplete_timeout &&
+          {VERBIO_SPEECH_INCOMPLETE_TIMEOUT: verbioOptions.speech_incomplete_timeout}),
+      };
+    } else if (vendor.startsWith('custom:')) {
       let {options = {}} = rOpts;
       const {auth_token, custom_stt_url} = sttCredentials;
       options = {
@@ -740,7 +829,7 @@ module.exports = (logger) => {
 
       opts = {
         ...opts,
-        JAMBONZ_STT_API_KEY: auth_token,
+        ...(auth_token && {JAMBONZ_STT_API_KEY: auth_token}),
         JAMBONZ_STT_URL: custom_stt_url,
         ...(Object.keys(options).length > 0 && {JAMBONZ_STT_OPTIONS: JSON.stringify(options)}),
       };
@@ -752,48 +841,6 @@ module.exports = (logger) => {
     return opts;
   };
 
-  const removeSpeechListeners = (ep) => {
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance);
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(AwsTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(AwsTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(AzureTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected);
-    ep.removeCustomEventListener(AzureTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.TranscriptionComplete);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.StartOfSpeech);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(DeepgramTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(DeepgramTranscriptionEvents.Connect);
-    ep.removeCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure);
-
-    ep.removeCustomEventListener(SonioxTranscriptionEvents.Transcription);
-
-    ep.removeCustomEventListener(CobaltTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(CobaltTranscriptionEvents.CompileContext);
-
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.TranscriptionComplete);
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.StartOfSpeech);
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.Connect);
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.ConnectFailure);
-
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.Error);
-
-    ep.removeCustomEventListener(AssemblyAiTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(AssemblyAiTranscriptionEvents.Connect);
-    ep.removeCustomEventListener(AssemblyAiTranscriptionEvents.ConnectFailure);
-  };
-
   const setSpeechCredentialsAtRuntime = (recognizer) => {
     if (!recognizer) return;
     if (recognizer.vendor === 'nuance') {
@@ -832,7 +879,6 @@ module.exports = (logger) => {
   return {
     normalizeTranscription,
     setChannelVarsForStt,
-    removeSpeechListeners,
     setSpeechCredentialsAtRuntime,
     compileSonioxTranscripts,
     consolidateTranscripts

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 {
@@ -55,6 +56,12 @@ class WsRequestor extends BaseRequestor {
     }
 
     if (type === 'session:new') this.call_sid = params.callSid;
+    if (type === 'session:reconnect') {
+      this._reconnectPromise = new Promise((resolve, reject) => {
+        this._reconnectResolve = resolve;
+        this._reconnectReject = reject;
+      });
+    }
 
     /* if we have an absolute url, and it is http then do a standard webhook */
     if (this._isAbsoluteUrl(url) && url.startsWith('http')) {
@@ -70,20 +77,23 @@ class WsRequestor extends BaseRequestor {
     }
 
     /* connect if necessary */
+    const queueMsg = () => {
+      this.logger.debug(
+        `WsRequestor:request(${this.id}) - queueing ${type} message since we are connecting`);
+      if (wantsAck) {
+        const p = new Promise((resolve, reject) => {
+          this.queuedMsg.push({type, hook, params, httpHeaders, promise: {resolve, reject}});
+        });
+        return p;
+      }
+      else {
+        this.queuedMsg.push({type, hook, params, httpHeaders});
+      }
+      return;
+    };
     if (!this.ws) {
       if (this.connectInProgress) {
-        this.logger.debug(
+        return queueMsg();
-          `WsRequestor:request(${this.id}) - queueing ${type} message since we are connecting`);
-        if (wantsAck) {
-          const p = new Promise((resolve, reject) => {
-            this.queuedMsg.push({type, hook, params, httpHeaders, promise: {resolve, reject}});
-          });
-          return p;
-        }
-        else {
-          this.queuedMsg.push({type, hook, params, httpHeaders});
-        }
-        return;
       }
       this.connectInProgress = true;
       this.logger.debug(`WsRequestor:request(${this.id}) - connecting since we do not have a connection for ${type}`);
@@ -101,6 +111,10 @@ class WsRequestor extends BaseRequestor {
         return Promise.reject(err);
       }
     }
+    // If jambonz wait for ack from reconnect, queue the msg until reconnect is acked
+    if (type !== 'session:reconnect' && this._reconnectPromise) {
+      return queueMsg();
+    }
     assert(this.ws);
 
     /* prepare and send message */
@@ -118,7 +132,7 @@ class WsRequestor extends BaseRequestor {
       type,
       msgid,
       call_sid: this.call_sid,
-      hook: type === 'verb:hook' ? url : undefined,
+      hook: ['verb:hook', 'session:redirect'].includes(type) ? url : undefined,
       data: {...payload},
       ...b3
     };
@@ -138,6 +152,18 @@ class WsRequestor extends BaseRequestor {
       }
     };
 
+    const rejectQueuedMsgs = (err) => {
+      if (this.queuedMsg.length > 0) {
+        for (const {promise} of this.queuedMsg) {
+          this.logger.debug(`WsRequestor:request - preparing queued ${type} for rejectQueuedMsgs`);
+          if (promise) {
+            promise.reject(err);
+          }
+        }
+        this.queuedMsg.length = 0;
+      }
+    };
+
     //this.logger.debug({obj}, `websocket: sending (${url})`);
 
     /* special case: reconnecting before we received ack to session:new */
@@ -178,16 +204,37 @@ class WsRequestor extends BaseRequestor {
           this.logger.debug({response}, `WsRequestor:request ${url} succeeded in ${rtt}ms`);
           this.stats.histogram('app.hook.ws_response_time', rtt, ['hook_type:app']);
           resolve(response);
+          if (this._reconnectResolve) {
+            this._reconnectResolve();
+          }
         },
         failure: (err) => {
+          if (this._reconnectReject) {
+            this._reconnectReject(err);
+          }
           clearTimeout(timer);
           reject(err);
         }
       });
 
       /* send the message */
-      this.ws.send(JSON.stringify(obj), () => {
+      this.ws.send(JSON.stringify(obj), async() => {
         this.logger.debug({obj}, `WsRequestor:request websocket: sent (${url})`);
+        // If session:reconnect is waiting for ack, hold here until ack to send queuedMsgs
+        if (this._reconnectPromise) {
+          try {
+            await this._reconnectPromise;
+          } catch (err) {
+            // bad thing happened to session:recconnect
+            rejectQueuedMsgs(err);
+            this.emit('reconnect-error');
+            return;
+          } finally {
+            this._reconnectPromise = null;
+            this._reconnectResolve = null;
+            this._reconnectReject = null;
+          }
+        }
         sendQueuedMsgs();
       });
     });
@@ -228,6 +275,9 @@ class WsRequestor extends BaseRequestor {
         maxRedirects: 2,
         handshakeTimeout,
         maxPayload: JAMBONES_WS_MAX_PAYLOAD ? parseInt(JAMBONES_WS_MAX_PAYLOAD) : 24 * 1024,
+        headers: {
+          ...(HTTP_USER_AGENT_HEADER && {'user-agent' : HTTP_USER_AGENT_HEADER})
+        }
       };
       if (this.username && this.password) opts = {...opts, auth: `${this.username}:${this.password}`};
 
@@ -322,7 +372,9 @@ class WsRequestor extends BaseRequestor {
           'WsRequestor:_onSocketClosed time to reconnect');
         if (!this.ws && !this.connectInProgress) {
           this.connectInProgress = true;
-          this._connect().catch((err) => this.connectInProgress = false);
+          return this._connect()
+            .catch((err) => this.logger.error('WsRequestor:_onSocketClosed There is error while reconnect', err))
+            .finally(() => this.connectInProgress = false);
         }
       }, this.backoffMs);
       this.backoffMs = this.backoffMs < 2000 ? this.backoffMs * 2 : (this.backoffMs + 2000);
@@ -340,7 +392,9 @@ class WsRequestor extends BaseRequestor {
     /* messages must be JSON format */
     try {
       const obj = JSON.parse(content);
-      const {type, msgid, command, call_sid = this.call_sid, queueCommand = false, data} = obj;
+      //const {type, msgid, command, call_sid = this.call_sid, queueCommand = false, data} = obj;
+      const {type, msgid, command, queueCommand = false, data} = obj;
+      const call_sid = obj.callSid || this.call_sid;
 
       //this.logger.debug({obj}, 'WsRequestor:request websocket: received');
       assert.ok(type, 'type property not supplied');

package.json

@@ -1,9 +1,9 @@
 {
   "name": "jambonz-feature-server",
-  "version": "0.8.5",
+  "version": "0.9.0",
   "main": "app.js",
   "engines": {
-    "node": ">= 10.16.0"
+    "node": ">= 18.x"
   },
   "keywords": [
     "sip",
@@ -25,57 +25,57 @@
     "jslint:fix": "eslint app.js tracer.js lib --fix"
   },
   "dependencies": {
-    "@aws-sdk/client-auto-scaling": "^3.360.0",
+    "@aws-sdk/client-auto-scaling": "^3.549.0",
-    "@aws-sdk/client-sns": "^3.360.0",
+    "@aws-sdk/client-sns": "^3.549.0",
-    "@jambonz/db-helpers": "^0.9.1",
+    "@jambonz/db-helpers": "^0.9.3",
     "@jambonz/http-health-check": "^0.0.1",
-    "@jambonz/mw-registrar": "^0.2.4",
+    "@jambonz/mw-registrar": "^0.2.7",
-    "@jambonz/realtimedb-helpers": "^0.8.7",
+    "@jambonz/realtimedb-helpers": "^0.8.8",
-    "@jambonz/speech-utils": "^0.0.26",
+    "@jambonz/speech-utils": "^0.1.3",
-    "@jambonz/stats-collector": "^0.1.9",
+    "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.46",
+    "@jambonz/verb-specifications": "^0.0.72",
-    "@opentelemetry/api": "^1.4.0",
+    "@opentelemetry/api": "^1.8.0",
-    "@opentelemetry/exporter-jaeger": "^1.9.0",
+    "@opentelemetry/exporter-jaeger": "^1.23.0",
-    "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
+    "@opentelemetry/exporter-trace-otlp-http": "^0.50.0",
-    "@opentelemetry/exporter-zipkin": "^1.9.0",
+    "@opentelemetry/exporter-zipkin": "^1.23.0",
-    "@opentelemetry/instrumentation": "^0.35.0",
+    "@opentelemetry/instrumentation": "^0.50.0",
-    "@opentelemetry/resources": "^1.9.0",
+    "@opentelemetry/resources": "^1.23.0",
-    "@opentelemetry/sdk-trace-base": "^1.9.0",
+    "@opentelemetry/sdk-trace-base": "^1.23.0",
-    "@opentelemetry/sdk-trace-node": "^1.9.0",
+    "@opentelemetry/sdk-trace-node": "^1.23.0",
-    "@opentelemetry/semantic-conventions": "^1.9.0",
+    "@opentelemetry/semantic-conventions": "^1.23.0",
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.27",
+    "drachtio-fsmrf": "^3.0.43",
-    "drachtio-srf": "^4.5.31",
+    "drachtio-srf": "^4.5.35",
-    "express": "^4.18.2",
+    "express": "^4.19.2",
     "express-validator": "^7.0.1",
-    "ip": "^1.1.8",
+    "ip": "^2.0.1",
-    "moment": "^2.29.4",
+    "moment": "^2.30.1",
-    "parse-url": "^8.1.0",
+    "parse-url": "^9.2.0",
-    "pino": "^8.8.0",
+    "pino": "^8.20.0",
     "polly-ssml-split": "^0.1.0",
     "proxyquire": "^2.1.3",
-    "sdp-transform": "^2.14.1",
+    "sdp-transform": "^2.14.2",
-    "short-uuid": "^4.2.2",
+    "short-uuid": "^5.1.0",
-    "sinon": "^15.0.1",
+    "sinon": "^17.0.1",
     "to-snake-case": "^1.0.0",
-    "undici": "^5.26.2",
+    "undici": "^6.15.0",
     "uuid-random": "^1.3.2",
     "verify-aws-sns-signature": "^0.1.0",
-    "ws": "^8.9.0",
+    "ws": "^8.17.0",
     "xml2js": "^0.6.2"
   },
   "devDependencies": {
     "clear-module": "^4.1.2",
-    "eslint": "^7.32.0",
+    "eslint": "7.32.0",
-    "eslint-plugin-promise": "^4.3.1",
+    "eslint-plugin-promise": "^6.1.1",
     "nyc": "^15.1.0",
-    "tape": "^5.6.1"
+    "tape": "^5.7.5"
   },
   "optionalDependencies": {
-    "bufferutil": "^4.0.6",
+    "bufferutil": "^4.0.8",
-    "utf-8-validate": "^5.0.8"
+    "utf-8-validate": "^6.0.3"
   }
 }

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}`);

test/docker-compose-testbed.yaml

@@ -42,7 +42,7 @@ services:
         ipv4_address: 172.38.0.7
 
   drachtio:
-    image: drachtio/drachtio-server:0.8.22
+    image: drachtio/drachtio-server:0.8.25-rc8
     restart: always
     command: drachtio --contact "sip:*;transport=udp" --mtu 4096 --address 0.0.0.0 --port 9022
     ports:
@@ -57,7 +57,7 @@ services:
         condition: service_healthy
 
   freeswitch:
-    image: drachtio/drachtio-freeswitch-mrf:0.4.33
+    image: drachtio/drachtio-freeswitch-mrf:0.7.3
     restart: always
     command: freeswitch --rtp-range-start 20000 --rtp-range-end 20100
     environment:

test/hangup-test.js

@@ -0,0 +1,53 @@
+const test = require('tape');
+const { sippUac } = require('./sipp')('test_fs');
+const clearModule = require('clear-module');
+const {provisionCallHook, provisionCustomHook} = require('./utils')
+const bent = require('bent');
+const getJSON = bent('json')
+
+process.on('unhandledRejection', (reason, p) => {
+  console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
+});
+
+function connect(connectable) {
+  return new Promise((resolve, reject) => {
+    connectable.on('connect', () => {
+      return resolve();
+    });
+  });
+}
+
+test('\'hangup\' custom headers', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const verbs = [
+      {
+        verb: 'play',
+        url: 'https://example.com/example.mp3'
+      },
+      {
+        "verb": "hangup",
+        "headers": {
+          "X-Reason" : "maximum call duration exceeded"
+        }
+      }
+    ];
+
+    const from = 'hangup_custom_headers';
+    await provisionCallHook(from, verbs)
+
+    // THEN
+    await sippUac('uac-success-received-bye.xml', '172.38.0.10', from);
+    t.pass('play: succeeds when using single link');
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});

test/index.js

@@ -16,6 +16,8 @@ require('./listen-tests');
 require('./config-test');
 require('./queue-test');
 require('./in-dialog-test');
+require('./hangup-test');
+require('./sdp-utils-test');
 require('./http-proxy-test');
 require('./remove-test-db');
 require('./docker_stop');

test/play-tests.js

@@ -188,7 +188,7 @@ test('\'play\' tests with seekOffset and actionHook', async(t) => {
     const seconds = parseInt(obj.body.playback_seconds);
     const milliseconds = parseInt(obj.body.playback_milliseconds);
     const lastOffsetPos = parseInt(obj.body.playback_last_offset_pos);
-    //console.log({obj}, 'lastRequest');
+    console.log({obj}, 'lastRequest');
     t.ok(obj.body.reason === "playCompleted", "play: actionHook success received");
     t.ok(seconds === 2, "playback_seconds: actionHook success received");
     t.ok(milliseconds === 2048, "playback_milliseconds: actionHook success received");

test/sdp-utils-test.js

@@ -0,0 +1,26 @@
+const test = require('tape');
+const {makeOpusFirst, isOpusFirst} = require('../lib/utils/sdp-utils');
+const sdpTransform = require('sdp-transform');
+
+test('test opus first', (t) => {
+    const sdp = 'v=0\r\no=- 3348584794228993675 2 IN IP4 127.0.0.1\r\ns=-\r\nt=0 0\r\na=group:BUNDLE 0\r\na=extmap-allow-mixed\r\na=msid-semantic: WMS caca8b77-5ae5-4e73-a4d5-de1fce930335\r\nm=audio 57088 UDP/TLS/RTP/SAVPF 111 63 9 0 8 13 110 126\r\nc=IN IP4 14.238.89.50\r\na=rtcp:9 IN IP4 0.0.0.0\r\na=candidate:1401281302 1 udp 2122260223 10.231.36.146 57088 typ host generation 0 network-id 1 network-cost 10\r\na=candidate:2173263513 1 udp 1686052607 14.238.89.50 57088 typ srflx raddr 10.231.36.146 rport 57088 generation 0 network-id 1 network-cost 10\r\na=ice-ufrag:k5nc\r\na=ice-pwd:J0qwMs6HrIcFNZbDG5m8Kqpk\r\na=ice-options:trickle\r\na=fingerprint:sha-256 66:DE:9A:76:CE:11:2D:65:C4:08:C7:87:B4:90:7E:F1:8D:07:B9:F4:FF:E3:81:D7:E7:7D:C6:56:47:01:6E:55\r\na=setup:actpass\r\na=mid:0\r\na=extmap:1 urn:ietf:params:rtp-hdrext:ssrc-audio-level\r\na=extmap:2 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time\r\na=extmap:3 http://www.ietf.org/id/draft-holmer-rmcat-transport-wide-cc-extensions-01\r\na=extmap:4 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=sendrecv\r\na=msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\na=rtcp-mux\r\na=rtcp-fb:111 transport-cc\r\na=fmtp:111 minptime=10;useinbandfec=1\r\na=rtpmap:63 red/48000/2\r\na=fmtp:63 111/111\r\na=rtpmap:9 G722/8000\r\na=rtpmap:0 PCMU/8000\r\na=rtpmap:8 PCMA/8000\r\na=rtpmap:13 CN/8000\r\na=rtpmap:111 opus/48000/2\r\na=rtpmap:110 telephone-event/48000\r\na=rtpmap:126 telephone-event/8000\r\na=ssrc:3207459321 cname:4nyPJ6KXvseBUIhu\r\na=ssrc:3207459321 msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\n';
+    const opusSdp = makeOpusFirst(sdp);
+    const parsedSdp = sdpTransform.parse(opusSdp);
+    const opusIndex = parsedSdp.media[0].rtp.findIndex((entry) => entry.codec === 'opus');
+    t.ok(opusIndex === 0, 'succesffuly move opus to be first offer')
+    t.end();
+});
+
+
+test('test is opus first', (t) => {
+
+    const sdp = 'v=0\r\no=- 3348584794228993675 2 IN IP4 127.0.0.1\r\ns=-\r\nt=0 0\r\na=group:BUNDLE 0\r\na=extmap-allow-mixed\r\na=msid-semantic: WMS caca8b77-5ae5-4e73-a4d5-de1fce930335\r\nm=audio 57088 UDP/TLS/RTP/SAVPF 111 63 9 0 8 13 110 126\r\nc=IN IP4 14.238.89.50\r\na=rtcp:9 IN IP4 0.0.0.0\r\na=candidate:1401281302 1 udp 2122260223 10.231.36.146 57088 typ host generation 0 network-id 1 network-cost 10\r\na=candidate:2173263513 1 udp 1686052607 14.238.89.50 57088 typ srflx raddr 10.231.36.146 rport 57088 generation 0 network-id 1 network-cost 10\r\na=ice-ufrag:k5nc\r\na=ice-pwd:J0qwMs6HrIcFNZbDG5m8Kqpk\r\na=ice-options:trickle\r\na=fingerprint:sha-256 66:DE:9A:76:CE:11:2D:65:C4:08:C7:87:B4:90:7E:F1:8D:07:B9:F4:FF:E3:81:D7:E7:7D:C6:56:47:01:6E:55\r\na=setup:actpass\r\na=mid:0\r\na=extmap:1 urn:ietf:params:rtp-hdrext:ssrc-audio-level\r\na=extmap:2 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time\r\na=extmap:3 http://www.ietf.org/id/draft-holmer-rmcat-transport-wide-cc-extensions-01\r\na=extmap:4 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=sendrecv\r\na=msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\na=rtcp-mux\r\na=rtpmap:111 opus/48000/2\r\na=rtcp-fb:111 transport-cc\r\na=fmtp:111 minptime=10;useinbandfec=1\r\na=rtpmap:63 red/48000/2\r\na=fmtp:63 111/111\r\na=rtpmap:9 G722/8000\r\na=rtpmap:0 PCMU/8000\r\na=rtpmap:8 PCMA/8000\r\na=rtpmap:13 CN/8000\r\na=rtpmap:110 telephone-event/48000\r\na=rtpmap:126 telephone-event/8000\r\na=ssrc:3207459321 cname:4nyPJ6KXvseBUIhu\r\na=ssrc:3207459321 msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\n';
+    t.ok(isOpusFirst(sdp), "opus is first offer");
+
+    const sdp2 = 'v=0\r\no=- 3348584794228993675 2 IN IP4 127.0.0.1\r\ns=-\r\nt=0 0\r\na=group:BUNDLE 0\r\na=extmap-allow-mixed\r\na=msid-semantic: WMS caca8b77-5ae5-4e73-a4d5-de1fce930335\r\nm=audio 57088 UDP/TLS/RTP/SAVPF 111 63 9 0 8 13 110 126\r\nc=IN IP4 14.238.89.50\r\na=rtcp:9 IN IP4 0.0.0.0\r\na=candidate:1401281302 1 udp 2122260223 10.231.36.146 57088 typ host generation 0 network-id 1 network-cost 10\r\na=candidate:2173263513 1 udp 1686052607 14.238.89.50 57088 typ srflx raddr 10.231.36.146 rport 57088 generation 0 network-id 1 network-cost 10\r\na=ice-ufrag:k5nc\r\na=ice-pwd:J0qwMs6HrIcFNZbDG5m8Kqpk\r\na=ice-options:trickle\r\na=fingerprint:sha-256 66:DE:9A:76:CE:11:2D:65:C4:08:C7:87:B4:90:7E:F1:8D:07:B9:F4:FF:E3:81:D7:E7:7D:C6:56:47:01:6E:55\r\na=setup:actpass\r\na=mid:0\r\na=extmap:1 urn:ietf:params:rtp-hdrext:ssrc-audio-level\r\na=extmap:2 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time\r\na=extmap:3 http://www.ietf.org/id/draft-holmer-rmcat-transport-wide-cc-extensions-01\r\na=extmap:4 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=sendrecv\r\na=msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\na=rtcp-mux\r\na=rtcp-fb:111 transport-cc\r\na=fmtp:111 minptime=10;useinbandfec=1\r\na=rtpmap:63 red/48000/2\r\na=fmtp:63 111/111\r\na=rtpmap:9 G722/8000\r\na=rtpmap:0 PCMU/8000\r\na=rtpmap:8 PCMA/8000\r\na=rtpmap:13 CN/8000\r\na=rtpmap:111 opus/48000/2\r\na=rtpmap:110 telephone-event/48000\r\na=rtpmap:126 telephone-event/8000\r\na=ssrc:3207459321 cname:4nyPJ6KXvseBUIhu\r\na=ssrc:3207459321 msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\n';
+    t.ok(!isOpusFirst(sdp2), "opus is not first offer")
+
+    const sdp3 = 'v=0\r\no=xhoaluu2 1314 1504 IN IP4 192.168.1.4\r\ns=Talk\r\nc=IN IP4 192.168.1.4\r\nt=0 0\r\na=ice-pwd:397d063ea23fdc05164e3ee4\r\na=ice-ufrag:16c449a3\r\na=rtcp-xr:rcvr-rtt=all:10000 stat-summary=loss,dup,jitt,TTL voip-metrics\r\na=group:BUNDLE as\r\na=record:off\r\nm=audio 56542 RTP/AVPF 0 8\r\nc=IN IP4 14.226.233.151\r\na=rtcp-mux\r\na=mid:as\r\na=extmap:1 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=rtcp:63076 IN IP4 192.168.1.4\r\na=candidate:1 1 UDP 2130706303 192.168.1.4 56542 typ host\r\na=candidate:1 2 UDP 2130706302 192.168.1.4 63076 typ host\r\na=candidate:2 1 UDP 2130706431 2001:ee0:d744:dcf0:c1d3:d73f:7a93:dc9f 56542 typ host\r\na=candidate:2 2 UDP 2130706430 2001:ee0:d744:dcf0:c1d3:d73f:7a93:dc9f 63076 typ host\r\na=candidate:3 1 UDP 2130706431 2001:ee0:d744:dcf0:15:6be3:8e6b:b736 56542 typ host\r\na=candidate:3 2 UDP 2130706430 2001:ee0:d744:dcf0:15:6be3:8e6b:b736 63076 typ host\r\na=candidate:4 1 UDP 1694498687 14.226.233.151 56542 typ srflx raddr 192.168.1.4 rport 56542\r\na=rtcp-fb:* trr-int 1000\r\na=rtcp-fb:* ccm tmmbr';
+    t.ok(!isOpusFirst(sdp2), "opus is not first offer")
+    t.end();
+});

test/transcribe-tests.js

@@ -52,6 +52,7 @@ test('\'transcribe\' test - google', 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));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using google credentials');
 
@@ -89,6 +90,7 @@ test('\'transcribe\' test - microsoft', 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));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using  microsoft credentials');
 
@@ -126,6 +128,7 @@ test('\'transcribe\' test - aws', 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));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using aws credentials');
 
@@ -137,6 +140,71 @@ test('\'transcribe\' test - aws', async(t) => {
   }
 });
 
+
+test('\'transcribe\' test - deepgram config options', async(t) => {
+  if (!DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "config",
+        "recognizer": {
+          "vendor": "deepgram",
+          "language": "en-US",
+          "altLanguages": [
+            "en-US"
+          ],
+          "deepgramOptions": {
+            "model": "2-ea",
+            "tier": "nova",
+            "numerals": true,
+            "ner": true,
+            "vadTurnoff": 10,
+            "keywords": [
+              "CPT"
+            ]
+          }
+        }
+      },
+      {
+        "verb": "transcribe",
+        "transcriptionHook": "/transcriptionHook",
+        "recognizer": {
+          "vendor": "deepgram",
+          "altLanguages": [
+            "en-AU"
+          ],
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "deepgramOptions": {
+            "apiKey": DEEPGRAM_API_KEY,
+          }
+        }
+      }
+    ];
+    let from = "gather_success";
+    await 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().includes('like to speak to customer support'),
+      'transcribe: succeeds when using deepgram credentials');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
 test('\'transcribe\' test - deepgram', async(t) => {
   if (!DEEPGRAM_API_KEY ) {
     t.pass('skipping deepgram tests');
@@ -166,6 +234,7 @@ test('\'transcribe\' test - deepgram', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
       'transcribe: succeeds when using deepgram credentials');
 
@@ -245,9 +314,131 @@ test('\'transcribe\' test - google with asrTimeout', 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));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using google credentials');
 
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+test('\'transcribe\' test - deepgram config options altLanguages', async(t) => {
+  if (!DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "config",
+        "recognizer": {
+          "vendor": "deepgram",
+          "language": "en-US",
+          "altLanguages": [
+            "en-US"
+          ],
+          "deepgramOptions": {
+            "model": "nova-2",
+            "numerals": true,
+            "ner": true,
+            "vadTurnoff": 10,
+            "keywords": [
+              "CPT"
+            ]
+          }
+        }
+      },
+      {
+        "verb": "transcribe",
+        "transcriptionHook": "/transcriptionHook",
+        "recognizer": {
+          "vendor": "deepgram",
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "deepgramOptions": {
+            "apiKey": DEEPGRAM_API_KEY,
+          }
+        }
+      }
+    ];
+    let from = "gather_success_no_altLanguages";
+    await provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
+      'transcribe: succeeds when using deepgram credentials');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+test('\'transcribe\' test - deepgram config options altLanguages', async(t) => {
+  if (!DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "config",
+        "recognizer": {
+          "vendor": "deepgram",
+          "language": "en-US",
+          "altLanguages": [
+            "en-US"
+          ],
+          "deepgramOptions": {
+            "model": "nova-2",
+            "numerals": true,
+            "ner": true,
+            "vadTurnoff": 10,
+            "keywords": [
+              "CPT"
+            ]
+          }
+        }
+      },
+      {
+        "verb": "transcribe",
+        "transcriptionHook": "/transcriptionHook",
+        "recognizer": {
+          "vendor": "deepgram",
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "altLanguages": [],
+          "deepgramOptions": {
+            "apiKey": DEEPGRAM_API_KEY,
+          }
+        }
+      }
+    ];
+    let from = "gather_success_has_altLanguages";
+    await provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
+      'transcribe: succeeds when using deepgram credentials');
+
     disconnect();
   } catch (err) {
     console.log(`error received: ${err}`);