fix: do not run snake case for customer data

README.md

@@ -21,7 +21,6 @@ Configuration is provided via environment variables:
 |ENCRYPTION_SECRET| secret for credential encryption(JWT_SECRET is deprecated) |yes|
 |GOOGLE_APPLICATION_CREDENTIALS| path to gcp service key file|yes|
 |HTTP_PORT| tcp port to listen on for API requests from jambonz-api-server|yes|
-|HTTP_IP| IP Address for API requests from jambonz-api-server |no|
 |JAMBONES_GATHER_EARLY_HINTS_MATCH| if true and hints are provided, gather will opportunistically review interim transcripts if possible to reduce ASR latency |no|
 |JAMBONES_FREESWITCH| IP:port:secret for Freeswitch server (e.g. '127.0.0.1:8021:JambonzR0ck$'|yes|
 |JAMBONES_LOGLEVEL| log level for application, 'info' or 'debug'|no|

lib/config.js

@@ -73,7 +73,6 @@ const JAMBONES_LOGLEVEL = process.env.JAMBONES_LOGLEVEL || 'info';
 const JAMBONES_INJECT_CONTENT = process.env.JAMBONES_INJECT_CONTENT;
 
 const PORT = parseInt(process.env.HTTP_PORT, 10) || 3000;
-const HTTP_IP = process.env.HTTP_IP;
 const HTTP_PORT_MAX = parseInt(process.env.HTTP_PORT_MAX, 10);
 
 const K8S = process.env.K8S;
@@ -171,7 +170,6 @@ module.exports = {
   JAMBONES_CLUSTER_ID,
   PORT,
   HTTP_PORT_MAX,
-  HTTP_IP,
   K8S,
   K8S_SBC_SIP_SERVICE_NAME,
   JAMBONES_SUBNET,

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

@@ -97,8 +97,7 @@ router.post('/',
         'X-Trace-ID': rootSpan.traceId,
         ...(req.body?.application_sid && {'X-Application-Sid': req.body.application_sid}),
         ...(restDial.fromHost && {'X-Preferred-From-Host': restDial.fromHost}),
-        ...(record_all_calls && {'X-Record-All-Calls': recordOutputFormat}),
-        ...target.headers
+        ...(record_all_calls && {'X-Record-All-Calls': recordOutputFormat})
       };
 
       switch (target.type) {
@@ -219,7 +218,7 @@ router.post('/',
       }
       if (!app.notifier && app.call_status_hook) {
         app.notifier = new HttpRequestor(logger, account.account_sid, app.call_status_hook, account.webhook_secret);
-        logger.debug({call_status_hook: app.call_status_hook}, 'creating http client for call status hook');
+        logger.debug({call_hook: app.call_hook}, 'creating http client for call status hook');
       }
       else if (!app.notifier) {
         logger.debug('creating null call status hook');

lib/middleware.js

@@ -354,13 +354,11 @@ module.exports = function(srf, logger) {
       });
       // if transferred call contains callInfo, let update original data to newly created callInfo in this instance.
       if (app.transferredCall && app.callInfo) {
-        const {direction, callerName, from, to, originatingSipIp, originatingSipTrunkName} = app.callInfo;
-        req.locals.callInfo.direction = direction;
-        req.locals.callInfo.callerName = callerName;
-        req.locals.callInfo.from = from;
-        req.locals.callInfo.to = to;
-        req.locals.callInfo.originatingSipIp = originatingSipIp;
-        req.locals.callInfo.originatingSipTrunkName = originatingSipTrunkName;
+        req.locals.callInfo.callerName = app.callInfo.callerName;
+        req.locals.callInfo.from = app.callInfo.from;
+        req.locals.callInfo.to = app.callInfo.to;
+        req.locals.callInfo.originatingSipIp = app.callInfo.originatingSipIp;
+        req.locals.callInfo.originatingSipTrunkName = app.callInfo.originatingSipTrunkName;
         delete app.callInfo;
       }
       next();

lib/session/call-session.js

@@ -8,8 +8,7 @@ const {
   KillReason,
   RecordState,
   AllowedSipRecVerbs,
-  AllowedConfirmSessionVerbs,
-  TranscribeStatus
+  AllowedConfirmSessionVerbs
 } = require('../utils/constants');
 const moment = require('moment');
 const assert = require('assert');
@@ -21,7 +20,6 @@ const listTaskNames = require('../utils/summarize-tasks');
 const HttpRequestor = require('../utils/http-requestor');
 const WsRequestor = require('../utils/ws-requestor');
 const ActionHookDelayProcessor = require('../utils/action-hook-delay');
-const {parseUri} = require('drachtio-srf');
 const {
   JAMBONES_INJECT_CONTENT,
   JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
@@ -30,7 +28,6 @@ const {
 } = require('../config');
 const bent = require('bent');
 const BackgroundTaskManager = require('../utils/background-task-manager');
-const { isOnhold } = require('../utils/sdp-utils');
 const BADPRECONDITIONS = 'preconditions not met';
 const CALLER_CANCELLED_ERR_MSG = 'Response not sent due to unknown transaction';
 
@@ -342,18 +339,6 @@ class CallSession extends Emitter {
     this.application.fallback_speech_recognizer_language = language;
   }
 
-  /**
-   * global referHook
-   */
-
-  set referHook(hook) {
-    this._referHook = hook;
-  }
-
-  get referHook() {
-    return this._referHook;
-  }
-
   /**
    * Vad
    */
@@ -535,6 +520,15 @@ class CallSession extends Emitter {
     this._actionHookDelayRetries = e;
   }
 
+  // Getter/setter for current tts vendor
+  get currentTtsVendor() {
+    return this._currentTtsVendor;
+  }
+
+  set currentTtsVendor(vendor) {
+    this._currentTtsVendor = vendor;
+  }
+
   get actionHookDelayProcessor() {
     return this._actionHookDelayProcessor;
   }
@@ -559,17 +553,6 @@ class CallSession extends Emitter {
             this.wakeupResolver = null;
           }
         });
-        this._actionHookDelayProcessor.on('giveupWithTasks', (tasks) => {
-          this.logger.info('CallSession: ActionHookDelayProcessor: giveupWithTasks event');
-          const giveUpTasks = normalizeJambones(this.logger, tasks).map((tdata) => makeTask(this.logger, tdata));
-          this.logger.info({tasks: listTaskNames(giveUpTasks)}, 'CallSession:giveupWithTasks task list');
-
-          // we need to clear the ahd, as we do not want to execute actionHookDelay actions again
-          this.clearActionHookDelayProcessor();
-
-          // replace the application with giveUpTasks
-          this.replaceApplication(giveUpTasks);
-        });
       } catch (err) {
         this.logger.error({err}, 'CallSession: Error creating ActionHookDelayProcessor');
       }
@@ -583,7 +566,9 @@ class CallSession extends Emitter {
         //this.logger.debug('CallSession:clearOrRestoreActionHookDelayProcessor - ahd settings');
         //await this.clearActionHookDelayProcessor();
       }
-      this.logger.debug('CallSession:clearOrRestoreActionHookDelayProcessor - say or play action completed');
+      else {
+        this.logger.debug('CallSession:clearOrRestoreActionHookDelayProcessor - restore ahd after gather override');
+      }
     }
   }
 
@@ -1084,7 +1069,7 @@ class CallSession extends Emitter {
         try {
           await this._awaitCommandsOrHangup();
 
-          //await this.clearOrRestoreActionHookDelayProcessor();
+          await this.clearOrRestoreActionHookDelayProcessor();
 
           //TODO: remove filler noise code and simply create as action hook delay
           if (this._isPlayingFillerNoise) {
@@ -1338,39 +1323,6 @@ class CallSession extends Emitter {
     listenTask.updateListen(opts.listen_status);
   }
 
-  // Jambonz only update transcribe task when it's received RE-INVITE:
-  // 1. Pause transcribe task if the task is running and hold RE-INVITE is received
-  // 2. Resume transcribe task if the task was paused by jambonz due to hold and unhold RE-INVITE is received
-  async pauseOrResumeTranscribeTaskByReInvite(req) {
-    const task = this.currentTask;
-    const transcribeTasks = [];
-    // Gather running transcribe tasks
-    if (this.backgroundTaskManager.isTaskRunning('transcribe')) {
-      transcribeTasks.push(this.backgroundTaskManager.getTask('transcribe'));
-    }
-
-    // Include current task if it is transcribe-related
-    if (task && [TaskName.Dial, TaskName.Transcribe].includes(task.name)) {
-      const transcribeTask = task.name === TaskName.Transcribe ? task : task.transcribeTask;
-      if (transcribeTask) transcribeTasks.push(transcribeTask);
-    }
-
-    const isHold = isOnhold(req.body);
-    const transcribe_status = isHold ? TranscribeStatus.Pause : TranscribeStatus.Resume;
-    this.logger.debug(`updating transcribe_status for ${transcribeTasks.length} transcribe tasks.`);
-    transcribeTasks.forEach((t) => {
-      if (isHold && !t.paused) {
-        t.updateTranscribe(transcribe_status);
-        t.pausedByHold = true;
-      } else if (!isHold && t.pausedByHold) {
-        if (t.paused) {
-          t.updateTranscribe(transcribe_status);
-        }
-        t.pausedByHold = false;
-      }
-    });
-  }
-
   /**
    * perform live call control -- change Transcribe status
    * @param {object} opts
@@ -2122,9 +2074,6 @@ Duration=${duration} `
     // When this call kicked out from conference, session need to replace endpoint
     // but this.ms might be undefined/null at this case.
     this.ms = this.ms || this.getMS();
-    // Destroy previous ep if it's still running.
-    if (this.ep?.connected) this.ep.destroy();
-
     this.ep = await this.ms.createEndpoint({remoteSdp: this.dlg.remote.sdp});
     this._configMsEndpoint();
 
@@ -2276,67 +2225,23 @@ Duration=${duration} `
         this.logger.info('got reINVITE but no endpoint and media has not been released');
         res.send(488);
       }
-
-      await this.pauseOrResumeTranscribeTaskByReInvite(req);
     } catch (err) {
       this.logger.error(err, 'Error handling reinvite');
     }
   }
 
   /**
-   * Handle incoming REFER
+   * Handle incoming REFER if we are in a dial task
    * @param {*} req
    * @param {*} res
    */
   _onRefer(req, res) {
     const task = this.currentTask;
     const sd = task.sd;
-    if (task && TaskName.Dial === task.name && sd && task.referHook) {
+    if (task && TaskName.Dial === task.name && sd) {
       task.handleRefer(this, req, res);
     }
     else {
-      this._handleRefer(req, res);
-    }
-  }
-
-  async _handleRefer(req, res) {
-    if (this._referHook) {
-      try {
-        const to = parseUri(req.getParsedHeader('Refer-To').uri);
-        const by = parseUri(req.getParsedHeader('Referred-By').uri);
-        const b3 = this.b3;
-        const httpHeaders = b3 && {b3};
-        const json = await this.requestor.request('verb:hook', this._referHook, {
-          ...(this.callInfo.toJSON()),
-          refer_details: {
-            sip_refer_to: req.get('Refer-To'),
-            sip_referred_by: req.get('Referred-By'),
-            sip_user_agent: req.get('User-Agent'),
-            refer_to_user: to.scheme === 'tel' ? to.number : to.user,
-            referred_by_user: by.scheme === 'tel' ? by.number : by.user,
-            referring_call_sid: this.callSid,
-            referred_call_sid: null,
-          }
-        }, httpHeaders);
-
-        if (json && Array.isArray(json)) {
-          const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
-          if (tasks && tasks.length > 0) {
-            this.logger.info('CallSession:handleRefer received REFER, get new tasks');
-            this.replaceApplication(tasks);
-            if (this.wakeupResolver) {
-              this.wakeupResolver({reason: 'CallSession: referHook new taks'});
-              this.wakeupResolver = null;
-            }
-          }
-        }
-        res.send(202);
-        this.logger.info('CallSession:handleRefer - sent 202 Accepted');
-      } catch (err) {
-        this.logger.error({err}, 'CallSession:handleRefer - error while asking referHook');
-        res.send(err.statusCode || 501);
-      }
-    } else {
       res.send(501);
     }
   }

lib/session/rest-call-session.js

@@ -1,6 +1,10 @@
 const CallSession = require('./call-session');
 const {CallStatus} = require('../utils/constants');
 const moment = require('moment');
+const {parseUri} = require('drachtio-srf');
+const { normalizeJambones } = require('@jambonz/verb-specifications');
+const makeTask = require('../tasks/make_task');
+
 /**
  * @classdesc Subclass of CallSession.  This represents a CallSession that is
  * created for an outbound call that is initiated via the REST API.
@@ -42,9 +46,61 @@ class RestCallSession extends CallSession {
     this.dlg = dlg;
     dlg.on('destroy', this._callerHungup.bind(this));
     dlg.on('refer', this._onRefer.bind(this));
-    dlg.on('modify', this._onReinvite.bind(this));
     this.wrapDialog(dlg);
   }
+
+  /**
+   * global referHook
+   */
+
+  set referHook(hook) {
+    this._referHook = hook;
+  }
+
+  /**
+   * This is invoked when the called party sends REFER to Jambonz.
+   */
+  async _onRefer(req, res) {
+    if (this._referHook) {
+      try {
+        const to = parseUri(req.getParsedHeader('Refer-To').uri);
+        const by = parseUri(req.getParsedHeader('Referred-By').uri);
+        const b3 = this.b3;
+        const httpHeaders = b3 && {b3};
+        const json = await this.requestor.request('verb:hook', this._referHook, {
+          ...(this.callInfo.toJSON()),
+          refer_details: {
+            sip_refer_to: req.get('Refer-To'),
+            sip_referred_by: req.get('Referred-By'),
+            sip_user_agent: req.get('User-Agent'),
+            refer_to_user: to.scheme === 'tel' ? to.number : to.user,
+            referred_by_user: by.scheme === 'tel' ? by.number : by.user,
+            referring_call_sid: this.callSid,
+            referred_call_sid: null,
+          }
+        }, httpHeaders);
+
+        if (json && Array.isArray(json)) {
+          const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
+          if (tasks && tasks.length > 0) {
+            this.logger.info('RestCallSession:handleRefer received REFER, get new tasks');
+            this.replaceApplication(tasks);
+            if (this.wakeupResolver) {
+              this.wakeupResolver({reason: 'RestCallSession: referHook new taks'});
+              this.wakeupResolver = null;
+            }
+          }
+        }
+        res.send(202);
+        this.logger.info('RestCallSession:handleRefer - sent 202 Accepted');
+      } catch (err) {
+        this.logger.error({err}, 'RestCallSession:handleRefer - error while asking referHook');
+        res.send(err.statusCode || 501);
+      }
+    } else {
+      res.send(501);
+    }
+  }
   /**
    * This is invoked when the called party hangs up, in order to calculate the call duration.
    */

lib/tasks/conference.js

@@ -118,9 +118,7 @@ class Conference extends Task {
     this.emitter.emit('kill');
     await this._doFinalMemberCheck(cs);
     if (this.ep && this.ep.connected) {
-      // drachtio-fsmrf override esl::event::CUSTOM to conference join listerner, After finish the conference
-      // the application need to reset the esl::event::CUSTOM for another use on the same endpoint
-      this.ep.resetEslCustomEvent();
+      this.ep.conn.removeAllListeners('esl::event::CUSTOM::*');
       this.ep.api(`conference ${this.confName} kick ${this.memberId}`)
         .catch((err) => this.logger.info({err}, 'Error kicking participant'));
     }
@@ -546,13 +544,6 @@ class Conference extends Task {
     } while (!this.killed && this.conf_hold_status === 'hold');
   }
 
-  /**
-   * mute or unmute side of the call
-   */
-  mute(callSid, doMute) {
-    this.doConferenceMute(this.callSession, {conf_mute_status: doMute});
-  }
-
   /**
    * Add ourselves to the waitlist of sessions to be notified once
    * the conference starts

lib/tasks/config.js

@@ -34,8 +34,7 @@ class TaskConfig extends Task {
         'finishOnKey', 'input', 'numDigits', 'minDigits', 'maxDigits',
         'interDigitTimeout', 'bargein', 'dtmfBargein', 'minBargeinWordCount', 'actionHook'
       ].forEach((k) => {
-        const val = this.bargeIn[k];
-        if (val !== undefined && val !== null) this.gatherOpts[k] = val;
+        if (this.bargeIn[k]) this.gatherOpts[k] = this.bargeIn[k];
       });
     }
     if (this.transcribe?.enable) {
@@ -74,7 +73,6 @@ class TaskConfig extends Task {
   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 hasReferHook() { return Object.keys(this.data).includes('referHook'); }
 
   get summary() {
     const phrase = [];
@@ -84,13 +82,13 @@ class TaskConfig extends Task {
 
     if (this.bargeIn.enable) phrase.push('enable barge-in');
     if (this.hasSynthesizer) {
-      const {vendor:v, language:l, voice, label} = this.synthesizer;
-      const s = `{${v},${l},${voice},${label || 'None'}}`;
+      const {vendor:v, language:l, voice} = this.synthesizer;
+      const s = `{${v},${l},${voice}}`;
       phrase.push(`set synthesizer${s}`);
     }
     if (this.hasRecognizer) {
-      const {vendor:v, language:l, label} = this.recognizer;
-      const s = `{${v},${l},${label || 'None'}}`;
+      const {vendor:v, language:l} = this.recognizer;
+      const s = `{${v},${l}}`;
       phrase.push(`set recognizer${s}`);
     }
     if (this.hasRecording) phrase.push(this.record.action);
@@ -105,7 +103,6 @@ class TaskConfig extends Task {
     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 ${this.data.boostAudioSignal}`);
-    if (this.hasReferHook) phrase.push('set referHook');
     return `${this.name}{${phrase.join(',')}}`;
   }
 
@@ -298,13 +295,9 @@ class TaskConfig extends Task {
         voiceMs: this.vad.voiceMs || 250,
         silenceMs: this.vad.silenceMs || 150,
         strategy: this.vad.strategy || 'one-shot',
-        mode: (this.vad.mode !== undefined && this.vad.mode !== null) ? this.vad.mode : 2
+        mode: this.vad.mod || 2
       };
     }
-
-    if (this.hasReferHook) {
-      cs.referHook = this.data.referHook;
-    }
   }
 
   async kill(cs) {

lib/tasks/dial.js

@@ -82,8 +82,6 @@ function filterAndLimit(logger, tasks) {
   return unique;
 }
 
-const sleepFor = (ms) => new Promise((resolve) => setTimeout(() => resolve(), ms));
-
 class TaskDial extends Task {
   constructor(logger, opts) {
     super(logger, opts);
@@ -205,16 +203,7 @@ class TaskDial extends Task {
       else {
         this.epOther = cs.ep;
         if (this.dialMusic && this.epOther && this.epOther.connected) {
-          (async() => {
-            do {
-              try {
-                await this.epOther.play(this.dialMusic);
-              } catch (err) {
-                this.logger.error(err, `TaskDial:exec error playing ${this.dialMusic}`);
-                await sleepFor(1000);
-              }
-            } while (!this.killed || !this.bridged);
-          })();
+          this.epOther.play(this.dialMusic).catch((err) => {});
         }
       }
       if (!this.killed) await this._attemptCalls(cs);

lib/tasks/gather.js

@@ -22,7 +22,6 @@ const {
 const makeTask = require('./make_task');
 const assert = require('assert');
 const SttTask = require('./stt-task');
-const { SpeechCredentialError } = require('../utils/error');
 
 class TaskGather extends SttTask {
   constructor(logger, opts, parentTask) {
@@ -123,20 +122,7 @@ class TaskGather extends SttTask {
     return s;
   }
 
-  async exec(cs, obj) {
-    try {
-      await this.handling(cs, obj);
-    } catch (error) {
-      if (error instanceof SpeechCredentialError) {
-        this.logger.info('Gather failed due to SpeechCredentialError, finished!');
-        this.notifyTaskDone();
-        return;
-      }
-      throw error;
-    }
-  }
-
-  async handling(cs, {ep}) {
+  async exec(cs, {ep}) {
     this.logger.debug({options: this.data}, 'Gather:exec');
     await super.exec(cs, {ep});
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
@@ -191,10 +177,8 @@ class TaskGather extends SttTask {
     this._startVad();
 
     const startDtmfListener = () => {
-      assert(!this._dtmfListenerStarted);
       if (this.input.includes('digits') || this.dtmfBargein || this.asrDtmfTerminationDigit) {
         ep.on('dtmf', this._onDtmf.bind(this, cs, ep));
-        this._dtmfListenerStarted = true;
       }
     };
 
@@ -297,8 +281,7 @@ class TaskGather extends SttTask {
         }
       }
 
-      // https://github.com/jambonz/jambonz-feature-server/issues/913
-      if (this.listenDuringPrompt || (!this.sayTask && !this.playTask)) {
+      if (this.listenDuringPrompt) {
         startDtmfListener();
       }
 
@@ -908,9 +891,8 @@ class TaskGather extends SttTask {
       if (this._timeoutTimer) {
         this._startTimer();
       }
-      /* restart asr timer if we get a partial transcript (only if the asr timer is already running) */
-      /* note: https://github.com/jambonz/jambonz-feature-server/issues/866 */
-      if (this.isContinuousAsr && this._asrTimer) this._startAsrTimer();
+      /* restart asr timer if we get a partial transcript */
+      if (this.isContinuousAsr) this._startAsrTimer();
     }
   }
   _onEndOfUtterance(cs, ep) {
@@ -1070,7 +1052,6 @@ class TaskGather extends SttTask {
 
     this.span.setAttributes({
       channel: 1,
-      'stt.label': this.label || 'None',
       'stt.resolve': reason,
       'stt.result': JSON.stringify(evt)
     });

lib/tasks/rest_dial.js

@@ -39,9 +39,9 @@ class TaskRestDial extends Task {
 
     if (this.data.amd) {
       this.startAmd = cs.startAmd;
+      this.stopAmd = cs.stopAmd;
       this.on('amd', this._onAmdEvent.bind(this, cs));
     }
-    this.stopAmd = cs.stopAmd;
 
     this._setCallTimer();
     await this.awaitTaskDone();

lib/tasks/say.js

@@ -1,7 +1,6 @@
 const TtsTask = require('./tts-task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
 const pollySSMLSplit = require('polly-ssml-split');
-const { SpeechCredentialError } = require('../utils/error');
 
 const breakLengthyTextIfNeeded = (logger,  text) => {
   const chunkSize = 1000;
@@ -62,23 +61,148 @@ class TaskSay extends TtsTask {
     }
   }
 
-  async exec(cs, obj) {
-    try {
-      await this.handling(cs, obj);
-      this.emit('playDone');
-    } catch (error) {
-      if (error instanceof SpeechCredentialError) {
-        // if say failed due to speech credentials, alarm is writtern and error notification is sent
-        // finished this say to move to next task.
-        this.logger.info('Say failed due to SpeechCredentialError, finished!');
-        this.emit('playDone');
-        return;
+  async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
+    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 || cs.synthesizer?.engine || 'neural';
+    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];
       }
-      throw error;
+    } 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.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._disableTracing) this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
+    try {
+      if (!credentials) {
+        writeAlerts({
+          account_sid,
+          alert_type: AlertType.TTS_NOT_PROVISIONED,
+          vendor,
+          target_sid: cs.callSid
+        }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
+        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._disableTracing)  {
+          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,
+            preCache
+          });
+          if (!filePath.startsWith('say:')) {
+            this.logger.debug(`Say: file ${filePath}, served from cache ${servedFromCache}`);
+            if (filePath) cs.trackTmpFile(filePath);
+            if (this.otelSpan) {
+              this.otelSpan.setAttributes({'tts.cached': servedFromCache});
+              this.otelSpan.end();
+              this.otelSpan = null;
+            }
+            if (!servedFromCache && !lastUpdated) {
+              lastUpdated = true;
+              updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
+            }
+            if (!servedFromCache && rtt && !preCache  && !this._disableTracing) {
+              this.notifyStatus({
+                event: 'synthesized-audio',
+                vendor,
+                language,
+                characters: text.length,
+                elapsedTime: rtt
+              });
+            }
+          }
+          else {
+            this.logger.debug('Say: a streaming tts api will be used');
+            const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${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,
+            target_sid: cs.callSid
+          }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
+          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;
+    }
+
   }
 
-  async handling(cs, {ep}) {
+  async exec(cs, {ep}) {
     const {srf, accountSid:account_sid, callSid:target_sid} = cs;
     const {writeAlerts, AlertType} = srf.locals;
     const {addFileToCache} = srf.locals.dbHelpers;
@@ -96,7 +220,10 @@ class TaskSay extends TtsTask {
     let voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
       this.synthesizer.voice :
       cs.speechSynthesisVoice;
-    let label = this.taskInlcudeSynthesizer ? this.synthesizer.label : cs.speechSynthesisLabel;
+    // label can be null/empty in synthesizer config, just use application level label if it's default
+    let label = this.synthesizer.label === 'default' ?
+      cs.speechSynthesisLabel :
+      this.synthesizer.label;
 
     const fallbackVendor = this.synthesizer.fallbackVendor && this.synthesizer.fallbackVendor !== 'default' ?
       this.synthesizer.fallbackVendor :
@@ -107,8 +234,10 @@ class TaskSay extends TtsTask {
     const fallbackVoice =  this.synthesizer.fallbackVoice && this.synthesizer.fallbackVoice !== 'default' ?
       this.synthesizer.fallbackVoice :
       cs.fallbackSpeechSynthesisVoice;
-    const fallbackLabel = this.taskInlcudeSynthesizer ?
-      this.synthesizer.fallbackLabel : cs.fallbackSpeechSynthesisLabel;
+    // label can be null/empty in synthesizer config, just use application level label if it's default
+    const fallbackLabel = this.synthesizer.fallbackLabel === 'default' ?
+      cs.fallbackSpeechSynthesisLabel :
+      this.synthesizer.fallbackLabel;
 
     if (cs.hasFallbackTts) {
       vendor = fallbackVendor;
@@ -134,7 +263,7 @@ class TaskSay extends TtsTask {
       } else {
         this.notifyError(
           { msg: 'TTS error', details:`TTS vendor ${vendor} error: ${error}`, failover: 'not available'});
-        throw new SpeechCredentialError(error.message);
+        throw error;
       }
     };
     let filepath;
@@ -153,12 +282,12 @@ class TaskSay extends TtsTask {
           await this.playToConfMember(ep, memberId, confName, confUuid, filepath[segment]);
         }
         else {
-          const isStreaming = filepath[segment].startsWith('say:{');
-          if (isStreaming) {
+          let tts_cache_filename;
+          if (filepath[segment].startsWith('say:{')) {
             const arr = /^say:\{.*\}\s*(.*)$/.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)}`);
+          else this.logger.debug(`Say:exec sending  ${filepath[segment].substring(0, 64)}`);
           ep.once('playback-start', (evt) => {
             this.logger.debug({evt}, 'Say got playback-start');
             if (this.otelSpan) {
@@ -166,19 +295,17 @@ class TaskSay extends TtsTask {
               this.otelSpan.end();
               this.otelSpan = null;
               if (evt.variable_tts_cache_filename) {
+                tts_cache_filename = evt.variable_tts_cache_filename;
                 cs.trackTmpFile(evt.variable_tts_cache_filename);
               }
+              else {
+                this.logger.info('No tts_cache_filename in playback-start event');
+              }
             }
           });
           ep.once('playback-stop', (evt) => {
-            this.logger.debug({evt}, 'Say 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 (!tts_cache_filename || evt.variable_tts_cache_filename !== tts_cache_filename) {
+              this.logger.info({evt}, 'Say: discarding playback-stop from other say verb');
             }
             else {
               this.logger.debug({evt}, 'Say got playback-stop');
@@ -223,7 +350,6 @@ class TaskSay extends TtsTask {
               continue;
             } catch (err) {
               this.logger.info({err}, 'Error waiting for playback-stop event');
-              throw err;
             }
           } finally {
             this._playPromise = null;
@@ -241,6 +367,7 @@ class TaskSay extends TtsTask {
         segment++;
       }
     }
+    this.emit('playDone');
   }
 
   async kill(cs) {

lib/tasks/stt-task.js

@@ -2,7 +2,6 @@ const Task = require('./task');
 const assert = require('assert');
 const crypto = require('crypto');
 const { TaskPreconditions, CobaltTranscriptionEvents } = require('../utils/constants');
-const { SpeechCredentialError } = require('../utils/error');
 
 class SttTask extends Task {
 
@@ -25,12 +24,6 @@ class SttTask extends Task {
     this.consolidateTranscripts = consolidateTranscripts;
     this.eventHandlers = [];
     this.isHandledByPrimaryProvider = true;
-    /**
-     * Task use taskIncludeRecognizer to identify
-     * if taskIncludeRecognizer === true, use label from verb.recognizer, even it's empty
-     * if taskIncludeRecognizer === false, use label from application.recognizer
-     */
-    this.taskIncludeRecognizer = !!this.data.recognizer;
     if (this.data.recognizer) {
       const recognizer = this.data.recognizer;
       this.vendor = recognizer.vendor;
@@ -40,6 +33,7 @@ class SttTask extends Task {
       //fallback
       this.fallbackVendor = recognizer.fallbackVendor || 'default';
       this.fallbackLanguage = recognizer.fallbackLanguage || 'default';
+      // label can be empty and should not have default value.
       this.fallbackLabel = recognizer.fallbackLabel;
 
       /* let credentials be supplied in the recognizer object at runtime */
@@ -88,7 +82,8 @@ class SttTask extends Task {
       this.language = cs.speechRecognizerLanguage;
       if (this.data.recognizer) this.data.recognizer.language = this.language;
     }
-    if (!this.taskIncludeRecognizer) {
+    // label can be empty, should not assign application level label
+    if ('default' === this.label) {
       this.label = cs.speechRecognizerLabel;
       if (this.data.recognizer) this.data.recognizer.label = this.label;
     }
@@ -101,21 +96,17 @@ class SttTask extends Task {
       this.fallbackLanguage = cs.fallbackSpeechRecognizerLanguage;
       if (this.data.recognizer) this.data.recognizer.fallbackLanguage = this.fallbackLanguage;
     }
-    if (!this.taskIncludeRecognizer) {
+    // label can be empty, should not assign application level label
+    if ('default' === 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) {
-      if (this.taskIncludeRecognizer) {
-        // reset fallback ASR from previous run if this verb contains data.recognizer.
-        cs.hasFallbackAsr = false;
-      } else {
-        this.logger.debug('Call session has fallback to 2nd ASR, use 2nd recognizer configuration');
-        this.vendor = this.fallbackVendor;
-        this.language = this.fallbackLanguage;
-        this.label = this.fallbackLabel;
-      }
+      this.vendor = this.fallbackVendor;
+      this.language = this.fallbackLanguage;
+      this.label = this.fallbackLabel;
     }
     if (!this.data.recognizer.vendor) {
       this.data.recognizer.vendor = this.vendor;
@@ -190,8 +181,8 @@ class SttTask extends Task {
         vendor,
         target_sid: cs.callSid
       }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
-      // the ASR might have fallback configuration, should not done task here.
-      throw new SpeechCredentialError(`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`);
     }
 
     if (vendor === 'nuance' && credentials.client_id) {
@@ -232,12 +223,12 @@ class SttTask extends Task {
 
   async _initFallback() {
     assert(this.fallbackVendor, 'fallback failed without fallbackVendor configuration');
-    this.logger.info(`Failed to use primary STT provider, fallback to ${this.fallbackVendor}`);
     this.isHandledByPrimaryProvider = false;
     this.cs.hasFallbackAsr = true;
-    this.vendor = this.cs.fallbackSpeechRecognizerVendor = this.fallbackVendor;
-    this.language = this.cs.fallbackSpeechRecognizerLanguage = this.fallbackLanguage;
-    this.label = this.cs.fallbackSpeechRecognizerLabel = this.fallbackLabel;
+    this.logger.info(`Failed to use primary STT provider, fallback to ${this.fallbackVendor}`);
+    this.vendor = this.fallbackVendor;
+    this.language = this.fallbackLanguage;
+    this.label = this.fallbackLabel;
     this.data.recognizer.vendor = this.vendor;
     this.data.recognizer.language = this.language;
     this.data.recognizer.label = this.label;

lib/tasks/transcribe.js

@@ -16,7 +16,6 @@ const {
 } = require('../utils/constants.json');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 const SttTask = require('./stt-task');
-const { SpeechCredentialError } = require('../utils/error');
 
 const STT_LISTEN_SPAN_NAME = 'stt-listen';
 
@@ -74,20 +73,7 @@ class TaskTranscribe extends SttTask {
     return this.channel === 2 || this.separateRecognitionPerChannel && this.ep2;
   }
 
-  async exec(cs, obj) {
-    try {
-      await this.handling(cs, obj);
-    } catch (error) {
-      if (error instanceof SpeechCredentialError) {
-        this.logger.info('Transcribe failed due to SpeechCredentialError, finished!');
-        this.notifyTaskDone();
-        return;
-      }
-      throw error;
-    }
-  }
-
-  async handling(cs, {ep, ep2}) {
+  async exec(cs, {ep, ep2}) {
     await super.exec(cs, {ep, ep2});
 
     if (this.data.recognizer.vendor === 'nuance') {
@@ -477,7 +463,6 @@ class TaskTranscribe extends SttTask {
       if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
         this.childSpan[channel - 1].span.setAttributes({
           channel,
-          'stt.label': this.label || 'None',
           'stt.resolve': 'transcript',
           'stt.result': JSON.stringify(evt)
         });
@@ -531,8 +516,7 @@ class TaskTranscribe extends SttTask {
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
-        'stt.resolve': 'timeout',
-        'stt.label': this.label || 'None',
+        'stt.resolve': 'timeout'
       });
       this.childSpan[channel - 1].span.end();
     }
@@ -549,8 +533,7 @@ class TaskTranscribe extends SttTask {
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
-        'stt.resolve': 'max duration exceeded',
-        'stt.label': this.label || 'None',
+        'stt.resolve': 'max duration exceeded'
       });
       this.childSpan[channel - 1].span.end();
     }
@@ -634,8 +617,7 @@ class TaskTranscribe extends SttTask {
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
-        'stt.resolve': 'connection failure',
-        'stt.label': this.label || 'None',
+        'stt.resolve': 'connection failure'
       });
       this.childSpan[channel - 1].span.end();
     }

lib/tasks/tts-task.js

@@ -1,6 +1,5 @@
 const Task = require('./task');
 const { TaskPreconditions } = require('../utils/constants');
-const { SpeechCredentialError } = require('../utils/error');
 
 class TtsTask extends Task {
 
@@ -11,12 +10,6 @@ class TtsTask extends Task {
     this.preconditions = TaskPreconditions.Endpoint;
 
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
-    /**
-     * Task use taskInlcudeSynthesizer to identify
-     * if taskInlcudeSynthesizer === true, use label from verb.synthesizer, even it's empty
-     * if taskInlcudeSynthesizer === false, use label from application.synthesizer
-     */
-    this.taskInlcudeSynthesizer = !!this.data.synthesizer;
     this.synthesizer = this.data.synthesizer || {};
     this.disableTtsCache = this.data.disableTtsCache;
     this.options = this.synthesizer.options || {};
@@ -24,26 +17,16 @@ class TtsTask extends Task {
 
   async exec(cs) {
     super.exec(cs);
-    if (cs.synthesizer) {
-      this.options = {...cs.synthesizer.options, ...this.options};
-      this.data.synthesizer = this.data.synthesizer || {};
-      for (const k in cs.synthesizer) {
-        const newValue = this.data.synthesizer && this.data.synthesizer[k] !== undefined ?
-          this.data.synthesizer[k] :
-          cs.synthesizer[k];
-
-        if (Array.isArray(newValue)) {
-          this.data.synthesizer[k] = [...(this.data.synthesizer[k] || []), ...cs.synthesizer[k]];
-        } else if (typeof newValue === 'object' && newValue !== null) {
-          this.data.synthesizer[k] = { ...(this.data.synthesizer[k] || {}), ...cs.synthesizer[k] };
-        } else {
-          this.data.synthesizer[k] = newValue;
-        }
-      }
-    }
   }
 
-  async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
+  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;
@@ -52,10 +35,6 @@ class TtsTask extends Task {
     const salt = cs.callSid;
 
     let credentials = cs.getSpeechCredentials(vendor, 'tts', label);
-    if (!credentials) {
-      throw new SpeechCredentialError(
-        `No text-to-speech service credentials for ${vendor} with labels: ${label} have been configured`);
-    }
     /* parse Nuance voices into name and model */
     let model;
     if (vendor === 'nuance' && voice) {
@@ -83,50 +62,27 @@ class TtsTask extends Task {
       credentials.optimize_streaming_latency = this.options.optimize_streaming_latency
       || credentials.optimize_streaming_latency;
       voice = this.options.voice_id || voice;
-    } else if (vendor === 'rimelabs') {
-      credentials = credentials || {};
-      credentials.model_id = this.options.model_id || credentials.model_id;
-    } else if (vendor === 'whisper') {
-      credentials = credentials || {};
-      credentials.model_id = this.options.model_id || credentials.model_id;
-    } else if (vendor === 'verbio') {
-      credentials = credentials || {};
-      credentials.engine_version = this.options.engine_version || credentials.engine_version;
-    } else if (vendor === 'playht') {
-      credentials = credentials || {};
-      credentials.voice_engine = this.options.voice_engine || credentials.voice_engine;
     }
 
-    /**
-     * note on cache_speech_handles.  This was found to be risky.
-     * It can cause a crash in the following sequence on a single call:
-     * 1. Stream tts on vendor A with cache_speech_handles=1, then
-     * 2. Stream tts on vendor B with cache_speech_handles=1
-     *
-     * we previously tried to track when vendors were switched and manage the flag accordingly,
-     * but it difficult to track all the scenarios and the benefit (slightly faster start to tts playout)
-     * is probably minimal.  DH.
-     */
     ep.set({
-      tts_engine: vendor.startsWith('custom:') ? 'custom' : vendor,
+      tts_engine: vendor,
       tts_voice: voice,
-      //cache_speech_handles: !cs.currentTtsVendor || cs.currentTtsVendor === vendor ? 1 : 0,
-      cache_speech_handles: 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;
+      cache_speech_handles: 1,
+    }).catch((err) => this.logger.info({err}, `${this.name}: Error setting tts_engine on endpoint`));
 
-    if (!preCache && !this._disableTracing) this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
+    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,
-          target_sid: cs.callSid
+          vendor
         }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-        throw new SpeechCredentialError('no provisioned speech credentials for 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;
@@ -137,12 +93,11 @@ class TtsTask extends Task {
         if (text.startsWith('silence_stream://')) return text;
 
         /* otel: trace time for tts */
-        if (!preCache && !this._disableTracing)  {
+        if (!preCache && !this.parentTask)  {
           const {span} = this.startChildSpan('tts-generation', {
             'tts.vendor': vendor,
             'tts.language': language,
-            'tts.voice': voice,
-            'tts.label': label || 'None',
+            'tts.voice': voice
           });
           this.otelSpan = span;
         }
@@ -159,10 +114,11 @@ class TtsTask extends Task {
             credentials,
             options: this.options,
             disableTtsCache : this.disableTtsCache,
-            renderForCaching: preCache
+            disableTtsStreaming,
+            preCache
           });
           if (!filePath.startsWith('say:')) {
-            this.logger.debug(`Say: file ${filePath}, served from cache ${servedFromCache}`);
+            this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
             if (filePath) cs.trackTmpFile(filePath);
             if (this.otelSpan) {
               this.otelSpan.setAttributes({'tts.cached': servedFromCache});
@@ -173,7 +129,7 @@ class TtsTask extends Task {
               lastUpdated = true;
               updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
             }
-            if (!servedFromCache && rtt && !preCache  && !this._disableTracing) {
+            if (!servedFromCache && rtt && !preCache) {
               this.notifyStatus({
                 event: 'synthesized-audio',
                 vendor,
@@ -184,7 +140,7 @@ class TtsTask extends Task {
             }
           }
           else {
-            this.logger.debug('Say: a streaming tts api will be used');
+            this.logger.debug('a streaming tts api will be used');
             const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${ep.uuid},`);
             return modifiedPath;
           }
@@ -199,6 +155,7 @@ class TtsTask extends Task {
             detail: err.message,
             target_sid: cs.callSid
           }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
+          this.notifyError({msg: 'TTS error', details: err.message || err});
           throw err;
         }
       };
@@ -209,7 +166,6 @@ class TtsTask extends Task {
       this.logger.info(err, 'TaskSay:exec error');
       throw err;
     }
-
   }
 
   _validateURL(urlString) {

lib/utils/action-hook-delay.js

@@ -2,6 +2,7 @@ const makeTask = require('../tasks/make_task');
 const Emitter = require('events');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 const {TaskName} = require('../utils/constants');
+const assert = require('assert');
 
 /**
  * ActionHookDelayProcessor
@@ -50,9 +51,8 @@ class ActionHookDelayProcessor extends Emitter {
 
     this.actions = opts.actions;
     this.retries = opts.retries || 0;
-    this.noResponseTimeout = opts.noResponseTimeout;
+    this.noResponseTimeout = opts.noResponseTimeout || 0;
     this.noResponseGiveUpTimeout = opts.noResponseGiveUpTimeout;
-    this.giveUpActions = opts.giveUpActions;
 
     // return false if these options actually disable the ahdp
     return ('enable' in opts && opts.enable === true) ||
@@ -66,16 +66,11 @@ class ActionHookDelayProcessor extends Emitter {
       this.logger.debug('ActionHookDelayProcessor#start: already started due to prior gather which is continuing');
       return;
     }
+    assert(!this._noResponseTimer);
     this._active = true;
     this._retryCount = 0;
-    if (this.noResponseTimeout > 0) {
-      const timeoutMs =  this.noResponseTimeout * 1000;
-      this._noResponseTimer = setTimeout(this._onNoResponseTimer.bind(this), timeoutMs);
-    } else {
-      this.logger.debug(
-        'ActionHookDelayProcessor#start: noResponseTimeout is 0 or undefined hence not calling _onNoResponseTimer'
-      );
-    }
+    const timeoutMs =  this.noResponseTimeout === 0 ? 1 : this.noResponseTimeout * 1000;
+    this._noResponseTimer = setTimeout(this._onNoResponseTimer.bind(this), timeoutMs);
 
     if (this.noResponseGiveUpTimeout > 0) {
       const timeoutMs = this.noResponseGiveUpTimeout * 1000;
@@ -84,6 +79,7 @@ class ActionHookDelayProcessor extends Emitter {
   }
 
   async stop() {
+    this.logger.debug('ActionHookDelayProcessor#stop');
     this._active = false;
 
     if (this._noResponseTimer) {
@@ -95,19 +91,25 @@ class ActionHookDelayProcessor extends Emitter {
       this._noResponseGiveUpTimer = null;
     }
     if (this._taskInProgress) {
-      this.logger.debug(`ActionHookDelayProcessor#stop: stopping ${this._taskInProgress.name}`);
+      this.logger.debug(`ActionHookDelayProcessor#stop: killing task in progress: ${this._taskInProgress.name}`);
 
-      this._sayResolver = () => {
-        this.logger.debug('ActionHookDelayProcessor#stop: play/say is done, continue on..');
-        //this._taskInProgress.kill(this.cs);
-        this._taskInProgress = null;
-      };
-
-      /* we let Say finish, but interrupt Play */
-      if (TaskName.Play === this._taskInProgress.name) {
-        await this._taskInProgress.kill(this.cs);
+      /** if we are doing a play, kill it immediately
+       * if we are doing a say, wait for it to finish
+       */
+      if (TaskName.Say === this._taskInProgress.name) {
+        this._sayResolver = () => {
+          this.logger.debug('ActionHookDelayProcessor#stop: say is done, continue on..');
+          this._taskInProgress.kill(this.cs);
+          this._taskInProgress = null;
+        };
+        this.logger.debug('ActionHookDelayProcessor#stop returning promise');
+        return new Promise((resolve) => this._sayResolver = resolve);
+      }
+      else {
+        /* play */
+        this._taskInProgress.kill(this.cs);
+        this._taskInProgress = null;
       }
-      return new Promise((resolve) => this._sayResolver = resolve);
     }
     this.logger.debug('ActionHookDelayProcessor#stop returning');
   }
@@ -135,9 +137,7 @@ class ActionHookDelayProcessor extends Emitter {
       this.logger.debug({evt}, 'got playback-start');
       if (!this._active) {
         this.logger.info({evt}, 'ActionHookDelayProcessor#_onNoResponseTimer: killing audio immediately');
-
-        /* note: in race condition we may have just hung up and cs.ep cleared */
-        this.ep?.api('uuid_break', this.ep?.uuid)
+        this.ep.api('uuid_break', this.ep.uuid)
           .catch((err) => this.logger.info(err,
             'ActionHookDelayProcessor#_onNoResponseTimer Error killing audio'));
       }
@@ -147,7 +147,7 @@ class ActionHookDelayProcessor extends Emitter {
       this._taskInProgress = null;
       if (this._sayResolver) {
         /* we were waiting for the play to finish before continuing to next task */
-        this.logger.debug({evt}, 'ActionHookDelayProcessor#_onNoResponseTimer got playback-stop');
+        this.logger.debug({evt}, 'got playback-stop');
         this._sayResolver();
         this._sayResolver = null;
       }
@@ -166,14 +166,9 @@ class ActionHookDelayProcessor extends Emitter {
 
   _onNoResponseGiveUpTimer() {
     this._active = false;
-    if (!this.giveUpActions) {
-      this.logger.info('ActionHookDelayProcessor#_onNoResponseGiveUpTimer');
-      this.stop().catch((err) => {});
-      this.emit('giveup');
-    } else {
-      this.logger.info('ActionHookDelayProcessor#_onNoResponseGiveUpTimer - giveUpActions');
-      this.emit('giveupWithTasks', this.giveUpActions);
-    }
+    this.logger.info('ActionHookDelayProcessor#_onNoResponseGiveUpTimer');
+    this.stop().catch((err) => {});
+    this.emit('giveup');
   }
 }
 

lib/utils/amd-utils.js

@@ -246,10 +246,7 @@ module.exports = (logger) => {
     const amd = ep.amd = new Amd(logger, cs, opts);
     const {vendor, language} = amd;
     let sttCredentials = amd.sttCredentials;
-    // hints from configuration might be too long for specific language and vendor that make transcribe freeswitch
-    // modules cannot connect to the vendor. hints is used in next step to validate if the transcription
-    // matchs voice mail hints.
-    const hints = [];
+    const hints = voicemailHints[language] || [];
 
     if (vendor === 'nuance' && sttCredentials.client_id) {
       /* get nuance access token */

lib/utils/error.js

@@ -1,9 +0,0 @@
-class SpeechCredentialError extends Error {
-  constructor(msg) {
-    super(msg);
-  }
-}
-
-module.exports = {
-  SpeechCredentialError
-};

lib/utils/install-srf-locals.js

@@ -12,7 +12,6 @@ const {
   JAMBONES_TIME_SERIES_HOST,
   JAMBONES_ESL_LISTEN_ADDRESS,
   PORT,
-  HTTP_IP,
   NODE_ENV,
 } = require('../config');
 const Registrar = require('@jambonz/mw-registrar');
@@ -194,8 +193,7 @@ function installSrfLocals(srf, logger) {
 
   let localIp;
   try {
-    // Either use the configured IP address or call ip.address() to find it
-    localIp = HTTP_IP || ip.address();
+    localIp = ip.address();
   } catch (err) {
     logger.error({err}, 'installSrfLocals - error detecting local ipv4 address');
   }

lib/utils/place-outdial.js

@@ -447,8 +447,6 @@ class SingleDialer extends Emitter {
     });
 
     cs.req = this.req;
-    // fixed hangup an adulting session does not send status callback Completed
-    cs.wrapDialog(this.dlg);
     cs.exec().catch((err) => newLogger.error({err}, 'doAdulting: error executing session'));
     return cs;
   }

lib/utils/transcription-utils.js

@@ -142,6 +142,7 @@ const optimalDeepramModels = {
   tr: ['nova-2', 'nova-2'],
   uk: ['nova-2', 'nova-2']
 };
+
 const selectDefaultDeepgramModel = (task, language) => {
   if (language in optimalDeepramModels) {
     const [gather, transcribe] = optimalDeepramModels[language];
@@ -150,24 +151,6 @@ const selectDefaultDeepgramModel = (task, language) => {
   return 'base';
 };
 
-const optimalGoogleModels = {
-  'v1' : {
-    'en-IN':['telephony', 'latest_long']
-  },
-  'v2' : {
-    'en-IN':['telephony', 'long']
-  }
-};
-const selectDefaultGoogleModel = (task, language, version) => {
-  const useV2 = version === 'v2';
-  if (language in optimalGoogleModels[version]) {
-    const [gather, transcribe] = optimalGoogleModels[version][language];
-    return task.name === TaskName.Gather ? gather : transcribe;
-  }
-  return task.name === TaskName.Gather ?
-    (useV2 ? 'telephony_short' : 'command_and_search') :
-    (useV2 ? 'long' : 'latest_long');
-};
 const consolidateTranscripts = (bufferedTranscripts, channel, language, vendor) => {
   if (bufferedTranscripts.length === 1) return bufferedTranscripts[0];
   let totalConfidence = 0;
@@ -514,9 +497,9 @@ module.exports = (logger) => {
 
     if ('google' === vendor) {
       const useV2 = rOpts.googleOptions?.serviceVersion === 'v2';
-      const version = useV2 ? 'v2' : 'v1';
-      let {model} = rOpts;
-      model = model || selectDefaultGoogleModel(task, language, version);
+      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)}),
@@ -856,7 +839,7 @@ module.exports = (logger) => {
           {VERBIO_SPEECH_INCOMPLETE_TIMEOUT: verbioOptions.speech_incomplete_timeout}),
       };
     } else if (vendor.startsWith('custom:')) {
-      let {options = {}} = rOpts.customOptions || {};
+      let {options = {}} = rOpts;
       const {sampleRate} = rOpts.customOptions || {};
       const {auth_token, custom_stt_url} = sttCredentials;
       options = {

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "jambonz-feature-server",
-  "version": "0.9.2",
+  "version": "0.9.0",
   "lockfileVersion": 2,
   "requires": true,
   "packages": {
     "": {
       "name": "jambonz-feature-server",
-      "version": "0.9.2",
+      "version": "0.9.0",
       "license": "MIT",
       "dependencies": {
         "@aws-sdk/client-auto-scaling": "^3.549.0",
@@ -15,10 +15,10 @@
         "@jambonz/http-health-check": "^0.0.1",
         "@jambonz/mw-registrar": "^0.2.7",
         "@jambonz/realtimedb-helpers": "^0.8.8",
-        "@jambonz/speech-utils": "^0.1.16",
+        "@jambonz/speech-utils": "^0.1.13",
         "@jambonz/stats-collector": "^0.1.10",
         "@jambonz/time-series": "^0.2.9",
-        "@jambonz/verb-specifications": "^0.0.81",
+        "@jambonz/verb-specifications": "^0.0.76",
         "@opentelemetry/api": "^1.8.0",
         "@opentelemetry/exporter-jaeger": "^1.23.0",
         "@opentelemetry/exporter-trace-otlp-http": "^0.50.0",
@@ -31,7 +31,7 @@
         "bent": "^7.3.12",
         "debug": "^4.3.4",
         "deepcopy": "^2.1.0",
-        "drachtio-fsmrf": "^3.0.45",
+        "drachtio-fsmrf": "^3.0.43",
         "drachtio-srf": "^4.5.35",
         "express": "^4.19.2",
         "express-validator": "^7.0.1",
@@ -45,10 +45,10 @@
         "short-uuid": "^5.1.0",
         "sinon": "^17.0.1",
         "to-snake-case": "^1.0.0",
-        "undici": "^6.19.4",
+        "undici": "^6.19.2",
         "uuid-random": "^1.3.2",
         "verify-aws-sns-signature": "^0.1.0",
-        "ws": "^8.18.0",
+        "ws": "^8.17.1",
         "xml2js": "^0.6.2"
       },
       "devDependencies": {
@@ -1536,9 +1536,10 @@
       }
     },
     "node_modules/@jambonz/speech-utils": {
-      "version": "0.1.16",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.16.tgz",
-      "integrity": "sha512-en1ZwDbjDJkeQJ9Ro1l3arYDPYrLnK9OTXQQ7ZjTfqzUipgu0mkTnaUCTzW9taozaBJ4+YelSA1PNg2MtLzVAw==",
+      "version": "0.1.13",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.13.tgz",
+      "integrity": "sha512-QeVmNFLtJGPGQfmp7jXpy742AyJIv2EteelDmNTqWGFEwTBj88q8GLP51hUsIR2ZbE5n/ZmZb/ytT6Y6LIQSDg==",
+      "license": "MIT",
       "dependencies": {
         "@aws-sdk/client-polly": "^3.496.0",
         "@aws-sdk/client-sts": "^3.496.0",
@@ -1574,9 +1575,9 @@
       }
     },
     "node_modules/@jambonz/verb-specifications": {
-      "version": "0.0.81",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.81.tgz",
-      "integrity": "sha512-lufYMLpwaejF3zl8FNy1SvmsxFHXNXU8p7I+GoiwbgOHmFeOX4bQla4RD7bNLFvfjSXPRc2u9h5fceicVa9dcg==",
+      "version": "0.0.76",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.76.tgz",
+      "integrity": "sha512-7s61qAsG07xLLaEAHW236rSYzEoh9Qg0aRWHPbTfxCsuTKDNeq+5EwGAShDU5R5ZpjgweZJLhArQm8Ym+4xJ2A==",
       "dependencies": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"
@@ -3830,10 +3831,9 @@
       }
     },
     "node_modules/drachtio-fsmrf": {
-      "version": "3.0.45",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.45.tgz",
-      "integrity": "sha512-YsrevTwGvg9v0OQXB4gZedlmzegB83fyd3NX3X2x7NXsKa5jO6TZCvZVHtBf/jR/ELE3B0aVLpxHw2YviRMIuQ==",
-      "license": "MIT",
+      "version": "3.0.43",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.43.tgz",
+      "integrity": "sha512-ZDgO4WCzZssmOpf+ng20P8r5pu7demWFiECqwRivbWadXqAm3LNabZwWOred2MdCRDvzgHWUwBZoJPzCqKBL5g==",
       "dependencies": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",
@@ -8831,9 +8831,10 @@
       }
     },
     "node_modules/undici": {
-      "version": "6.19.4",
-      "resolved": "https://registry.npmjs.org/undici/-/undici-6.19.4.tgz",
-      "integrity": "sha512-i3uaEUwNdkRq2qtTRRJb13moW5HWqviu7Vl7oYRYz++uPtGHJj+x7TGjcEuwS5Mt2P4nA0U9dhIX3DdB6JGY0g==",
+      "version": "6.19.2",
+      "resolved": "https://registry.npmjs.org/undici/-/undici-6.19.2.tgz",
+      "integrity": "sha512-JfjKqIauur3Q6biAtHJ564e3bWa8VvT+7cSiOJHFbX4Erv6CLGDpg8z+Fmg/1OI/47RA+GI2QZaF48SSaLvyBA==",
+      "license": "MIT",
       "engines": {
         "node": ">=18.17"
       }
@@ -9192,9 +9193,10 @@
       }
     },
     "node_modules/ws": {
-      "version": "8.18.0",
-      "resolved": "https://registry.npmjs.org/ws/-/ws-8.18.0.tgz",
-      "integrity": "sha512-8VbfWfHLbbwu3+N6OKsOMpBdT4kXPDDB9cJk2bJ6mh9ucxdlnNvH1e+roYkKmN9Nxw2yjz7VzeO9oOz2zJ04Pw==",
+      "version": "8.17.1",
+      "resolved": "https://registry.npmjs.org/ws/-/ws-8.17.1.tgz",
+      "integrity": "sha512-6XQFvXTkbfUOZOKKILFG1PDK2NDQs4azKQl26T0YS5CxqWLgXajbPZ+h4gZekJyRqFU8pvnbAbbs/3TgRPy+GQ==",
+      "license": "MIT",
       "engines": {
         "node": ">=10.0.0"
       },
@@ -10499,9 +10501,9 @@
       }
     },
     "@jambonz/speech-utils": {
-      "version": "0.1.16",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.16.tgz",
-      "integrity": "sha512-en1ZwDbjDJkeQJ9Ro1l3arYDPYrLnK9OTXQQ7ZjTfqzUipgu0mkTnaUCTzW9taozaBJ4+YelSA1PNg2MtLzVAw==",
+      "version": "0.1.13",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.1.13.tgz",
+      "integrity": "sha512-QeVmNFLtJGPGQfmp7jXpy742AyJIv2EteelDmNTqWGFEwTBj88q8GLP51hUsIR2ZbE5n/ZmZb/ytT6Y6LIQSDg==",
       "requires": {
         "@aws-sdk/client-polly": "^3.496.0",
         "@aws-sdk/client-sts": "^3.496.0",
@@ -10537,9 +10539,9 @@
       }
     },
     "@jambonz/verb-specifications": {
-      "version": "0.0.81",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.81.tgz",
-      "integrity": "sha512-lufYMLpwaejF3zl8FNy1SvmsxFHXNXU8p7I+GoiwbgOHmFeOX4bQla4RD7bNLFvfjSXPRc2u9h5fceicVa9dcg==",
+      "version": "0.0.76",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.76.tgz",
+      "integrity": "sha512-7s61qAsG07xLLaEAHW236rSYzEoh9Qg0aRWHPbTfxCsuTKDNeq+5EwGAShDU5R5ZpjgweZJLhArQm8Ym+4xJ2A==",
       "requires": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"
@@ -12253,9 +12255,9 @@
       }
     },
     "drachtio-fsmrf": {
-      "version": "3.0.45",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.45.tgz",
-      "integrity": "sha512-YsrevTwGvg9v0OQXB4gZedlmzegB83fyd3NX3X2x7NXsKa5jO6TZCvZVHtBf/jR/ELE3B0aVLpxHw2YviRMIuQ==",
+      "version": "3.0.43",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.43.tgz",
+      "integrity": "sha512-ZDgO4WCzZssmOpf+ng20P8r5pu7demWFiECqwRivbWadXqAm3LNabZwWOred2MdCRDvzgHWUwBZoJPzCqKBL5g==",
       "requires": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",
@@ -16054,9 +16056,9 @@
       }
     },
     "undici": {
-      "version": "6.19.4",
-      "resolved": "https://registry.npmjs.org/undici/-/undici-6.19.4.tgz",
-      "integrity": "sha512-i3uaEUwNdkRq2qtTRRJb13moW5HWqviu7Vl7oYRYz++uPtGHJj+x7TGjcEuwS5Mt2P4nA0U9dhIX3DdB6JGY0g=="
+      "version": "6.19.2",
+      "resolved": "https://registry.npmjs.org/undici/-/undici-6.19.2.tgz",
+      "integrity": "sha512-JfjKqIauur3Q6biAtHJ564e3bWa8VvT+7cSiOJHFbX4Erv6CLGDpg8z+Fmg/1OI/47RA+GI2QZaF48SSaLvyBA=="
     },
     "undici-types": {
       "version": "5.26.5",
@@ -16327,9 +16329,9 @@
       }
     },
     "ws": {
-      "version": "8.18.0",
-      "resolved": "https://registry.npmjs.org/ws/-/ws-8.18.0.tgz",
-      "integrity": "sha512-8VbfWfHLbbwu3+N6OKsOMpBdT4kXPDDB9cJk2bJ6mh9ucxdlnNvH1e+roYkKmN9Nxw2yjz7VzeO9oOz2zJ04Pw==",
+      "version": "8.17.1",
+      "resolved": "https://registry.npmjs.org/ws/-/ws-8.17.1.tgz",
+      "integrity": "sha512-6XQFvXTkbfUOZOKKILFG1PDK2NDQs4azKQl26T0YS5CxqWLgXajbPZ+h4gZekJyRqFU8pvnbAbbs/3TgRPy+GQ==",
       "requires": {}
     },
     "xml2js": {

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jambonz-feature-server",
-  "version": "0.9.2",
+  "version": "0.9.0",
   "main": "app.js",
   "engines": {
     "node": ">= 18.x"
@@ -31,10 +31,10 @@
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.8",
-    "@jambonz/speech-utils": "^0.1.16",
+    "@jambonz/speech-utils": "^0.1.13",
     "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.9",
-    "@jambonz/verb-specifications": "^0.0.81",
+    "@jambonz/verb-specifications": "^0.0.76",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.50.0",
@@ -47,7 +47,7 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.45",
+    "drachtio-fsmrf": "^3.0.43",
     "drachtio-srf": "^4.5.35",
     "express": "^4.19.2",
     "express-validator": "^7.0.1",
@@ -61,10 +61,10 @@
     "short-uuid": "^5.1.0",
     "sinon": "^17.0.1",
     "to-snake-case": "^1.0.0",
-    "undici": "^6.19.4",
+    "undici": "^6.19.2",
     "uuid-random": "^1.3.2",
     "verify-aws-sns-signature": "^0.1.0",
-    "ws": "^8.18.0",
+    "ws": "^8.17.1",
     "xml2js": "^0.6.2"
   },
   "devDependencies": {

test/docker-compose-testbed.yaml

@@ -57,7 +57,7 @@ services:
         condition: service_healthy
 
   freeswitch:
-    image: drachtio/drachtio-freeswitch-mrf:latest
+    image: drachtio/drachtio-freeswitch-mrf:0.7.3
     restart: always
     command: freeswitch --rtp-range-start 20000 --rtp-range-end 20100
     environment: