Feat/dialogflow cx (#7) (#1516)

* wip

* wip

* wip

* wip

* logging

* wip

* wip

* wip

* update docker env to latest freeswitch

* wip

* lint

* wip

* support dialogflowcx

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

---------

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

lib/session/call-session.js

@@ -1082,7 +1082,7 @@ class CallSession extends Emitter {
             const cred = JSON.parse(credential.service_key.replace(/\n/g, '\\n'));
             return {
               speech_credential_sid: credential.speech_credential_sid,
-              credentials: cred
+              credentials: cred,
             };
           } catch (err) {
             const sid = this.accountInfo.account.account_sid;
@@ -2028,7 +2028,7 @@ Duration=${duration} `
         return this._lccDub(opts.dub, callSid);
       }
       else if (opts.boostAudioSignal) {
-        return this._lccBoostAudioSignal(opts, callSid);
+        return this._lccBoostAudioSignal(opts.boostAudioSignal, callSid);
       }
       else if (opts.media_path) {
         return this._lccMediaPath(opts.media_path, callSid);

lib/tasks/dial.js

@@ -195,6 +195,9 @@ class TaskDial extends Task {
   async exec(cs) {
     await super.exec(cs);
 
+    /* capture whether A leg was already answered before this dial task started */
+    this._aLegAlreadyAnswered = !!cs.dlg;
+
     if (this.data.anchorMedia && this.data.exitMediaPath) {
       this.logger.info('Dial:exec - incompatible anchorMedia and exitMediaPath are both set, will obey anchorMedia');
       delete this.data.exitMediaPath;
@@ -550,7 +553,7 @@ class TaskDial extends Task {
     let sbcAddress = this.proxy || getSBC();
     const teamsInfo = {};
     let fqdn;
-    const forwardPAI = this.forwardPAI ?? JAMBONZ_DIAL_PAI_HEADER; // dial verb overides env var
+    const forwardPAI = this.forwardPAI ?? !JAMBONZ_DIAL_PAI_HEADER; // dial verb overides env var
     this.logger.debug(forwardPAI, 'forwardPAI value');
     if (!sbcAddress) throw new Error('no SBC found for outbound call');
     this.headers = {
@@ -872,8 +875,12 @@ class TaskDial extends Task {
     this.sd = sd;
     this.callSid = sd.callSid;
     if (this.earlyMedia) {
-      debug('Dial:_selectSingleDial propagating answer supervision on A leg now that B is connected');
+      if (this._aLegAlreadyAnswered) {
-      await cs.propagateAnswer();
+        debug('Dial:_selectSingleDial A leg was already answered, skipping propagateAnswer');
+      } else {
+        debug('Dial:_selectSingleDial propagating answer supervision on A leg now that B is connected');
+        await cs.propagateAnswer();
+      }
     }
     if (this.timeLimit) {
       this.timerMaxCallDuration = setTimeout(this._onMaxCallDuration.bind(this, cs), this.timeLimit * 1000);

lib/tasks/dialogflow/index.js

@@ -1,3 +1,4 @@
+const assert = require('assert');
 const Task = require('../task');
 const {TaskName, TaskPreconditions} = require('../../utils/constants');
 const Intent = require('./intent');
@@ -10,19 +11,27 @@ class Dialogflow extends Task {
     super(logger, opts);
     this.preconditions = TaskPreconditions.Endpoint;
     this.credentials = this.data.credentials;
+    this.project = this.data.project;
+    this.agent = this.data.agent;
+    this.region = this.data.region || 'us-central1';
+    this.model = this.data.model || 'es';
 
-    /* set project id with environment and region (optionally) */
+    assert(this.agent || !this.isCX, 'agent is required for dialogflow cx');
-    if (this.data.environment && this.data.region) {
+    assert(this.credentials, 'dialogflow credentials are required');
-      this.project = `${this.data.project}:${this.data.environment}:${this.data.region}`;
+
-    }
+    if (this.isCX) {
-    else if (this.data.environment) {
+      this.environment = this.data.environment || 'none';
-      this.project = `${this.data.project}:${this.data.environment}`;
-    }
-    else if (this.data.region) {
-      this.project = `${this.data.project}::${this.data.region}`;
     }
     else {
-      this.project = this.data.project;
+      if (this.data.environment && this.data.region) {
+        this.project = `${this.data.project}:${this.data.environment}:${this.data.region}`;
+      }
+      else if (this.data.environment) {
+        this.project = `${this.data.project}:${this.data.environment}`;
+      }
+      else if (this.data.region) {
+        this.project = `${this.data.project}::${this.data.region}`;
+      }
     }
 
     this.lang = this.data.lang || 'en-US';
@@ -39,7 +48,6 @@ class Dialogflow extends Task {
       this.events = this.data.events;
     }
     else if (this.eventHook) {
-      // send all events by default - except interim transcripts
       this.events = [
         'intent',
         'transcription',
@@ -60,38 +68,33 @@ class Dialogflow extends Task {
       this.voice = this.data.tts.voice || 'default';
       this.speechSynthesisLabel = this.data.tts.label;
 
-      // fallback tts
       this.fallbackVendor = this.data.tts.fallbackVendor || 'default';
       this.fallbackLanguage = this.data.tts.fallbackLanguage || 'default';
-      this.fallbackVoice = this.data.tts.fallbackLanguage || 'default';
+      this.fallbackVoice = this.data.tts.fallbackVoice || 'default';
       this.fallbackLabel = this.data.tts.fallbackLabel;
     }
     this.bargein = this.data.bargein;
+
+    this.cmd = this.isCX ? 'dialogflow_cx_start' : 'dialogflow_start';
+    this.cmdStop = this.isCX ? 'dialogflow_cx_stop' : 'dialogflow_stop';
+
+    // CX-specific state
+    this._suppressNextCXAudio = false;
+    this._cxAudioHandled = false;
   }
 
   get name() { return TaskName.Dialogflow; }
 
+  get isCX() { return this.model === 'cx'; }
+
+  get isES() { return !this.isCX; }
+
   async exec(cs, {ep}) {
     await super.exec(cs);
 
     try {
       await this.init(cs, ep);
-
+      await this.startBot('default');
-      this.logger.debug(`starting dialogflow bot ${this.project}`);
-
-      // kick it off
-      const baseArgs = `${this.ep.uuid} ${this.project} ${this.lang} ${this.welcomeEvent}`;
-      if (this.welcomeEventParams) {
-        this.ep.api('dialogflow_start', `${baseArgs} '${JSON.stringify(this.welcomeEventParams)}'`);
-      }
-      else if (this.welcomeEvent.length) {
-        this.ep.api('dialogflow_start', baseArgs);
-      }
-      else {
-        this.ep.api('dialogflow_start', `${this.ep.uuid} ${this.project} ${this.lang}`);
-      }
-      this.logger.debug(`started dialogflow bot ${this.project}`);
-
       await this.awaitTaskDone();
     } catch (err) {
       this.logger.error({err}, 'Dialogflow:exec error');
@@ -108,6 +111,12 @@ class Dialogflow extends Task {
       this.ep.removeCustomEventListener('dialogflow::end_of_utterance');
       this.ep.removeCustomEventListener('dialogflow::error');
 
+      this.ep.removeCustomEventListener('dialogflow_cx::intent');
+      this.ep.removeCustomEventListener('dialogflow_cx::transcription');
+      this.ep.removeCustomEventListener('dialogflow_cx::audio_provided');
+      this.ep.removeCustomEventListener('dialogflow_cx::end_of_utterance');
+      this.ep.removeCustomEventListener('dialogflow_cx::error');
+
       this._clearNoinputTimer();
 
       if (!this.reportedFinalAction) this.performAction({dialogflowResult: 'caller hungup'})
@@ -141,6 +150,12 @@ class Dialogflow extends Task {
       this.ep.addCustomEventListener('dialogflow::end_of_utterance', this._onEndOfUtterance.bind(this, ep, cs));
       this.ep.addCustomEventListener('dialogflow::error', this._onError.bind(this, ep, cs));
 
+      this.ep.addCustomEventListener('dialogflow_cx::intent', this._onIntent.bind(this, ep, cs));
+      this.ep.addCustomEventListener('dialogflow_cx::transcription', this._onTranscription.bind(this, ep, cs));
+      this.ep.addCustomEventListener('dialogflow_cx::audio_provided', this._onAudioProvided.bind(this, ep, cs));
+      this.ep.addCustomEventListener('dialogflow_cx::end_of_utterance', this._onEndOfUtterance.bind(this, ep, cs));
+      this.ep.addCustomEventListener('dialogflow_cx::error', this._onError.bind(this, ep, cs));
+
       const obj = typeof this.credentials === 'string' ? JSON.parse(this.credentials) : this.credentials;
       const creds = JSON.stringify(obj);
       await this.ep.set('GOOGLE_APPLICATION_CREDENTIALS', creds);
@@ -151,56 +166,113 @@ class Dialogflow extends Task {
     }
   }
 
+  async startBot(intent) {
+    if (this.isCX) {
+      const event = this.welcomeEvent || intent;
+      const args = this._buildStartArgs({
+        event: event && event !== 'default' ? event : undefined
+      });
+      this.logger.info({args}, 'starting dialogflow CX bot');
+      await this.ep.api(this.cmd, args);
+    }
+    else {
+      await this._startBotES();
+    }
+  }
+
+  async _startBotES() {
+    this.logger.info('starting dialogflow ES bot');
+    const baseArgs = `${this.ep.uuid} ${this.project} ${this.lang} ${this.welcomeEvent}`;
+    if (this.welcomeEventParams) {
+      await this.ep.api(this.cmd, `${baseArgs} '${JSON.stringify(this.welcomeEventParams)}'`);
+    }
+    else if (this.welcomeEvent.length) {
+      await this.ep.api(this.cmd, baseArgs);
+    }
+    else {
+      await this.ep.api(this.cmd, `${this.ep.uuid} ${this.project} ${this.lang}`);
+    }
+  }
+
+  /**
+   * Build the start command args string for either ES or CX.
+   * @param {object} opts - options
+   * @param {string} opts.event - optional event to send
+   * @param {string} opts.text - optional text to send
+   * @param {number} opts.singleUtterance - 1 or 0 (CX only, default 1)
+   * @returns {string} command args string
+   */
+  _buildStartArgs({event, text, singleUtterance = 1} = {}) {
+    if (this.isCX) {
+      const args = [
+        this.ep.uuid,
+        this.project,
+        this.region,
+        this.agent,
+        this.environment || 'none',
+        this.lang,
+        event || 'none',
+        text ? `'${text}'` : 'none',
+        singleUtterance ? '1' : '0',
+      ];
+      return args.join(' ');
+    }
+    // ES
+    const args = [this.ep.uuid, this.project, this.lang];
+    if (event) {
+      args.push(event);
+    }
+    if (text) {
+      if (!event) args.push('none');
+      args.push(`'${text}'`);
+    }
+    return args.join(' ');
+  }
+
   /**
    * An intent has been returned.  Since we are using SINGLE_UTTERANCE on the dialogflow side,
    * we may get an empty intent, signified by the lack of a 'response_id' attribute.
    * In such a case, we just start another StreamingIntentDetectionRequest.
    * @param {*} ep -  media server endpoint
+   * @param {*} cs - call session
    * @param {*} evt - event data
    */
   async _onIntent(ep, cs, evt) {
     const intent = new Intent(this.logger, evt);
 
     if (intent.isEmpty) {
-      /**
-       * An empty intent is returned in 3 conditions:
-       * 1. Our no-input timer fired
-       * 2. We collected dtmf that needs to be fed to dialogflow
-       * 3. A normal dialogflow timeout
-       */
       if (this.noinput && this.greetingPlayed) {
         this.logger.info('no input timer fired, reprompting..');
         this.noinput = false;
-        ep.api('dialogflow_start', `${ep.uuid} ${this.project} ${this.lang} ${this.noInputEvent}`);
+        ep.api(this.cmd, this._buildStartArgs({event: this.noInputEvent}));
       }
       else if (this.dtmfEntry && this.greetingPlayed) {
         this.logger.info('dtmf detected, reprompting..');
-        ep.api('dialogflow_start', `${ep.uuid} ${this.project} ${this.lang} none \'${this.dtmfEntry}\'`);
+        ep.api(this.cmd, this._buildStartArgs({text: this.dtmfEntry}));
         this.dtmfEntry = null;
       }
-      else if (this.greetingPlayed) {
-        this.logger.info('starting another intent');
-        ep.api('dialogflow_start', `${ep.uuid} ${this.project} ${this.lang}`);
-      }
       else {
-        this.logger.info('got empty intent');
+        this.logger.info('got empty intent, restarting');
-        ep.api('dialogflow_start', `${ep.uuid} ${this.project} ${this.lang}`);
+        ep.api(this.cmd, this._buildStartArgs());
       }
       return;
     }
 
+    // For CX: suppress NO_INPUT "I didn't get that" audio and silently restart
+    if (this.isCX && intent.isNoInput && this.greetingPlayed) {
+      this.logger.info('CX returned NO_INPUT after greeting, suppressing and restarting');
+      this._suppressNextCXAudio = true;
+      return;
+    }
+
     if (this.events.includes('intent')) {
       this._performHook(cs, this.eventHook, {event: 'intent', data: evt});
     }
 
-    // clear the no-input timer and the digit buffer
     this._clearNoinputTimer();
     if (this.digitBuffer) this.digitBuffer.flush();
 
-    /* hang up (or tranfer call) after playing next audio file? */
     if (intent.saysEndInteraction) {
-      //  if 'end_interaction' is true, end the dialog after playing the final prompt
-      //  (or in 1 second if there is no final prompt)
       this.hangupAfterPlayDone = true;
       this.waitingForPlayStart = true;
       setTimeout(() => {
@@ -211,8 +283,6 @@ class Dialogflow extends Task {
         }
       }, 1000);
     }
-
-    /* collect digits? */
     else if (intent.saysCollectDtmf || this.enableDtmfAlways) {
       const opts = Object.assign({
         idt: this.opts.interDigitTimeout
@@ -221,68 +291,44 @@ class Dialogflow extends Task {
       this.digitBuffer.once('fulfilled', this._onDtmfEntryComplete.bind(this, ep));
     }
 
-    /* if we are using tts and a message was provided, play it out */
+    // If we have a TTS vendor and fulfillment text, synthesize and play
     if (this.vendor && intent.fulfillmentText && intent.fulfillmentText.length > 0) {
-      const {srf} = cs;
-      const {stats} = srf.locals;
-      const {synthAudio} = srf.locals.dbHelpers;
       this.waitingForPlayStart = false;
 
-      // start a new intent, (we want to continue to listen during the audio playback)
+      // ES: start a new intent during playback so we continue to listen
-      // _unless_ we are transferring or ending the session
+      if (!this.hangupAfterPlayDone && this.isES) {
-      if (!this.hangupAfterPlayDone) {
+        ep.api(this.cmd, this._buildStartArgs());
-        ep.api('dialogflow_start', `${ep.uuid} ${this.project} ${this.lang}`);
       }
 
       try {
+        const {srf} = cs;
+        const {stats} = srf.locals;
+        const {synthAudio} = srf.locals.dbHelpers;
         const {filePath} = await this._fallbackSynthAudio(cs, intent, stats, synthAudio);
         if (filePath) cs.trackTmpFile(filePath);
-
+        await this._playAndHandlePostPlay(ep, cs, filePath);
-        if (this.playInProgress) {
-          await ep.api('uuid_break', ep.uuid).catch((err) => this.logger.info(err, 'Error killing audio'));
-        }
-        this.playInProgress = true;
-        this.curentAudioFile = filePath;
-
-        this.logger.debug(`starting to play tts ${filePath}`);
-
-        if (this.events.includes('start-play')) {
-          this._performHook(cs, this.eventHook, {event: 'start-play', data: {path: filePath}});
-        }
-        await ep.play(filePath);
-        if (this.events.includes('stop-play')) {
-          this._performHook(cs, this.eventHook, {event: 'stop-play', data: {path: filePath}});
-        }
-        this.logger.debug(`finished ${filePath}`);
-
-        if (this.curentAudioFile === filePath) {
-          this.playInProgress = false;
-          if (this.queuedTasks) {
-            this.logger.debug('finished playing audio and we have queued tasks');
-            this._redirect(cs, this.queuedTasks);
-            return;
-          }
-        }
-        this.greetingPlayed = true;
-
-        if (this.hangupAfterPlayDone) {
-          this.logger.info('hanging up since intent was marked end interaction and we completed final prompt');
-          this.performAction({dialogflowResult: 'completed'});
-          this.notifyTaskDone();
-        }
-        else {
-          // every time we finish playing a prompt, start the no-input timer
-          this._startNoinputTimer(ep, cs);
-        }
       } catch (err) {
         this.logger.error({err}, 'Dialogflow:_onIntent - error playing tts');
       }
     }
+    else if (this.isCX && !this.hangupAfterPlayDone) {
+      // CX intent with no TTS — _onAudioProvided may handle playback.
+      // If not, restart CX after a short delay.
+      this.greetingPlayed = true;
+      this._cxAudioHandled = false;
+      setTimeout(() => {
+        if (!this._cxAudioHandled && !this.playInProgress) {
+          this.logger.info('CX: no TTS and no audio provided, restarting to listen');
+          ep.api(this.cmd, this._buildStartArgs());
+          this._startNoinputTimer(ep, cs);
+        }
+      }, 500);
+    }
   }
 
   async _fallbackSynthAudio(cs, intent, stats, synthAudio) {
     try {
-      const obj = {
+      return await synthAudio(stats, {
         account_sid: cs.accountSid,
         text: intent.fulfillmentText,
         vendor: this.vendor,
@@ -290,17 +336,13 @@ class Dialogflow extends Task {
         voice: this.voice,
         salt: cs.callSid,
         credentials: this.ttsCredentials
-      };
+      });
-      this.logger.debug({obj}, 'Dialogflow:_onIntent - playing message via tts');
-
-      return await synthAudio(stats, obj);
     } catch (error) {
       this.logger.info({error}, 'Failed to synthesize audio from primary vendor');
-
+      if (this.fallbackVendor) {
-      try {
+        try {
-        if (this.fallbackVendor) {
           const credentials = cs.getSpeechCredentials(this.fallbackVendor, 'tts', this.fallbackLabel);
-          const obj = {
+          return await synthAudio(stats, {
             account_sid: cs.accountSid,
             text: intent.fulfillmentText,
             vendor: this.fallbackVendor,
@@ -308,24 +350,20 @@ class Dialogflow extends Task {
             voice: this.fallbackVoice,
             salt: cs.callSid,
             credentials
-          };
+          });
-          this.logger.debug({obj}, 'Dialogflow:_onIntent - playing message via fallback tts');
+        } catch (err) {
-          return await synthAudio(stats, obj);
+          this.logger.info({err}, 'Failed to synthesize audio from fallback vendor');
+          throw err;
         }
-      } catch (err) {
-        this.logger.info({err}, 'Failed to synthesize audio from falllback vendor');
-        throw err;
       }
       throw error;
     }
   }
 
   /**
-   * A transcription - either interim or final - has been returned.
+   * A transcription has been returned.
-   * If we are doing barge-in based on hotword detection, check for the hotword or phrase.
+   * @param {*} ep -  media server endpoint
-   * If we are playing a filler sound, like typing, during the fullfillment phase, start that
+   * @param {*} cs - call session
-   * if this is a final transcript.
-   * @param {*} ep  -  media server endpoint
    * @param {*} evt - event data
    */
   async _onTranscription(ep, cs, evt) {
@@ -338,13 +376,11 @@ class Dialogflow extends Task {
       this._performHook(cs, this.eventHook, {event: 'transcription', data: evt});
     }
 
-    // if a final transcription, start a typing sound
     if (this.thinkingMusic && !transcription.isEmpty && transcription.isFinal &&
       transcription.confidence > 0.8) {
       ep.play(this.data.thinkingMusic).catch((err) => this.logger.info(err, 'Error playing typing sound'));
     }
 
-    // interrupt playback on speaking if bargein = true
     if (this.bargein && this.playInProgress) {
       this.logger.debug('terminating playback due to speech bargein');
       this.playInProgress = false;
@@ -353,17 +389,21 @@ class Dialogflow extends Task {
   }
 
   /**
-   * The caller has just finished speaking.  No action currently taken.
+   * The caller has just finished speaking.
+   * @param {*} ep -  media server endpoint
+   * @param {*} cs - call session
    * @param {*} evt - event data
    */
-  _onEndOfUtterance(cs, evt) {
+  _onEndOfUtterance(ep, cs, evt) {
     if (this.events.includes('end-utterance')) {
       this._performHook(cs, this.eventHook, {event: 'end-utterance'});
     }
   }
 
   /**
-   * Dialogflow has returned an error of some kind.
+   * Dialogflow has returned an error.
+   * @param {*} ep -  media server endpoint
+   * @param {*} cs - call session
    * @param {*} evt - event data
    */
   _onError(ep, cs, evt) {
@@ -372,70 +412,87 @@ class Dialogflow extends Task {
 
   /**
    * Audio has been received from dialogflow and written to a temporary disk file.
-   * Start playing the audio, after killing any filler sound that might be playing.
+   * Play the audio, then restart or hang up as appropriate.
-   * When the audio completes, start the no-input timer.
    * @param {*} ep -  media server endpoint
+   * @param {*} cs - call session
    * @param {*} evt - event data
    */
   async _onAudioProvided(ep, cs, evt) {
-
+    // For CX: suppress NO_INPUT reprompt audio and silently restart
-    if (this.vendor) return;
+    if (this._suppressNextCXAudio) {
-
+      this._suppressNextCXAudio = false;
-    this.waitingForPlayStart = false;
+      ep.api(this.cmd, this._buildStartArgs());
-
+      return;
-    // kill filler audio
-    await ep.api('uuid_break', ep.uuid);
-
-    // start a new intent, (we want to continue to listen during the audio playback)
-    // _unless_ we are transferring or ending the session
-    if (/*this.greetingPlayed &&*/ !this.hangupAfterPlayDone) {
-      ep.api('dialogflow_start', `${ep.uuid} ${this.project} ${this.lang}`);
     }
 
-    this.playInProgress = true;
+    if (this.vendor) {
-    this.curentAudioFile = evt.path;
+      if (this.isCX && !this.playInProgress) {
-
+        // CX audio arrived but TTS didn't play — fall through to use CX audio
-    this.logger.info(`starting to play ${evt.path}`);
+        this.logger.info('CX audio provided, TTS vendor did not play - using CX audio');
-    if (this.events.includes('start-play')) {
+      } else {
-      this._performHook(cs, this.eventHook, {event: 'start-play', data: {path: evt.path}});
-    }
-    await ep.play(evt.path);
-    if (this.events.includes('stop-play')) {
-      this._performHook(cs, this.eventHook, {event: 'stop-play', data: {path: evt.path}});
-    }
-    this.logger.info(`finished ${evt.path}, queued tasks: ${(this.queuedTasks || []).length}`);
-
-    if (this.curentAudioFile === evt.path) {
-      this.playInProgress = false;
-      if (this.queuedTasks) {
-        this.logger.debug('finished playing audio and we have queued tasks');
-        this._redirect(cs, this.queuedTasks);
-        this.queuedTasks.length = 0;
         return;
       }
     }
-    /*
+
-    if (!this.inbound && !this.greetingPlayed) {
+    this._cxAudioHandled = true;
-      this.logger.info('finished greeting on outbound call, starting new intent');
+    this.waitingForPlayStart = false;
-      this.ep.api('dialogflow_start', `${ep.uuid} ${this.project} ${this.lang}`);
+
+    await ep.api('uuid_break', ep.uuid);
+
+    // ES: start a new intent during playback so we continue to listen
+    if (!this.hangupAfterPlayDone && this.isES) {
+      ep.api(this.cmd, this._buildStartArgs());
     }
-    */
+
+    await this._playAndHandlePostPlay(ep, cs, evt.path);
+  }
+
+  /**
+   * Shared post-play logic for both TTS (_onIntent) and CX audio (_onAudioProvided).
+   * Plays audio, then either hangs up, redirects, or restarts the dialog.
+   */
+  async _playAndHandlePostPlay(ep, cs, filePath) {
+    if (this.playInProgress) {
+      await ep.api('uuid_break', ep.uuid).catch((err) => this.logger.info(err, 'Error killing audio'));
+    }
+    this.playInProgress = true;
+    this.curentAudioFile = filePath;
+
+    if (this.events.includes('start-play')) {
+      this._performHook(cs, this.eventHook, {event: 'start-play', data: {path: filePath}});
+    }
+    await ep.play(filePath);
+    if (this.events.includes('stop-play')) {
+      this._performHook(cs, this.eventHook, {event: 'stop-play', data: {path: filePath}});
+    }
+
+    if (this.curentAudioFile === filePath) {
+      this.playInProgress = false;
+      if (this.queuedTasks) {
+        this._redirect(cs, this.queuedTasks);
+        this.queuedTasks = null;
+        return;
+      }
+    }
+
     this.greetingPlayed = true;
 
     if (this.hangupAfterPlayDone) {
-      this.logger.info('hanging up since intent was marked end interaction and we completed final prompt');
+      this.logger.info('hanging up after end interaction prompt');
       this.performAction({dialogflowResult: 'completed'});
       this.notifyTaskDone();
     }
     else {
-      // every time we finish playing a prompt, start the no-input timer
+      // CX: restart to listen for the next utterance
+      if (this.isCX) {
+        ep.api(this.cmd, this._buildStartArgs());
+      }
       this._startNoinputTimer(ep, cs);
     }
   }
 
   /**
-   * receive a dmtf entry from the caller.
+   * Receive a DTMF entry from the caller.
-   * If we have active dtmf instructions, collect and process accordingly.
    */
   _onDtmf(ep, cs, evt) {
     if (this.digitBuffer) this.digitBuffer.process(evt.dtmf);
@@ -444,41 +501,48 @@ class Dialogflow extends Task {
     }
   }
 
-  _onDtmfEntryComplete(ep, dtmfEntry) {
+  async _onDtmfEntryComplete(ep, dtmfEntry) {
     this.logger.info(`collected dtmf entry: ${dtmfEntry}`);
-    this.dtmfEntry = dtmfEntry;
     this.digitBuffer = null;
-    // if a final transcription, start a typing sound
     if (this.thinkingMusic) {
       ep.play(this.thinkingMusic).catch((err) => this.logger.info(err, 'Error playing typing sound'));
     }
 
-    // kill the current dialogflow, which will result in us getting an immediate intent
+    if (this.isCX) {
-    ep.api('dialogflow_stop', `${ep.uuid}`)
+      try {
-      .catch((err) => this.logger.info(`dialogflow_stop failed: ${err.message}`));
+        await ep.api(this.cmdStop, ep.uuid);
+      } catch (err) {
+        this.logger.info(err, 'dialogflow_cx_stop failed');
+      }
+      ep.api(this.cmd, this._buildStartArgs({text: dtmfEntry}));
+    } else {
+      this.dtmfEntry = dtmfEntry;
+      ep.api(this.cmdStop, `${ep.uuid}`)
+        .catch((err) => this.logger.info(`dialogflow_stop failed: ${err.message}`));
+    }
   }
 
-  /**
+  async _onNoInput(ep, cs) {
-   * The user has not provided any input for some time.
+    this.logger.info('no-input timer fired');
-   * Set the 'noinput' member to true and kill the current dialogflow.
-   * This will result in us re-prompting with an event indicating no input.
-   * @param {*} ep
-   */
-  _onNoInput(ep, cs) {
-    this.noinput = true;
 
     if (this.events.includes('no-input')) {
-      this._performHook(cs, this.eventHook,  {event: 'no-input'});
+      this._performHook(cs, this.eventHook, {event: 'no-input'});
     }
 
-    // kill the current dialogflow, which will result in us getting an immediate intent
+    if (this.isCX) {
-    ep.api('dialogflow_stop', `${ep.uuid}`)
+      try {
-      .catch((err) => this.logger.info(`dialogflow_stop failed: ${err.message}`));
+        await ep.api(this.cmdStop, ep.uuid);
+      } catch (err) {
+        this.logger.info(err, 'dialogflow_cx_stop failed');
+      }
+      ep.api(this.cmd, this._buildStartArgs({event: this.noInputEvent}));
+    } else {
+      this.noinput = true;
+      ep.api(this.cmdStop, `${ep.uuid}`)
+        .catch((err) => this.logger.info(`dialogflow_stop failed: ${err.message}`));
+    }
   }
 
-  /**
-   * Stop the no-input timer, if it is running
-   */
   _clearNoinputTimer() {
     if (this.noinputTimer) {
       clearTimeout(this.noinputTimer);
@@ -486,10 +550,6 @@ class Dialogflow extends Task {
     }
   }
 
-  /**
-   * Start the no-input timer.  The duration is set in the configuration file.
-   * @param {*} ep
-   */
   _startNoinputTimer(ep, cs) {
     if (!this.noInputTimeout) return;
     this._clearNoinputTimer();
@@ -507,7 +567,7 @@ class Dialogflow extends Task {
       if (tasks && tasks.length > 0) {
         if (this.playInProgress) {
           this.queuedTasks = tasks;
-          this.logger.info({tasks: tasks},
+          this.logger.info({tasks},
             `${this.name} replacing application with ${tasks.length} tasks after play completes`);
           return;
         }
@@ -517,7 +577,7 @@ class Dialogflow extends Task {
   }
 
   _redirect(cs, tasks) {
-    this.logger.info({tasks: tasks}, `${this.name} replacing application with ${tasks.length} tasks`);
+    this.logger.info({tasks}, `${this.name} replacing application with ${tasks.length} tasks`);
     this.performAction({dialogflowResult: 'redirect'}, false);
     this.reportedFinalAction = true;
     cs.replaceApplication(tasks);

lib/tasks/dialogflow/intent.js

@@ -3,20 +3,44 @@ class Intent {
     this.logger = logger;
     this.evt = evt;
 
-    this.logger.debug({evt}, 'intent');
+    this.qr = this.isCX ? evt.detect_intent_response.query_result : evt.query_result;
-    this.dtmfRequest = checkIntentForDtmfEntry(logger, evt);
+    this.dtmfRequest = this._checkIntentForDtmfEntry();
+  }
+
+  get response_id() {
+    return this.isCX ? this.evt.detect_intent_response.response_id : this.evt.response_id;
   }
 
   get isEmpty() {
-    return this.evt.response_id.length === 0;
+    return !(this.response_id?.length > 0);
   }
 
   get fulfillmentText() {
-    return this.evt.query_result.fulfillment_text;
+    if (this.isCX) {
+      if (this.qr && this.qr.response_messages) {
+        for (const msg of this.qr.response_messages) {
+          if (msg.text && msg.text.text && msg.text.text.length > 0) {
+            return msg.text.text.join('\n');
+          }
+          if (msg.output_audio_text) {
+            if (msg.output_audio_text.text) return msg.output_audio_text.text;
+            if (msg.output_audio_text.ssml) return msg.output_audio_text.ssml;
+          }
+        }
+      }
+      return undefined;
+    }
+    return this.qr.fulfillment_text;
   }
 
   get saysEndInteraction() {
-    return this.evt.query_result.intent.end_interaction ;
+    if (this.isCX) {
+      if (!this.qr || !this.qr.response_messages) return false;
+      const end_interaction = this.qr.response_messages
+        .find((m) => typeof m === 'object' && 'end_interaction' in m)?.end_interaction;
+      return end_interaction && Object.keys(end_interaction).length > 0;
+    }
+    return this.qr.intent.end_interaction;
   }
 
   get saysCollectDtmf() {
@@ -28,7 +52,23 @@ class Intent {
   }
 
   get name() {
-    if (!this.isEmpty) return this.evt.query_result.intent.display_name;
+    if (!this.isEmpty) {
+      if (this.isCX) {
+        return this.qr.match?.intent?.display_name;
+      }
+      return this.qr.intent.display_name;
+    }
+  }
+
+  get isCX() {
+    return typeof this.evt.detect_intent_response === 'object';
+  }
+
+  get isNoInput() {
+    if (this.isCX && this.qr && this.qr.match) {
+      return this.qr.match.match_type === 'NO_INPUT';
+    }
+    return false;
   }
 
   toJSON() {
@@ -38,52 +78,48 @@ class Intent {
     };
   }
 
+  /**
+   * Parse a returned intent for DTMF entry information (ES only).
+   * CX does not use fulfillment_messages or output_contexts.
+   *
+   * allow-dtmf-x-y-z
+   * x = min number of digits
+   * y = optional, max number of digits
+   * z = optional, terminating character
+   */
+  _checkIntentForDtmfEntry() {
+    if (this.isCX) return;
+
+    const qr = this.qr;
+    if (!qr || !qr.fulfillment_messages || !qr.output_contexts) {
+      return;
+    }
+
+    // check for custom payloads with a gather verb
+    const custom = qr.fulfillment_messages.find((f) => f.payload && f.payload.verb === 'gather');
+    if (custom) {
+      this.logger.info({custom}, 'found dtmf custom payload');
+      return {
+        max: custom.payload.numDigits,
+        term: custom.payload.finishOnKey,
+        template: custom.payload.responseTemplate
+      };
+    }
+
+    // check for an output context with a specific naming convention
+    const context = qr.output_contexts.find((oc) => oc.name.includes('/contexts/allow-dtmf-'));
+    if (context) {
+      const arr = /allow-dtmf-(\d+)(?:-(\d+))?(?:-(.*))?/.exec(context.name);
+      if (arr) {
+        this.logger.info('found dtmf output context');
+        return {
+          min: parseInt(arr[1]),
+          max: arr.length > 2 ? parseInt(arr[2]) : null,
+          term: arr.length > 3 ? arr[3] : null
+        };
+      }
+    }
+  }
 }
 
 module.exports = Intent;
-
-/**
- * Parse a returned intent for DTMF entry information
- * i.e.
- * allow-dtmf-x-y-z
- * x = min number of digits
- * y = optional, max number of digits
- * z = optional, terminating character
- * e.g.
- * allow-dtmf-5 :     collect 5 digits
- * allow-dtmf-1-4 :   collect between 1 to 4 (inclusive) digits
- * allow-dtmf-1-4-# : collect 1-4 digits, terminating if '#' is entered
- * @param {*} intent - dialogflow intent
- */
-const checkIntentForDtmfEntry = (logger, intent) => {
-  const qr = intent.query_result;
-  if (!qr || !qr.fulfillment_messages || !qr.output_contexts) {
-    logger.info({f: qr.fulfillment_messages, o: qr.output_contexts}, 'no dtmfs');
-    return;
-  }
-
-  // check for custom payloads with a gather verb
-  const custom = qr.fulfillment_messages.find((f) => f.payload && f.payload.verb === 'gather');
-  if (custom && custom.payload && custom.payload.verb === 'gather') {
-    logger.info({custom}, 'found dtmf custom payload');
-    return {
-      max: custom.payload.numDigits,
-      term: custom.payload.finishOnKey,
-      template: custom.payload.responseTemplate
-    };
-  }
-
-  // check for an output context with a specific naming convention
-  const context = qr.output_contexts.find((oc) => oc.name.includes('/contexts/allow-dtmf-'));
-  if (context) {
-    const arr = /allow-dtmf-(\d+)(?:-(\d+))?(?:-(.*))?/.exec(context.name);
-    if (arr) {
-      logger.info({custom}, 'found dtmf output context');
-      return {
-        min: parseInt(arr[1]),
-        max: arr.length > 2 ? parseInt(arr[2]) : null,
-        term: arr.length > 3 ? arr[3] : null
-      };
-    }
-  }
-};

lib/tasks/listen.js

@@ -152,9 +152,17 @@ class TaskListen extends Task {
 
   async _startListening(cs, ep) {
     this._initListeners(ep);
-    const ci = this.nested ? this.parentTask.sd.callInfo : cs.callInfo.toJSON();
+    const tempci = this.nested ? this.parentTask.sd.callInfo : cs.callInfo.toJSON();
+    const ci = structuredClone(tempci);
     if (this._ignoreCustomerData) {
       delete ci.customerData;
+    } else {
+      for (const key in ci.customerData) {
+        if (ci.customerData.hasOwnProperty(key)) {
+          const value = ci.customerData[key];
+          ci.customerData[key] = typeof value === 'string' ? escapeString(value) : value;
+        }
+      }
     }
     const metadata = Object.assign(
       {sampleRate: this.sampleRate, mixType: this.mixType},

lib/tasks/llm/llms/google_s2s.js

@@ -36,6 +36,9 @@ class TaskLlmGoogle_S2S extends Task {
     this.model = this.parent.model || 'models/gemini-2.0-flash-live-001';
     this.auth = this.parent.auth;
     this.connectionOptions = this.parent.connectOptions;
+    const {host, version} = this.connectionOptions || {};
+    this.host = host;
+    this.version = version;
 
     const {apiKey} = this.auth || {};
     if (!apiKey) throw new Error('auth.apiKey is required for Google S2S');
@@ -46,7 +49,7 @@ class TaskLlmGoogle_S2S extends Task {
     this.eventHook = this.data.eventHook;
     this.toolHook = this.data.toolHook;
 
-    const {setup} = this.data.llmOptions;
+    const {setup, sessionResumption} = this.data.llmOptions;
 
     if (typeof setup !== 'object') {
       throw new Error('llmOptions with an initial setup is required for Google S2S');
@@ -54,6 +57,7 @@ class TaskLlmGoogle_S2S extends Task {
     this.setup = {
       ...setup,
       model: this.model,
+      ...(sessionResumption && {sessionResumption}),
       // make sure output is always audio
       generationConfig: {
         ...(setup.generationConfig || {}),
@@ -138,6 +142,10 @@ class TaskLlmGoogle_S2S extends Task {
 
     try {
       const args = [ep.uuid, 'session.create', this.apiKey];
+      if (this.host) {
+        args.push(this.host);
+        if (this.version) args.push(this.version);
+      }
       await this._api(ep, args);
     } catch (err) {
       this.logger.error({err}, 'TaskLlmGoogle_S2S:_startListening');

lib/tasks/redirect.js

@@ -1,7 +1,6 @@
 const Task = require('./task');
 const {TaskName} = require('../utils/constants');
 const WsRequestor = require('../utils/ws-requestor');
-const URL = require('url');
 const HttpRequestor = require('../utils/http-requestor');
 
 /**
@@ -10,6 +9,7 @@ const HttpRequestor = require('../utils/http-requestor');
 class TaskRedirect extends Task {
   constructor(logger, opts) {
     super(logger, opts);
+    this.statusHook = opts.statusHook || false;
   }
 
   get name() { return TaskName.Redirect; }
@@ -47,6 +47,30 @@ class TaskRedirect extends Task {
         }
       }
     }
+
+    /* update the notifier if a new statusHook was provided */
+    if (this.statusHook) {
+      this.logger.info(`TaskRedirect updating statusHook to ${this.statusHook}`);
+      try {
+        const oldNotifier = cs.application.notifier;
+        const isStatusHookAbsolute = cs.notifier?._isAbsoluteUrl(this.statusHook);
+        if (isStatusHookAbsolute) {
+          if (cs.notifier instanceof WsRequestor) {
+            cs.application.notifier = new WsRequestor(this.logger, cs.accountSid, {url: this.statusHook},
+              cs.accountInfo.account.webhook_secret);
+          } else {
+            cs.application.notifier = new HttpRequestor(this.logger, cs.accountSid, {url: this.statusHook},
+              cs.accountInfo.account.webhook_secret);
+          }
+          if (oldNotifier?.close) oldNotifier.close();
+        }
+        /* update the call_status_hook URL that gets passed to the notifier */
+        cs.application.call_status_hook = this.statusHook;
+      } catch (err) {
+        this.logger.info(err, `TaskRedirect error updating statusHook to ${this.statusHook}`);
+      }
+    }
+
     await this.performAction();
   }
 }

lib/tasks/tts-task.js

@@ -31,8 +31,9 @@ class TtsTask extends Task {
     this.synthesizer = this.data.synthesizer || {};
     this.disableTtsCache = this.data.disableTtsCache;
     this.options = this.synthesizer.options || {};
-    this.instructions = this.data.instructions;
+    this.instructions = this.data.instructions || this.options.instructions;
     this.playbackIds = [];
+    this.useGeminiTts = this.options.useGeminiTts;
   }
 
   getPlaybackId(offset) {
@@ -156,6 +157,13 @@ class TtsTask extends Task {
           ...(reduceLatency && {RIMELABS_TTS_STREAMING_REDUCE_LATENCY: reduceLatency})
         };
         break;
+      case 'google':
+        obj = {
+          GOOGLE_TTS_LANGUAGE_CODE: language,
+          GOOGLE_TTS_VOICE_NAME: voice,
+          GOOGLE_APPLICATION_CREDENTIALS: JSON.stringify(credentials.credentials)
+        };
+        break;
       default:
         if (vendor.startsWith('custom:')) {
           const use_tls = custom_tts_streaming_url.startsWith('wss://');
@@ -242,6 +250,8 @@ class TtsTask extends Task {
       }
     } else if (vendor === 'cartesia') {
       credentials.model_id = this.options.model_id || credentials.model_id;
+    } else if (vendor === 'google') {
+      this.model = this.options.model || credentials.credentials.model_id;
     }
 
     this.model_id = credentials.model_id;

lib/utils/action-hook-delay.js

@@ -118,6 +118,13 @@ class ActionHookDelayProcessor extends Emitter {
     this.logger.debug('ActionHookDelayProcessor#_onNoResponseTimer');
     this._noResponseTimer = null;
 
+    /* check if endpoint is still available (call may have ended) */
+    if (!this.ep) {
+      this.logger.debug('ActionHookDelayProcessor#_onNoResponseTimer: endpoint is null, call may have ended');
+      this._active = false;
+      return;
+    }
+
     /* get the next play or say action */
     const verb = this.actions[this._retryCount % this.actions.length];
 
@@ -129,8 +136,8 @@ class ActionHookDelayProcessor extends Emitter {
       this._taskInProgress.exec(this.cs, {ep: this.ep}).catch((err) => {
         this.logger.info(`ActionHookDelayProcessor#_onNoResponseTimer: error playing file: ${err.message}`);
         this._taskInProgress = null;
-        this.ep.removeAllListeners('playback-start');
+        this.ep?.removeAllListeners('playback-start');
-        this.ep.removeAllListeners('playback-stop');
+        this.ep?.removeAllListeners('playback-stop');
       });
     } catch (err) {
       this.logger.info(err, 'ActionHookDelayProcessor#_onNoResponseTimer: error starting action');

lib/utils/constants.json

@@ -311,6 +311,11 @@
     "ConnectFailure": "deepgram_tts_streaming::connect_failed",
     "Connect": "deepgram_tts_streaming::connect"
   },
+  "GoogleTtsStreamingEvents": {
+    "Empty": "google_tts_streaming::empty",
+    "ConnectFailure": "google_tts_streaming::connect_failed",
+    "Connect": "google_tts_streaming::connect"
+  },
   "CartesiaTtsStreamingEvents": {
     "Empty": "cartesia_tts_streaming::empty",
     "ConnectFailure": "cartesia_tts_streaming::connect_failed",

lib/utils/transcription-utils.js

@@ -1310,6 +1310,9 @@ module.exports = (logger) => {
           ...(openaiOptions.turn_detection.silence_duration_ms && {
             OPENAI_TURN_DETECTION_SILENCE_DURATION_MS: openaiOptions.turn_detection.silence_duration_ms
           }),
+          ...(openaiOptions.turn_detection.eagerness && {
+            OPENAI_TURN_DETECTION_EAGERNESS: openaiOptions.turn_detection.eagerness
+          })
         };
       }
     }
@@ -1375,7 +1378,9 @@ module.exports = (logger) => {
             speechmaticsOptions.transcription_config.audio_filtering_config.volume_threshold}),
         ...(speechmaticsOptions.transcription_config?.transcript_filtering_config?.remove_disfluencies &&
           {SPEECHMATICS_REMOVE_DISFLUENCIES:
-            speechmaticsOptions.transcription_config.transcript_filtering_config.remove_disfluencies})
+            speechmaticsOptions.transcription_config.transcript_filtering_config.remove_disfluencies}),
+        SPEECHMATICS_END_OF_UTTERANCE_SILENCE_TRIGGER:
+          speechmaticsOptions.transcription_config?.conversation_config?.end_of_utterance_silence_trigger || 0.5
       };
     }
     else if (vendor.startsWith('custom:')) {

lib/utils/tts-streaming-buffer.js

@@ -421,6 +421,7 @@ class TtsStreamingBuffer extends Emitter {
       'cartesia',
       'elevenlabs',
       'rimelabs',
+      'google',
       'custom'
     ].forEach((vendor) => {
       const eventClassName = `${vendor.charAt(0).toUpperCase() + vendor.slice(1)}TtsStreamingEvents`;

package.json

@@ -31,10 +31,10 @@
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.15",
-    "@jambonz/speech-utils": "^0.2.26",
+    "@jambonz/speech-utils": "^0.2.30",
     "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.15",
-    "@jambonz/verb-specifications": "^0.0.123",
+    "@jambonz/verb-specifications": "^0.0.125",
     "@modelcontextprotocol/sdk": "^1.9.0",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",
@@ -49,7 +49,7 @@
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
     "drachtio-fsmrf": "^4.1.2",
-    "drachtio-srf": "^5.0.14",
+    "drachtio-srf": "^5.0.18",
     "express": "^4.19.2",
     "express-validator": "^7.0.1",
     "moment": "^2.30.1",