wip (#830)

* wip

* wip

* wip

* wip

* update deps

* update test to use latest freeswitch image

.github/workflows/build.yml

@@ -12,6 +12,11 @@ jobs:
           node-version: 20
       - run: npm ci
       - run: npm run jslint
+      - name: Install Docker Compose
+        run: |
+          sudo curl -L "https://github.com/docker/compose/releases/download/1.29.2/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose
+          sudo chmod +x /usr/local/bin/docker-compose
+          docker-compose --version
       - run: docker pull drachtio/sipp
       - run: npm test
         env:

README.md

@@ -21,6 +21,7 @@ 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

@@ -73,6 +73,7 @@ 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;
@@ -170,6 +171,7 @@ 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

@@ -218,7 +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_hook: app.call_hook}, 'creating http client for call status hook');
+        logger.debug({call_status_hook: app.call_status_hook}, 'creating http client for call status hook');
       }
       else if (!app.notifier) {
         logger.debug('creating null call status hook');

lib/middleware.js

@@ -354,11 +354,13 @@ 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) {
-        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;
+        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;
         delete app.callInfo;
       }
       next();

lib/session/call-session.js

@@ -566,9 +566,7 @@ class CallSession extends Emitter {
         //this.logger.debug('CallSession:clearOrRestoreActionHookDelayProcessor - ahd settings');
         //await this.clearActionHookDelayProcessor();
       }
-      else {
-        this.logger.debug('CallSession:clearOrRestoreActionHookDelayProcessor - restore ahd after gather override');
-      }
+      this.logger.debug('CallSession:clearOrRestoreActionHookDelayProcessor - say or play action completed');
     }
   }
 
@@ -873,7 +871,8 @@ class CallSession extends Emitter {
             writeAlerts({
               alert_type: AlertType.TTS_FAILURE,
               account_sid: this.accountSid,
-              vendor
+              vendor,
+              target_sid: this.callSid
             }).catch((err) => this.logger.error({err}, 'Error writing tts alert'));
           }
         }
@@ -919,6 +918,7 @@ class CallSession extends Emitter {
             speech_credential_sid: credential.speech_credential_sid,
             api_key: credential.api_key,
             deepgram_stt_uri: credential.deepgram_stt_uri,
+            deepgram_tts_uri: credential.deepgram_tts_uri,
             deepgram_stt_use_tls: credential.deepgram_stt_use_tls
           };
         }
@@ -996,7 +996,8 @@ class CallSession extends Emitter {
         writeAlerts({
           alert_type: AlertType.STT_NOT_PROVISIONED,
           account_sid: this.accountSid,
-          vendor
+          vendor,
+          target_sid: this.callSid
         }).catch((err) => this.logger.error({err}, 'Error writing tts alert'));
       }
     }
@@ -1066,7 +1067,7 @@ 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) {
@@ -1806,7 +1807,8 @@ Duration=${duration} `
       await writeAlerts({
         alert_type: AlertType.WEBHOOK_CONNECTION_FAILURE,
         account_sid: this.accountSid,
-        detail: `Session:reconnect error ${err}`
+        detail: `Session:reconnect error ${err}`,
+        url: this.application.call_hook.url,
       });
     } catch (error) {
       this.logger.error({error}, 'Error writing WEBHOOK_CONNECTION_FAILURE alert');
@@ -2067,6 +2069,12 @@ Duration=${duration} `
       this.logger.error('CallSession:replaceEndpoint cannot be called without stable dlg');
       return;
     }
+    // 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();
 

lib/session/rest-call-session.js

@@ -46,6 +46,7 @@ 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);
   }
 

lib/tasks/conference.js

@@ -135,15 +135,10 @@ class Conference extends Task {
    * @param {SipDialog} dlg
    */
   async _init(cs, dlg) {
-    const friendlyName = this.confName;
     const {createHash, retrieveHash} = cs.srf.locals.dbHelpers;
+    this.friendlyName = this.confName;
     this.confName = `conf:${cs.accountSid}:${this.confName}`;
 
-    this.statusParams = Object.assign({
-      conferenceSid: this.confName,
-      friendlyName
-    }, cs.callInfo);
-
     // check if conference is in progress
     const obj = await retrieveHash(this.confName);
     if (obj) {
@@ -493,7 +488,7 @@ class Conference extends Task {
   }
 
   async doConferenceParticipantAction(cs, opts) {
-    const {action, tag} = opts;
+    const {action, tag, wait_hook } = opts;
 
     switch (action) {
       case 'tag':
@@ -509,7 +504,10 @@ class Conference extends Task {
         await this.clearCoachMode();
         break;
       case 'hold':
-        this.doConferenceHold(cs, {conf_hold_status: 'hold'});
+        this.doConferenceHold(cs, {
+          conf_hold_status: 'hold',
+          ...(wait_hook && {wait_hook})
+        });
         break;
       case 'unhold':
         this.doConferenceHold(cs, {conf_hold_status: 'unhold'});
@@ -546,6 +544,13 @@ 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
@@ -604,7 +609,7 @@ class Conference extends Task {
      * when we hang up as the last member, the current member count = 1
      * when we are kicked out of the call when the moderator leaves, the member count = 0
      */
-    if (this.participantCount === 0) {
+    if (this.participantCount === 0 || this.endConferenceOnExit) {
       const {deleteKey} = cs.srf.locals.dbHelpers;
       try {
         this._notifyConferenceEvent(cs, 'end');
@@ -612,7 +617,8 @@ class Conference extends Task {
         this.logger.info(`conf ${this.confName} deprovisioned: ${removed ? 'success' : 'failure'}`);
       }
       catch (err) {
-        this.logger.error(err, `Error deprovisioning conference ${this.confName}`);
+        this.logger.error(err, `Error deprovisioning conference ${this.confName},
+          might be the conference already cleaned by another moderator`);
       }
     }
   }
@@ -689,8 +695,24 @@ class Conference extends Task {
       if (!params.time) params.time = (new Date()).toISOString();
       if (!params.members && typeof this.participantCount === 'number') params.members = this.participantCount;
       cs.application.requestor
-        .request('verb:hook', this.statusHook, Object.assign(params, this.statusParams, httpHeaders))
-        .catch((err) => this.logger.info(err, 'Conference:notifyConferenceEvent - error'));
+        .request(
+          'verb:hook',
+          this.statusHook,
+          Object.assign(
+            params,
+            Object.assign(
+              {
+                conferenceSid: this.confName,
+                friendlyName: this.friendlyName,
+              },
+              cs.callInfo.toJSON()
+            ),
+            httpHeaders
+          )
+        )
+        .catch((err) =>
+          this.logger.info(err, 'Conference:notifyConferenceEvent - error')
+        );
     }
   }
 

lib/tasks/gather.js

@@ -322,6 +322,9 @@ class TaskGather extends SttTask {
     clearTimeout(this.interDigitTimer);
     let resolved = false;
     if (this.dtmfBargein) {
+      if (!this.playComplete) {
+        this.notifyStatus({event: 'dtmf-bargein-detected', ...evt});
+      }
       this._killAudio(cs);
       this.emit('dtmf', evt);
     }
@@ -543,7 +546,8 @@ class TaskGather extends SttTask {
         account_sid: this.cs.accountSid,
         alert_type: AlertType.STT_FAILURE,
         vendor: this.vendor,
-        detail: err.message
+        detail: err.message,
+        target_sid: this.cs.callSid
       });
     }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
   }
@@ -699,6 +703,7 @@ class TaskGather extends SttTask {
       this.playTask.kill(cs);
       this.playTask = null;
     }
+    this.playComplete = true;
   }
 
   _onTranscription(cs, ep, evt, fsEvent) {
@@ -862,6 +867,7 @@ class TaskGather extends SttTask {
           if (!this.playComplete) {
             this.logger.debug({transcript: evt.alternatives[0].transcript}, 'killing audio due to speech');
             this.emit('vad');
+            this.notifyStatus({event: 'speech-bargein-detected', ...evt});
           }
           this._killAudio(cs);
         }
@@ -876,15 +882,18 @@ class TaskGather extends SttTask {
           this.cs.callInfo, httpHeaders));
       }
       if (this.vendor === 'soniox') {
-        this._clearTimer();
         if (evt.vendor.finalWords.length) {
           this.logger.debug({evt}, 'TaskGather:_onTranscription - buffering soniox transcript');
           this._sonioxTranscripts.push(evt.vendor.finalWords);
         }
       }
-
-      /* restart asr timer if we get a partial transcript */
-      if (this.isContinuousAsr) this._startAsrTimer();
+      // If transcription received, reset timeout timer.
+      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();
     }
   }
   _onEndOfUtterance(cs, ep) {
@@ -966,6 +975,7 @@ class TaskGather extends SttTask {
       alert_type: AlertType.STT_FAILURE,
       message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
     if (!(await this._startFallback(cs, ep, evt))) {
       this.notifyTaskDone();

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

@@ -61,147 +61,8 @@ class TaskSay extends TtsTask {
     }
   }
 
-  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];
-      }
-    } 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
-        }).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
-          }).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 exec(cs, {ep}) {
-    const {srf, accountSid:account_sid} = cs;
+    const {srf, accountSid:account_sid, callSid:target_sid} = cs;
     const {writeAlerts, AlertType} = srf.locals;
     const {addFileToCache} = srf.locals.dbHelpers;
     const engine = this.synthesizer.engine || cs.synthesizer?.engine || 'neural';
@@ -280,12 +141,12 @@ class TaskSay extends TtsTask {
           await this.playToConfMember(ep, memberId, confName, confUuid, filepath[segment]);
         }
         else {
-          let tts_cache_filename;
-          if (filepath[segment].startsWith('say:{')) {
+          const isStreaming = filepath[segment].startsWith('say:{');
+          if (isStreaming) {
             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) {
@@ -293,17 +154,19 @@ 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) => {
-            if (!tts_cache_filename || evt.variable_tts_cache_filename !== tts_cache_filename) {
-              this.logger.info({evt}, 'Say: discarding playback-stop from other say verb');
+            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'));
             }
             else {
               this.logger.debug({evt}, 'Say got playback-stop');
@@ -312,7 +175,8 @@ class TaskSay extends TtsTask {
                   account_sid,
                   alert_type: AlertType.TTS_FAILURE,
                   vendor,
-                  detail: evt.variable_tts_error
+                  detail: evt.variable_tts_error,
+                  target_sid
                 }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
               }
               if (evt.variable_tts_cache_filename && !this.killed) {

lib/tasks/stt-task.js

@@ -178,7 +178,8 @@ class SttTask extends Task {
       writeAlerts({
         account_sid: cs.accountSid,
         alert_type: AlertType.STT_NOT_PROVISIONED,
-        vendor
+        vendor,
+        target_sid: cs.callSid
       }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
       this.notifyTaskDone();
       throw new Error(`No speech-to-text service credentials for ${vendor} have been configured`);
@@ -309,6 +310,7 @@ class SttTask extends Task {
       message: 'STT failure reported by vendor',
       detail: evt.error,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
   }
 
@@ -321,6 +323,7 @@ class SttTask extends Task {
       alert_type: AlertType.STT_FAILURE,
       message: `Failed connecting to ${this.vendor} speech recognizer: ${reason}`,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
   }
 }

lib/tasks/transcribe.js

@@ -605,6 +605,7 @@ class TaskTranscribe extends SttTask {
       alert_type: AlertType.STT_FAILURE,
       message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
     if (!(await this._startFallback(cs, _ep, evt))) {
       this.notifyTaskDone();

lib/tasks/tts-task.js

@@ -17,16 +17,26 @@ 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,
-    disableTtsStreaming,
-    preCache
-  }) {
+  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;
@@ -65,23 +75,23 @@ class TtsTask extends Task {
     }
 
     ep.set({
-      tts_engine: vendor,
+      tts_engine: vendor.startsWith('custom:') ? 'custom' : vendor,
       tts_voice: voice,
-      cache_speech_handles: 1,
-    }).catch((err) => this.logger.info({err}, `${this.name}: Error setting tts_engine on endpoint`));
+      cache_speech_handles: !cs.currentTtsVendor || cs.currentTtsVendor === vendor ? 1 : 0,
+    }).catch((err) => this.logger.info({err}, 'Error setting tts_engine on endpoint'));
+    // set the current vendor on the call session
+    // If vendor is changed from the previous one, then reset the cache_speech_handles flag
+    cs.currentTtsVendor = vendor;
 
-    if (!preCache) this.logger.info({vendor, language, voice, model}, `${this.name}:exec`);
+    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
+          vendor,
+          target_sid: cs.callSid
         }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-        this.notifyError({
-          msg: 'TTS error',
-          details:`No speech credentials provisioned for selected vendor ${vendor}`
-        });
         throw new Error('no provisioned speech credentials for TTS');
       }
       // synthesize all of the text elements
@@ -93,7 +103,7 @@ class TtsTask extends Task {
         if (text.startsWith('silence_stream://')) return text;
 
         /* otel: trace time for tts */
-        if (!preCache && !this.parentTask)  {
+        if (!preCache && !this._disableTracing)  {
           const {span} = this.startChildSpan('tts-generation', {
             'tts.vendor': vendor,
             'tts.language': language,
@@ -114,11 +124,10 @@ class TtsTask extends Task {
             credentials,
             options: this.options,
             disableTtsCache : this.disableTtsCache,
-            disableTtsStreaming,
-            preCache
+            renderForCaching: preCache
           });
           if (!filePath.startsWith('say:')) {
-            this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
+            this.logger.debug(`Say: file ${filePath}, served from cache ${servedFromCache}`);
             if (filePath) cs.trackTmpFile(filePath);
             if (this.otelSpan) {
               this.otelSpan.setAttributes({'tts.cached': servedFromCache});
@@ -129,7 +138,7 @@ class TtsTask extends Task {
               lastUpdated = true;
               updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
             }
-            if (!servedFromCache && rtt && !preCache) {
+            if (!servedFromCache && rtt && !preCache  && !this._disableTracing) {
               this.notifyStatus({
                 event: 'synthesized-audio',
                 vendor,
@@ -140,7 +149,7 @@ class TtsTask extends Task {
             }
           }
           else {
-            this.logger.debug('a streaming tts api will be used');
+            this.logger.debug('Say: a streaming tts api will be used');
             const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${ep.uuid},`);
             return modifiedPath;
           }
@@ -152,9 +161,9 @@ class TtsTask extends Task {
             account_sid: cs.accountSid,
             alert_type: AlertType.TTS_FAILURE,
             vendor,
-            detail: err.message
+            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;
         }
       };
@@ -165,6 +174,7 @@ class TtsTask extends Task {
       this.logger.info(err, 'TaskSay:exec error');
       throw err;
     }
+
   }
 
   _validateURL(urlString) {

lib/utils/action-hook-delay.js

@@ -79,7 +79,6 @@ class ActionHookDelayProcessor extends Emitter {
   }
 
   async stop() {
-    this.logger.debug('ActionHookDelayProcessor#stop');
     this._active = false;
 
     if (this._noResponseTimer) {
@@ -91,25 +90,19 @@ class ActionHookDelayProcessor extends Emitter {
       this._noResponseGiveUpTimer = null;
     }
     if (this._taskInProgress) {
-      this.logger.debug(`ActionHookDelayProcessor#stop: killing task in progress: ${this._taskInProgress.name}`);
+      this.logger.debug(`ActionHookDelayProcessor#stop: stopping ${this._taskInProgress.name}`);
 
-      /** 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._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);
       }
+      return new Promise((resolve) => this._sayResolver = resolve);
     }
     this.logger.debug('ActionHookDelayProcessor#stop returning');
   }
@@ -147,7 +140,7 @@ 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}, 'got playback-stop');
+        this.logger.debug({evt}, 'ActionHookDelayProcessor#_onNoResponseTimer got playback-stop');
         this._sayResolver();
         this._sayResolver = null;
       }

lib/utils/amd-utils.js

@@ -210,7 +210,8 @@ module.exports = (logger) => {
         account_sid: cs.accountSid,
         alert_type: AlertType.STT_FAILURE,
         vendor: vendor,
-        detail: err.message
+        detail: err.message,
+        target_sid: cs.callSid
       });
     }).catch((err) => logger.info({err}, 'Error generating alert for tts failure'));
 
@@ -245,7 +246,10 @@ module.exports = (logger) => {
     const amd = ep.amd = new Amd(logger, cs, opts);
     const {vendor, language} = amd;
     let sttCredentials = amd.sttCredentials;
-    const hints = voicemailHints[language] || [];
+    // 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 = [];
 
     if (vendor === 'nuance' && sttCredentials.client_id) {
       /* get nuance access token */

lib/utils/db-utils.js

@@ -77,6 +77,7 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
       obj.deepgram_stt_uri = o.deepgram_stt_uri;
+      obj.deepgram_tts_uri = o.deepgram_tts_uri;
       obj.deepgram_stt_use_tls = o.deepgram_stt_use_tls;
     }
     else if ('soniox' === obj.vendor) {

lib/utils/install-srf-locals.js

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

lib/utils/place-outdial.js

@@ -441,7 +441,7 @@ class SingleDialer extends Emitter {
     });
     app.requestor.request('session:adulting', '/adulting', {
       ...cs.callInfo.toJSON(),
-      parentCallInfo: this.parentCallInfo
+      parentCallInfo: this.parentCallInfo.toJSON()
     }).catch((err) => {
       newLogger.error({err}, 'doAdulting: error sending adulting request');
     });

lib/utils/transcription-utils.js

@@ -45,7 +45,8 @@ const stickyVars = {
     'DEEPGRAM_SPEECH_ENDPOINTING',
     'DEEPGRAM_SPEECH_UTTERANCE_END_MS',
     'DEEPGRAM_SPEECH_VAD_TURNOFF',
-    'DEEPGRAM_SPEECH_TAG'
+    'DEEPGRAM_SPEECH_TAG',
+    'DEEPGRAM_SPEECH_MODEL_VERSION'
   ],
   aws: [
     'AWS_VOCABULARY_NAME',
@@ -316,8 +317,10 @@ const normalizeIbm = (evt, channel, language) => {
 
 const normalizeGoogle = (evt, channel, language) => {
   const copy = JSON.parse(JSON.stringify(evt));
+  const language_code = evt.language_code || language;
+
   return {
-    language_code: language,
+    language_code: language_code,
     channel_tag: channel,
     is_final: evt.is_final,
     alternatives: [evt.alternatives[0]],
@@ -707,7 +710,9 @@ module.exports = (logger) => {
         ...(deepgramOptions.vadTurnoff) &&
           {DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.vadTurnoff},
         ...(deepgramOptions.tag) &&
-          {DEEPGRAM_SPEECH_TAG: deepgramOptions.tag}
+          {DEEPGRAM_SPEECH_TAG: deepgramOptions.tag},
+        ...(deepgramOptions.version) &&
+          {DEEPGRAM_SPEECH_MODEL_VERSION: deepgramOptions.version}
       };
     }
     else if ('soniox' === vendor) {
@@ -835,6 +840,7 @@ module.exports = (logger) => {
       };
     } else if (vendor.startsWith('custom:')) {
       let {options = {}} = rOpts;
+      const {sampleRate} = rOpts.customOptions || {};
       const {auth_token, custom_stt_url} = sttCredentials;
       options = {
         ...options,
@@ -842,14 +848,15 @@ module.exports = (logger) => {
         {hints: rOpts.hints}),
         ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
         {hints: JSON.stringify(rOpts.hints)}),
-        ...(typeof rOpts.hintsBoost === 'number' && {hintsBoost: rOpts.hintsBoost})
+        ...(typeof rOpts.hintsBoost === 'number' && {hintsBoost: rOpts.hintsBoost}),
+        ...(task.cs?.callSid && {callSid: task.cs.callSid})
       };
-
       opts = {
         ...opts,
         ...(auth_token && {JAMBONZ_STT_API_KEY: auth_token}),
         JAMBONZ_STT_URL: custom_stt_url,
         ...(Object.keys(options).length > 0 && {JAMBONZ_STT_OPTIONS: JSON.stringify(options)}),
+        ...(sampleRate && {JAMBONZ_STT_SAMPLING: sampleRate})
       };
     }
 

package.json

@@ -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": "^0.1.11",
+    "@jambonz/speech-utils": "^0.1.15",
     "@jambonz/stats-collector": "^0.1.10",
-    "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.74",
+    "@jambonz/time-series": "^0.2.9",
+    "@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",
@@ -61,10 +61,10 @@
     "short-uuid": "^5.1.0",
     "sinon": "^17.0.1",
     "to-snake-case": "^1.0.0",
-    "undici": "^6.19.2",
+    "undici": "^6.19.4",
     "uuid-random": "^1.3.2",
     "verify-aws-sns-signature": "^0.1.0",
-    "ws": "^8.17.1",
+    "ws": "^8.18.0",
     "xml2js": "^0.6.2"
   },
   "devDependencies": {

test/docker-compose-testbed.yaml

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