Support sipindialog for conference (#1050)

* fix: add _onRequestWithinDialog catch block

* feat: support sipindialog for conference

* fix: remove any existing listener before adding new

README.md

@@ -13,7 +13,7 @@ Configuration is provided via environment variables:
 |AWS_ACCESS_KEY_ID| aws access key id, used for TTS/STT as well SNS notifications|no|
 |AWS_REGION| aws region| no|
 |AWS_SECRET_ACCESS_KEY| aws secret access key, used per above|no|
-|AWS_SNS_TOPIC_ARM| aws sns topic arn that scale-in lifecycle notifications will be published to|no|
+|AWS_SNS_TOPIC_ARN| aws sns topic arn that scale-in lifecycle notifications will be published to|no|
 |DRACHTIO_HOST| ip address of drachtio server (typically '127.0.0.1')|yes|
 |DRACHTIO_PORT| listening port of drachtio server for control connections (typically 9022)|yes|
 |DRACHTIO_SECRET| shared secret|yes|
@@ -72,7 +72,7 @@ module.exports = {
       STATS_PORT: 8125,
       STATS_PROTOCOL: 'tcp',
       STATS_TELEGRAF: 1,
-      AWS_SNS_TOPIC_ARM: 'arn:aws:sns:us-west-1:xxxxxxxxxxx:terraform-20201107200347128600000002',
+      AWS_SNS_TOPIC_ARN: 'arn:aws:sns:us-west-1:xxxxxxxxxxx:terraform-20201107200347128600000002',
       JAMBONES_NETWORK_CIDR: '172.31.0.0/16',
       JAMBONES_MYSQL_HOST: 'aurora-cluster-jambonz.cluster-yyyyyyyyyyy.us-west-1.rds.amazonaws.com',
       JAMBONES_MYSQL_USER: 'admin',

lib/config.js

@@ -93,7 +93,7 @@ const AWS_REGION = process.env.AWS_REGION;
 const AWS_ACCESS_KEY_ID = process.env.AWS_ACCESS_KEY_ID;
 const AWS_SECRET_ACCESS_KEY = process.env.AWS_SECRET_ACCESS_KEY;
 const AWS_SNS_PORT = parseInt(process.env.AWS_SNS_PORT, 10) || 3001;
-const AWS_SNS_TOPIC_ARM = process.env.AWS_SNS_TOPIC_ARM;
+const AWS_SNS_TOPIC_ARN = process.env.AWS_SNS_TOPIC_ARN;
 const AWS_SNS_PORT_MAX = parseInt(process.env.AWS_SNS_PORT_MAX, 10) || 3005;
 
 const GCP_JSON_KEY = process.env.GCP_JSON_KEY;
@@ -189,7 +189,7 @@ module.exports = {
   AWS_ACCESS_KEY_ID,
   AWS_SECRET_ACCESS_KEY,
   AWS_SNS_PORT,
-  AWS_SNS_TOPIC_ARM,
+  AWS_SNS_TOPIC_ARN,
   AWS_SNS_PORT_MAX,
 
   ANCHOR_MEDIA_ALWAYS,

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

@@ -100,6 +100,7 @@ router.post('/',
         ...(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.proxy && {'X-SIP-Proxy': target.proxy}),
         ...target.headers
       };
 

lib/session/call-session.js

@@ -1084,6 +1084,12 @@ class CallSession extends Emitter {
             api_key: credential.api_key
           };
         }
+        else if ('voxist' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            api_key: credential.api_key
+          };
+        }
         else if ('whisper' === vendor) {
           return {
             api_key: credential.api_key,
@@ -1793,62 +1799,66 @@ Duration=${duration} `
   async updateCall(opts, callSid) {
     this.logger.debug(opts, 'CallSession:updateCall');
 
-    if (opts.call_status) {
-      return this._lccCallStatus(opts);
-    }
-    if (opts.call_hook || opts.child_call_hook) {
-      return await this._lccCallHook(opts);
-    }
-    if (opts.listen_status || opts.stream_status) {
-      await this._lccListenStatus(opts);
-    }
-    if (opts.transcribe_status) {
-      await this._lccTranscribeStatus(opts);
-    }
-    else if (opts.mute_status) {
-      await this._lccMuteStatus(opts.mute_status === 'mute', callSid);
-    }
-    else if (opts.conf_hold_status) {
-      await this._lccConfHoldStatus(opts);
-    }
-    else if (opts.conf_mute_status) {
-      await this._lccConfMuteStatus(opts);
-    }
-    else if (opts.sip_request) {
-      const res = await this._lccSipRequest(opts, callSid);
-      return {status: res.status, reason: res.reason};
-    } else if (opts.dtmf) {
-      await this._lccDtmf(opts, callSid);
-    }
-    else if (opts.record) {
-      await this.notifyRecordOptions(opts.record);
-    }
-    else if (opts.tag) {
-      return this._lccTag(opts);
-    }
-    else if (opts.conferenceParticipantAction) {
-      return this._lccConferenceParticipantAction(opts.conferenceParticipantAction);
-    }
-    else if (opts.dub) {
-      return this._lccDub(opts.dub, callSid);
-    }
-    else if (opts.boostAudioSignal) {
-      return this._lccBoostAudioSignal(opts, callSid);
-    }
-    else if (opts.media_path) {
-      return this._lccMediaPath(opts.media_path, callSid);
-    }
-    else if (opts.llm_tool_output) {
-      return this._lccToolOutput(opts.tool_call_id, opts.llm_tool_output, callSid);
-    }
-    else if (opts.llm_update) {
-      return this._lccLlmUpdate(opts.llm_update, callSid);
-    }
+    try {
+      if (opts.call_status) {
+        return this._lccCallStatus(opts);
+      }
+      if (opts.call_hook || opts.child_call_hook) {
+        return await this._lccCallHook(opts);
+      }
+      if (opts.listen_status || opts.stream_status) {
+        await this._lccListenStatus(opts);
+      }
+      if (opts.transcribe_status) {
+        await this._lccTranscribeStatus(opts);
+      }
+      else if (opts.mute_status) {
+        await this._lccMuteStatus(opts.mute_status === 'mute', callSid);
+      }
+      else if (opts.conf_hold_status) {
+        await this._lccConfHoldStatus(opts);
+      }
+      else if (opts.conf_mute_status) {
+        await this._lccConfMuteStatus(opts);
+      }
+      else if (opts.sip_request) {
+        const res = await this._lccSipRequest(opts, callSid);
+        return {status: res.status, reason: res.reason};
+      } else if (opts.dtmf) {
+        await this._lccDtmf(opts, callSid);
+      }
+      else if (opts.record) {
+        await this.notifyRecordOptions(opts.record);
+      }
+      else if (opts.tag) {
+        return this._lccTag(opts);
+      }
+      else if (opts.conferenceParticipantAction) {
+        return this._lccConferenceParticipantAction(opts.conferenceParticipantAction);
+      }
+      else if (opts.dub) {
+        return this._lccDub(opts.dub, callSid);
+      }
+      else if (opts.boostAudioSignal) {
+        return this._lccBoostAudioSignal(opts, callSid);
+      }
+      else if (opts.media_path) {
+        return this._lccMediaPath(opts.media_path, callSid);
+      }
+      else if (opts.llm_tool_output) {
+        return this._lccToolOutput(opts.tool_call_id, opts.llm_tool_output, callSid);
+      }
+      else if (opts.llm_update) {
+        return this._lccLlmUpdate(opts.llm_update, callSid);
+      }
 
-    // whisper may be the only thing we are asked to do, or it may that
-    // we are doing a whisper after having muted, paused recording etc..
-    if (opts.whisper) {
-      return this._lccWhisper(opts, callSid);
+      // whisper may be the only thing we are asked to do, or it may that
+      // we are doing a whisper after having muted, paused recording etc..
+      if (opts.whisper) {
+        return this._lccWhisper(opts, callSid);
+      }
+    } catch (err) {
+      this.logger.info({err, opts, callSid}, 'CallSession:updateCall - error updating call');
     }
   }
 
@@ -2468,12 +2478,14 @@ Duration=${duration} `
     } else if (sip_method === 'MESSAGE') {
       res.send(202);
     } else {
-      this.logger.info(`CallSession:_onRequestWithinDialog unsported method: ${req.method}`);
+      this.logger.warn(`CallSession:_onRequestWithinDialog unsupported method: ${req.method}`);
       res.send(501);
       return;
     }
     const params = {sip_method, sip_body: req.body, sip_headers: req.headers};
-    this.currentTask.performHook(this, this.sipRequestWithinDialogHook, params);
+    this.currentTask.performHook(this, this.sipRequestWithinDialogHook, params).catch((err) => {
+      this.logger.error({err}, 'CallSession:_onRequestWithinDialog - error calling sipRequestWithinDialogHook');
+    });
   }
 
   async _onReinvite(req, res) {
@@ -2484,7 +2496,7 @@ Duration=${duration} `
           res.send(200, {body: this.ep.local.sdp});
         }
         else {
-          if (this.currentTask.name === TaskName.Dial && this.currentTask.isOnHoldEnabled) {
+          if (this.currentTask && this.currentTask.name === TaskName.Dial && this.currentTask.isOnHoldEnabled) {
             this.logger.info('onholdMusic reINVITE after media has been released');
             await this.currentTask.handleReinviteAfterMediaReleased(req, res);
           } else {

lib/tasks/conference.js

@@ -83,7 +83,11 @@ class Conference extends Task {
     // reset answer time if we were transferred from another feature server
     if (this.connectTime) dlg.connectTime = this.connectTime;
 
-
+    if (cs.sipRequestWithinDialogHook) {
+      /* remove any existing listener to escape from duplicating events */
+      this._removeSipIndialogRequestListener(this.dlg);
+      this._initSipIndialogRequestListener(cs, dlg);
+    }
     this.ep.on('destroy', this._kicked.bind(this, cs, dlg));
 
     try {
@@ -103,6 +107,7 @@ class Conference extends Task {
 
       this.logger.debug(`Conference:exec - conference ${this.confName} is over`);
       if (this.callMoved !== false) await this.performAction(this.results);
+      this._removeSipIndialogRequestListener(dlg);
     } catch (err) {
       this.logger.info(err, `TaskConference:exec - error in conference ${this.confName}`);
     }
@@ -416,6 +421,20 @@ class Conference extends Task {
     }
   }
 
+  _initSipIndialogRequestListener(cs, dlg) {
+    dlg.on('info', this._onRequestWithinDialog.bind(this, cs));
+    dlg.on('message', this._onRequestWithinDialog.bind(this, cs));
+  }
+
+  _removeSipIndialogRequestListener(dlg) {
+    dlg && dlg.removeAllListeners('message');
+    dlg && dlg.removeAllListeners('info');
+  }
+
+  _onRequestWithinDialog(cs, req, res) {
+    cs._onRequestWithinDialog(req, res);
+  }
+
   /**
    * The conference we have been waiting for has started.
    * It may be on this server or a different one, and we are

lib/tasks/config.js

@@ -187,18 +187,20 @@ class TaskConfig extends Task {
         : cs.speechRecognizerVendor;
       cs.speechRecognizerLabel = this.recognizer.label === 'default'
         ? cs.speechRecognizerLabel : this.recognizer.label;
-      cs.speechRecognizerLanguage = this.recognizer.language !== 'default'
+      cs.speechRecognizerLanguage = this.recognizer.language !== undefined && this.recognizer.language !== 'default'
         ? this.recognizer.language
         : cs.speechRecognizerLanguage;
 
       //fallback
-      cs.fallbackSpeechRecognizerVendor = this.recognizer.fallbackVendor !== 'default'
+      cs.fallbackSpeechRecognizerVendor = this.recognizer.fallbackVendor !== undefined &&
+        this.recognizer.fallbackVendor !== 'default'
         ? this.recognizer.fallbackVendor
         : cs.fallbackSpeechRecognizerVendor;
       cs.fallbackSpeechRecognizerLabel = this.recognizer.fallbackLabel === 'default' ?
         cs.fallbackSpeechRecognizerLabel :
         this.recognizer.fallbackLabel;
-      cs.fallbackSpeechRecognizerLanguage = this.recognizer.fallbackLanguage !== 'default'
+      cs.fallbackSpeechRecognizerLanguage = this.recognizer.fallbackLanguage !== undefined &&
+        this.recognizer.fallbackLanguage !== 'default'
         ? this.recognizer.fallbackLanguage
         : cs.fallbackSpeechRecognizerLanguage;
 

lib/tasks/dial.js

@@ -230,10 +230,10 @@ class TaskDial extends Task {
               try {
                 await this.epOther.play(this.dialMusic);
               } catch (err) {
-                this.logger.error(err, `TaskDial:exec error playing ${this.dialMusic}`);
+                this.logger.error(err, `TaskDial:exec error playing dialMusic ${this.dialMusic}`);
                 await sleepFor(1000);
               }
-            } while (!this.killed || !this.bridged);
+            } while (!this.killed && !this.bridged && this._mediaPath === MediaPath.FullMedia);
           })();
         }
       }
@@ -499,7 +499,7 @@ class TaskDial extends Task {
     dlg && dlg.removeAllListeners('info');
   }
 
-  async _onRequestWithinDialog(cs, req, res) {
+  _onRequestWithinDialog(cs, req, res) {
     cs._onRequestWithinDialog(req, res);
   }
 
@@ -871,7 +871,11 @@ class TaskDial extends Task {
 
     if (this.parentDtmfCollector) this._installDtmfDetection(cs, cs.dlg);
     if (this.childDtmfCollector) this._installDtmfDetection(cs, this.dlg);
-    if (cs.sipRequestWithinDialogHook) this._initSipIndialogRequestListener(cs, this.dlg);
+    if (cs.sipRequestWithinDialogHook) {
+      /* remove any existing listener to escape from duplicating events */
+      this._removeSipIndialogRequestListener(this.dlg);
+      this._initSipIndialogRequestListener(cs, this.dlg);
+    }
 
     if (this.transcribeTask) this.transcribeTask.exec(cs, {ep: this.epOther, ep2:this.ep});
     if (this.listenTask) this.listenTask.exec(cs, {ep: this.listenTask.channel === 2 ? this.ep : this.epOther});

lib/tasks/gather.js

@@ -11,6 +11,7 @@ const {
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
   AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
   VadDetection,
   VerbioTranscriptionEvents,
   SpeechmaticsTranscriptionEvents
@@ -524,6 +525,17 @@ class TaskGather extends SttTask {
           this._onVendorConnectFailure.bind(this, cs, ep));
         break;
 
+      case 'voxist':
+        this.bugname = `${this.bugname_prefix}voxist_transcribe`;
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, VoxistTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        break;
+
       case 'speechmatics':
         this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
         this.addCustomEventListener(
@@ -624,6 +636,13 @@ class TaskGather extends SttTask {
     this._asrTimer = setTimeout(() => {
       this.logger.debug('_startAsrTimer - asr timer went off');
       const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
+
+      /* special case for speechmatics - keep listening if we dont have any transcripts */
+      if (this.vendor === 'speechmatics' && this._bufferedTranscripts.length === 0) {
+        this.logger.debug('Gather:_startAsrTimer - speechmatics, no transcripts yet, keep listening');
+        this._startAsrTimer();
+        return;
+      }
       this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
     }, this.asrTimeout);
     this.logger.debug(`_startAsrTimer: set for ${this.asrTimeout}ms`);
@@ -766,10 +785,16 @@ class TaskGather extends SttTask {
         this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram but no buffered transcripts');
       }
       else {
-        this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram, return buffered transcript');
-        evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
-        this._bufferedTranscripts = [];
-        this._resolve('speech', evt);
+        const utteranceTime = evt.last_word_end;
+        if (utteranceTime && this._dgTimeOfLastUnprocessedWord && utteranceTime < this._dgTimeOfLastUnprocessedWord) {
+          this.logger.debug('Gather:_onTranscription - got UtteranceEnd with unprocessed words, continue listening');
+        }
+        else {
+          this.logger.debug('Gather:_onTranscription - got UtteranceEnd from deepgram, return buffered transcript');
+          evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
+          this._bufferedTranscripts = [];
+          this._resolve('speech', evt);
+        }
       }
       return;
     }
@@ -780,7 +805,7 @@ class TaskGather extends SttTask {
 
     evt = this.normalizeTranscription(evt, this.vendor, 1, this.language,
       this.shortUtterance, this.data.recognizer.punctuation);
-    this.logger.debug({evt, bugname, finished, vendor: this.vendor}, 'Gather:_onTranscription normalized transcript');
+    //this.logger.debug({evt, bugname, finished, vendor: this.vendor}, 'Gather:_onTranscription normalized transcript');
 
     if (evt.alternatives.length === 0) {
       this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, continue listening');
@@ -788,8 +813,6 @@ class TaskGather extends SttTask {
     }
     const confidence = evt.alternatives[0].confidence;
     const minConfidence = this.data.recognizer?.minConfidence;
-    this.logger.debug({evt},
-      `TaskGather:_onTranscription - confidence (${confidence}), minConfidence (${minConfidence})`);
     if (confidence && minConfidence && confidence < minConfidence) {
       this.logger.info({evt},
         'TaskGather:_onTranscription - Transcript confidence ' +
@@ -905,8 +928,21 @@ class TaskGather extends SttTask {
         if (originalEvent.is_final && evt.alternatives[0].transcript !== '') {
           this.logger.debug({evt}, 'Gather:_onTranscription - buffering a completed (partial) deepgram transcript');
           this._bufferedTranscripts.push(evt);
+          this._dgTimeOfLastUnprocessedWord = null;
+        }
+        if (evt.alternatives[0].transcript === '') {
+          emptyTranscript = true;
+        }
+        else if (!originalEvent.is_final) {
+          /* Deepgram: we have unprocessed words-save last word end time so we can later compare to UtteranceEnd */
+          const words = originalEvent.channel.alternatives[0].words;
+          if (words?.length > 0) {
+            this._dgTimeOfLastUnprocessedWord = words.slice(-1)[0].end;
+            this.logger.debug(
+              `TaskGather:_onTranscription - saving word end time: ${this._dgTimeOfLastUnprocessedWord}`);
+          }
+
         }
-        if (evt.alternatives[0].transcript === '') emptyTranscript = true;
       }
 
       if (!emptyTranscript) {
@@ -1176,7 +1212,7 @@ class TaskGather extends SttTask {
         if (this.parentTask) this.parentTask.emit('stt-low-confidence', evt);
         else {
           this.emit('stt-low-confidence', evt);
-          returnedVerbs = await this.performAction({reason: 'stt-low-confidence'});
+          returnedVerbs = await this.performAction({speech:evt, reason: 'stt-low-confidence'});
         }
       }
     } catch (err) {  /*already logged error*/ }

lib/tasks/llm/index.js

@@ -3,6 +3,7 @@ const {TaskPreconditions} = require('../../utils/constants');
 const TaskLlmOpenAI_S2S = require('./llms/openai_s2s');
 const TaskLlmVoiceAgent_S2S = require('./llms/voice_agent_s2s');
 const TaskLlmUltravox_S2S = require('./llms/ultravox_s2s');
+const TaskLlmElevenlabs_S2S = require('./llms/elevenlabs_s2s');
 
 class TaskLlm extends Task {
   constructor(logger, opts) {
@@ -54,6 +55,10 @@ class TaskLlm extends Task {
         llm = new TaskLlmUltravox_S2S(this.logger, this.data, this);
         break;
 
+      case 'elevenlabs':
+        llm = new TaskLlmElevenlabs_S2S(this.logger, this.data, this);
+        break;
+
       default:
         throw new Error(`Unsupported vendor ${this.vendor} for LLM`);
     }

lib/tasks/llm/llms/elevenlabs_s2s.js

@@ -0,0 +1,302 @@
+const Task = require('../../task');
+const TaskName = 'Llm_Elevenlabs_s2s';
+const {LlmEvents_Elevenlabs} = require('../../../utils/constants');
+const {request} = require('undici');
+const ClientEvent = 'client.event';
+const SessionDelete = 'session.delete';
+
+const elevenlabs_server_events = [
+  'conversation_initiation_metadata',
+  'user_transcript',
+  'agent_response',
+  'client_tool_call'
+];
+
+const expandWildcards = (events) => {
+  const expandedEvents = [];
+
+  events.forEach((evt) => {
+    if (evt.endsWith('.*')) {
+      const prefix = evt.slice(0, -2); // Remove the wildcard ".*"
+      const matchingEvents = elevenlabs_server_events.filter((e) => e.startsWith(prefix));
+      expandedEvents.push(...matchingEvents);
+    } else {
+      expandedEvents.push(evt);
+    }
+  });
+
+  return expandedEvents;
+};
+
+class TaskLlmElevenlabs_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.auth = this.parent.auth;
+
+    const {agent_id, api_key} = this.auth || {};
+    if (!agent_id) throw new Error('auth.agent_id is required for Elevenlabs S2S');
+
+    this.agent_id = agent_id;
+    this.api_key = api_key;
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+    const {
+      conversation_initiation_client_data,
+      input_sample_rate = 16000,
+      output_sample_rate = 16000
+    } = this.data.llmOptions;
+    this.conversation_initiation_client_data = conversation_initiation_client_data;
+    this.input_sample_rate = input_sample_rate;
+    this.output_sample_rate = output_sample_rate;
+    this.results = {
+      completionReason: 'normal conversation end'
+    };
+
+    /**
+     * only one of these will have items,
+     * if includeEvents, then these are the events to include
+     * if excludeEvents, then these are the events to exclude
+     */
+    this.includeEvents = [];
+    this.excludeEvents = [];
+
+    /* default to all events if user did not specify */
+    this._populateEvents(this.data.events || elevenlabs_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  async getSignedUrl() {
+    if (!this.api_key) {
+      return {
+        host: 'api.elevenlabs.io',
+        path: `/v1/convai/conversation?agent_id=${this.agent_id}`,
+      };
+    }
+
+    const {statusCode, body} = await request(
+      `https://api.elevenlabs.io/v1/convai/conversation/get_signed_url?agent_id=${this.agent_id}`, {
+        method: 'GET',
+        headers: {
+          'xi-api-key': this.api_key
+        },
+      }
+    );
+    const data = await body.json();
+    if (statusCode !== 200 || !data?.signed_url) {
+      this.logger.error({statusCode, data}, 'Elevenlabs Error registering call');
+      throw new Error(`Elevenlabs  Error registering call: ${data.message}`);
+    }
+
+    const url = new URL(data.signed_url);
+    return {
+      host: url.hostname,
+      path: url.pathname + url.search,
+    };
+  }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_elevenlabs_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error({args}, `Error calling uuid_elevenlabs_s2s: ${res.body}`);
+    }
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs);
+    await this._startListening(cs, ep);
+
+    await this.awaitTaskDone();
+
+    /* note: the parent llm verb started the span, which is why this is necessary */
+    await this.parent.performAction(this.results);
+
+    this._unregisterHandlers();
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+
+    this._api(cs.ep, [cs.ep.uuid, SessionDelete])
+      .catch((err) => this.logger.info({err}, 'TaskLlmElevenlabs_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  /**
+   * Send function call output to the Elevenlabs server in the form of conversation.item.create
+   * per https://elevenlabs.io/docs/conversational-ai/api-reference/conversational-ai/websocket
+   */
+  async processToolOutput(ep, tool_call_id, rawData) {
+    try {
+      const {data} = rawData;
+      this.logger.debug({tool_call_id, data}, 'TaskLlmElevenlabs_S2S:processToolOutput');
+
+      if (!data.type || data.type !== 'client_tool_result') {
+        this.logger.info({data},
+          'TaskLlmElevenlabs_S2S:processToolOutput - invalid tool output, must be client_tool_result');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmElevenlabs_S2S:processToolOutput');
+    }
+  }
+
+  /**
+   * Send a session.update to the Elevenlabs server
+   * Note: creating and deleting conversation items also supported as well as interrupting the assistant
+   */
+  async processLlmUpdate(ep, data, _callSid) {
+    this.logger.debug({data, _callSid}, 'TaskLlmElevenlabs_S2S:processLlmUpdate, ignored');
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    try {
+      const {host, path} = await this.getSignedUrl();
+      const args = [ep.uuid, 'session.create', this.input_sample_rate, this.output_sample_rate, host, path];
+      await this._api(ep, args);
+    } catch (err) {
+      this.logger.error({err}, 'TaskLlmElevenlabs_S2S:_startListening');
+      this.notifyTaskDone();
+    }
+  }
+
+  async _sendClientEvent(ep, obj) {
+    let ok = true;
+    this.logger.debug({obj}, 'TaskLlmElevenlabs_S2S:_sendClientEvent');
+    try {
+      const args = [ep.uuid, ClientEvent, JSON.stringify(obj)];
+      await this._api(ep, args);
+    } catch (err) {
+      ok = false;
+      this.logger.error({err}, 'TaskLlmElevenlabs_S2S:_sendClientEvent - Error');
+    }
+    return ok;
+  }
+
+  async _sendInitialMessage(ep) {
+    if (this.conversation_initiation_client_data) {
+      if (!await this._sendClientEvent(ep, {
+        type: 'conversation_initiation_client_data',
+        conversation_initiation_client_data: this.conversation_initiation_client_data
+      })) {
+        this.notifyTaskDone();
+      }
+    }
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _onError(ep, evt) {
+    this.logger.info({evt}, 'TaskLlmElevenlabs_S2S:_onError');
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.debug('TaskLlmElevenlabs_S2S:_onConnect');
+    this._sendInitialMessage(ep);
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmElevenlabs_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmElevenlabs_S2S:_onConnectFailure');
+    this.results = {completionReason: 'disconnect from remote end'};
+    this.notifyTaskDone();
+  }
+  async _onServerEvent(ep, evt) {
+    let endConversation = false;
+    const type = evt.type;
+    this.logger.info({evt}, 'TaskLlmElevenlabs_S2S:_onServerEvent');
+
+    if (type === 'error') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server error',
+        error: evt.error
+      };
+    }
+
+    /* tool calls */
+    else if (type === 'client_tool_call') {
+      this.logger.debug({evt}, 'TaskLlmElevenlabs_S2S:_onServerEvent - function_call');
+      if (!this.toolHook) {
+        this.logger.warn({evt}, 'TaskLlmElevenlabs_S2S:_onServerEvent - no toolHook defined!');
+      }
+      else {
+        const {client_tool_call} = evt;
+        const {tool_name: name, tool_call_id: call_id, parameters: args} = client_tool_call;
+
+        try {
+          await this.parent.sendToolHook(call_id, {name, args});
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmElevenlabs_S2S - error calling function');
+          this.results = {
+            completionReason: 'client error calling function',
+            error: err
+          };
+          endConversation = true;
+        }
+      }
+    }
+
+    /* check whether we should notify on this event */
+    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
+      this.parent.sendEventHook(evt)
+        .catch((err) => this.logger.info({err},
+          'TaskLlmElevenlabs_S2S:_onServerEvent - error sending event hook'));
+    }
+
+    if (endConversation) {
+      this.logger.info({results: this.results},
+        'TaskLlmElevenlabs_S2S:_onServerEvent - ending conversation due to error');
+      this.notifyTaskDone();
+    }
+  }
+
+  _populateEvents(events) {
+    if (events.includes('all')) {
+      /* work by excluding specific events */
+      const exclude = events
+        .filter((evt) => evt.startsWith('-'))
+        .map((evt) => evt.slice(1));
+      if (exclude.length === 0) this.includeEvents = elevenlabs_server_events;
+      else this.excludeEvents = expandWildcards(exclude);
+    }
+    else {
+      /* work by including specific events */
+      const include = events
+        .filter((evt) => !evt.startsWith('-'));
+      this.includeEvents = expandWildcards(include);
+    }
+
+    this.logger.debug({
+      includeEvents: this.includeEvents,
+      excludeEvents: this.excludeEvents
+    }, 'TaskLlmElevenlabs_S2S:_populateEvents');
+  }
+}
+
+module.exports = TaskLlmElevenlabs_S2S;

lib/tasks/llm/llms/ultravox_s2s.js

@@ -4,6 +4,7 @@ const {request} = require('undici');
 const {LlmEvents_Ultravox} = require('../../../utils/constants');
 
 const ultravox_server_events = [
+  'createCall',
   'pong',
   'state',
   'transcript',
@@ -88,7 +89,7 @@ class TaskLlmUltravox_S2S extends Task {
       throw new Error(`Ultravox  Error registering call: ${data.message}`);
     }
     this.logger.info({joinUrl: data.joinUrl}, 'Ultravox Call registered');
-    return data.joinUrl;
+    return data;
   }
 
   _unregisterHandlers() {
@@ -105,13 +106,19 @@ class TaskLlmUltravox_S2S extends Task {
   async _startListening(cs, ep) {
     this._registerHandlers(ep);
 
-    const joinUrl = await this.createCall();
+    const data = await this.createCall();
+    const {joinUrl} = data;
     // split the joinUrl into host and path
     const {host, pathname, search} = new URL(joinUrl);
 
     try {
       const args = [ep.uuid, 'session.create', host, pathname + search];
       await this._api(ep, args);
+      // Notify the application that the session has been created with detail information
+      this._sendLlmEvent('createCall', {
+        type: 'createCall',
+        ...data
+      });
     } catch (err) {
       this.logger.error({err}, 'TaskLlmUltraVox_S2S:_startListening');
       this.notifyTaskDone();
@@ -190,11 +197,7 @@ class TaskLlmUltravox_S2S extends Task {
       }
     }
 
-    /* check whether we should notify on this event */
-    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
-      this.parent.sendEventHook(evt)
-        .catch((err) => this.logger.info({err}, 'TaskLlmUltravox_S2S:_onServerEvent - error sending event hook'));
-    }
+    this._sendLlmEvent(type, evt);
 
     if (endConversation) {
       this.logger.info({results: this.results},
@@ -203,6 +206,14 @@ class TaskLlmUltravox_S2S extends Task {
     }
   }
 
+  _sendLlmEvent(type, evt) {
+    /* check whether we should notify on this event */
+    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
+      this.parent.sendEventHook(evt)
+        .catch((err) => this.logger.info({err}, 'TaskLlmUltravox_S2S:_onServerEvent - error sending event hook'));
+    }
+  }
+
   async processToolOutput(ep, tool_call_id, data) {
     try {
       this.logger.debug({tool_call_id, data}, 'TaskLlmUltravox_S2S:processToolOutput');

lib/tasks/say.js

@@ -213,7 +213,7 @@ class TaskSay extends TtsTask {
           ep.once('playback-start', (evt) => {
             this.logger.debug({evt}, 'Say got playback-start');
             if (this.otelSpan) {
-              this._addStreamingTtsAttributes(this.otelSpan, evt);
+              this._addStreamingTtsAttributes(this.otelSpan, evt, vendor);
               this.otelSpan.end();
               this.otelSpan = null;
               if (evt.variable_tts_cache_filename) {
@@ -240,6 +240,7 @@ class TaskSay extends TtsTask {
                 language,
                 voice,
                 engine,
+                model: this.model || this.model_id,
                 text
               }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
             }
@@ -307,7 +308,7 @@ class TaskSay extends TtsTask {
     this.notifyTaskDone();
   }
 
-  _addStreamingTtsAttributes(span, evt) {
+  _addStreamingTtsAttributes(span, evt, vendor) {
     const attrs = {'tts.cached': false};
     for (const [key, value] of Object.entries(evt)) {
       if (key.startsWith('variable_tts_')) {
@@ -321,6 +322,9 @@ class TaskSay extends TtsTask {
           .replace('elevenlabs_', 'elevenlabs.');
         if (spanMapping[newKey]) newKey = spanMapping[newKey];
         attrs[newKey] = value;
+        if (key === 'variable_tts_time_to_first_byte_ms' && value) {
+          this.cs.srf.locals.stats.histogram('tts.response_time', value, [`vendor:${vendor}`]);
+        }
       }
     }
     delete attrs['cache_filename']; //no value in adding this to the span

lib/tasks/transcribe.js

@@ -13,6 +13,7 @@ const {
   JambonzTranscriptionEvents,
   TranscribeStatus,
   AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
   VerbioTranscriptionEvents,
   SpeechmaticsTranscriptionEvents
 } = require('../utils/constants.json');
@@ -300,6 +301,17 @@ class TaskTranscribe extends SttTask {
           this._onVendorConnectFailure.bind(this, cs, ep, channel));
         break;
 
+      case 'voxist':
+        this.bugname = `${this.bugname_prefix}voxist_transcribe`;
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep,
+          VoxistTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
+        break;
+
       case 'speechmatics':
         this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
         this.addCustomEventListener(

lib/tasks/tts-task.js

@@ -143,16 +143,16 @@ class TtsTask extends Task {
         `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) {
       const arr = /([A-Za-z-]*)\s+-\s+(enhanced|standard)/.exec(voice);
       if (arr) {
         voice = arr[1];
-        model = arr[2];
+        this.model = arr[2];
       }
     } else if (vendor === 'deepgram') {
-      model = voice;
+      this.model = voice;
     }
+    this.model_id = credentials.model_id;
 
     /* allow for microsoft custom region voice and api_key to be specified as an override */
     if (vendor === 'microsoft' && this.options.deploymentId) {
@@ -215,7 +215,8 @@ class TtsTask extends Task {
     // 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');
+    if (!preCache && !this._disableTracing)
+      this.logger.info({vendor, language, voice, model: this.model}, 'TaskSay:exec');
     try {
       if (!credentials) {
         writeAlerts({
@@ -250,7 +251,7 @@ class TtsTask extends Task {
             language,
             voice,
             engine,
-            model,
+            model: this.model,
             salt,
             credentials,
             options: this.options,

lib/utils/aws-sns-lifecycle.js

@@ -4,7 +4,7 @@ const assert = require('assert');
 const {
   AWS_REGION,
   AWS_SNS_PORT: PORT,
-  AWS_SNS_TOPIC_ARM,
+  AWS_SNS_TOPIC_ARN,
   AWS_SNS_PORT_MAX,
 } = require('../config');
 const {LifeCycleEvents} = require('./constants');
@@ -55,12 +55,12 @@ class SnsNotifier extends Emitter {
   async _handlePost(req, res) {
     try {
       const parsedBody = JSON.parse(req.body);
-      this.logger.debug({headers: req.headers, body: parsedBody}, 'Received HTTP POST from AWS');
+      this.logger.info({headers: req.headers, body: parsedBody}, 'Received HTTP POST from AWS');
       if (!validatePayload(parsedBody)) {
         this.logger.info('incoming AWS SNS HTTP POST failed signature validation');
         return res.sendStatus(403);
       }
-      this.logger.debug('incoming HTTP POST passed validation');
+      this.logger.info('incoming HTTP POST passed validation');
       res.sendStatus(200);
 
       switch (parsedBody.Type) {
@@ -74,7 +74,18 @@ class SnsNotifier extends Emitter {
             subscriptionRequestId: this.subscriptionRequestId
           }, 'response from SNS SubscribeURL');
           const data = await this.describeInstance();
-          this.lifecycleState = data.AutoScalingGroups[0].Instances[0].LifecycleState;
+
+          const group = data.AutoScalingGroups.find((group) =>
+            group.Instances && group.Instances.some((instance) => instance.InstanceId === this.instanceId)
+          );
+          if (!group) {
+            this.logger.error('Current instance not found in any Auto Scaling group', data);
+          } else {
+            const instance = group.Instances.find((instance) => instance.InstanceId === this.instanceId);
+            this.lifecycleState = instance.LifecycleState;
+          }
+
+          //this.lifecycleState = data.AutoScalingGroups[0].Instances[0].LifecycleState;
           this.emit('SubscriptionConfirmation', {publicIp: this.publicIp});
           break;
 
@@ -94,7 +105,7 @@ class SnsNotifier extends Emitter {
               this.unsubscribe();
             }
             else {
-              this.logger.debug(`SnsNotifier - instance ${msg.EC2InstanceId} is scaling in (not us)`);
+              this.logger.info(`SnsNotifier - instance ${msg.EC2InstanceId} is scaling in (not us)`);
             }
           }
           break;
@@ -111,7 +122,7 @@ class SnsNotifier extends Emitter {
 
   async init() {
     try {
-      this.logger.debug('SnsNotifier: retrieving instance data');
+      this.logger.info('SnsNotifier: retrieving instance data');
       this.instanceId = await getString('http://169.254.169.254/latest/meta-data/instance-id');
       this.publicIp = await getString('http://169.254.169.254/latest/meta-data/public-ipv4');
       this.logger.info({
@@ -142,13 +153,13 @@ class SnsNotifier extends Emitter {
     try {
       const params = {
         Protocol: 'http',
-        TopicArn: AWS_SNS_TOPIC_ARM,
+        TopicArn: AWS_SNS_TOPIC_ARN,
         Endpoint: this.snsEndpoint
       };
       const response = await snsClient.send(new SubscribeCommand(params));
-      this.logger.info({response}, `response to SNS subscribe to ${AWS_SNS_TOPIC_ARM}`);
+      this.logger.info({response}, `response to SNS subscribe to ${AWS_SNS_TOPIC_ARN}`);
     } catch (err) {
-      this.logger.error({err}, `Error subscribing to SNS topic arn ${AWS_SNS_TOPIC_ARM}`);
+      this.logger.error({err}, `Error subscribing to SNS topic arn ${AWS_SNS_TOPIC_ARN}`);
     }
   }
 
@@ -159,9 +170,9 @@ class SnsNotifier extends Emitter {
         SubscriptionArn: this.subscriptionArn
       };
       const response = await snsClient.send(new UnsubscribeCommand(params));
-      this.logger.info({response}, `response to SNS unsubscribe to ${AWS_SNS_TOPIC_ARM}`);
+      this.logger.info({response}, `response to SNS unsubscribe to ${AWS_SNS_TOPIC_ARN}`);
     } catch (err) {
-      this.logger.error({err}, `Error unsubscribing to SNS topic arn ${AWS_SNS_TOPIC_ARM}`);
+      this.logger.error({err}, `Error unsubscribing to SNS topic arn ${AWS_SNS_TOPIC_ARN}`);
     }
   }
 

lib/utils/constants.json

@@ -149,6 +149,12 @@
     "ConnectFailure": "assemblyai_transcribe::connect_failed",
     "Connect": "assemblyai_transcribe::connect"
   },
+  "VoxistTranscriptionEvents": {
+    "Transcription": "voxist_transcribe::transcription",
+    "Error": "voxist_transcribe::error",
+    "ConnectFailure": "voxist_transcribe::connect_failed",
+    "Connect": "voxist_transcribe::connect"
+  },
   "VadDetection": {
     "Detection": "vad_detect:detection"
   },
@@ -176,6 +182,13 @@
     "Disconnect": "openai_s2s::disconnect",
     "ServerEvent": "openai_s2s::server_event"
   },
+  "LlmEvents_Elevenlabs": {
+    "Error": "error",
+    "Connect": "elevenlabs_s2s::connect",
+    "ConnectFailure": "elevenlabs_s2s::connect_failed",
+    "Disconnect": "elevenlabs_s2s::disconnect",
+    "ServerEvent": "elevenlabs_s2s::server_event"
+  },
   "LlmEvents_VoiceAgent": {
     "Error": "error",
     "Connect": "voice_agent_s2s::connect",

lib/utils/db-utils.js

@@ -122,6 +122,10 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
     }
+    else if ('voxist' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+    }
     else if ('whisper' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;

lib/utils/sbc-pinger.js

@@ -8,7 +8,7 @@ const {
   JAMBONES_SBCS,
   K8S,
   K8S_SBC_SIP_SERVICE_NAME,
-  AWS_SNS_TOPIC_ARM,
+  AWS_SNS_TOPIC_ARN,
   OPTIONS_PING_INTERVAL,
   AWS_REGION,
   NODE_ENV,
@@ -35,7 +35,7 @@ module.exports = (logger) => {
   // listen for SNS lifecycle changes
   let lifecycleEmitter = new Emitter();
   let dryUpCalls = false;
-  if (AWS_SNS_TOPIC_ARM && AWS_REGION) {
+  if (AWS_SNS_TOPIC_ARN && AWS_REGION) {
 
     (async function() {
       try {

lib/utils/transcription-utils.js

@@ -30,6 +30,7 @@ const stickyVars = {
     'DEEPGRAM_SPEECH_TIER',
     'DEEPGRAM_SPEECH_MODEL',
     'DEEPGRAM_SPEECH_ENABLE_SMART_FORMAT',
+    'DEEPGRAM_SPEECH_ENABLE_NO_DELAY',
     'DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION',
     'DEEPGRAM_SPEECH_PROFANITY_FILTER',
     'DEEPGRAM_SPEECH_REDACT',
@@ -44,7 +45,8 @@ const stickyVars = {
     'DEEPGRAM_SPEECH_VAD_TURNOFF',
     'DEEPGRAM_SPEECH_TAG',
     'DEEPGRAM_SPEECH_MODEL_VERSION',
-    'DEEPGRAM_SPEECH_FILLER_WORDS'
+    'DEEPGRAM_SPEECH_FILLER_WORDS',
+    'DEEPGRAM_SPEECH_KEYTERMS',
   ],
   aws: [
     'AWS_VOCABULARY_NAME',
@@ -105,6 +107,9 @@ const stickyVars = {
     'ASSEMBLYAI_API_KEY',
     'ASSEMBLYAI_WORD_BOOST'
   ],
+  voxist: [
+    'VOXIST_API_KEY',
+  ],
   speechmatics: [
     'SPEECHMATICS_API_KEY',
     'SPEECHMATICS_HOST',
@@ -517,6 +522,25 @@ const normalizeAssemblyAi = (evt, channel, language) => {
   };
 };
 
+const normalizeVoxist = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.type === 'final',
+    alternatives: [
+      {
+        confidence: 1.00,
+        transcript: evt.text,
+      }
+    ],
+    vendor: {
+      name: 'voxist',
+      evt: copy
+    }
+  };
+};
+
 const normalizeSpeechmatics = (evt, channel, language) => {
   const copy = JSON.parse(JSON.stringify(evt));
   const is_final = evt.message === 'AddTranscript';
@@ -567,6 +591,8 @@ module.exports = (logger) => {
         return normalizeCobalt(evt, channel, language);
       case 'assemblyai':
         return normalizeAssemblyAi(evt, channel, language, shortUtterance);
+      case 'voxist':
+        return normalizeVoxist(evt, channel, language);
       case 'verbio':
         return normalizeVerbio(evt, channel, language);
       case 'speechmatics':
@@ -705,6 +731,8 @@ module.exports = (logger) => {
         //azureSttEndpointId overrides sttCredentials.custom_stt_endpoint
         ...(rOpts.azureSttEndpointId &&
           {AZURE_SERVICE_ENDPOINT_ID: rOpts.azureSttEndpointId}),
+        ...(azureOptions.speechRecognitionMode &&
+          {AZURE_RECOGNITION_MODE: azureOptions.speechRecognitionMode}),
       };
     }
     else if ('nuance' === vendor) {
@@ -778,6 +806,8 @@ module.exports = (logger) => {
           {DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION: 1},
         ...(deepgramOptions.smartFormatting) &&
           {DEEPGRAM_SPEECH_ENABLE_SMART_FORMAT: 1},
+        ...(deepgramOptions.noDelay) &&
+          {DEEPGRAM_SPEECH_ENABLE_NO_DELAY: 1},
         ...(deepgramOptions.profanityFilter) &&
           {DEEPGRAM_SPEECH_PROFANITY_FILTER: 1},
         ...(deepgramOptions.redact) &&
@@ -815,7 +845,9 @@ module.exports = (logger) => {
         ...(deepgramOptions.version) &&
           {DEEPGRAM_SPEECH_MODEL_VERSION: deepgramOptions.version},
         ...(deepgramOptions.fillerWords) &&
-          {DEEPGRAM_SPEECH_FILLER_WORDS: deepgramOptions.fillerWords}
+          {DEEPGRAM_SPEECH_FILLER_WORDS: deepgramOptions.fillerWords},
+        ...((Array.isArray(deepgramOptions.keyterms) && deepgramOptions.keyterms.length > 0) &&
+          {DEEPGRAM_SPEECH_KEYTERMS: deepgramOptions.keyterms.join(',')})
       };
     }
     else if ('soniox' === vendor) {
@@ -924,6 +956,13 @@ module.exports = (logger) => {
           {ASSEMBLYAI_WORD_BOOST: JSON.stringify(rOpts.hints)})
       };
     }
+    else if ('voxist' === vendor) {
+      opts = {
+        ...opts,
+        ...(sttCredentials.api_key) &&
+          {VOXIST_API_KEY: sttCredentials.api_key},
+      };
+    }
     else if ('verbio' === vendor) {
       const {verbioOptions = {}} = rOpts;
       opts = {

lib/utils/tts-streaming-buffer.js

@@ -4,37 +4,50 @@ const {
   TtsStreamingEvents,
   TtsStreamingConnectionStatus
 } = require('../utils/constants');
+
 const MAX_CHUNK_SIZE = 1800;
 const HIGH_WATER_BUFFER_SIZE = 1000;
 const LOW_WATER_BUFFER_SIZE = 200;
 const TIMEOUT_RETRY_MSECS = 3000;
 
+
+const isWhitespace = (str) => /^\s*$/.test(str);
+
+/**
+ * Each queue item is an object:
+ *   - { type: 'text', value: '…' } for text tokens.
+ *   - { type: 'flush' } for a flush command.
+ */
 class TtsStreamingBuffer extends Emitter {
   constructor(cs) {
     super();
     this.cs = cs;
     this.logger = cs.logger;
 
-    this.tokens = '';
+    // Use an array to hold our structured items.
+    this.queue = [];
+    // Track total number of characters in text items.
+    this.bufferedLength = 0;
     this.eventHandlers = [];
     this._isFull = false;
     this._connectionStatus = TtsStreamingConnectionStatus.NotConnected;
-    this._flushPending = false;
     this.timer = null;
+    // Record the last time the text buffer was updated.
+    this.lastUpdateTime = 0;
   }
 
   get isEmpty() {
-    return this.tokens.length === 0;
+    return this.queue.length === 0;
+  }
+
+  get size() {
+    return this.bufferedLength;
   }
 
   get isFull() {
     return this._isFull;
   }
 
-  get size() {
-    return this.tokens.length;
-  }
-
   get ep() {
     return this.cs?.ep;
   }
@@ -42,7 +55,8 @@ class TtsStreamingBuffer extends Emitter {
   async start() {
     assert.ok(
       this._connectionStatus === TtsStreamingConnectionStatus.NotConnected,
-      'TtsStreamingBuffer:start already started, or has failed');
+      'TtsStreamingBuffer:start already started, or has failed'
+    );
 
     this.vendor = this.cs.getTsStreamingVendor();
     if (!this.vendor) {
@@ -55,9 +69,9 @@ class TtsStreamingBuffer extends Emitter {
     this._connectionStatus = TtsStreamingConnectionStatus.Connecting;
     try {
       if (this.eventHandlers.length === 0) this._initHandlers(this.ep);
-      await  this._api(this.ep, [this.ep.uuid, 'connect']);
+      await this._api(this.ep, [this.ep.uuid, 'connect']);
     } catch (err) {
-      this.logger.info({err}, 'TtsStreamingBuffer:start Error connecting to TTS streaming');
+      this.logger.info({ err }, 'TtsStreamingBuffer:start Error connecting to TTS streaming');
       this._connectionStatus = TtsStreamingConnectionStatus.Failed;
     }
   }
@@ -67,204 +81,319 @@ class TtsStreamingBuffer extends Emitter {
     this.removeCustomEventListeners();
     if (this.ep) {
       this._api(this.ep, [this.ep.uuid, 'close'])
-        .catch((err) => this.logger.info({err}, 'TtsStreamingBuffer:kill Error closing TTS streaming'));
+        .catch((err) =>
+          this.logger.info({ err }, 'TtsStreamingBuffer:stop Error closing TTS streaming')
+        );
     }
     this.timer = null;
-    this.tokens = '';
+    this.queue = [];
+    this.bufferedLength = 0;
     this._connectionStatus = TtsStreamingConnectionStatus.NotConnected;
   }
 
   /**
-   * Add tokens to the buffer and start feeding them to the endpoint if necessary.
+   * Buffer new text tokens.
    */
   async bufferTokens(tokens) {
-
     if (this._connectionStatus === TtsStreamingConnectionStatus.Failed) {
       this.logger.info('TtsStreamingBuffer:bufferTokens TTS streaming connection failed, rejecting request');
-      return {status: 'failed', reason: `connection to ${this.vendor} failed`};
+      return { status: 'failed', reason: `connection to ${this.vendor} failed` };
+    }
+
+    if (0 === this.bufferedLength && isWhitespace(tokens)) {
+      this.logger.debug({tokens}, 'TtsStreamingBuffer:bufferTokens discarded whitespace tokens');
+      return { status: 'ok' };
     }
 
     const displayedTokens = tokens.length <= 40 ? tokens : tokens.substring(0, 40);
     const totalLength = tokens.length;
 
-    /* if we crossed the high water mark, reject the request */
-    if (this.tokens.length + totalLength > HIGH_WATER_BUFFER_SIZE) {
+    if (this.bufferedLength + totalLength > HIGH_WATER_BUFFER_SIZE) {
       this.logger.info(
-        `TtsStreamingBuffer throttling: buffer is full, rejecting request to buffer ${totalLength} tokens`);
-
+        `TtsStreamingBuffer throttling: buffer is full, rejecting request to buffer ${totalLength} tokens`
+      );
       if (!this._isFull) {
         this._isFull = true;
         this.emit(TtsStreamingEvents.Pause);
       }
-      return {status: 'failed', reason: 'full'};
+      return { status: 'failed', reason: 'full' };
     }
 
     this.logger.debug(
-      `TtsStreamingBuffer:bufferTokens "${displayedTokens}" (length: ${totalLength}), starting? ${this.isEmpty}`
+      `TtsStreamingBuffer:bufferTokens "${displayedTokens}" (length: ${totalLength})`
     );
-    this.tokens += (tokens || '');
+    this.queue.push({ type: 'text', value: tokens });
+    this.bufferedLength += totalLength;
+    // Update the last update time each time new text is buffered.
+    this.lastUpdateTime = Date.now();
 
-    await this._feedTokens();
-
-    return {status: 'ok'};
+    await this._feedQueue();
+    return { status: 'ok' };
   }
 
+  /**
+   * Insert a flush command. If no text is queued, flush immediately.
+   * Otherwise, append a flush marker so that all text preceding it will be sent
+   * (regardless of sentence boundaries) before the flush is issued.
+   */
   flush() {
-    this.logger.debug('TtsStreamingBuffer:flush');
     if (this._connectionStatus === TtsStreamingConnectionStatus.Connecting) {
       this.logger.debug('TtsStreamingBuffer:flush TTS stream is not quite ready - wait for connect');
-      this._flushPending = true;
+      if (this.queue.length === 0 || this.queue[this.queue.length - 1].type !== 'flush') {
+        this.queue.push({ type: 'flush' });
+      }
       return;
     }
     else if (this._connectionStatus === TtsStreamingConnectionStatus.Connected) {
-
-      if (this.size === 0) {
+      if (this.isEmpty) {
         this._doFlush();
       }
       else {
-        /* we have tokens queued, so flush after they have been sent */
-        this._pendingFlush = true;
+        if (this.queue[this.queue.length - 1].type !== 'flush') {
+          this.queue.push({ type: 'flush' });
+          this.logger.debug('TtsStreamingBuffer:flush added flush marker to queue');
+        }
       }
     }
+    else {
+      this.logger.debug(
+        `TtsStreamingBuffer:flush TTS stream is not connected, status: ${this._connectionStatus}`
+      );
+    }
   }
 
   clear() {
     this.logger.debug('TtsStreamingBuffer:clear');
-
     if (this._connectionStatus !== TtsStreamingConnectionStatus.Connected) return;
     clearTimeout(this.timer);
-    this._api(this.ep, [this.ep.uuid, 'clear'])
-      .catch((err) => this.logger.info({err}, 'TtsStreamingBuffer:clear Error clearing TTS streaming'));
-    this.tokens = '';
+    this._api(this.ep, [this.ep.uuid, 'clear']).catch((err) =>
+      this.logger.info({ err }, 'TtsStreamingBuffer:clear Error clearing TTS streaming')
+    );
+    this.queue = [];
+    this.bufferedLength = 0;
     this.timer = null;
     this._isFull = false;
   }
 
   /**
-   * Send tokens to the TTS engine in sentence chunks for best playout
+   * Process the queue in two phases.
+   *
+   * Phase 1: Look for flush markers. When a flush marker is found (even if not at the very front),
+   *   send all text tokens that came before it immediately (ignoring sentence boundaries)
+   *   and then send the flush command. Repeat until there are no flush markers left.
+   *
+   * Phase 2: With the remaining queue (now containing only text items), accumulate text
+   *   up to MAX_CHUNK_SIZE and use sentence-boundary logic to determine a chunk.
+   *   Then, remove the exact tokens (or portions thereof) that were consumed.
    */
-  async _feedTokens(handlingTimeout = false) {
-    this.logger.debug({tokens: this.tokens}, '_feedTokens');
-
+  async _feedQueue(handlingTimeout = false) {
+    this.logger.debug({ queue: this.queue }, 'TtsStreamingBuffer:_feedQueue');
     try {
-
-      /* are we in a state where we can feed tokens to the TTS? */
-      if (!this.cs.isTtsStreamOpen || !this.ep || !this.tokens) {
-        this.logger.debug('TTS stream is not open or no tokens to send');
-        return this.tokens?.length || 0;
+      if (!this.cs.isTtsStreamOpen || !this.ep) {
+        this.logger.debug('TtsStreamingBuffer:_feedQueue TTS stream is not open or no endpoint available');
+        return;
       }
-
-      if (this._connectionStatus === TtsStreamingConnectionStatus.NotConnected ||
-        this._connectionStatus === TtsStreamingConnectionStatus.Failed) {
-        this.logger.debug('TtsStreamingBuffer:_feedTokens TTS stream is not connected');
+      if (
+        this._connectionStatus === TtsStreamingConnectionStatus.NotConnected ||
+        this._connectionStatus === TtsStreamingConnectionStatus.Failed
+      ) {
+        this.logger.debug('TtsStreamingBuffer:_feedQueue TTS stream is not connected');
         return;
       }
 
-      if (this._connectionStatus === TtsStreamingConnectionStatus.Connecting) {
-        this.logger.debug('TtsStreamingBuffer:_feedTokens TTS stream is not ready, waiting for connect');
+      // --- Phase 1: Process flush markers ---
+      // Process any flush marker that isn’t in the very first position.
+      let flushIndex = this.queue.findIndex((item, idx) => item.type === 'flush' && idx > 0);
+      while (flushIndex !== -1) {
+        let flushText = '';
+        // Accumulate all text tokens preceding the flush marker.
+        for (let i = 0; i < flushIndex; i++) {
+          if (this.queue[i].type === 'text') {
+            flushText += this.queue[i].value;
+          }
+        }
+        // Remove those text items.
+        for (let i = 0; i < flushIndex; i++) {
+          const item = this.queue.shift();
+          if (item.type === 'text') {
+            this.bufferedLength -= item.value.length;
+          }
+        }
+        // Remove the flush marker (now at the front).
+        if (this.queue.length > 0 && this.queue[0].type === 'flush') {
+          this.queue.shift();
+        }
+        // Immediately send all accumulated text (ignoring sentence boundaries).
+        if (flushText.length > 0) {
+          const modifiedFlushText = flushText.replace(/\n\n/g, '\n \n');
+          try {
+            await this._api(this.ep, [this.ep.uuid, 'send', modifiedFlushText]);
+          } catch (err) {
+            this.logger.info({ err, flushText }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
+          }
+        }
+        // Send the flush command.
+        await this._doFlush();
+
+        flushIndex = this.queue.findIndex((item, idx) => item.type === 'flush' && idx > 0);
+      }
+
+      // If a flush marker is at the very front, process it.
+      while (this.queue.length > 0 && this.queue[0].type === 'flush') {
+        this.queue.shift();
+        await this._doFlush();
+      }
+
+      // --- Phase 2: Process remaining text tokens ---
+      if (this.queue.length === 0) {
+        this._removeTimer();
         return;
       }
 
-      /* must send at least one sentence */
-      const limit = Math.min(MAX_CHUNK_SIZE, this.tokens.length);
-      let chunkEnd = findSentenceBoundary(this.tokens, limit);
+      // Accumulate contiguous text tokens (from the front) up to MAX_CHUNK_SIZE.
+      let combinedText = '';
+      for (const item of this.queue) {
+        if (item.type !== 'text') break;
+        combinedText += item.value;
+        if (combinedText.length >= MAX_CHUNK_SIZE) break;
+      }
+      if (combinedText.length === 0) {
+        this._removeTimer();
+        return;
+      }
 
+      const limit = Math.min(MAX_CHUNK_SIZE, combinedText.length);
+      let chunkEnd = findSentenceBoundary(combinedText, limit);
       if (chunkEnd <= 0) {
         if (handlingTimeout) {
-          /* on a timeout we've left some tokens sitting around, so be more aggressive now in sending them */
-          chunkEnd = findWordBoundary(this.tokens, limit);
+          chunkEnd = findWordBoundary(combinedText, limit);
           if (chunkEnd <= 0) {
-            this.logger.debug('TtsStreamingBuffer:_feedTokens: no word boundary found');
             this._setTimerIfNeeded();
             return;
           }
-        }
-        else {
-          /* if we just received tokens, we wont send unless we have at least a full sentence */
-          this.logger.debug('TtsStreamingBuffer:_feedTokens: no sentence boundary found');
+        } else {
           this._setTimerIfNeeded();
           return;
         }
       }
+      const chunk = combinedText.slice(0, chunkEnd);
 
-      const chunk = this.tokens.slice(0, chunkEnd);
-      this.tokens = this.tokens.slice(chunkEnd);
+      // Now we iterate over the queue items
+      // and deduct their lengths until we've accounted for chunkEnd characters.
+      let remaining = chunkEnd;
+      let tokensProcessed = 0;
+      for (let i = 0; i < this.queue.length; i++) {
+        const token = this.queue[i];
+        if (token.type !== 'text') break;
+        if (remaining >= token.value.length) {
+          remaining -= token.value.length;
+          tokensProcessed = i + 1;
+        } else {
+          // Partially consumed token: update its value to remove the consumed part.
+          token.value = token.value.slice(remaining);
+          tokensProcessed = i;
+          remaining = 0;
+          break;
+        }
+      }
+      // Remove the fully consumed tokens from the front of the queue.
+      this.queue.splice(0, tokensProcessed);
+      this.bufferedLength -= chunkEnd;
 
-      /* freeswitch looks for sequence of 2 newlines to determine end of message, so insert a space */
       const modifiedChunk = chunk.replace(/\n\n/g, '\n \n');
-      await this._api(this.ep, [this.ep.uuid, 'send', modifiedChunk]);
-      this.logger.debug(`TtsStreamingBuffer:_feedTokens: sent ${chunk.length}, remaining: ${this.tokens.length}`);
+      this.logger.debug(`TtsStreamingBuffer:_feedQueue sending chunk to tts: ${modifiedChunk}`);
 
-      if (this._pendingFlush) {
-        this._doFlush();
-        this._pendingFlush = false;
+      try {
+        await this._api(this.ep, [this.ep.uuid, 'send', modifiedChunk]);
+      } catch (err) {
+        this.logger.info({ err, chunk }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
       }
 
-      if (this.isFull && this.tokens.length <= LOW_WATER_BUFFER_SIZE) {
-        this.logger.info('TtsStreamingBuffer throttling: TTS streaming buffer is no longer full - resuming');
+      if (this._isFull && this.bufferedLength <= LOW_WATER_BUFFER_SIZE) {
+        this.logger.info('TtsStreamingBuffer throttling: buffer is no longer full - resuming');
         this._isFull = false;
         this.emit(TtsStreamingEvents.Resume);
       }
-    } catch (err) {
-      this.logger.info({err}, 'TtsStreamingBuffer:_feedTokens Error sending TTS chunk');
-      this.tokens = '';
-    }
 
-    return;
+      return this._feedQueue();
+    } catch (err) {
+      this.logger.info({ err }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
+      this.queue = [];
+      this.bufferedLength = 0;
+    }
   }
 
   async _api(ep, args) {
     const apiCmd = `uuid_${this.vendor.startsWith('custom:') ? 'custom' : this.vendor}_tts_streaming`;
     const res = await ep.api(apiCmd, `^^|${args.join('|')}`);
     if (!res.body?.startsWith('+OK')) {
-      this.logger.info({args}, `Error calling ${apiCmd}: ${res.body}`);
+      this.logger.info({ args }, `Error calling ${apiCmd}: ${res.body}`);
       throw new Error(`Error calling ${apiCmd}: ${res.body}`);
     }
   }
 
-  _onConnectFailure(vendor) {
-    this.logger.info(`streaming tts connection failed to ${vendor}`);
-    this._connectionStatus = TtsStreamingConnectionStatus.Failed;
-    this.tokens = '';
-    this.emit(TtsStreamingEvents.ConnectFailure, {vendor});
-  }
-
   _doFlush() {
-    this._api(this.ep, [this.ep.uuid, 'flush'])
-      .catch((err) => this.logger.info({err},
-        `TtsStreamingBuffer:_doFlush Error flushing TTS streaming: ${JSON.stringify(err)}`));
+    return this._api(this.ep, [this.ep.uuid, 'flush'])
+      .then(() => this.logger.debug('TtsStreamingBuffer:_doFlush sent flush command'))
+      .catch((err) =>
+        this.logger.info(
+          { err },
+          `TtsStreamingBuffer:_doFlush Error flushing TTS streaming: ${JSON.stringify(err)}`
+        )
+      );
   }
 
   async _onConnect(vendor) {
-    this.logger.info(`streaming tts connection made to ${vendor}`);
+    this.logger.info(`TtsStreamingBuffer:_onConnect streaming tts connection made to ${vendor} successful`);
     this._connectionStatus = TtsStreamingConnectionStatus.Connected;
-    if (this.tokens.length > 0) {
-      await this._feedTokens();
-    }
-    if (this._flushPending) {
-      this.flush();
-      this._flushPending = false;
+    if (this.queue.length > 0) {
+      await this._feedQueue();
     }
   }
 
+  _onConnectFailure(vendor) {
+    this.logger.info(`TtsStreamingBuffer:_onConnectFailure streaming tts connection failed to ${vendor}`);
+    this._connectionStatus = TtsStreamingConnectionStatus.Failed;
+    this.queue = [];
+    this.bufferedLength = 0;
+    this.emit(TtsStreamingEvents.ConnectFailure, { vendor });
+  }
+
   _setTimerIfNeeded() {
-    if (this.tokens.length > 0 && !this.timer) {
+    if (this.bufferedLength > 0 && !this.timer) {
+      this.logger.debug({queue: this.queue},
+        `TtsStreamingBuffer:_setTimerIfNeeded setting timer because ${this.bufferedLength} buffered`);
       this.timer = setTimeout(this._onTimeout.bind(this), TIMEOUT_RETRY_MSECS);
     }
   }
 
+  _removeTimer() {
+    if (this.timer) {
+      this.logger.debug('TtsStreamingBuffer:_removeTimer clearing timer');
+      clearTimeout(this.timer);
+      this.timer = null;
+    }
+  }
+
   _onTimeout() {
-    this.logger.info('TtsStreamingBuffer:_onTimeout');
+    this.logger.debug('TtsStreamingBuffer:_onTimeout Timeout waiting for sentence boundary');
+    // Check if new text has been added since the timer was set.
+    const now = Date.now();
+    if (now - this.lastUpdateTime < TIMEOUT_RETRY_MSECS) {
+      this.logger.debug('TtsStreamingBuffer:_onTimeout New text received recently; postponing flush.');
+      this._setTimerIfNeeded();
+      return;
+    }
     this.timer = null;
-    this._feedTokens(true);
+    this._feedQueue(true);
   }
 
   _onTtsEmpty(vendor) {
-    this.emit(TtsStreamingEvents.Empty, {vendor});
+    this.emit(TtsStreamingEvents.Empty, { vendor });
   }
 
   addCustomEventListener(ep, event, handler) {
-    this.eventHandlers.push({ep, event, handler});
+    this.eventHandlers.push({ ep, event, handler });
     ep.addCustomEventListener(event, handler);
   }
 
@@ -274,7 +403,6 @@ class TtsStreamingBuffer extends Emitter {
 
   _initHandlers(ep) {
     [
-      // DH: add other vendors here as modules are added
       'deepgram',
       'cartesia',
       'elevenlabs',
@@ -293,23 +421,21 @@ class TtsStreamingBuffer extends Emitter {
 }
 
 const findSentenceBoundary = (text, limit) => {
-  // Match traditional sentence boundaries or double newlines
+  // Look for punctuation or double newline that signals sentence end.
   const sentenceEndRegex = /[.!?](?=\s|$)|\n\n/g;
   let lastSentenceBoundary = -1;
   let match;
-
   while ((match = sentenceEndRegex.exec(text)) && match.index < limit) {
-    const precedingText = text.slice(0, match.index).trim(); // Extract text before the match and trim whitespace
-    if (precedingText.length > 0) { // Check if there's actual content
+    const precedingText = text.slice(0, match.index).trim();
+    if (precedingText.length > 0) {
       if (
-        match[0] === '\n\n' || // It's a double newline
-        (match.index === 0 || !/\d$/.test(text[match.index - 1])) // Standard punctuation rules
+        match[0] === '\n\n' ||
+        (match.index === 0 || !/\d$/.test(text[match.index - 1]))
       ) {
-        lastSentenceBoundary = match.index + (match[0] === '\n\n' ? 2 : 1); // Include the boundary
+        lastSentenceBoundary = match.index + (match[0] === '\n\n' ? 2 : 1);
       }
     }
   }
-
   return lastSentenceBoundary;
 };
 
@@ -317,7 +443,6 @@ const findWordBoundary = (text, limit) => {
   const wordBoundaryRegex = /\s+/g;
   let lastWordBoundary = -1;
   let match;
-
   while ((match = wordBoundaryRegex.exec(text)) && match.index < limit) {
     lastWordBoundary = match.index;
   }

lib/utils/ws-requestor.js

@@ -431,6 +431,21 @@ class WsRequestor extends BaseRequestor {
       }
     } catch (err) {
       this.logger.info({err, content}, 'WsRequestor:_onMessage - invalid incoming message');
+      const params = {
+        msg: 'InvalidMessage',
+        details: err.message,
+        content: Buffer.from(content).toString('utf-8')
+      };
+      const {writeAlerts, AlertType} = this.Alerter;
+      writeAlerts({
+        account_sid: this.account_sid,
+        alert_type: AlertType.INVALID_APP_PAYLOAD,
+        target_sid: this.call_sid,
+        message: err.message,
+
+      }).catch((err) => this.logger.info({err}, 'Error generating alert for invalid message'));
+      this.request('jambonz:error', '/error', params)
+        .catch((err) => this.logger.debug({err}, 'WsRequestor:_onMessage - Error sending'));
     }
   }
 

package.json

@@ -31,9 +31,9 @@
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.8",
-    "@jambonz/speech-utils": "^0.2.1",
+    "@jambonz/speech-utils": "^0.2.3",
     "@jambonz/stats-collector": "^0.1.10",
-    "@jambonz/verb-specifications": "^0.0.94",
+    "@jambonz/verb-specifications": "^0.0.97",
     "@jambonz/time-series": "^0.2.13",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",
@@ -48,7 +48,7 @@
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
     "drachtio-fsmrf": "^4.0.1",
-    "drachtio-srf": "^5.0.1",
+    "drachtio-srf": "^5.0.2",
     "express": "^4.19.2",
     "express-validator": "^7.0.1",
     "moment": "^2.30.1",