update drachtio-srf

* support ultravox sends createCall response to app

* update type issue

Co-authored-by: Matt Hertogs <matthertogs@gmail.com>

---------

Co-authored-by: Matt Hertogs <matthertogs@gmail.com>

.gitignore

@@ -42,4 +42,5 @@ ecosystem.config.js
 test/credentials/*.json
 run-tests.sh
 run-coverage.sh
-.vscode
+.vscode
+.env

Dockerfile

@@ -1,4 +1,4 @@
-FROM --platform=linux/amd64 node:18.15-alpine3.16 as base
+FROM --platform=linux/amd64 node:20-alpine as base
 
 RUN apk --update --no-cache add --virtual .builds-deps build-base python3
 

app.js

@@ -25,10 +25,22 @@ const opts = {
 };
 const pino = require('pino');
 const logger = pino(opts, pino.destination({sync: false}));
-const {LifeCycleEvents, FS_UUID_SET_NAME} = require('./lib/utils/constants');
+const {LifeCycleEvents, FS_UUID_SET_NAME, SystemState, FEATURE_SERVER} = require('./lib/utils/constants');
 const installSrfLocals = require('./lib/utils/install-srf-locals');
 installSrfLocals(srf, logger);
 
+const writeSystemAlerts = srf.locals?.writeSystemAlerts;
+if (writeSystemAlerts) {
+  writeSystemAlerts({
+    system_component: FEATURE_SERVER,
+    state : SystemState.Online,
+    fields : {
+      detail: `feature-server with process_id ${process.pid} started`,
+      host: srf.locals?.ipv4
+    }
+  });
+}
+
 const {
   initLocals,
   createRootSpan,
@@ -100,12 +112,19 @@ createHttpListener(logger, srf)
   });
 
 
-setInterval(async() => {
+const monInterval = setInterval(async() => {
   srf.locals.stats.gauge('fs.sip.calls.count', sessionTracker.count);
-  // Checking system log level
-  const systemInformation = await srf.locals.dbHelpers.lookupSystemInformation();
-  if (systemInformation && systemInformation.log_level) {
-    logger.level = systemInformation.log_level;
+  try {
+    const systemInformation = await srf.locals.dbHelpers.lookupSystemInformation();
+    if (systemInformation && systemInformation.log_level) {
+      logger.level = systemInformation.log_level;
+    }
+  } catch (err) {
+    if (process.env.NODE_ENV === 'test') {
+      clearInterval(monInterval);
+      logger.error('all tests complete');
+    }
+    else logger.error({err}, 'Error checking system log level in database');
   }
 }, 20000);
 
@@ -117,13 +136,25 @@ const disconnect = () => {
     srf.locals.mediaservers?.forEach((ms) => ms.disconnect());
   });
 };
-
 process.on('SIGTERM', handle);
+process.on('SIGINT', handle);
 
-function handle(signal) {
+async function handle(signal) {
   const {removeFromSet} = srf.locals.dbHelpers;
   srf.locals.disabled = true;
   logger.info(`got signal ${signal}`);
+  const writeSystemAlerts = srf.locals?.writeSystemAlerts;
+  if (writeSystemAlerts) {
+    // it has to be synchronous call, or else by the time system saves the app terminates
+    await writeSystemAlerts({
+      system_component: FEATURE_SERVER,
+      state : SystemState.Offline,
+      fields : {
+        detail: `feature-server with process_id ${process.pid} stopped, signal ${signal}`,
+        host: srf.locals?.ipv4
+      }
+    });
+  }
   const setName = `${(JAMBONES_CLUSTER_ID || 'default')}:active-fs`;
   const fsServiceUrlSetName = `${(JAMBONES_CLUSTER_ID || 'default')}:fs-service-url`;
   if (setName && srf.locals.localSipAddress) {

lib/config.js

@@ -137,6 +137,8 @@ const JAMBONES_EAGERLY_PRE_CACHE_AUDIO = parseInt(process.env.JAMBONES_EAGERLY_P
 
 const JAMBONES_USE_FREESWITCH_TIMER_FD = process.env.JAMBONES_USE_FREESWITCH_TIMER_FD;
 const JAMBONES_DIAL_SBC_FOR_REGISTERED_USER = process.env.JAMBONES_DIAL_SBC_FOR_REGISTERED_USER || false;
+const JAMBONES_MEDIA_TIMEOUT_MS = process.env.JAMBONES_MEDIA_TIMEOUT_MS || 0;
+const JAMBONES_MEDIA_HOLD_TIMEOUT_MS = process.env.JAMBONES_MEDIA_HOLD_TIMEOUT_MS || 0;
 
 module.exports = {
   JAMBONES_MYSQL_HOST,
@@ -223,5 +225,7 @@ module.exports = {
   JAMBONZ_DISABLE_DIAL_PAI_HEADER,
   JAMBONES_DISABLE_DIRECT_P2P_CALL,
   JAMBONES_USE_FREESWITCH_TIMER_FD,
-  JAMBONES_DIAL_SBC_FOR_REGISTERED_USER
+  JAMBONES_DIAL_SBC_FOR_REGISTERED_USER,
+  JAMBONES_MEDIA_TIMEOUT_MS,
+  JAMBONES_MEDIA_HOLD_TIMEOUT_MS
 };

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
       };
 
@@ -208,10 +209,13 @@ router.post('/',
 
       /**
      * create our application object -
-     * not from the database as per an inbound call,
-     * but from the provided params in the request
+     * we merge the inbound call application,
+     * with the provided app params from the request body
      */
-      const app = req.body;
+      const app = {
+        ...application,
+        ...req.body
+      };
 
       /**
      * attach our requestor and notifier objects

lib/middleware.js

@@ -1,5 +1,5 @@
 const uuidv4 = require('uuid-random');
-const {CallDirection, AllowedSipRecVerbs} = require('./utils/constants');
+const {CallDirection, AllowedSipRecVerbs, WS_CLOSE_CODES} = require('./utils/constants');
 const {parseSiprecPayload} = require('./utils/siprec-utils');
 const CallInfo = require('./session/call-info');
 const HttpRequestor = require('./utils/http-requestor');
@@ -460,7 +460,7 @@ module.exports = function(srf, logger) {
       }).catch((err) => this.logger.info({err}, 'Error generating alert for parsing application'));
       logger.info({err}, `Error retrieving or parsing application: ${err?.message}`);
       res.send(480, {headers: {'X-Reason': err?.message || 'unknown'}});
-      app.requestor.close();
+      app.requestor.close(WS_CLOSE_CODES.GoingAway);
     }
   }
 

lib/session/adulting-call-session.js

@@ -45,8 +45,10 @@ class AdultingCallSession extends CallSession {
     return this.sd.ep;
   }
 
-  /* see note above */
-  set ep(newEp) {}
+  // When adulting session kicked from conference, replaceEndpoint is a must
+  set ep(newEp) {
+    this.sd.ep = newEp;
+  }
 
   get callSid() {
     return this.callInfo.callSid;

lib/session/call-session.js

@@ -2,13 +2,15 @@ const Emitter = require('events');
 const fs = require('fs');
 const {
   CallDirection,
+  MediaPath,
   TaskPreconditions,
   CallStatus,
   TaskName,
   KillReason,
   RecordState,
   AllowedSipRecVerbs,
-  AllowedConfirmSessionVerbs
+  AllowedConfirmSessionVerbs,
+  TtsStreamingEvents
 } = require('../utils/constants');
 const moment = require('moment');
 const assert = require('assert');
@@ -20,18 +22,22 @@ const listTaskNames = require('../utils/summarize-tasks');
 const HttpRequestor = require('../utils/http-requestor');
 const WsRequestor = require('../utils/ws-requestor');
 const ActionHookDelayProcessor = require('../utils/action-hook-delay');
+const TtsStreamingBuffer = require('../utils/tts-streaming-buffer');
 const {parseUri} = require('drachtio-srf');
 const {
   JAMBONES_INJECT_CONTENT,
   JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   AWS_REGION,
-  JAMBONES_USE_FREESWITCH_TIMER_FD
+  JAMBONES_USE_FREESWITCH_TIMER_FD,
+  JAMBONES_MEDIA_TIMEOUT_MS,
+  JAMBONES_MEDIA_HOLD_TIMEOUT_MS
 } = require('../config');
 const bent = require('bent');
 const BackgroundTaskManager = require('../utils/background-task-manager');
+const dbUtils = require('../utils/db-utils');
 const BADPRECONDITIONS = 'preconditions not met';
 const CALLER_CANCELLED_ERR_MSG = 'Response not sent due to unknown transaction';
-
+const { NonFatalTaskError} = require('../utils/error');
 const sqlRetrieveQueueEventHook = `SELECT * FROM webhooks 
 WHERE webhook_sid = 
 (
@@ -398,33 +404,41 @@ class CallSession extends Emitter {
     return this.application.transferredCall === true;
   }
 
+  /**
+   * returns true if this session is an inbound call session
+   */
+  get isInboundCallSession() {
+    return this.constructor.name === 'InboundCallSession';
+  }
+
   /**
    * returns true if this session is a ConfirmCallSession
    */
   get isAdultingCallSession() {
     return this.constructor.name === 'AdultingCallSession';
   }
-
-  /**
-   * returns true if this session is a ConfirmCallSession
-   */
   get isConfirmCallSession() {
     return this.constructor.name === 'ConfirmCallSession';
   }
-
-  /**
-   * returns true if this session is a SipRecCallSession
-   */
   get isSipRecCallSession() {
     return this.constructor.name === 'SipRecCallSession';
   }
-
-  /**
-   * returns true if this session is a SmsCallSession
-   */
   get isSmsCallSession() {
     return this.constructor.name === 'SmsCallSession';
   }
+  get isRestCallSession() {
+    return this.constructor.name === 'RestCallSession';
+  }
+  get InboundCallSession() {
+    return this.constructor.name === 'InboundCallSession';
+  }
+  get isNormalCallSession() {
+    return this.constructor.name === 'InboundCallSession' || this.constructor.name === 'RestCallSession';
+  }
+
+  get is3pccInvite() {
+    return this.isInboundCallSession && this.req?.body?.length === 0;
+  }
 
   get webhook_secret() {
     return this.accountInfo?.account?.webhook_secret;
@@ -438,6 +452,10 @@ class CallSession extends Emitter {
     return this.backgroundTaskManager.isTaskRunning('bargeIn');
   }
 
+  get isTtsStreamEnabled() {
+    return this.backgroundTaskManager.isTaskRunning('ttsStream');
+  }
+
   get isListenEnabled() {
     return this.backgroundTaskManager.isTaskRunning('listen');
   }
@@ -500,6 +518,10 @@ class CallSession extends Emitter {
     this._sipRequestWithinDialogHook = url;
   }
 
+  get isTtsStreamOpen() {
+    return this.currentTask?.isStreamingTts ||
+      this.backgroundTaskManager.getTask('ttsStream')?.isStreamingTts;
+  }
   // Bot Delay (actionHook delayed)
   get actionHookDelayEnabled() {
     return this._actionHookDelayEnabled;
@@ -550,7 +572,18 @@ class CallSession extends Emitter {
         this._actionHookDelayProcessor = new ActionHookDelayProcessor(this.logger, opts, this, this.ep);
         this._actionHookDelayProcessor.on('giveup', () => {
           this.logger.info('CallSession: ActionHookDelayProcessor: giveup event - hanging up call');
-          this._jambonzHangup();
+          const {writeAlerts} = this.srf.locals;
+          try {
+            writeAlerts({
+              alert_type: 'bot-action-delay-giveup',
+              account_sid: this.accountSid,
+              message: 'Call terminated due to bot action delay timeout',
+              target_sid: this.callSid
+            });
+          } catch (err) {
+            this.logger.error({err}, 'Error writing bot-action-delay-giveup alert');
+          }
+          this._jambonzHangup('bot-action-delay-giveup');
           if (this.wakeupResolver) {
             this.logger.debug('CallSession: Giveup timer expired - waking up');
             this.wakeupResolver({reason: 'noResponseGiveUp'});
@@ -574,6 +607,25 @@ class CallSession extends Emitter {
     }
   }
 
+  getTsStreamingVendor() {
+    let v;
+    if (this.currentTask?.isStreamingTts) {
+      const {vendor} = this.currentTask.getTtsVendorData(this);
+      v = vendor;
+    }
+    else if (this.backgroundTaskManager.getTask('ttsStream')?.isStreamingTts) {
+      const {vendor} = this.backgroundTaskManager.getTask('ttsStream').getTtsVendorData(this);
+      v = vendor;
+    }
+    return v;
+  }
+
+  get appIsUsingWebsockets() {
+    return this.requestor instanceof WsRequestor;
+  }
+
+  /* end of getters and setters */
+
   async clearOrRestoreActionHookDelayProcessor() {
     if (this._actionHookDelayProcessor) {
       await this._actionHookDelayProcessor.stop();
@@ -791,6 +843,38 @@ class CallSession extends Emitter {
     }
   }
 
+  async enableBackgroundTtsStream(say) {
+    try {
+      if (this.isTtsStreamEnabled) {
+        this.logger.debug('CallSession:enableBackgroundTtsStream - ttsStream currently enabled, ignoring request');
+      } else if (this.appIsUsingWebsockets && this.isNormalCallSession) {
+        await this.backgroundTaskManager.newTask('ttsStream', say);
+        this.logger.debug('CallSession:enableBackgroundTtsStream - ttsStream enabled');
+      } else {
+        this.logger.debug(
+          'CallSession:enableBackgroundTtsStream - ignoring request as call does not have required conditions');
+      }
+    }  catch (err) {
+      this.logger.info({err, say}, 'CallSession:enableBackgroundTtsStream - Error creating background tts stream task');
+    }
+  }
+
+  disableTtsStream() {
+    if (this.isTtsStreamEnabled) {
+      this.backgroundTaskManager.stop('ttsStream');
+      this.logger.debug('CallSession:disableTtsStream - ttsStream disabled');
+    }
+  }
+  clearTtsStream() {
+    this.requestor?.request('tts:streaming-event', '/streaming-event', {event_type: 'user_interruption'})
+      .catch((err) => this.logger.info({err}, 'CallSession:clearTtsStream - Error sending user_interruption'));
+    this.ttsStreamingBuffer?.clear();
+  }
+
+  startTtsStream() {
+    this.ttsStreamingBuffer?.start();
+  }
+
   async enableBotMode(gather, autoEnable) {
     try {
       let task;
@@ -979,6 +1063,14 @@ class CallSession extends Emitter {
             options: credential.options
           };
         }
+        else if ('cartesia' === vendor) {
+          return {
+            api_key: credential.api_key,
+            model_id: credential.model_id,
+            embedding: credential.embedding,
+            options: credential.options
+          };
+        }
         else if ('rimelabs' === vendor) {
           return {
             api_key: credential.api_key,
@@ -992,6 +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,
@@ -1017,7 +1115,8 @@ class CallSession extends Emitter {
             speech_credential_sid: credential.speech_credential_sid,
             auth_token: credential.auth_token,
             custom_stt_url: credential.custom_stt_url,
-            custom_tts_url: credential.custom_tts_url
+            custom_tts_url: credential.custom_tts_url,
+            custom_tts_streaming_url: credential.custom_tts_streaming_url
           };
         }
       }
@@ -1040,6 +1139,27 @@ class CallSession extends Emitter {
   async exec() {
     this.logger.info({tasks: listTaskNames(this.tasks)}, `CallSession:exec starting ${this.tasks.length} tasks`);
 
+    // calculate if inbandDTMF tone is used
+    const voip_carrier_sid = this.req?.has('X-Voip-Carrier-Sid') ? this.req.get('X-Voip-Carrier-Sid') :
+      this.req?.has('X-Requested-Carrier-Sid') ? this.req.get('X-Requested-Carrier-Sid') : null;
+
+    if (voip_carrier_sid) {
+      const {lookupVoipCarrierBySid} = dbUtils(this.logger, this.srf);
+      const [voipCarrier] = await lookupVoipCarrierBySid(voip_carrier_sid);
+      this.inbandDtmfEnabled = voipCarrier?.dtmf_type === 'tones';
+    }
+
+    if (this.isNormalCallSession) {
+      this.ttsStreamingBuffer = new TtsStreamingBuffer(this);
+      this.ttsStreamingBuffer.on(TtsStreamingEvents.Empty, this._onTtsStreamingEmpty.bind(this));
+      this.ttsStreamingBuffer.on(TtsStreamingEvents.Pause, this._onTtsStreamingPause.bind(this));
+      this.ttsStreamingBuffer.on(TtsStreamingEvents.Resume, this._onTtsStreamingResume.bind(this));
+      this.ttsStreamingBuffer.on(TtsStreamingEvents.ConnectFailure, this._onTtsStreamingConnectFailure.bind(this));
+    }
+    else {
+      this.logger.info(`CallSession:exec - not a normal call session: ${this.constructor.name}`);
+    }
+
     while (this.tasks.length && !this.callGone) {
       const taskNum = ++this.taskIdx;
       const stackNum = this.stackIdx;
@@ -1080,6 +1200,8 @@ class CallSession extends Emitter {
         this.currentTask = null;
         if (err.message?.includes(BADPRECONDITIONS)) {
           this.logger.info(`CallSession:exec task #${stackNum}:${taskNum}: ${task.name}: ${err.message}`);
+        } else if (err instanceof NonFatalTaskError) {
+          this.logger.error(err, `Error executing task  #${stackNum}:${taskNum}: ${task.name}`);
         }
         else {
           this.logger.error(err, `Error executing task  #${stackNum}:${taskNum}: ${task.name}`);
@@ -1172,6 +1294,11 @@ class CallSession extends Emitter {
       this.wakeupResolver({reason: 'session ended'});
       this.wakeupResolver = null;
     }
+
+    if (this._maxCallDurationTimer) {
+      clearTimeout(this._maxCallDurationTimer);
+      this._maxCallDurationTimer = null;
+    }
   }
 
   /**
@@ -1347,7 +1474,7 @@ class CallSession extends Emitter {
     if (!listenTask) {
       return this.logger.info('CallSession:_lccListenStatus - invalid listen_status: Dial does not have a listen');
     }
-    listenTask.updateListen(opts.listen_status);
+    listenTask.updateListen(opts.listen_status || opts.stream_status);
   }
 
   /**
@@ -1489,7 +1616,7 @@ Duration=${duration} `
     // this whole thing requires us to be in a Dial verb
     const task = this.currentTask;
     if (!task || ![TaskName.Dial, TaskName.Listen].includes(task.name)) {
-      return this.logger.info('CallSession:_lccWhisper - invalid command since we are not in a dial or listen');
+      return this.logger.info('CallSession:_lccWhisper - invalid command since we are not in a dial or stream/listen');
     }
 
     // allow user to provide a url object, a url string, an array of tasks, or a single task
@@ -1589,6 +1716,15 @@ Duration=${duration} `
     this.logger.info({response}, '_lccBoostAudioSignal: response from freeswitch');
   }
 
+  async _lccMediaPath(desiredPath) {
+    const task = this.currentTask;
+    if (!task || task.name !== TaskName.Dial) {
+      return this.logger.info('CallSession:_lccMediaPath - invalid command since we are not in a dial verb');
+    }
+    task.updateMediaPath(desiredPath)
+      .catch((err) => this.logger.error(err, 'CallSession:_lccMediaPath'));
+  }
+
   _lccToolOutput(tool_call_id, opts, callSid) {
     // only valid if we are in an LLM verb
     const task = this.currentTask;
@@ -1612,6 +1748,39 @@ Duration=${duration} `
       .catch((err) => this.logger.error(err, 'CallSession:_lccLlmUpdate'));
   }
 
+  async _lccTtsTokens(opts) {
+    const {id, tokens} = opts;
+
+    if (id === undefined) {
+      this.logger.info({opts}, 'CallSession:_lccTtsTokens - invalid command since id is missing');
+      return;
+    }
+    else if (tokens === undefined) {
+      this.logger.info({opts}, 'CallSession:_lccTtsTokens - invalid command since id is missing');
+      return this.requestor.request('tts:tokens-result', '/tokens-result', {
+        id,
+        status: 'failed',
+        reason: 'missing tokens'
+      }).catch((err) => this.logger.debug({err}, 'CallSession:_notifyTaskStatus - Error sending'));
+    }
+    let res;
+    try {
+      res = await this.ttsStreamingBuffer?.bufferTokens(tokens);
+      this.logger.info({id, res}, 'CallSession:_lccTtsTokens - tts:tokens-result');
+    } catch (err) {
+      this.logger.info(err, 'CallSession:_lccTtsTokens');
+    }
+    this.requestor.request('tts:tokens-result', '/tokens-result', {id, ...res})
+      .catch((err) => this.logger.debug({err}, 'CallSession:_notifyTaskStatus - Error sending'));
+  }
+
+  _lccTtsFlush(opts) {
+    this.ttsStreamingBuffer?.flush(opts);
+  }
+
+  _lccTtsClear(opts) {
+    this.ttsStreamingBuffer?.clear(opts);
+  }
 
   /**
    * perform call hangup by jambonz
@@ -1630,59 +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) {
-      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.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');
     }
   }
 
@@ -1943,6 +2119,7 @@ Duration=${duration} `
         break;
 
       case 'listen:status':
+      case 'stream:status':
         this._lccListenStatus(data);
         break;
 
@@ -1975,6 +2152,13 @@ Duration=${duration} `
           });
         break;
 
+      case 'media:path':
+        this._lccMediaPath(data, call_sid)
+          .catch((err) => {
+            this.logger.info({err, data}, 'CallSession:_onCommand - error setting media path');
+          });
+        break;
+
       case 'llm:tool-output':
         this._lccToolOutput(tool_call_id, data, call_sid);
         break;
@@ -1983,6 +2167,18 @@ Duration=${duration} `
         this._lccLlmUpdate(data, call_sid);
         break;
 
+      case 'tts:tokens':
+        this._lccTtsTokens(data);
+        break;
+
+      case 'tts:flush':
+        this._lccTtsFlush(data);
+        break;
+
+      case 'tts:clear':
+        this._lccTtsClear(data);
+        break;
+
       default:
         this.logger.info(`CallSession:_onCommand - invalid command ${command}`);
     }
@@ -2040,6 +2236,8 @@ Duration=${duration} `
         };
       }
 
+      this._enableInbandDtmfIfRequired(this.ep);
+
       // we are going from an early media connection to answer
       if (this.direction === CallDirection.Inbound) {
         // only do this for inbound call.
@@ -2063,7 +2261,6 @@ Duration=${duration} `
       //ep.cs = this;
       this.ep = ep;
       this.logger.info(`allocated endpoint ${ep.uuid}`);
-
       this._configMsEndpoint();
 
       this.ep.on('destroy', () => {
@@ -2176,6 +2373,8 @@ Duration=${duration} `
     // close all background tasks
     this.backgroundTaskManager.stopAll();
     this.clearOrRestoreActionHookDelayProcessor().catch((err) => {});
+
+    this.ttsStreamingBuffer?.stop();
   }
 
   /**
@@ -2225,22 +2424,43 @@ Duration=${duration} `
         },
         localSdp: this.ep.local.sdp
       });
-      this.logger.debug('answered call');
-      this.dlg.on('destroy', this._callerHungup.bind(this));
-      this.wrapDialog(this.dlg);
-      this.dlg.callSid = this.callSid;
-      this.emit('callStatusChange', {sipStatus: 200, sipReason: 'OK', callStatus: CallStatus.InProgress});
 
-      if (this.recordOptions && this.recordState === RecordState.RecordingOff) {
-        this.startRecording();
+      const tidyUp = () => {
+        this.dlg.on('destroy', this._callerHungup.bind(this));
+        this.wrapDialog(this.dlg);
+        this.dlg.callSid = this.callSid;
+        this.emit('callStatusChange', {sipStatus: 200, sipReason: 'OK', callStatus: CallStatus.InProgress});
+
+        if (this.recordOptions && this.recordState === RecordState.RecordingOff) {
+          this.startRecording();
+        }
+        this.dlg.on('modify', this._onReinvite.bind(this));
+        this.dlg.on('refer', this._onRefer.bind(this));
+        if (this.sipRequestWithinDialogHook) {
+          this.dlg.on('info', this._onRequestWithinDialog.bind(this));
+          this.dlg.on('message', this._onRequestWithinDialog.bind(this));
+        }
+        this.logger.debug(`CallSession:propagateAnswer - answered callSid ${this.callSid}`);
+      };
+
+      if (this.is3pccInvite) {
+        return new Promise((resolve, reject) => {
+          this.dlg.once('ack', (ackRequest) => {
+            this.logger.debug({body: ackRequest.body}, 'received ACK for 3pcc invite');
+            this.ep.modify(ackRequest.body)
+              .then(() => {
+                tidyUp();
+                return resolve();
+              })
+              .catch((err) => {
+                this.logger.info({err}, 'Error modifying endpoint with ACK');
+                reject(err);
+              });
+          });
+          this.logger.debug('this is a 3pcc invite');
+        });
       }
-      this.dlg.on('modify', this._onReinvite.bind(this));
-      this.dlg.on('refer', this._onRefer.bind(this));
-      if (this.sipRequestWithinDialogHook) {
-        this.dlg.on('info', this._onRequestWithinDialog.bind(this));
-        this.dlg.on('message', this._onRequestWithinDialog.bind(this));
-      }
-      this.logger.debug(`CallSession:propagateAnswer - answered callSid ${this.callSid}`);
+      tidyUp();
     }
     else {
       this.logger.debug('CallSession:propagateAnswer - call already answered - re-anchor media with a reinvite');
@@ -2274,7 +2494,7 @@ Duration=${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 {
@@ -2318,6 +2538,12 @@ Duration=${duration} `
       try {
         const to = parseUri(req.getParsedHeader('Refer-To').uri);
         const by = parseUri(req.getParsedHeader('Referred-By').uri);
+        const customHeaders = Object.keys(req.headers)
+          .filter((h) => h.toLowerCase().startsWith('x-'))
+          .reduce((acc, h) => {
+            acc[h] = req.get(h);
+            return acc;
+          }, {});
         const b3 = this.b3;
         const httpHeaders = b3 && {b3};
         const json = await this.requestor.request('verb:hook', this._referHook, {
@@ -2330,6 +2556,7 @@ Duration=${duration} `
             referred_by_user: by.scheme === 'tel' ? by.number : by.user,
             referring_call_sid: this.callSid,
             referred_call_sid: null,
+            ...customHeaders
           }
         }, httpHeaders);
 
@@ -2505,27 +2732,35 @@ Duration=${duration} `
     };
   }
 
-  async releaseMediaToSBC(remoteSdp) {
+  async releaseMediaToSBC(remoteSdp, releaseMediaEntirely) {
     assert(this.dlg && this.dlg.connected && this.ep && typeof remoteSdp === 'string');
     await this.dlg.modify(remoteSdp, {
       headers: {
-        'X-Reason': 'release-media'
+        'X-Reason': releaseMediaEntirely ? 'release-media-entirely' : 'release-media'
       }
     });
-    this.ep.destroy()
-      .then(() => this.ep = null)
-      .catch((err) => this.logger.error({err}, 'CallSession:releaseMediaToSBC: Error destroying endpoint'));
+    try {
+      await this.ep.destroy();
+    } catch (err) {
+      this.logger.error({err}, 'CallSession:releaseMediaToSBC: Error destroying endpoint');
+    }
+    this.ep = null;
   }
 
-  async reAnchorMedia() {
+  async reAnchorMedia(currentMediaRoute = MediaPath.PartialMedia) {
     assert(this.dlg && this.dlg.connected && !this.ep);
+
     this.ep = await this.ms.createEndpoint({remoteSdp: this.dlg.remote.sdp});
+    this._configMsEndpoint();
     await this.dlg.modify(this.ep.local.sdp, {
       headers: {
         'X-Reason': 'anchor-media'
       }
     });
-    this._configMsEndpoint();
+
+    if (currentMediaRoute === MediaPath.NoMedia) {
+      await this.ep.modify(this.dlg.remote.sdp);
+    }
   }
 
   async handleReinviteAfterMediaReleased(req, res) {
@@ -2592,15 +2827,40 @@ Duration=${duration} `
   }
 
   _configMsEndpoint() {
+    this._enableInbandDtmfIfRequired(this.ep);
+    this.ep.once('destroy', this._handleMediaTimeout.bind(this));
     const opts = {
       ...(this.onHoldMusic && {holdMusic: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`}),
-      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'})
+      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'}),
+      ...(JAMBONES_MEDIA_TIMEOUT_MS && {media_timeout: JAMBONES_MEDIA_TIMEOUT_MS}),
+      ...(JAMBONES_MEDIA_HOLD_TIMEOUT_MS && {media_hold_timeout: JAMBONES_MEDIA_HOLD_TIMEOUT_MS})
     };
     if (Object.keys(opts).length > 0) {
       this.ep.set(opts);
     }
   }
 
+  async _handleMediaTimeout(evt) {
+    if (evt.reason === 'MEDIA_TIMEOUT' && !this.callGone) {
+      this.logger.info('CallSession:_handleMediaTimeout: received MEDIA_TIMEOUT, hangup the call');
+      this._jambonzHangup('Media Timeout');
+    }
+  }
+
+  async _enableInbandDtmfIfRequired(ep) {
+    if (ep.inbandDtmfEnabled) return;
+    // only enable inband dtmf detection if voip carrier dtmf_type === tones
+    if (this.inbandDtmfEnabled) {
+      // https://developer.signalwire.com/freeswitch/FreeSWITCH-Explained/Modules/mod-dptools/6587132/#0-about
+      try {
+        ep.execute('start_dtmf');
+        ep.inbandDtmfEnabled = true;
+      } catch (err) {
+        this.logger.info(err, 'CallSession:_enableInbandDtmf - error enable inband DTMF');
+      }
+    }
+  }
+
   /**
    * notifyTaskError - only used when websocket connection is used instead of webhooks
    */
@@ -2680,6 +2940,58 @@ Duration=${duration} `
       this.verbHookSpan = null;
     }
   }
+
+  _onTtsStreamingEmpty() {
+    const task = this.currentTask;
+    if (task && TaskName.Say === task.name) {
+      task.notifyTtsStreamIsEmpty();
+    }
+  }
+
+  _onTtsStreamingPause() {
+    this.requestor?.request('tts:streaming-event', '/streaming-event', {event_type: 'stream_paused'})
+      .catch((err) => this.logger.info({err}, 'CallSession:_onTtsStreamingPause - Error sending'));
+  }
+
+  _onTtsStreamingResume() {
+    this.requestor?.request('tts:streaming-event', 'streaming-event', {event_type: 'stream_resumed'})
+      .catch((err) => this.logger.info({err}, 'CallSession:_onTtsStreamingResume - Error sending'));
+  }
+
+  async _onTtsStreamingConnectFailure(vendor) {
+    const {writeAlerts, AlertType} = this.srf.locals;
+    try {
+      await writeAlerts({
+        alert_type: AlertType.TTS_STREAMING_CONNECTION_FAILURE,
+        account_sid: this.accountSid,
+        vendor
+      });
+    } catch (error) {
+      this.logger.error({error}, 'Error writing WEBHOOK_CONNECTION_FAILURE alert');
+    }
+    this.logger.info({vendor}, 'CallSession:_onTtsStreamingConnectFailure - tts streaming connect failure');
+  }
+
+  async startMaxCallDurationTimer(timeLimit) {
+    if (!this._maxCallDurationTimer && timeLimit > 0) {
+      this.timeLimit = timeLimit;
+      this._maxCallDurationTimer = setTimeout(this._onMaxCallDuration.bind(this), timeLimit * 1000);
+      this.logger.debug(`CallSession:startMaxCallDurationTimer - started max call duration timer for ${timeLimit}s`);
+    }
+  }
+
+  /**
+   * _onMaxCallDuration - called when the call has reached the maximum duration
+   */
+  _onMaxCallDuration() {
+    this.logger.info(`callSession:_onMaxCallDuration tearing down call as it has reached ${this.timeLimit}s`);
+    if (!this.dlg) {
+      this.logger.debug('CallSession:_onMaxCallDuration - no dialog, call already gone');
+      return;
+    }
+    this._jambonzHangup('Max Call Duration');
+    this._maxCallDurationTimer = null;
+  }
 }
 
 module.exports = CallSession;

lib/session/confirm-call-session.js

@@ -8,7 +8,7 @@ const CallSession = require('./call-session');
 
  */
 class ConfirmCallSession extends CallSession {
-  constructor({logger, application, dlg, ep, tasks, callInfo, accountInfo, memberId, confName, rootSpan}) {
+  constructor({logger, application, dlg, ep, tasks, callInfo, accountInfo, memberId, confName, rootSpan, req}) {
     super({
       logger,
       application,
@@ -23,6 +23,7 @@ class ConfirmCallSession extends CallSession {
     });
     this.dlg = dlg;
     this.ep = ep;
+    this.req = req;
   }
 
   /**

lib/session/inbound-call-session.js

@@ -70,8 +70,12 @@ class InboundCallSession extends CallSession {
     this._hangup('caller');
   }
 
-  _jambonzHangup() {
-    this.dlg?.destroy();
+  _jambonzHangup(reason) {
+    this.dlg?.destroy({
+      headers: {
+        ...(reason && {'X-Reason': reason})
+      }
+    });
     // kill current task or wakeup the call session.
     this._callReleased();
   }

lib/tasks/conference.js

@@ -653,7 +653,8 @@ class Conference extends Task {
         memberId: this.memberId,
         confName: this.confName,
         tasks,
-        rootSpan: cs.rootSpan
+        rootSpan: cs.rootSpan,
+        req: cs.req
       });
       await this._playSession.exec();
       this._playSession = null;

lib/tasks/config.js

@@ -16,7 +16,8 @@ class TaskConfig extends Task {
       'fillerNoise',
       'actionHookDelayAction',
       'boostAudioSignal',
-      'vad'
+      'vad',
+      'ttsStream'
     ].forEach((k) => this[k] = this.data[k] || {});
 
     if ('notifyEvents' in this.data) {
@@ -45,6 +46,12 @@ class TaskConfig extends Task {
       };
       delete this.transcribeOpts.enable;
     }
+    if (this.ttsStream.enable) {
+      this.sayOpts = {
+        verb: 'say',
+        stream: true
+      };
+    }
 
     if (this.data.reset) {
       if (typeof this.data.reset === 'string') this.data.reset = [this.data.reset];
@@ -75,6 +82,7 @@ class TaskConfig extends Task {
   get hasVad() { return Object.keys(this.vad).length; }
   get hasFillerNoise() { return Object.keys(this.fillerNoise).length; }
   get hasReferHook() { return Object.keys(this.data).includes('referHook'); }
+  get hasTtsStream() { return Object.keys(this.ttsStream).length; }
 
   get summary() {
     const phrase = [];
@@ -106,6 +114,9 @@ class TaskConfig extends Task {
     if (this.onHoldMusic) phrase.push(`onHoldMusic: ${this.onHoldMusic}`);
     if ('boostAudioSignal' in this.data) phrase.push(`setGain ${this.data.boostAudioSignal}`);
     if (this.hasReferHook) phrase.push('set referHook');
+    if (this.hasTtsStream) {
+      phrase.push(`${this.ttsStream.enable ? 'enable' : 'disable'} ttsStream`);
+    }
     return `${this.name}{${phrase.join(',')}}`;
   }
 
@@ -176,18 +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;
 
@@ -305,6 +318,22 @@ class TaskConfig extends Task {
     if (this.hasReferHook) {
       cs.referHook = this.data.referHook;
     }
+
+    if (this.ttsStream.enable && this.sayOpts) {
+      this.sayOpts.synthesizer = this.hasSynthesizer ? this.synthesizer : {
+        vendor: cs.speechSynthesisVendor,
+        language: cs.speechSynthesisLanguage,
+        voice: cs.speechSynthesisVoice,
+        ...(cs.speechSynthesisLabel && {
+          label: cs.speechSynthesisLabel
+        })
+      };
+      this.logger.info({opts: this.gatherOpts}, 'Config: enabling ttsStream');
+      cs.enableBackgroundTtsStream(this.sayOpts);
+    } else if (!this.ttsStream.enable) {
+      this.logger.info('Config: disabling ttsStream');
+      cs.disableTtsStream();
+    }
   }
 
   async kill(cs) {

lib/tasks/dial.js

@@ -6,6 +6,7 @@ const {
   TaskName,
   TaskPreconditions,
   MAX_SIMRINGS,
+  MediaPath,
   KillReason
 } = require('../utils/constants');
 const assert = require('assert');
@@ -108,6 +109,7 @@ class TaskDial extends Task {
     this.proxy = this.data.proxy;
     this.tag = this.data.tag;
     this.boostAudioSignal = this.data.boostAudioSignal;
+    this._mediaPath = MediaPath.FullMedia;
 
     if (this.dtmfHook) {
       const {parentDtmfCollector, childDtmfCollector} = parseDtmfOptions(logger, this.data.dtmfCapture || {});
@@ -119,8 +121,9 @@ class TaskDial extends Task {
       }
     }
 
-    if (this.data.listen) {
-      this.listenTask = makeTask(logger, {'listen': this.data.listen}, this);
+    const listenData = this.data.listen || this.data.stream;
+    if (listenData) {
+      this.listenTask = makeTask(logger, {'listen': listenData }, this);
     }
     if (this.data.transcribe) {
       this.transcribeTask = makeTask(logger, {'transcribe' : this.data.transcribe}, this);
@@ -155,17 +158,21 @@ class TaskDial extends Task {
 
   get canReleaseMedia() {
     const keepAnchor = this.data.anchorMedia ||
-    this.cs.isBackGroundListen ||
-    this.cs.onHoldMusic ||
-    ANCHOR_MEDIA_ALWAYS ||
-    this.listenTask ||
-    this.dubTasks ||
-    this.transcribeTask ||
-    this.startAmd;
+      this.cs.isBackGroundListen ||
+      this.cs.onHoldMusic ||
+      ANCHOR_MEDIA_ALWAYS ||
+      this.listenTask ||
+      this.dubTasks ||
+      this.transcribeTask ||
+      this.startAmd;
 
     return !keepAnchor;
   }
 
+  get shouldExitMediaPathEntirely() {
+    return this.data.exitMediaPath;
+  }
+
   get summary() {
     if (this.target.length === 1) {
       const target = this.target[0];
@@ -186,6 +193,16 @@ class TaskDial extends Task {
 
   async exec(cs) {
     await super.exec(cs);
+
+    if (this.data.anchorMedia && this.data.exitMediaPath) {
+      this.logger.info('Dial:exec - incompatible anchorMedia and exitMediaPath are both set, will obey anchorMedia');
+      delete this.data.exitMediaPath;
+    }
+    if (!this.canReleaseMedia && this.data.exitMediaPath) {
+      this.logger.info(
+        'Dial:exec - exitMediaPath is set so features such as transcribe and record will not work on this call');
+    }
+
     try {
       if (this.listenTask) {
         const {span, ctx} = this.startChildSpan(`nested:${this.listenTask.summary}`);
@@ -213,10 +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);
           })();
         }
       }
@@ -257,7 +274,9 @@ class TaskDial extends Task {
     this._removeDtmfDetection(this.dlg);
     await this._killOutdials();
     if (this.sd) {
-      this.sd.kill();
+      const byeReasonHeader = this.killReason === KillReason.MediaTimeout ? 'Media Timeout' : undefined;
+      this.sd.kill(byeReasonHeader);
+      this.sd.ep?.removeListener('destroy', this._handleMediaTimeout.bind(this));
       this.sd.removeAllListeners();
       this.sd = null;
     }
@@ -303,7 +322,7 @@ class TaskDial extends Task {
       if (!cs.callGone && this.epOther) {
 
         /* if we can release the media back to the SBC, do so now */
-        if (this.canReleaseMedia) this._releaseMedia(cs, this.sd);
+        if (this.canReleaseMedia) this._releaseMedia(cs, this.sd, this.shouldExitMediaPathEntirely);
         else this.epOther.bridge(this.ep);
       }
     } catch (err) {
@@ -345,6 +364,12 @@ class TaskDial extends Task {
         const by = parseUri(req.getParsedHeader('Referred-By').uri);
         const referredBy = req.get('Referred-By');
         const userAgent = req.get('User-Agent');
+        const customHeaders = Object.keys(req.headers)
+          .filter((h) => h.toLowerCase().startsWith('x-'))
+          .reduce((acc, h) => {
+            acc[h] = req.get(h);
+            return acc;
+          }, {});
         this.logger.info({to}, 'refer to parsed');
         const json = await cs.requestor.request('verb:hook', this.referHook, {
           ...(callInfo.toJSON()),
@@ -355,7 +380,8 @@ class TaskDial extends Task {
             ...(userAgent && {sip_user_agent: userAgent}),
             ...(by && {referred_by_user: by.scheme === 'tel' ? by.number : by.user}),
             referring_call_sid,
-            referred_call_sid
+            referred_call_sid,
+            ...customHeaders
           }
         }, httpHeaders);
         if (json && Array.isArray(json)) {
@@ -496,7 +522,7 @@ class TaskDial extends Task {
     const {req, callInfo, direction, srf} = cs;
     const {getSBC} = srf.locals;
     const {lookupTeamsByAccount, lookupAccountBySid} = srf.locals.dbHelpers;
-    const {lookupCarrier, lookupCarrierByPhoneNumber} = dbUtils(this.logger, cs.srf);
+    const {lookupCarrier, lookupCarrierByPhoneNumber, lookupVoipCarrierBySid} = dbUtils(this.logger, cs.srf);
     let sbcAddress = this.proxy || getSBC();
     const teamsInfo = {};
     let fqdn;
@@ -515,6 +541,8 @@ class TaskDial extends Task {
       ...this.headers
     };
 
+    // default to inband dtmf if not specified
+    this.inbandDtmfEnabled = cs.inbandDtmfEnabled;
     // get calling user from From header
     const parsedFrom = req.getParsedHeader('from');
     const fromUri = parseUri(parsedFrom.uri);
@@ -592,10 +620,17 @@ class TaskDial extends Task {
           const str = this.callerId || req.callingNumber || '';
           const callingNumber = str.startsWith('+') ? str.substring(1) : str;
           const voip_carrier_sid = await lookupCarrierByPhoneNumber(cs.accountSid, callingNumber);
+          const req_voip_carrier_sid = req.has('X-Voip-Carrier-Sid') ? req.get('X-Voip-Carrier-Sid') : null;
           if (voip_carrier_sid) {
             this.logger.info(
               `Dial:_attemptCalls: selected voip_carrier_sid ${voip_carrier_sid} for callingNumber: ${callingNumber}`);
             opts.headers['X-Requested-Carrier-Sid'] = voip_carrier_sid;
+
+            // Checking if outbound carrier is different from inbound carrier and has dtmf type tones
+            if (voip_carrier_sid !== req_voip_carrier_sid) {
+              const [voipCarrier] = await lookupVoipCarrierBySid(voip_carrier_sid);
+              this.inbandDtmfEnabled = voipCarrier?.dtmf_type === 'tones';
+            }
           }
         }
 
@@ -752,7 +787,7 @@ class TaskDial extends Task {
             // Offhold, time to release media
             const newSdp = await this.ep.modify(req.body);
             await res.send(200, {body: newSdp});
-            await this._releaseMedia(this.cs, this.sd);
+            await this._releaseMedia(this.cs, this.sd, this.shouldExitMediaPathEntirely);
             this.isOutgoingLegHold = false;
           } else {
             this.logger.debug('Dial: _onReinvite receive unhold Request, update media server');
@@ -839,7 +874,7 @@ class TaskDial extends Task {
     if (cs.sipRequestWithinDialogHook) 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.epOther});
+    if (this.listenTask) this.listenTask.exec(cs, {ep: this.listenTask.channel === 2 ? this.ep : this.epOther});
     if (this.startAmd) {
       try {
         this.startAmd(cs, this.ep, this, this.data.amd);
@@ -861,7 +896,17 @@ class TaskDial extends Task {
     }
 
     /* if we can release the media back to the SBC, do so now */
-    if (this.canReleaseMedia) setTimeout(this._releaseMedia.bind(this, cs, sd), 200);
+    if (this.canReleaseMedia || this.shouldExitMediaPathEntirely) {
+      setTimeout(this._releaseMedia.bind(this, cs, sd, this.shouldExitMediaPathEntirely), 200);
+    }
+
+    this.sd.ep.once('destroy', this._handleMediaTimeout.bind(this));
+  }
+
+  _handleMediaTimeout(evt) {
+    if (evt.reason === 'MEDIA_TIMEOUT' && this.sd && this.bridged) {
+      this.kill(this.cs, KillReason.MediaTimeout);
+    }
   }
 
   _bridgeEarlyMedia(sd) {
@@ -873,22 +918,57 @@ class TaskDial extends Task {
     }
   }
 
+  /* public api */
+  async updateMediaPath(desiredPath) {
+    this.logger.info(`Dial:updateMediaPath - ${this._mediaPath} => ${desiredPath}`);
+    switch (desiredPath) {
+      case MediaPath.NoMedia:
+        assert(this._mediaPath !== MediaPath.NoMedia, 'updateMediaPath: already no-media');
+        await this._releaseMedia(this.cs, this.sd, true);
+        break;
+
+      case MediaPath.PartialMedia:
+        assert(this._mediaPath !== MediaPath.PartialMedia, 'updateMediaPath: already partial-media');
+        if (this._mediaPath === MediaPath.FullMedia) {
+          await this._releaseMedia(this.cs, this.sd, false);
+        }
+        else {
+          // to go from no-media to partial-media we need to go through full-media first
+          await this.reAnchorMedia(this.cs, this.sd);
+          await this._releaseMedia(this.cs, this.sd, false);
+        }
+        assert(!this.epOther, 'updateMediaPath: epOther should be null');
+        assert(!this.ep, 'updateMediaPath: ep should be null');
+
+        break;
+      case MediaPath.FullMedia:
+        assert(this._mediaPath !== MediaPath.FullMedia, 'updateMediaPath: already full-media');
+        await this.reAnchorMedia(this.cs, this.sd);
+        break;
+
+      default:
+        assert(false, `updateMediaPath: invalid path request ${desiredPath}`);
+    }
+  }
+
   /**
    * Release the media from freeswitch
    * @param {*} cs
    * @param {*} sd
    */
-  async _releaseMedia(cs, sd) {
+  async _releaseMedia(cs, sd, releaseEntirely = false) {
     assert(cs.ep && sd.ep);
 
     try {
       // Wait until we got new SDP from B leg to ofter to A Leg
       const aLegSdp = cs.ep.remote.sdp;
-      await sd.releaseMediaToSBC(aLegSdp, cs.ep.local.sdp);
+      await sd.releaseMediaToSBC(aLegSdp, cs.ep.local.sdp, releaseEntirely);
       const bLegSdp = sd.dlg.remote.sdp;
-      await cs.releaseMediaToSBC(bLegSdp);
+      await cs.releaseMediaToSBC(bLegSdp, releaseEntirely);
       this.epOther = null;
-      this.logger.info('Dial:_releaseMedia - successfully released media from freewitch');
+      this._mediaPath = releaseEntirely ? MediaPath.NoMedia : MediaPath.PartialMedia;
+      this.logger.info(
+        `Dial:_releaseMedia - successfully released media from freewitch, media path is now ${this._mediaPath}`);
     } catch (err) {
       this.logger.info({err}, 'Dial:_releaseMedia error');
     }
@@ -898,8 +978,14 @@ class TaskDial extends Task {
     if (cs.ep && sd.ep) return;
 
     this.logger.info('Dial:reAnchorMedia - re-anchoring media to freewitch');
-    await Promise.all([sd.reAnchorMedia(), cs.reAnchorMedia()]);
+    await Promise.all([sd.reAnchorMedia(this._mediaPath), cs.reAnchorMedia(this._mediaPath)]);
     this.epOther = cs.ep;
+
+    this.epOther.bridge(this.ep);
+
+    this._mediaPath = MediaPath.FullMedia;
+    this.logger.info(
+      `Dial:_releaseMedia - successfully re-anchored media to freewitch, media path is now ${this._mediaPath}`);
   }
 
   // Handle RE-INVITE hold from caller leg.
@@ -918,11 +1004,12 @@ class TaskDial extends Task {
         }
         this._onHoldHook(req);
       } else if (!isOnhold(req.body)) {
-        if (this.epOther && this.ep && this.isIncomingLegHold && this.canReleaseMedia) {
+        if (this.epOther && this.ep && this.isIncomingLegHold &&
+          (this.canReleaseMedia || this.shouldExitMediaPathEntirely)) {
           // Offhold, time to release media
           const newSdp = await this.epOther.modify(req.body);
           await res.send(200, {body: newSdp});
-          await this._releaseMedia(this.cs, this.sd);
+          await this._releaseMedia(this.cs, this.sd, this.shouldExitMediaPathEntirely);
           isHandled = true;
         }
         this.isIncomingLegHold = false;
@@ -981,7 +1068,8 @@ class TaskDial extends Task {
               callInfo: this.cs.callInfo,
               accountInfo: this.cs.accountInfo,
               tasks,
-              rootSpan: this.cs.rootSpan
+              rootSpan: this.cs.rootSpan,
+              req: this.cs.req
             });
             await this._onHoldSession.exec();
             this._onHoldSession = null;

lib/tasks/enqueue.js

@@ -369,7 +369,8 @@ class TaskEnqueue extends Task {
         callInfo: cs.callInfo,
         accountInfo: cs.accountInfo,
         tasks: tasksToRun,
-        rootSpan: cs.rootSpan
+        rootSpan: cs.rootSpan,
+        req: cs.req
       });
       await this._playSession.exec();
       this._playSession = null;

lib/tasks/gather.js

@@ -11,6 +11,7 @@ const {
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
   AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
   VadDetection,
   VerbioTranscriptionEvents,
   SpeechmaticsTranscriptionEvents
@@ -24,6 +25,7 @@ const makeTask = require('./make_task');
 const assert = require('assert');
 const SttTask = require('./stt-task');
 const { SpeechCredentialError } = require('../utils/error');
+const SPEECHMATICS_DEFAULT_ASR_TIMEOUT = 1200;
 
 class TaskGather extends SttTask {
   constructor(logger, opts, parentTask) {
@@ -155,12 +157,23 @@ class TaskGather extends SttTask {
       const {hints, hintsBoost} = cs.globalSttHints;
       const setOfHints = new Set((this.data.recognizer.hints || [])
         .concat(hints)
-        .filter((h) => typeof h === 'string' && h.length > 0));
+        // allow for hints to be an array of object
+        .filter((h) => (typeof h === 'string' && h.length > 0) || (typeof h === 'object')));
       this.data.recognizer.hints = [...setOfHints];
       if (!this.data.recognizer.hintsBoost && hintsBoost) this.data.recognizer.hintsBoost = hintsBoost;
       this.logger.debug({hints: this.data.recognizer.hints, hintsBoost: this.data.recognizer.hintsBoost},
         'Gather:exec - applying global sttHints');
     }
+
+    // specials case for speechmatics: they dont do endpointing so we need to enable continuous ASR
+    if (this.vendor === 'speechmatics' && !this.isContinuousAsr) {
+      const maxDelay = this.recognizer?.speechmaticsOptions?.transcription_config?.max_delay;
+      if (maxDelay) this.asrTimeout = Math.min(SPEECHMATICS_DEFAULT_ASR_TIMEOUT, maxDelay * 1000);
+      else this.asrTimeout = SPEECHMATICS_DEFAULT_ASR_TIMEOUT;
+      this.isContinuousAsr = true;
+      this.logger.debug(`Gather:exec - auto-enabling continuous ASR for speechmatics w/ timeout ${this.asrTimeout}`);
+    }
+
     if (!this.isContinuousAsr && cs.isContinuousAsr) {
       this.isContinuousAsr = true;
       this.asrTimeout = cs.asrTimeout * 1000;
@@ -342,6 +355,13 @@ class TaskGather extends SttTask {
       this._killAudio(cs);
       this.emit('dtmf', evt);
     }
+    if (this.isContinuousAsr && evt.dtmf === this.asrDtmfTerminationDigit && this._bufferedTranscripts.length > 0) {
+      this.logger.info(`continuousAsr triggered with dtmf ${this.asrDtmfTerminationDigit}`);
+      this._clearAsrTimer();
+      this._clearTimer();
+      this._startFinalAsrTimer();
+      return;
+    }
     if (evt.dtmf === this.finishOnKey && this.input.includes('digits')) {
       resolved = true;
       this._resolve('dtmf-terminator-key');
@@ -364,13 +384,6 @@ class TaskGather extends SttTask {
         this._resolve('dtmf-num-digits');
       }
     }
-    else if (this.isContinuousAsr && evt.dtmf === this.asrDtmfTerminationDigit) {
-      this.logger.info(`continuousAsr triggered with dtmf ${this.asrDtmfTerminationDigit}`);
-      this._clearAsrTimer();
-      this._clearTimer();
-      this._startFinalAsrTimer();
-      return;
-    }
     if (!resolved && this.interDigitTimeout > 0 && this.digitBuffer.length >= this.minDigits) {
       /* start interDigitTimer */
       const ms = this.interDigitTimeout * 1000;
@@ -512,6 +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(
@@ -612,6 +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`);
@@ -722,6 +753,7 @@ class TaskGather extends SttTask {
         this._fillerNoiseOn = false;  // in a race, if we just started audio it may sneak through here
         this.ep.api('uuid_break', this.ep.uuid)
           .catch((err) => this.logger.info(err, 'Error killing audio'));
+        cs.clearTtsStream();
       }
       return;
     }
@@ -753,10 +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;
     }
@@ -767,7 +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');
@@ -775,8 +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 ' +
@@ -830,7 +866,7 @@ class TaskGather extends SttTask {
         const t = evt.alternatives[0].transcript;
         if (t) {
           /* remove trailing punctuation */
-          if (/[,;:\.!\?]$/.test(t)) {
+          if (this.vendor !== 'speechmatics' && /[,;:\.!\?]$/.test(t)) {
             this.logger.debug('TaskGather:_onTranscription - removing trailing punctuation');
             evt.alternatives[0].transcript = t.slice(0, -1);
           }
@@ -846,7 +882,10 @@ class TaskGather extends SttTask {
         this._startAsrTimer();
 
         /* some STT engines will keep listening after a final response, so no need to restart */
-        if (!['soniox', 'aws', 'microsoft', 'deepgram'].includes(this.vendor)) this._startTranscribing(ep);
+        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'speechmatics'].includes(this.vendor) &&
+          !this.vendor.startsWith('custom')) {
+          this._startTranscribing(ep);
+        }
       }
       else {
         /* this was removed to fix https://github.com/jambonz/jambonz-feature-server/issues/783 */
@@ -889,8 +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) {
@@ -1160,13 +1212,12 @@ 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*/ }
 
     // Gather got response from hook, cancel actionHookDelay processing
-    this.logger.debug('TaskGather:_resolve - checking ahd');
     if (this.cs.actionHookDelayProcessor) {
       if (returnedVerbs) {
         this.logger.debug('TaskGather:_resolve - got response from action hook, cancelling actionHookDelay');

lib/tasks/listen.js

@@ -17,7 +17,7 @@ class TaskListen extends Task {
 
     [
       'action', 'auth', 'method', 'url', 'finishOnKey', 'maxLength', 'metadata', 'mixType', 'passDtmf', 'playBeep',
-      'sampleRate', 'timeout', 'transcribe', 'wsAuth', 'disableBidirectionalAudio'
+      'sampleRate', 'timeout', 'transcribe', 'wsAuth', 'disableBidirectionalAudio', 'channel'
     ].forEach((k) => this[k] = this.data[k]);
 
     this.mixType = this.mixType || 'mono';

lib/tasks/llm/index.js

@@ -1,6 +1,9 @@
 const Task = require('../task');
 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) {
@@ -40,10 +43,22 @@ class TaskLlm extends Task {
     switch (this.vendor) {
       case 'openai':
       case 'microsoft':
-        if (this.model.startsWith('gpt-4o-realtime')) {
-          llm = new TaskLlmOpenAI_S2S(this.logger, this.data, this);
-        }
+        llm = new TaskLlmOpenAI_S2S(this.logger, this.data, this);
         break;
+
+      case 'voiceagent':
+      case 'deepgram':
+        llm = new TaskLlmVoiceAgent_S2S(this.logger, this.data, this);
+        break;
+
+      case 'ultravox':
+        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/openai_s2s.js

@@ -59,7 +59,7 @@ class TaskLlmOpenAI_S2S extends Task {
     this.parent = parentTask;
 
     this.vendor = this.parent.vendor;
-    this.model = this.parent.model;
+    this.model = this.parent.model || 'gpt-4o-realtime-preview-2024-12-17';
     this.auth = this.parent.auth;
     this.connectionOptions = this.parent.connectOptions;
 
@@ -120,9 +120,9 @@ class TaskLlmOpenAI_S2S extends Task {
 
     switch (this.vendor) {
       case 'openai':
-        return 'v1/realtime?model=gpt-4o-realtime-preview-2024-10-01';
+        return `v1/realtime?model=${this.model}`;
       case 'microsoft':
-        return 'openai/realtime?api-version=2024-10-01-preview&deployment=gpt-4o-realtime-preview-1001&';
+        return `openai/realtime?api-version=2024-10-01-preview&deployment=${this.model}`;
     }
   }
 

lib/tasks/llm/llms/ultravox_s2s.js

@@ -0,0 +1,256 @@
+const Task = require('../../task');
+const TaskName = 'Llm_Ultravox_s2s';
+const {request} = require('undici');
+const {LlmEvents_Ultravox} = require('../../../utils/constants');
+
+const ultravox_server_events = [
+  'createCall',
+  'pong',
+  'state',
+  'transcript',
+  'conversationText',
+  'clientToolInvocation',
+  'playbackClearBuffer',
+];
+
+const ClientEvent = 'client.event';
+
+const expandWildcards = (events) => {
+  // no-op for deepgram
+  return events;
+};
+
+const SessionDelete = 'session.delete';
+
+class TaskLlmUltravox_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.model = this.parent.model || 'fixie-ai/ultravox';
+    this.auth = this.parent.auth;
+    this.connectionOptions = this.parent.connectOptions;
+
+    const {apiKey} = this.auth || {};
+    if (!apiKey) throw new Error('auth.apiKey is required for Vendor: Ultravox');
+    this.apiKey = apiKey;
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+
+    /**
+     * 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 || ultravox_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_ultravox_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error(`Error calling uuid_ultravox_s2s: ${JSON.stringify(res.body)}`);
+    }
+  }
+
+  async createCall() {
+    const payload = {
+      ...this.data.llmOptions,
+      model: this.model,
+      medium: {
+        ...(this.data.llmOptions.medium || {}),
+        serverWebSocket: {
+          inputSampleRate: 8000,
+          outputSampleRate: 8000,
+        }
+      }
+    };
+    const {statusCode, body} = await request('https://api.ultravox.ai/api/calls', {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+        'X-API-Key': this.apiKey
+      },
+      body: JSON.stringify(payload)
+    });
+    const data = await body.json();
+    if (statusCode !== 201 || !data?.joinUrl) {
+      this.logger.error({statusCode, data}, 'Ultravox Error registering call');
+      throw new Error(`Ultravox  Error registering call: ${data.message}`);
+    }
+    this.logger.info({joinUrl: data.joinUrl}, 'Ultravox Call registered');
+    return data;
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    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();
+    }
+  }
+
+  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}, 'TaskLlmUltravox_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.debug('TaskLlmUltravox_S2S:_onConnect');
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmUltravox_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmUltravox_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}, 'TaskLlmUltravox_S2S:_onServerEvent');
+
+    /* server errors of some sort */
+    if (type === 'error') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server error',
+        error: evt.error
+      };
+    }
+
+    /* tool calls */
+    else if (type === 'client_tool_invocation') {
+      this.logger.debug({evt}, 'TaskLlmUltravox_S2S:_onServerEvent - function_call');
+      if (!this.toolHook) {
+        this.logger.warn({evt}, 'TaskLlmUltravox_S2S:_onServerEvent - no toolHook defined!');
+      }
+      else {
+        const {toolName: name, invocationId: call_id, parameters: args} = evt;
+
+        try {
+          await this.parent.sendToolHook(call_id, {name, args});
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmUltravox_S2S - error calling function');
+          this.results = {
+            completionReason: 'client error calling function',
+            error: err
+          };
+          endConversation = true;
+        }
+      }
+    }
+
+    this._sendLlmEvent(type, evt);
+
+    if (endConversation) {
+      this.logger.info({results: this.results},
+        'TaskLlmUltravox_S2S:_onServerEvent - ending conversation due to error');
+      this.notifyTaskDone();
+    }
+  }
+
+  _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');
+
+      if (!data.type || data.type !== 'client_tool_result') {
+        this.logger.info({data},
+          'TaskLlmUltravox_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}, 'TaskLlmUltravox_S2S:processToolOutput');
+    }
+  }
+
+  _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 = ultravox_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
+    }, 'TaskLlmUltravox_S2S:_populateEvents');
+  }
+}
+
+module.exports = TaskLlmUltravox_S2S;

lib/tasks/llm/llms/voice_agent_s2s.js

@@ -0,0 +1,313 @@
+const Task = require('../../task');
+const TaskName = 'Llm_VoiceAgent_s2s';
+const {LlmEvents_VoiceAgent} = require('../../../utils/constants');
+const ClientEvent = 'client.event';
+const SessionDelete = 'session.delete';
+
+const va_server_events = [
+  'Error',
+  'Welcome',
+  'SettingsApplied',
+  'ConversationText',
+  'UserStartedSpeaking',
+  'EndOfThought',
+  'AgentThinking',
+  'FunctionCallRequest',
+  'FunctionCalling',
+  'AgentStartedSpeaking',
+  'AgentAudioDone',
+];
+
+const expandWildcards = (events) => {
+  // no-op for deepgram
+  return events;
+};
+
+class TaskLlmVoiceAgent_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.model = this.parent.model;
+    this.auth = this.parent.auth;
+    this.connectionOptions = this.parent.connectOptions;
+
+    const {apiKey} = this.auth || {};
+    if (!apiKey) throw new Error('auth.apiKey is required for VoiceAgent S2S');
+
+    this.apiKey = apiKey;
+    this.authType = 'bearer';
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+    const {settingsConfiguration} = this.data.llmOptions;
+
+    if (typeof settingsConfiguration !== 'object') {
+      throw new Error('llmOptions with an initial settingsConfiguration is required for VoiceAgent S2S');
+    }
+
+    // eslint-disable-next-line no-unused-vars
+    const {audio, ...rest} = settingsConfiguration;
+    const cfg = this.settingsConfiguration = rest;
+
+    if (!cfg.agent) throw new Error('llmOptions.settingsConfiguration.agent is required for VoiceAgent S2S');
+    if (!cfg.agent.think) {
+      throw new Error('llmOptions.settingsConfiguration.agent.think is required for VoiceAgent S2S');
+    }
+    if (!cfg.agent.think.model) {
+      throw new Error('llmOptions.settingsConfiguration.agent.think.model is required for VoiceAgent S2S');
+    }
+    if (!cfg.agent.think.provider?.type) {
+      throw new Error('llmOptions.settingsConfiguration.agent.think.provider.type is required for VoiceAgent S2S');
+    }
+
+    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 || va_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  get host() {
+    const {host} = this.connectionOptions || {};
+    return host || 'agent.deepgram.com';
+  }
+
+  get path() {
+    const {path} = this.connectionOptions || {};
+    if (path) return path;
+
+    return '/agent';
+  }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_voice_agent_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error(`Error calling uuid_voice_agent_s2s: ${JSON.stringify(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}, 'TaskLlmVoiceAgent_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  /**
+   * Send function call response to the VoiceAgent server
+   */
+  async processToolOutput(ep, tool_call_id, data) {
+    try {
+      const {data:response} = data;
+      this.logger.debug({tool_call_id, response}, 'TaskLlmVoiceAgent_S2S:processToolOutput');
+
+      if (!response.type || response.type !== 'FunctionCallResponse') {
+        this.logger.info({response},
+          'TaskLlmVoiceAgent_S2S:processToolOutput - invalid tool output, must be FunctionCallResponse');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(response)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmVoiceAgent_S2S:processToolOutput');
+    }
+  }
+
+  /**
+   * Send a session.update to the VoiceAgent server
+   * Note: creating and deleting conversation items also supported as well as interrupting the assistant
+   */
+  async processLlmUpdate(ep, data, _callSid) {
+    try {
+      this.logger.debug({data, _callSid}, 'TaskLlmVoiceAgent_S2S:processLlmUpdate');
+
+      if (!data.type || ![
+        'UpdateInstructions',
+        'UpdateSpeak',
+        'InjectAgentMessage',
+      ].includes(data.type)) {
+        this.logger.info({data}, 'TaskLlmVoiceAgent_S2S:processLlmUpdate - invalid mid-call request');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmVoiceAgent_S2S:processLlmUpdate');
+    }
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    try {
+      const args = [ep.uuid, 'session.create', this.host, this.path, this.authType, this.apiKey];
+      await this._api(ep, args);
+    } catch (err) {
+      this.logger.error({err}, `TaskLlmVoiceAgent_S2S:_startListening: ${JSON.stringify(err)}`);
+      this.notifyTaskDone();
+    }
+  }
+
+  async _sendClientEvent(ep, obj) {
+    let ok = true;
+    this.logger.debug({obj}, 'TaskLlmVoiceAgent_S2S:_sendClientEvent');
+    try {
+      const args = [ep.uuid, ClientEvent, JSON.stringify(obj)];
+      await this._api(ep, args);
+    } catch (err) {
+      ok = false;
+      this.logger.error({err}, 'TaskLlmVoiceAgent_S2S:_sendClientEvent - Error');
+    }
+    return ok;
+  }
+
+  async _sendInitialMessage(ep) {
+    if (!await this._sendClientEvent(ep, this.settingsConfiguration)) {
+      this.notifyTaskDone();
+    }
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _onError(_ep, evt) {
+    this.logger.info({evt}, 'TaskLlmVoiceAgent_S2S:_onError');
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.debug('TaskLlmVoiceAgent_S2S:_onConnect');
+    this._sendInitialMessage(ep);
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmVoiceAgent_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmVoiceAgent_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}, 'TaskLlmVoiceAgent_S2S:_onServerEvent');
+
+    /* check for failures, such as rate limit exceeded, that should terminate the conversation */
+    if (type === 'response.done' && evt.response.status === 'failed') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server failure',
+        error: evt.response.status_details?.error
+      };
+    }
+
+    /* server errors of some sort */
+    else if (type === 'error') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server error',
+        error: evt.error
+      };
+    }
+
+    /* tool calls */
+    else if (type === 'FunctionCallRequest') {
+      this.logger.debug({evt}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - function_call');
+      if (!this.toolHook) {
+        this.logger.warn({evt}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - no toolHook defined!');
+      }
+      else {
+        const {function_name:name, function_call_id:call_id} = evt;
+        const args = evt.input;
+
+        try {
+          await this.parent.sendToolHook(call_id, {name, args});
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmVoiceAgent - 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}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - error sending event hook'));
+    }
+
+    if (endConversation) {
+      this.logger.info({results: this.results},
+        'TaskLlmVoiceAgent_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 = va_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
+    }, 'TaskLlmVoiceAgent_S2S:_populateEvents');
+  }
+}
+
+module.exports = TaskLlmVoiceAgent_S2S;

lib/tasks/make_task.js

@@ -84,6 +84,7 @@ function makeTask(logger, obj, parent) {
       const TaskTranscribe = require('./transcribe');
       return new TaskTranscribe(logger, data, parent);
     case TaskName.Listen:
+    case TaskName.Stream:
       const TaskListen = require('./listen');
       return new TaskListen(logger, data, parent);
     case TaskName.Redirect:

lib/tasks/play.js

@@ -1,5 +1,6 @@
 const Task = require('./task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
+const { PlayFileNotFoundError } = require('../utils/error');
 
 class TaskPlay extends Task {
   constructor(logger, opts) {
@@ -66,8 +67,20 @@ class TaskPlay extends Task {
         }
       }
     } catch (err) {
-      if (timeout) clearTimeout(timeout);
-      this.logger.info(err, `TaskPlay:exec - error playing ${this.url}`);
+      this.logger.info(`TaskPlay:exec - error playing ${this.url}: ${err.message}`);
+      this.playComplete = true;
+      if (err.message === 'File Not Found') {
+        const {writeAlerts, AlertType} = cs.srf.locals;
+        await this.performAction({status: 'fail', reason: 'playFailed'}, !(this.parentTask || cs.isConfirmCallSession));
+        this.emit('playDone');
+        writeAlerts({
+          account_sid: cs.accountSid,
+          alert_type: AlertType.PLAY_FILENOTFOUND,
+          url: this.url,
+          target_sid: cs.callSid
+        });
+        throw new PlayFileNotFoundError(this.url);
+      }
     }
     this.emit('playDone');
   }

lib/tasks/redirect.js

@@ -1,5 +1,6 @@
 const Task = require('./task');
 const {TaskName} = require('../utils/constants');
+const WsRequestor = require('../utils/ws-requestor');
 
 /**
  * Redirects to a new application
@@ -13,6 +14,17 @@ class TaskRedirect extends Task {
 
   async exec(cs) {
     await super.exec(cs);
+
+    if (cs.requestor instanceof WsRequestor && cs.application.requestor._isAbsoluteUrl(this.actionHook)) {
+      this.logger.info(`Task:performAction redirecting to ${this.actionHook}, requires new ws connection`);
+      try {
+        this.cs.requestor.close();
+        const requestor = new WsRequestor(this.logger, cs.accountSid, {url: this.actionHook}, this.webhook_secret) ;
+        this.cs.application.requestor = requestor;
+      } catch (err) {
+        this.logger.info(err, `Task:performAction error redirecting to ${this.actionHook}`);
+      }
+    }
     await this.performAction();
   }
 }

lib/tasks/rest_dial.js

@@ -12,6 +12,7 @@ class TaskRestDial extends Task {
 
     this.from = this.data.from;
     this.callerName = this.data.callerName;
+    this.timeLimit = this.data.timeLimit;
     this.fromHost = this.data.fromHost;
     this.to = this.data.to;
     this.call_hook = this.data.call_hook;
@@ -66,6 +67,9 @@ class TaskRestDial extends Task {
     const cs = this.callSession;
     cs.setDialog(dlg);
     cs.referHook = this.referHook;
+    if (this.timeLimit) {
+      cs.startMaxCallDurationTimer(this.timeLimit);
+    }
     this.logger.debug('TaskRestDial:_onConnect - call connected');
     if (this.sipRequestWithinDialogHook) this._initSipRequestWithinDialogHandler(cs, dlg);
     try {
@@ -77,11 +81,13 @@ class TaskRestDial extends Task {
           synthesizer: {
             vendor: cs.speechSynthesisVendor,
             language: cs.speechSynthesisLanguage,
-            voice: cs.speechSynthesisVoice
+            voice: cs.speechSynthesisVoice,
+            label: cs.speechSynthesisLabel,
           },
           recognizer: {
             vendor:  cs.speechRecognizerVendor,
-            language: cs.speechRecognizerLanguage
+            language: cs.speechRecognizerLanguage,
+            label: cs.speechRecognizerLabel,
           }
         }
       };

lib/tasks/say.js

@@ -1,3 +1,4 @@
+const assert = require('assert');
 const TtsTask = require('./tts-task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
 const pollySSMLSplit = require('polly-ssml-split');
@@ -35,24 +36,40 @@ class TaskSay extends TtsTask {
     super(logger, opts, parentTask);
     this.preconditions = TaskPreconditions.Endpoint;
 
-    this.text = (Array.isArray(this.data.text) ? this.data.text : [this.data.text])
-      .map((t) => breakLengthyTextIfNeeded(this.logger, t))
-      .flat();
+    assert.ok((typeof this.data.text === 'string' || Array.isArray(this.data.text)) || this.data.stream === true,
+      'Say: either text or stream:true is required');
 
-    this.loop = this.data.loop || 1;
-    this.isHandledByPrimaryProvider = true;
+
+    if (this.data.stream === true) {
+      this._isStreamingTts = true;
+      this.closeOnStreamEmpty = this.data.closeOnStreamEmpty !== false;
+    }
+    else {
+      this._isStreamingTts = false;
+      this.text = (Array.isArray(this.data.text) ? this.data.text : [this.data.text])
+        .map((t) => breakLengthyTextIfNeeded(this.logger, t))
+        .flat();
+
+      this.loop = this.data.loop || 1;
+      this.isHandledByPrimaryProvider = true;
+    }
   }
 
   get name() { return TaskName.Say; }
 
   get summary() {
-    for (let i = 0; i < this.text.length; i++) {
-      if (this.text[i].startsWith('silence_stream')) continue;
-      return `${this.name}{text=${this.text[i].slice(0, 15)}${this.text[i].length > 15 ? '...' : ''}}`;
+    if (this.isStreamingTts) return `${this.name} streaming`;
+    else {
+      for (let i = 0; i < this.text.length; i++) {
+        if (this.text[i].startsWith('silence_stream')) continue;
+        return `${this.name}{text=${this.text[i].slice(0, 15)}${this.text[i].length > 15 ? '...' : ''}}`;
+      }
+      return `${this.name}{${this.text[0]}}`;
     }
-    return `${this.name}{${this.text[0]}}`;
   }
 
+  get isStreamingTts() { return this._isStreamingTts; }
+
   _validateURL(urlString) {
     try {
       new URL(urlString);
@@ -63,14 +80,19 @@ class TaskSay extends TtsTask {
   }
 
   async exec(cs, obj) {
+    if (this.isStreamingTts && !cs.appIsUsingWebsockets) {
+      throw new Error('Say: streaming say verb requires applications to use the websocket API');
+    }
+
     try {
-      await this.handling(cs, obj);
+      if (this.isStreamingTts) await this.handlingStreaming(cs, obj);
+      else await this.handling(cs, obj);
       this.emit('playDone');
     } catch (error) {
       if (error instanceof SpeechCredentialError) {
         // if say failed due to speech credentials, alarm is writtern and error notification is sent
         // finished this say to move to next task.
-        this.logger.info('Say failed due to SpeechCredentialError, finished!');
+        this.logger.info({error}, 'Say failed due to SpeechCredentialError, finished!');
         this.emit('playDone');
         return;
       }
@@ -78,6 +100,35 @@ class TaskSay extends TtsTask {
     }
   }
 
+  async handlingStreaming(cs, {ep}) {
+    const {vendor, language, voice, label} = this.getTtsVendorData(cs);
+    const  credentials = cs.getSpeechCredentials(vendor, 'tts', label);
+    if (!credentials) {
+      throw new SpeechCredentialError(
+        `No text-to-speech service credentials for ${vendor} with labels: ${label} have been configured`);
+    }
+
+    try {
+
+      await this.setTtsStreamingChannelVars(vendor, language, voice, credentials, ep);
+
+      await cs.startTtsStream();
+
+      cs.requestor?.request('tts:streaming-event', '/streaming-event', {event_type: 'stream_open'})
+        .catch((err) => this.logger.info({err}, 'TaskSay:handlingStreaming - Error sending'));
+    } catch (err) {
+      this.logger.info({err}, 'TaskSay:handlingStreaming - Error setting channel vars');
+      cs.requestor?.request('tts:streaming-event', '/streaming-event', {event_type: 'stream_closed'})
+        .catch((err) => this.logger.info({err}, 'TaskSay:handlingStreaming - Error sending'));
+
+      //TODO: send tts:streaming-event with error?
+      this.notifyTaskDone();
+    }
+
+    await this.awaitTaskDone();
+    this.logger.info('TaskSay:handlingStreaming - done');
+  }
+
   async handling(cs, {ep}) {
     const {srf, accountSid:account_sid, callSid:target_sid} = cs;
     const {writeAlerts, AlertType} = srf.locals;
@@ -96,7 +147,7 @@ class TaskSay extends TtsTask {
     let voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
       this.synthesizer.voice :
       cs.speechSynthesisVoice;
-    let label = this.taskInlcudeSynthesizer ? this.synthesizer.label : cs.speechSynthesisLabel;
+    let label = this.taskIncludeSynthesizer ? this.synthesizer.label : cs.speechSynthesisLabel;
 
     const fallbackVendor = this.synthesizer.fallbackVendor && this.synthesizer.fallbackVendor !== 'default' ?
       this.synthesizer.fallbackVendor :
@@ -107,7 +158,7 @@ class TaskSay extends TtsTask {
     const fallbackVoice =  this.synthesizer.fallbackVoice && this.synthesizer.fallbackVoice !== 'default' ?
       this.synthesizer.fallbackVoice :
       cs.fallbackSpeechSynthesisVoice;
-    const fallbackLabel = this.taskInlcudeSynthesizer ?
+    const fallbackLabel = this.taskIncludeSynthesizer ?
       this.synthesizer.fallbackLabel : cs.fallbackSpeechSynthesisLabel;
 
     if (cs.hasFallbackTts) {
@@ -162,7 +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) {
@@ -189,6 +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'));
             }
@@ -253,9 +305,10 @@ class TaskSay extends TtsTask {
         this._playResolve = null;
       }
     }
+    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_')) {
@@ -263,16 +316,27 @@ class TaskSay extends TtsTask {
           .replace('whisper_', 'whisper.')
           .replace('deepgram_', 'deepgram.')
           .replace('playht_', 'playht.')
+          .replace('cartesia_', 'cartesia.')
           .replace('rimelabs_', 'rimelabs.')
           .replace('verbio_', 'verbio.')
           .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
     span.setAttributes(attrs);
   }
+
+  notifyTtsStreamIsEmpty() {
+    if (this.isStreamingTts && this.closeOnStreamEmpty) {
+      this.logger.info('TaskSay:notifyTtsStreamIsEmpty - stream is empty, killing task');
+      this.notifyTaskDone();
+    }
+  }
 }
 
 const spanMapping = {
@@ -309,6 +373,11 @@ const spanMapping = {
   'playht.name_lookup_time_ms': 'name_lookup_ms',
   'playht.connect_time_ms': 'connect_ms',
   'playht.final_response_time_ms': 'final_response_ms',
+  // Cartesia
+  'cartesia.request_id': 'cartesia.req_id',
+  'cartesia.name_lookup_time_ms': 'name_lookup_ms',
+  'cartesia.connect_time_ms': 'connect_ms',
+  'cartesia.final_response_time_ms': 'final_response_ms',
   // Rimelabs
   'rimelabs.name_lookup_time_ms': 'name_lookup_ms',
   'rimelabs.connect_time_ms': 'connect_ms',

lib/tasks/sip_refer.js

@@ -12,6 +12,7 @@ class TaskSipRefer extends Task {
 
     this.referTo = this.data.referTo;
     this.referredBy = this.data.referredBy;
+    this.referredByDisplayName = this.data.referredByDisplayName;
     this.headers = this.data.headers || {};
     this.eventHook = this.data.eventHook;
   }
@@ -94,7 +95,10 @@ class TaskSipRefer extends Task {
         }
         if (status >= 200) {
           this.referSpan.setAttributes({'refer.finalNotify': status});
-          await this.performAction({refer_status: 202, final_referred_call_status: status});
+          await this.performAction({refer_status: 202, final_referred_call_status: status})
+            .catch((err) => {
+              this.logger.error(err, 'TaskSipRefer:exec - error performing action finalNotify');
+            });
           this.notifyTaskDone();
         }
       }
@@ -102,7 +106,7 @@ class TaskSipRefer extends Task {
   }
 
   _normalizeReferHeaders(cs, dlg) {
-    let {referTo, referredBy} = this;
+    let {referTo, referredBy, referredByDisplayName} = this;
 
     /* get IP address of the SBC to use as hostname if needed */
     const {host} = parseUri(dlg.remote.uri);
@@ -117,9 +121,12 @@ class TaskSipRefer extends Task {
       referredBy = cs.req?.callingNumber || dlg.local.uri;
       this.logger.info({referredBy}, 'setting referredby');
     }
+    if (!referredByDisplayName) {
+      referredByDisplayName = cs.req?.callingName;
+    }
     if (!referredBy.startsWith('<') && !referredBy.startsWith('sip') && !referredBy.startsWith('"')) {
       /* they may have only provided a phone number/user */
-      referredBy = `sip:${referredBy}@${host}`;
+      referredBy = `${referredByDisplayName ? `"${referredByDisplayName}"` : ''}<sip:${referredBy}@${host}>`;
     }
     return {referTo, referredBy};
   }

lib/tasks/stt-task.js

@@ -219,7 +219,8 @@ class SttTask extends Task {
           roleArn
         });
       this.logger.debug({roleArn}, `(roleArn) got aws access token ${servedFromCache ? 'from cache' : ''}`);
-      credentials = {...credentials, accessKeyId, secretAccessKey, sessionToken};
+      // from role ARN, we will get SessionToken, but feature server use it as securityToken.
+      credentials = {...credentials, accessKeyId, secretAccessKey, securityToken: sessionToken};
     }
     else if (vendor === 'verbio' && credentials.client_id && credentials.client_secret) {
       const {access_token, servedFromCache} = await getVerbioAccessToken(credentials);
@@ -229,9 +230,13 @@ class SttTask extends Task {
     }
     else if (vendor == 'aws' && !JAMBONES_AWS_TRANSCRIBE_USE_GRPC) {
       /* get AWS access token */
-      const {accessKeyId, secretAccessKey, securityToken, region } = credentials;
+      const {speech_credential_sid, accessKeyId, secretAccessKey, securityToken, region } = credentials;
       if (!securityToken) {
-        const { servedFromCache, ...newCredentials} = await getAwsAuthToken({accessKeyId, secretAccessKey, region});
+        const { servedFromCache, ...newCredentials} = await getAwsAuthToken({
+          speech_credential_sid,
+          accessKeyId,
+          secretAccessKey,
+          region});
         this.logger.debug({newCredentials}, `got aws security token ${servedFromCache ? 'from cache' : ''}`);
         credentials = {...newCredentials, region};
       }

lib/tasks/transcribe.js

@@ -13,6 +13,7 @@ const {
   JambonzTranscriptionEvents,
   TranscribeStatus,
   AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
   VerbioTranscriptionEvents,
   SpeechmaticsTranscriptionEvents
 } = require('../utils/constants.json');
@@ -27,6 +28,7 @@ class TaskTranscribe extends SttTask {
     super(logger, opts, parentTask);
 
     this.transcriptionHook = this.data.transcriptionHook;
+    this.translationHook = this.data.translationHook;
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
 
     if (this.data.recognizer) {
@@ -299,10 +301,23 @@ 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(
-          ep, SpeechmaticsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+          ep, SpeechmaticsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(
+          ep, SpeechmaticsTranscriptionEvents.Translation, this._onTranslation.bind(this, cs, ep, channel));
         this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.Info,
           this._onSpeechmaticsInfo.bind(this, cs, ep));
         this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.RecognitionStarted,
@@ -441,7 +456,7 @@ class TaskTranscribe extends SttTask {
         this._startAsrTimer(channel);
 
         /* some STT engines will keep listening after a final response, so no need to restart */
-        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google']
+        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google', 'speechmatics']
           .includes(this.vendor)) this._startTranscribing(cs, ep, channel);
       }
       else {
@@ -466,7 +481,7 @@ class TaskTranscribe extends SttTask {
         this._resolve(channel, evt);
 
         /* some STT engines will keep listening after a final response, so no need to restart */
-        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google'].includes(this.vendor) &&
+        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google', 'speechmatics'].includes(this.vendor) &&
           !this.vendor.startsWith('custom:')) {
           this.logger.debug('TaskTranscribe:_onTranscription - restarting transcribe');
           this._startTranscribing(cs, ep, channel);
@@ -492,6 +507,47 @@ class TaskTranscribe extends SttTask {
     }
   }
 
+  async _onTranslation(_cs, _ep, channel, evt, _fsEvent) {
+    this.logger.debug({evt}, 'TaskTranscribe:_onTranslation');
+    if (this.translationHook && evt.results?.length > 0) {
+      try {
+        const b3 = this.getTracingPropagation();
+        const httpHeaders = b3 && {b3};
+        const payload = {
+          ...this.cs.callInfo,
+          ...httpHeaders,
+          translation: {
+            channel,
+            language: evt.language,
+            translation: evt.results[0].content
+          }
+        };
+
+        this.logger.debug({payload}, 'sending translationHook');
+        const json = await this.cs.requestor.request('verb:hook', this.translationHook, payload);
+        this.logger.info({json}, 'completed translationHook');
+        if (json && Array.isArray(json) && !this.parentTask) {
+          const makeTask = require('./make_task');
+          const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
+          if (tasks && tasks.length > 0) {
+            this.logger.info({tasks: tasks}, `${this.name} replacing application with ${tasks.length} tasks`);
+            this.cs.replaceApplication(tasks);
+          }
+        }
+      } catch (err) {
+        this.logger.info(err, 'TranscribeTask:_onTranslation error');
+      }
+      if (this.parentTask) {
+        this.parentTask.emit('translation', evt);
+      }
+    }
+    if (this.killed) {
+      this.logger.debug('TaskTranscribe:_onTranslation exiting after receiving final transcription');
+      this._clearTimer();
+      this.notifyTaskDone();
+    }
+  }
+
   async _resolve(channel, evt) {
     if (evt.is_final) {
       /* we've got a final transcript, so end the otel child span for this channel */
@@ -671,7 +727,7 @@ class TaskTranscribe extends SttTask {
     this.logger.debug({evt}, 'TaskGather:_onSpeechmaticsInfo');
   }
 
-  async _onSpeechmaticsErrror(cs, _ep, evt) {
+  async _onSpeechmaticsError(cs, _ep, evt) {
     // eslint-disable-next-line no-unused-vars
     const {message, ...e} = evt;
     this._onVendorError(cs, _ep, {error: JSON.stringify(e)});

lib/tasks/tts-task.js

@@ -13,11 +13,11 @@ class TtsTask extends Task {
 
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
     /**
-     * Task use taskInlcudeSynthesizer to identify
-     * if taskInlcudeSynthesizer === true, use label from verb.synthesizer, even it's empty
-     * if taskInlcudeSynthesizer === false, use label from application.synthesizer
+     * Task use taskIncludeSynthesizer to identify
+     * if taskIncludeSynthesizer === true, use label from verb.synthesizer, even it's empty
+     * if taskIncludeSynthesizer === false, use label from application.synthesizer
      */
-    this.taskInlcudeSynthesizer = !!this.data.synthesizer;
+    this.taskIncludeSynthesizer = !!this.data.synthesizer;
     this.synthesizer = this.data.synthesizer || {};
     this.disableTtsCache = this.data.disableTtsCache;
     this.options = this.synthesizer.options || {};
@@ -44,6 +44,92 @@ class TtsTask extends Task {
     }
   }
 
+  getTtsVendorData(cs) {
+    const vendor = this.synthesizer.vendor && this.synthesizer.vendor !== 'default' ?
+      this.synthesizer.vendor :
+      cs.speechSynthesisVendor;
+    const language = this.synthesizer.language && this.synthesizer.language !== 'default' ?
+      this.synthesizer.language :
+      cs.speechSynthesisLanguage ;
+    const voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
+      this.synthesizer.voice :
+      cs.speechSynthesisVoice;
+    const label = this.taskIncludeSynthesizer ? this.synthesizer.label : cs.speechSynthesisLabel;
+    return {vendor, language, voice, label};
+  }
+
+  async setTtsStreamingChannelVars(vendor, language, voice, credentials, ep) {
+    const {api_key, model_id, custom_tts_streaming_url, auth_token} = credentials;
+    const {stability, similarity_boost, use_speaker_boost, style} = this.options;
+    let obj;
+
+    this.logger.debug({credentials},
+      `setTtsStreamingChannelVars: vendor: ${vendor}, language: ${language}, voice: ${voice}`);
+
+    switch (vendor) {
+      case 'deepgram':
+        obj = {
+          DEEPGRAM_API_KEY: api_key,
+          DEEPGRAM_TTS_STREAMING_MODEL: voice
+        };
+        break;
+      case 'cartesia':
+        obj = {
+          CARTESIA_API_KEY: api_key,
+          CARTESIA_TTS_STREAMING_MODEL_ID: model_id,
+          CARTESIA_TTS_STREAMING_VOICE_ID: voice,
+          CARTESIA_TTS_STREAMING_LANGUAGE: language || 'en',
+        };
+        break;
+      case 'elevenlabs':
+        obj = {
+          ELEVENLABS_API_KEY: api_key,
+          ELEVENLABS_TTS_STREAMING_MODEL_ID: model_id,
+          ELEVENLABS_TTS_STREAMING_VOICE_ID: voice,
+          // 20/12/2024 - only eleven_turbo_v2_5 support multiple language
+          ...(['eleven_turbo_v2_5'].includes(model_id) && {ELEVENLABS_TTS_STREAMING_LANGUAGE: language}),
+          ...(stability && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_STABILITY: stability}),
+          ...(similarity_boost && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_SIMILARITY_BOOST: similarity_boost}),
+          ...(use_speaker_boost && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_USE_SPEAKER_BOOST: use_speaker_boost}),
+          ...(style && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_STYLE: style})
+        };
+        break;
+      case 'rimelabs':
+        const {
+          pauseBetweenBrackets, phonemizeBetweenBrackets, inlineSpeedAlpha, speedAlpha, reduceLatency
+        } = this.options;
+        obj = {
+          RIMELABS_API_KEY: api_key,
+          RIMELABS_TTS_STREAMING_MODEL_ID: model_id,
+          RIMELABS_TTS_STREAMING_VOICE_ID: voice,
+          RIMELABS_TTS_STREAMING_LANGUAGE: language || 'en',
+          ...(pauseBetweenBrackets && {RIMELABS_TTS_STREAMING_PAUSE_BETWEEN_BRACKETS: pauseBetweenBrackets}),
+          ...(phonemizeBetweenBrackets &&
+            {RIMELABS_TTS_STREAMING_PHONEMIZE_BETWEEN_BRACKETS: phonemizeBetweenBrackets}),
+          ...(inlineSpeedAlpha && {RIMELABS_TTS_STREAMING_INLINE_SPEED_ALPHA: inlineSpeedAlpha}),
+          ...(speedAlpha && {RIMELABS_TTS_STREAMING_SPEED_ALPHA: speedAlpha}),
+          ...(reduceLatency && {RIMELABS_TTS_STREAMING_REDUCE_LATENCY: reduceLatency})
+        };
+        break;
+      default:
+        if (vendor.startsWith('custom:')) {
+          const use_tls = custom_tts_streaming_url.startsWith('wss://');
+          obj = {
+            CUSTOM_TTS_STREAMING_HOST: custom_tts_streaming_url.replace(/^(ws|wss):\/\//, ''),
+            CUSTOM_TTS_STREAMING_API_KEY: auth_token,
+            CUSTOM_TTS_STREAMING_VOICE_ID: voice,
+            CUSTOM_TTS_STREAMING_LANGUAGE: language || 'en',
+            CUSTOM_TTS_STREAMING_USE_TLS: use_tls
+          };
+        } else {
+          throw new Error(`vendor ${vendor} is not supported for tts streaming yet`);
+        }
+    }
+    this.logger.info({vendor, credentials, obj}, 'setTtsStreamingChannelVars');
+
+    await ep.set(obj);
+  }
+
   async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
     const {srf, accountSid:account_sid} = cs;
     const {writeAlerts, AlertType, stats} = srf.locals;
@@ -57,16 +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) {
@@ -129,7 +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({
@@ -164,7 +251,7 @@ class TtsTask extends Task {
             language,
             voice,
             engine,
-            model,
+            model: this.model,
             salt,
             credentials,
             options: this.options,

lib/utils/action-hook-delay.js

@@ -126,7 +126,12 @@ class ActionHookDelayProcessor extends Emitter {
     try {
       this._taskInProgress = makeTask(this.logger, t[0]);
       this._taskInProgress.disableTracing = true;
-      this._taskInProgress.exec(this.cs, {ep: this.ep});
+      this._taskInProgress.exec(this.cs, {ep: this.ep}).catch((err) => {
+        this.logger.info(`ActionHookDelayProcessor#_onNoResponseTimer: error playing file: ${err.message}`);
+        this._taskInProgress = null;
+        this.ep.removeAllListeners('playback-start');
+        this.ep.removeAllListeners('playback-stop');
+      });
     } catch (err) {
       this.logger.info(err, 'ActionHookDelayProcessor#_onNoResponseTimer: error starting action');
       this._taskInProgress = null;

lib/utils/amd-utils.js

@@ -153,7 +153,7 @@ class Amd extends Emitter {
       const wordCount = t.alternatives[0].transcript.split(' ').length;
       const final = t.is_final;
 
-      const foundHint = hints.find((h) =>  t.alternatives[0].transcript.includes(h));
+      const foundHint = hints.find((h) =>  t.alternatives[0].transcript.toLowerCase().includes(h.toLowerCase()));
       if (foundHint) {
         /* we detected a common voice mail greeting */
         this.logger.debug(`Amd:evaluateTranscription: found hint ${foundHint}`);

lib/utils/background-task-manager.js

@@ -46,6 +46,9 @@ class BackgroundTaskManager extends Emitter {
       case 'transcribe':
         task = await this._initTranscribe(opts);
         break;
+      case 'ttsStream':
+        task = await this._initTtsStream(opts);
+        break;
       default:
         break;
     }
@@ -100,6 +103,7 @@ class BackgroundTaskManager extends Emitter {
   async _initBargeIn(opts) {
     let task;
     try {
+      const copy = JSON.parse(JSON.stringify(opts));
       const t = normalizeJambones(this.logger, [opts]);
       task = makeTask(this.logger, t[0]);
       task
@@ -118,7 +122,7 @@ class BackgroundTaskManager extends Emitter {
           if (task.sticky && !this.cs.callGone && !this.cs._stopping) {
             this.logger.info('BackgroundTaskManager:_initBargeIn: restarting background bargeIn');
             this._bargeInHandled = false;
-            this.newTask('bargeIn', opts, true);
+            this.newTask('bargeIn', copy, true);
           }
           return;
         })
@@ -173,6 +177,25 @@ class BackgroundTaskManager extends Emitter {
     return task;
   }
 
+  // Initiate Tts Stream
+  async _initTtsStream(opts) {
+    let task;
+    try {
+      const t = normalizeJambones(this.logger, [opts]);
+      task = makeTask(this.logger, t[0]);
+      const resources = await this.cs._evaluatePreconditions(task);
+      const {span, ctx} = this.rootSpan.startChildSpan(`background-ttsStream:${task.summary}`);
+      task.span = span;
+      task.ctx = ctx;
+      task.exec(this.cs, resources)
+        .then(this._taskCompleted.bind(this, 'ttsStream', task))
+        .catch(this._taskError.bind(this, 'ttsStream', task));
+    } catch (err) {
+      this.logger.info(err, 'BackgroundTaskManager:_initTtsStream - Error creating ttsStream task');
+    }
+    return task;
+  }
+
   _taskCompleted(type, task) {
     this.logger.debug({type, task}, `BackgroundTaskManager:_taskCompleted: task completed, sticky: ${task.sticky}`);
     task.removeAllListeners();

lib/utils/constants.json

@@ -28,6 +28,7 @@
     "SipRedirect": "sip:redirect",
     "Say": "say",
     "SayLegacy": "say:legacy",
+    "Stream": "stream",
     "Tag": "tag",
     "Transcribe": "transcribe"
   },
@@ -129,6 +130,7 @@
   },
   "SpeechmaticsTranscriptionEvents": {
     "Transcription": "speechmatics_transcribe::transcription",
+    "Translation": "speechmatics_transcribe::translation",
     "Info": "speechmatics_transcribe::info",
     "RecognitionStarted": "speechmatics_transcribe::recognition_started",
     "ConnectFailure": "speechmatics_transcribe::connect_failed",
@@ -147,6 +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"
   },
@@ -174,6 +182,27 @@
     "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",
+    "ConnectFailure": "voice_agent_s2s::connect_failed",
+    "Disconnect": "voice_agent_s2s::disconnect",
+    "ServerEvent": "voice_agent_s2s::server_event"
+  },
+  "LlmEvents_Ultravox": {
+    "Error": "error",
+    "Connect": "ultravox_s2s::connect",
+    "ConnectFailure": "ultravox_s2s::connect_failed",
+    "Disconnect": "ultravox_s2s::disconnect",
+    "ServerEvent": "ultravox_s2s::server_event"
+  },
   "QueueResults": {
     "Bridged": "bridged",
     "Error": "error",
@@ -188,7 +217,8 @@
   },
   "KillReason": {
     "Hangup": "hangup",
-    "Replaced": "replaced"
+    "Replaced": "replaced",
+    "MediaTimeout": "media_timeout"
   },
   "HookMsgTypes": [
     "session:new",
@@ -202,6 +232,8 @@
     "verb:status",
     "llm:event",
     "llm:tool-call",
+    "tts:tokens-result",
+    "tts:streaming-event",
     "jambonz:error"
   ],
   "RecordState": {
@@ -220,7 +252,59 @@
     "ToneTimeout": "amd_tone_timeout",
     "Stopped": "amd_stopped"
   },
+  "MediaPath": {
+    "NoMedia": "no-media",
+    "PartialMedia": "partial-media",
+    "FullMedia": "full-media"
+  },
+  "DeepgramTtsStreamingEvents": {
+    "Empty": "deepgram_tts_streaming::empty",
+    "ConnectFailure": "deepgram_tts_streaming::connect_failed",
+    "Connect": "deepgram_tts_streaming::connect"
+  },
+  "CartesiaTtsStreamingEvents": {
+    "Empty": "cartesia_tts_streaming::empty",
+    "ConnectFailure": "cartesia_tts_streaming::connect_failed",
+    "Connect": "cartesia_tts_streaming::connect"
+  },
+  "ElevenlabsTtsStreamingEvents": {
+    "Empty": "elevenlabs_tts_streaming::empty",
+    "ConnectFailure": "elevenlabs_tts_streaming::connect_failed",
+    "Connect": "elevenlabs_tts_streaming::connect"
+  },
+  "RimelabsTtsStreamingEvents": {
+    "Empty": "rimelabs_tts_streaming::empty",
+    "ConnectFailure": "rimelabs_tts_streaming::connect_failed",
+    "Connect": "rimelabs_tts_streaming::connect"
+  },
+  "CustomTtsStreamingEvents": {
+    "Empty": "custom_tts_streaming::empty",
+    "ConnectFailure": "custom_tts_streaming::connect_failed",
+    "Connect": "custom_tts_streaming::connect"
+  },
+  "TtsStreamingEvents": {
+    "Empty": "tts_streaming::empty",
+    "Pause": "tts_streaming::pause",
+    "Resume": "tts_streaming::resume",
+    "ConnectFailure": "tts_streaming::connect_failed"
+  },
+  "TtsStreamingConnectionStatus": {
+    "NotConnected": "not_connected",
+    "Connected": "connected",
+    "Connecting": "connecting",
+    "Failed": "failed"
+  },
   "MAX_SIMRINGS": 10,
   "BONG_TONE": "tone_stream://v=-7;%(100,0,941.0,1477.0);v=-7;>=2;+=.1;%(1400,0,350,440)",
-  "FS_UUID_SET_NAME": "fsUUIDs"
+  "FS_UUID_SET_NAME": "fsUUIDs",
+  "SystemState" : {
+    "Online": "ONLINE",
+    "Offline": "OFFLINE",
+    "GracefulShutdownInProgress":"SHUTDOWN_IN_PROGRESS"
+  },
+  "FEATURE_SERVER" : "feature-server",
+  "WS_CLOSE_CODES": {
+    "NormalClosure": 1000,
+    "GoingAway": 1001
+  }
 }

lib/utils/db-utils.js

@@ -105,6 +105,13 @@ const speechMapper = (cred) => {
       obj.voice_engine = o.voice_engine;
       obj.options = o.options;
     }
+    else if ('cartesia' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.model_id = o.model_id;
+      obj.embedding = o.embedding;
+      obj.options = o.options;
+    }
     else if ('rimelabs' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
@@ -115,6 +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;
@@ -136,6 +147,7 @@ const speechMapper = (cred) => {
       obj.auth_token = o.auth_token;
       obj.custom_stt_url = o.custom_stt_url;
       obj.custom_tts_url = o.custom_tts_url;
+      obj.custom_tts_streaming_url = o.custom_tts_streaming_url;
     }
   } catch (err) {
     console.log(err);
@@ -215,11 +227,23 @@ module.exports = (logger, srf) => {
     }
   };
 
+  const lookupVoipCarrierBySid = async(sid) => {
+    const pp = pool.promise();
+    try {
+      const [r] = await pp.query('SELECT * FROM voip_carriers WHERE voip_carrier_sid = ?', [sid]);
+      return r;
+
+    } catch (err) {
+      logger.error({err}, `lookupVoipCarrierBySid: Error ${sid}`);
+    }
+  };
+
   return {
     lookupAccountDetails,
     updateSpeechCredentialLastUsed,
     lookupCarrier,
     lookupCarrierByPhoneNumber,
-    lookupGoogleCustomVoice
+    lookupGoogleCustomVoice,
+    lookupVoipCarrierBySid
   };
 };

lib/utils/error.js

@@ -1,9 +1,33 @@
-class SpeechCredentialError extends Error {
+class NonFatalTaskError extends Error {
   constructor(msg) {
     super(msg);
   }
 }
 
+class SpeechCredentialError extends NonFatalTaskError {
+  constructor(msg) {
+    super(msg);
+  }
+}
+
+class PlayFileNotFoundError extends NonFatalTaskError {
+  constructor(url) {
+    super('File not found');
+    this.url = url;
+  }
+}
+
+class HTTPResponseError extends Error {
+  constructor(statusCode) {
+    super('Unexpected HTTP Response');
+    delete this.stack;
+    this.statusCode = statusCode;
+  }
+}
+
 module.exports = {
-  SpeechCredentialError
+  SpeechCredentialError,
+  NonFatalTaskError,
+  PlayFileNotFoundError,
+  HTTPResponseError
 };

lib/utils/http-requestor.js

@@ -16,6 +16,7 @@ const {
   NODE_ENV,
   HTTP_USER_AGENT_HEADER,
 } = require('../config');
+const {HTTPResponseError} = require('./error');
 
 const toBase64 = (str) => Buffer.from(str || '', 'utf8').toString('base64');
 
@@ -190,8 +191,7 @@ class HttpRequestor extends BaseRequestor {
         followRedirects: false
       });
       if (![200, 202, 204].includes(statusCode)) {
-        const err = new Error();
-        err.statusCode = statusCode;
+        const err = new HTTPResponseError(statusCode);
         throw err;
       }
       if (headers['content-type']?.includes('application/json')) {

lib/utils/install-srf-locals.js

@@ -199,7 +199,8 @@ function installSrfLocals(srf, logger) {
   } = require('@jambonz/speech-utils')({}, logger);
   const {
     writeAlerts,
-    AlertType
+    AlertType,
+    writeSystemAlerts
   } = require('@jambonz/time-series')(logger, {
     host: JAMBONES_TIME_SERIES_HOST,
     commitSize: 50,
@@ -269,7 +270,8 @@ function installSrfLocals(srf, logger) {
     getFreeswitch,
     stats: stats,
     writeAlerts,
-    AlertType
+    AlertType,
+    writeSystemAlerts
   };
 
   if (localIp) {

lib/utils/place-outdial.js

@@ -1,5 +1,5 @@
 const Emitter = require('events');
-const {CallStatus} = require('./constants');
+const {CallStatus, MediaPath} = require('./constants');
 const SipError = require('drachtio-srf').SipError;
 const {TaskPreconditions, CallDirection} = require('../utils/constants');
 const CallInfo = require('../session/call-info');
@@ -17,7 +17,9 @@ const HttpRequestor = require('./http-requestor');
 const WsRequestor = require('./ws-requestor');
 const {makeOpusFirst} = require('./sdp-utils');
 const {
-  JAMBONES_USE_FREESWITCH_TIMER_FD
+  JAMBONES_USE_FREESWITCH_TIMER_FD,
+  JAMBONES_MEDIA_TIMEOUT_MS,
+  JAMBONES_MEDIA_HOLD_TIMEOUT_MS
 } = require('../config');
 
 class SingleDialer extends Emitter {
@@ -317,14 +319,19 @@ class SingleDialer extends Emitter {
   /**
    * kill the call in progress or the stable dialog, whichever we have
    */
-  async kill() {
+  async kill(Reason) {
     this.killed = true;
     if (this.inviteInProgress) await this.inviteInProgress.cancel();
     else if (this.dlg && this.dlg.connected) {
       const duration = moment().diff(this.dlg.connectTime, 'seconds');
       this.logger.debug('SingleDialer:kill hanging up called party');
       this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});
-      this.dlg.destroy();
+      const headers = {
+        ...(Reason && {'X-Reason': Reason})
+      };
+      this.dlg.destroy({
+        headers
+      });
     }
     if (this.ep) {
       this.logger.debug(`SingleDialer:kill - deleting endpoint ${this.ep.uuid}`);
@@ -335,11 +342,22 @@ class SingleDialer extends Emitter {
   _configMsEndpoint() {
     const opts = {
       ...(this.onHoldMusic && {holdMusic: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`}),
-      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'})
+      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'}),
+      ...(JAMBONES_MEDIA_TIMEOUT_MS && {media_timeout: JAMBONES_MEDIA_TIMEOUT_MS}),
+      ...(JAMBONES_MEDIA_HOLD_TIMEOUT_MS && {media_hold_timeout: JAMBONES_MEDIA_HOLD_TIMEOUT_MS})
     };
     if (Object.keys(opts).length > 0) {
       this.ep.set(opts);
     }
+    if (this.dialTask?.inbandDtmfEnabled && !this.ep.inbandDtmfEnabled) {
+      // https://developer.signalwire.com/freeswitch/FreeSWITCH-Explained/Modules/mod-dptools/6587132/#0-about
+      try {
+        this.ep.execute('start_dtmf');
+        this.ep.inbandDtmfEnabled = true;
+      } catch (err) {
+        this.logger.info(err, 'place-outdial:_configMsEndpoint - error enable inband DTMF');
+      }
+    }
   }
 
   /**
@@ -381,7 +399,8 @@ class SingleDialer extends Emitter {
         callInfo: this.callInfo,
         accountInfo: this.accountInfo,
         tasks,
-        rootSpan: this.rootSpan
+        rootSpan: this.rootSpan,
+        req: this.req
       });
       await cs.exec();
 
@@ -455,21 +474,26 @@ class SingleDialer extends Emitter {
     return cs;
   }
 
-  async releaseMediaToSBC(remoteSdp, localSdp) {
+  async releaseMediaToSBC(remoteSdp, localSdp, releaseMediaEntirely) {
     assert(this.dlg && this.dlg.connected && this.ep && typeof remoteSdp === 'string');
     const sdp = stripCodecs(this.logger, remoteSdp, localSdp) || remoteSdp;
     await this.dlg.modify(sdp, {
       headers: {
-        'X-Reason': 'release-media'
+        'X-Reason': releaseMediaEntirely ? 'release-media-entirely' : 'release-media'
       }
     });
-    this.ep.destroy()
-      .then(() => this.ep = null)
-      .catch((err) => this.logger.error({err}, 'SingleDialer:releaseMediaToSBC: Error destroying endpoint'));
+    try {
+      await this.ep.destroy();
+    } catch (err) {
+      this.logger.error({err}, 'SingleDialer:releaseMediaToSBC: Error destroying endpoint');
+    }
+    this.ep = null;
   }
 
-  async reAnchorMedia() {
+  async reAnchorMedia(currentMediaRoute = MediaPath.PartialMedia) {
     assert(this.dlg && this.dlg.connected && !this.ep);
+
+    this.logger.debug('SingleDialer:reAnchorMedia: re-anchoring media after partial media');
     this.ep = await this.ms.createEndpoint({remoteSdp: this.dlg.remote.sdp});
     this._configMsEndpoint();
     await this.dlg.modify(this.ep.local.sdp, {
@@ -477,6 +501,11 @@ class SingleDialer extends Emitter {
         'X-Reason': 'anchor-media'
       }
     });
+
+    if (currentMediaRoute === MediaPath.NoMedia) {
+      this.logger.debug('SingleDialer:reAnchorMedia: repoint endpoint after no media');
+      await this.ep.modify(this.dlg.remote.sdp);
+    }
   }
 
   _notifyCallStatusChange({callStatus, sipStatus, sipReason, duration}) {

lib/utils/sbc-pinger.js

@@ -46,12 +46,24 @@ module.exports = (logger) => {
             const {srf} = require('../..');
             srf.locals.publicIp = publicIp;
           })
-          .on(LifeCycleEvents.ScaleIn, () => {
+          .on(LifeCycleEvents.ScaleIn, async() => {
             logger.info('AWS scale-in notification: begin drying up calls');
             dryUpCalls = true;
             lifecycleEmitter.operationalState = LifeCycleEvents.ScaleIn;
 
             const {srf} = require('../..');
+            const {writeSystemAlerts} = srf.locals;
+            if (writeSystemAlerts) {
+              const {SystemState, FEATURE_SERVER} = require('./constants');
+              await writeSystemAlerts({
+                system_component: FEATURE_SERVER,
+                state : SystemState.GracefulShutdownInProgress,
+                fields : {
+                  detail: `feature-server with process_id ${process.pid} shutdown in progress`,
+                  host: srf.locals?.ipv4
+                }
+              });
+            }
             pingProxies(srf);
 
             // if we have zero calls, we can complete the scale-in right

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',
@@ -43,7 +44,9 @@ const stickyVars = {
     'DEEPGRAM_SPEECH_UTTERANCE_END_MS',
     'DEEPGRAM_SPEECH_VAD_TURNOFF',
     'DEEPGRAM_SPEECH_TAG',
-    'DEEPGRAM_SPEECH_MODEL_VERSION'
+    'DEEPGRAM_SPEECH_MODEL_VERSION',
+    'DEEPGRAM_SPEECH_FILLER_WORDS',
+    'DEEPGRAM_SPEECH_KEYTERMS',
   ],
   aws: [
     'AWS_VOCABULARY_NAME',
@@ -104,11 +107,16 @@ const stickyVars = {
     'ASSEMBLYAI_API_KEY',
     'ASSEMBLYAI_WORD_BOOST'
   ],
+  voxist: [
+    'VOXIST_API_KEY',
+  ],
   speechmatics: [
     'SPEECHMATICS_API_KEY',
     'SPEECHMATICS_HOST',
     'SPEECHMATICS_PATH',
     'SPEECHMATICS_SPEECH_HINTS',
+    'SPEECHMATICS_TRANSLATION_LANGUAGES',
+    'SPEECHMATICS_TRANSLATION_PARTIALS'
   ]
 };
 
@@ -183,7 +191,10 @@ const selectDefaultGoogleModel = (task, language, version) => {
     (useV2 ? 'long' : 'latest_long');
 };
 const consolidateTranscripts = (bufferedTranscripts, channel, language, vendor) => {
-  if (bufferedTranscripts.length === 1) return bufferedTranscripts[0];
+  if (bufferedTranscripts.length === 1) {
+    bufferedTranscripts[0].is_final = true;
+    return bufferedTranscripts[0];
+  }
   let totalConfidence = 0;
   const finalTranscript = bufferedTranscripts.reduce((acc, evt) => {
     totalConfidence += evt.alternatives[0].confidence;
@@ -203,7 +214,7 @@ const consolidateTranscripts = (bufferedTranscripts, channel, language, vendor)
     const lastChar = acc.alternatives[0].transcript.slice(-1);
     const firstChar = newTranscript.charAt(0);
 
-    if (lastChar.match(/\d/) && firstChar.match(/\d/)) {
+    if (vendor === 'speechmatics' || (lastChar.match(/\d/) && firstChar.match(/\d/))) {
       acc.alternatives[0].transcript += newTranscript;
     } else {
       acc.alternatives[0].transcript += ` ${newTranscript}`;
@@ -511,6 +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';
@@ -561,6 +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':
@@ -699,6 +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) {
@@ -772,6 +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) &&
@@ -807,7 +843,11 @@ module.exports = (logger) => {
         ...(deepgramOptions.tag) &&
           {DEEPGRAM_SPEECH_TAG: deepgramOptions.tag},
         ...(deepgramOptions.version) &&
-          {DEEPGRAM_SPEECH_MODEL_VERSION: deepgramOptions.version}
+          {DEEPGRAM_SPEECH_MODEL_VERSION: deepgramOptions.version},
+        ...(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) {
@@ -916,6 +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 = {
@@ -937,11 +984,48 @@ module.exports = (logger) => {
       };
     }
     else if ('speechmatics' === vendor) {
+      const {speechmaticsOptions = {}} = rOpts;
       opts = {
         ...opts,
         ...(sttCredentials.api_key) && {SPEECHMATICS_API_KEY: sttCredentials.api_key},
         ...(sttCredentials.speechmatics_stt_uri) && {SPEECHMATICS_HOST: sttCredentials.speechmatics_stt_uri},
         ...(rOpts.hints?.length > 0 && {SPEECHMATICS_SPEECH_HINTS: rOpts.hints.join(',')}),
+        ...(speechmaticsOptions.translation_config &&
+          {
+            SPEECHMATICS_TRANSLATION_LANGUAGES: speechmaticsOptions.translation_config.target_languages.join(','),
+            SPEECHMATICS_TRANSLATION_PARTIALS: speechmaticsOptions.translation_config.enable_partials ? 1 : 0
+          }
+        ),
+        ...(speechmaticsOptions.transcription_config?.domain &&
+          {SPEECHMATICS_DOMAIN: speechmaticsOptions.transcription_config.domain}),
+        ...{SPEECHMATICS_MAX_DELAY: speechmaticsOptions.transcription_config?.max_delay || 0.7},
+        ...{SPEECHMATICS_MAX_DELAY_MODE: speechmaticsOptions.transcription_config?.max_delay_mode || 'flexible'},
+        ...(speechmaticsOptions.transcription_config?.diarization &&
+            {SPEECHMATICS_DIARIZATION: speechmaticsOptions.transcription_config.diarization}),
+        ...(speechmaticsOptions.transcription_config?.speaker_diarization_config?.speaker_sensitivity &&
+          {SPEECHMATICS_DIARIZATION_SPEAKER_SENSITIVITY:
+            speechmaticsOptions.transcription_config.speaker_diarization_config.speaker_sensitivity}),
+        ...(speechmaticsOptions.transcription_config?.speaker_diarization_config?.max_speakers &&
+          {SPEECHMATICS_DIARIZATION_MAX_SPEAKERS:
+            speechmaticsOptions.transcription_config.speaker_diarization_config.max_speakers}),
+        ...(speechmaticsOptions.transcription_config?.output_locale &&
+          {SPEECHMATICS_OUTPUT_LOCALE: speechmaticsOptions.transcription_config.output_locale}),
+        ...(speechmaticsOptions.transcription_config?.punctuation_overrides?.permitted_marks &&
+          {SPEECHMATICS_PUNCTUATION_ALLOWED:
+            speechmaticsOptions.transcription_config.punctuation_overrides.permitted_marks.join(',')}),
+        ...(speechmaticsOptions.transcription_config?.punctuation_overrides?.sensitivity &&
+          {SPEECHMATICS_PUNCTUATION_SENSITIVITY:
+            speechmaticsOptions.transcription_config?.punctuation_overrides?.sensitivity}),
+        ...(speechmaticsOptions.transcription_config?.operating_point &&
+          {SPEECHMATICS_OPERATING_POINT: speechmaticsOptions.transcription_config.operating_point}),
+        ...(speechmaticsOptions.transcription_config?.enable_entities &&
+          {SPEECHMATICS_ENABLE_ENTTIES: speechmaticsOptions.transcription_config.enable_entities}),
+        ...(speechmaticsOptions.transcription_config?.audio_filtering_config?.volume_threshold &&
+          {SPEECHMATICS_VOLUME_THRESHOLD:
+            speechmaticsOptions.transcription_config.audio_filtering_config.volume_threshold}),
+        ...(speechmaticsOptions.transcription_config?.transcript_filtering_config?.remove_disfluencies &&
+          {SPEECHMATICS_REMOVE_DISFLUENCIES:
+            speechmaticsOptions.transcription_config.transcript_filtering_config.remove_disfluencies})
       };
     }
     else if (vendor.startsWith('custom:')) {

lib/utils/tts-streaming-buffer.js

@@ -0,0 +1,452 @@
+const Emitter = require('events');
+const assert = require('assert');
+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;
+
+    // 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.timer = null;
+    // Record the last time the text buffer was updated.
+    this.lastUpdateTime = 0;
+  }
+
+  get isEmpty() {
+    return this.queue.length === 0;
+  }
+
+  get size() {
+    return this.bufferedLength;
+  }
+
+  get isFull() {
+    return this._isFull;
+  }
+
+  get ep() {
+    return this.cs?.ep;
+  }
+
+  async start() {
+    assert.ok(
+      this._connectionStatus === TtsStreamingConnectionStatus.NotConnected,
+      'TtsStreamingBuffer:start already started, or has failed'
+    );
+
+    this.vendor = this.cs.getTsStreamingVendor();
+    if (!this.vendor) {
+      this.logger.info('TtsStreamingBuffer:start No TTS streaming vendor configured');
+      throw new Error('No TTS streaming vendor configured');
+    }
+
+    this.logger.info(`TtsStreamingBuffer:start Connecting to TTS streaming with vendor ${this.vendor}`);
+
+    this._connectionStatus = TtsStreamingConnectionStatus.Connecting;
+    try {
+      if (this.eventHandlers.length === 0) this._initHandlers(this.ep);
+      await this._api(this.ep, [this.ep.uuid, 'connect']);
+    } catch (err) {
+      this.logger.info({ err }, 'TtsStreamingBuffer:start Error connecting to TTS streaming');
+      this._connectionStatus = TtsStreamingConnectionStatus.Failed;
+    }
+  }
+
+  stop() {
+    clearTimeout(this.timer);
+    this.removeCustomEventListeners();
+    if (this.ep) {
+      this._api(this.ep, [this.ep.uuid, 'close'])
+        .catch((err) =>
+          this.logger.info({ err }, 'TtsStreamingBuffer:stop Error closing TTS streaming')
+        );
+    }
+    this.timer = null;
+    this.queue = [];
+    this.bufferedLength = 0;
+    this._connectionStatus = TtsStreamingConnectionStatus.NotConnected;
+  }
+
+  /**
+   * 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` };
+    }
+
+    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 (this.bufferedLength + totalLength > HIGH_WATER_BUFFER_SIZE) {
+      this.logger.info(
+        `TtsStreamingBuffer throttling: buffer is full, rejecting request to buffer ${totalLength} tokens`
+      );
+      if (!this._isFull) {
+        this._isFull = true;
+        this.emit(TtsStreamingEvents.Pause);
+      }
+      return { status: 'failed', reason: 'full' };
+    }
+
+    this.logger.debug(
+      `TtsStreamingBuffer:bufferTokens "${displayedTokens}" (length: ${totalLength})`
+    );
+    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._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() {
+    if (this._connectionStatus === TtsStreamingConnectionStatus.Connecting) {
+      this.logger.debug('TtsStreamingBuffer:flush TTS stream is not quite ready - wait for connect');
+      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.isEmpty) {
+        this._doFlush();
+      }
+      else {
+        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.queue = [];
+    this.bufferedLength = 0;
+    this.timer = null;
+    this._isFull = false;
+  }
+
+  /**
+   * 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 _feedQueue(handlingTimeout = false) {
+    this.logger.debug({ queue: this.queue }, 'TtsStreamingBuffer:_feedQueue');
+    try {
+      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:_feedQueue TTS stream is not connected');
+        return;
+      }
+
+      // --- 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;
+      }
+
+      // 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) {
+          chunkEnd = findWordBoundary(combinedText, limit);
+          if (chunkEnd <= 0) {
+            this._setTimerIfNeeded();
+            return;
+          }
+        } else {
+          this._setTimerIfNeeded();
+          return;
+        }
+      }
+      const chunk = combinedText.slice(0, 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;
+
+      const modifiedChunk = chunk.replace(/\n\n/g, '\n \n');
+      this.logger.debug(`TtsStreamingBuffer:_feedQueue sending chunk to tts: ${modifiedChunk}`);
+
+      try {
+        await this._api(this.ep, [this.ep.uuid, 'send', modifiedChunk]);
+      } catch (err) {
+        this.logger.info({ err, chunk }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
+      }
+
+      if (this._isFull && this.bufferedLength <= LOW_WATER_BUFFER_SIZE) {
+        this.logger.info('TtsStreamingBuffer throttling: buffer is no longer full - resuming');
+        this._isFull = false;
+        this.emit(TtsStreamingEvents.Resume);
+      }
+
+      return this._feedQueue();
+    } catch (err) {
+      this.logger.info({ err }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
+      this.queue = [];
+      this.bufferedLength = 0;
+    }
+  }
+
+  async _api(ep, args) {
+    const apiCmd = `uuid_${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}`);
+      throw new Error(`Error calling ${apiCmd}: ${res.body}`);
+    }
+  }
+
+  _doFlush() {
+    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(`TtsStreamingBuffer:_onConnect streaming tts connection made to ${vendor} successful`);
+    this._connectionStatus = TtsStreamingConnectionStatus.Connected;
+    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.bufferedLength > 0 && !this.timer) {
+      this.logger.debug({queue: this.queue},
+        `TtsStreamingBuffer:_setTimerIfNeeded setting timer because ${this.bufferedLength} buffered`);
+      this.timer = setTimeout(this._onTimeout.bind(this), TIMEOUT_RETRY_MSECS);
+    }
+  }
+
+  _removeTimer() {
+    if (this.timer) {
+      this.logger.debug('TtsStreamingBuffer:_removeTimer clearing timer');
+      clearTimeout(this.timer);
+      this.timer = null;
+    }
+  }
+
+  _onTimeout() {
+    this.logger.debug('TtsStreamingBuffer:_onTimeout Timeout waiting for sentence boundary');
+    // Check if new text has been added since the timer was set.
+    const now = Date.now();
+    if (now - this.lastUpdateTime < TIMEOUT_RETRY_MSECS) {
+      this.logger.debug('TtsStreamingBuffer:_onTimeout New text received recently; postponing flush.');
+      this._setTimerIfNeeded();
+      return;
+    }
+    this.timer = null;
+    this._feedQueue(true);
+  }
+
+  _onTtsEmpty(vendor) {
+    this.emit(TtsStreamingEvents.Empty, { vendor });
+  }
+
+  addCustomEventListener(ep, event, handler) {
+    this.eventHandlers.push({ ep, event, handler });
+    ep.addCustomEventListener(event, handler);
+  }
+
+  removeCustomEventListeners() {
+    this.eventHandlers.forEach((h) => h.ep.removeCustomEventListener(h.event, h.handler));
+  }
+
+  _initHandlers(ep) {
+    [
+      'deepgram',
+      'cartesia',
+      'elevenlabs',
+      'rimelabs',
+      'custom'
+    ].forEach((vendor) => {
+      const eventClassName = `${vendor.charAt(0).toUpperCase() + vendor.slice(1)}TtsStreamingEvents`;
+      const eventClass = require('../utils/constants')[eventClassName];
+      if (!eventClass) throw new Error(`Event class for vendor ${vendor} not found`);
+
+      this.addCustomEventListener(ep, eventClass.Connect, this._onConnect.bind(this, vendor));
+      this.addCustomEventListener(ep, eventClass.ConnectFailure, this._onConnectFailure.bind(this, vendor));
+      this.addCustomEventListener(ep, eventClass.Empty, this._onTtsEmpty.bind(this, vendor));
+    });
+  }
+}
+
+const findSentenceBoundary = (text, limit) => {
+  // Look for punctuation or double newline that signals sentence end.
+  const sentenceEndRegex = /[.!?](?=\s|$)|\n\n/g;
+  let lastSentenceBoundary = -1;
+  let match;
+  while ((match = sentenceEndRegex.exec(text)) && match.index < limit) {
+    const precedingText = text.slice(0, match.index).trim();
+    if (precedingText.length > 0) {
+      if (
+        match[0] === '\n\n' ||
+        (match.index === 0 || !/\d$/.test(text[match.index - 1]))
+      ) {
+        lastSentenceBoundary = match.index + (match[0] === '\n\n' ? 2 : 1);
+      }
+    }
+  }
+  return lastSentenceBoundary;
+};
+
+const findWordBoundary = (text, limit) => {
+  const wordBoundaryRegex = /\s+/g;
+  let lastWordBoundary = -1;
+  let match;
+  while ((match = wordBoundaryRegex.exec(text)) && match.index < limit) {
+    lastWordBoundary = match.index;
+  }
+  return lastWordBoundary;
+};
+
+module.exports = TtsStreamingBuffer;

lib/utils/ws-requestor.js

@@ -1,7 +1,7 @@
 const assert = require('assert');
 const BaseRequestor = require('./base-requestor');
 const short = require('short-uuid');
-const {HookMsgTypes} = require('./constants.json');
+const {HookMsgTypes, WS_CLOSE_CODES} = require('./constants.json');
 const Websocket = require('ws');
 const snakeCaseKeys = require('./snakecase-keys');
 const {
@@ -12,6 +12,20 @@ const {
   JAMBONES_WS_MAX_PAYLOAD,
   HTTP_USER_AGENT_HEADER
 } = require('../config');
+const MTYPE_WANTS_ACK = [
+  'call:status',
+  'verb:status',
+  'jambonz:error',
+  'llm:event',
+  'llm:tool-call',
+  'tts:streaming-event',
+  'tts:tokens-result',
+];
+const MTYPE_NO_DATA = [
+  'llm:tool-output',
+  'tts:flush',
+  'tts:clear'
+];
 
 class WsRequestor extends BaseRequestor {
   constructor(logger, account_sid, hook, secret) {
@@ -44,7 +58,7 @@ class WsRequestor extends BaseRequestor {
   async request(type, hook, params, httpHeaders = {}) {
     assert(HookMsgTypes.includes(type));
     const url = hook.url || hook;
-    const wantsAck = !['call:status', 'verb:status', 'jambonz:error', 'llm:event', 'llm:tool-call'].includes(type);
+    const wantsAck = !MTYPE_WANTS_ACK.includes(type);
 
     if (this.maliciousClient) {
       this.logger.info({url: this.url}, 'WsRequestor:request - discarding msg to malicious client');
@@ -247,13 +261,13 @@ class WsRequestor extends BaseRequestor {
     }
   }
 
-  close() {
+  close(code = WS_CLOSE_CODES.NormalClosure) {
     this.closedGracefully = true;
-    this.logger.debug('WsRequestor:close closing socket');
+    this.logger.debug(`WsRequestor:close closing socket with code ${code}`);
     this._stopPingTimer();
     try {
       if (this.ws) {
-        this.ws.close(1000);
+        this.ws.close(code);
         this.ws.removeAllListeners();
         this.ws = null;
       }
@@ -408,7 +422,7 @@ class WsRequestor extends BaseRequestor {
 
         case 'command':
           assert.ok(command, 'command property not supplied');
-          assert.ok(data || command === 'llm:tool-output', 'data property not supplied');
+          assert.ok(data || MTYPE_NO_DATA.includes(command), 'data property not supplied');
           this._recvCommand(msgid, command, call_sid, queueCommand, tool_call_id, data);
           break;
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jambonz-feature-server",
-  "version": "0.9.2",
+  "version": "0.9.3",
   "main": "app.js",
   "engines": {
     "node": ">= 18.x"
@@ -31,10 +31,10 @@
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.8",
-    "@jambonz/speech-utils": "^0.1.22",
+    "@jambonz/speech-utils": "^0.2.3",
     "@jambonz/stats-collector": "^0.1.10",
-    "@jambonz/time-series": "^0.2.9",
-    "@jambonz/verb-specifications": "^0.0.83",
+    "@jambonz/verb-specifications": "^0.0.97",
+    "@jambonz/time-series": "^0.2.13",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.50.0",
@@ -47,8 +47,8 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.45",
-    "drachtio-srf": "^4.5.35",
+    "drachtio-fsmrf": "^4.0.1",
+    "drachtio-srf": "^5.0.2",
     "express": "^4.19.2",
     "express-validator": "^7.0.1",
     "moment": "^2.30.1",
@@ -56,7 +56,7 @@
     "pino": "^8.20.0",
     "polly-ssml-split": "^0.1.0",
     "proxyquire": "^2.1.3",
-    "sdp-transform": "^2.14.2",
+    "sdp-transform": "^2.15.0",
     "short-uuid": "^5.1.0",
     "sinon": "^17.0.1",
     "to-snake-case": "^1.0.0",

test/create-call-test.js

@@ -222,3 +222,62 @@ test('test create-call app_json', async(t) => {
     t.error(err);
   }
 });
+
+test('test create-call timeLimit', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+
+    // GIVEN
+    let from = 'create-call-app-json';
+    let account_sid = 'bb845d4b-83a9-4cde-a6e9-50f3743bab3f';
+
+    // Give UAS app time to come up
+    const p = sippUac('uas.xml', '172.38.0.10', from);
+    await waitFor(1000);
+
+    const startTime = Date.now();
+
+    const app_json = `[
+      {
+        "verb": "pause",
+        "length": 7
+      }
+    ]`;
+
+    const post = bent('http://127.0.0.1:3000/', 'POST', 'json', 201);
+    post('v1/createCall', {
+      'account_sid':account_sid,
+      "call_hook": {
+        "url": "http://127.0.0.1:3100/",
+        "method": "POST",
+        "username": "username",
+        "password": "password"
+      },
+      app_json,
+      "from": from,
+      "to": {
+        "type": "phone",
+        "number": "15583084809"
+      },
+      "timeLimit": 1,
+      "speech_recognizer_vendor": "google",
+      "speech_recognizer_language": "en"
+    });
+
+    //THEN
+    await p;
+    const endTime = Date.now();
+
+    t.ok(endTime - startTime < 2000, 'create-call: timeLimit is respected');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});

test/db/jambones-sql.sql

@@ -351,6 +351,8 @@ speech_credential_sid CHAR(36) NOT NULL,
 model VARCHAR(512) NOT NULL,
 reported_usage ENUM('REPORTED_USAGE_UNSPECIFIED','REALTIME','OFFLINE') DEFAULT 'REALTIME',
 name VARCHAR(64) NOT NULL,
+voice_cloning_key MEDIUMTEXT,
+use_voice_cloning_key BOOLEAN DEFAULT false,
 PRIMARY KEY (google_custom_voice_sid)
 );