add startup logging

turn down some logging
wip
2026-01-25 02:07:56 +00:00 · 2025-08-21 13:15:36 -04:00 · 2025-08-21 13:00:23 -04:00 · 2025-08-20 18:16:36 -04:00 · 2025-08-20 15:15:56 -04:00 · 2025-08-20 11:02:03 -04:00
43 changed files with 4784 additions and 2891 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -2,6 +2,9 @@
 logs
 *.log

+.claude/
+CLAUDE.md 
+
 # Runtime data
 pids
 *.pid
--- a/app.js
+++ b/app.js
@@ -29,6 +29,12 @@ const {LifeCycleEvents, FS_UUID_SET_NAME, SystemState, FEATURE_SERVER} = require
 const installSrfLocals = require('./lib/utils/install-srf-locals');
 const createHttpListener = require('./lib/utils/http-listener');
 const healthCheck = require('@jambonz/http-health-check');
+const ProcessMonitor = require('./lib/utils/process-monitor');
+const monitor = new ProcessMonitor(logger);
+
+// Log startup
+monitor.logStartup();
+monitor.setupSignalHandlers();

 logger.on('level-change', (lvl, _val, prevLvl, _prevVal, instance) => {
  if (logger !== instance) {
--- a/lib/config.js
+++ b/lib/config.js
@@ -130,7 +130,7 @@ const OPTIONS_PING_INTERVAL = parseInt(process.env.OPTIONS_PING_INTERVAL, 10) ||
 const JAMBONZ_RECORD_WS_BASE_URL = process.env.JAMBONZ_RECORD_WS_BASE_URL || process.env.JAMBONES_RECORD_WS_BASE_URL;
 const JAMBONZ_RECORD_WS_USERNAME = process.env.JAMBONZ_RECORD_WS_USERNAME || process.env.JAMBONES_RECORD_WS_USERNAME;
 const JAMBONZ_RECORD_WS_PASSWORD = process.env.JAMBONZ_RECORD_WS_PASSWORD || process.env.JAMBONES_RECORD_WS_PASSWORD;
-const JAMBONZ_DISABLE_DIAL_PAI_HEADER = process.env.JAMBONZ_DISABLE_DIAL_PAI_HEADER || false;
+const JAMBONZ_DIAL_PAI_HEADER = process.env.JAMBONZ_DIAL_PAI_HEADER || false;
 const JAMBONES_DISABLE_DIRECT_P2P_CALL = process.env.JAMBONES_DISABLE_DIRECT_P2P_CALL || false;

 const JAMBONES_EAGERLY_PRE_CACHE_AUDIO = parseInt(process.env.JAMBONES_EAGERLY_PRE_CACHE_AUDIO, 10) || 0;
@@ -225,11 +225,11 @@ module.exports = {
  JAMBONZ_RECORD_WS_BASE_URL,
  JAMBONZ_RECORD_WS_USERNAME,
  JAMBONZ_RECORD_WS_PASSWORD,
-  JAMBONZ_DISABLE_DIAL_PAI_HEADER,
+  JAMBONZ_DIAL_PAI_HEADER,
  JAMBONES_DISABLE_DIRECT_P2P_CALL,
  JAMBONES_USE_FREESWITCH_TIMER_FD,
  JAMBONES_DIAL_SBC_FOR_REGISTERED_USER,
  JAMBONES_MEDIA_TIMEOUT_MS,
  JAMBONES_MEDIA_HOLD_TIMEOUT_MS,
-  JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS
+  JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS,
 };
--- a/lib/http-routes/api/create-call.js
+++ b/lib/http-routes/api/create-call.js
@@ -13,10 +13,11 @@ const WsRequestor = require('../../utils/ws-requestor');
 const RootSpan = require('../../utils/call-tracer');
 const dbUtils = require('../../utils/db-utils');
 const { decrypt } = require('../../utils/encrypt-decrypt');
-const { mergeSdpMedia, extractSdpMedia } = require('../../utils/sdp-utils');
+const { mergeSdpMedia, extractSdpMedia, removeVideoSdp } = require('../../utils/sdp-utils');
 const { createCallSchema, customSanitizeFunction } = require('../schemas/create-call');
 const { selectHostPort } = require('../../utils/network');
 const { JAMBONES_DIAL_SBC_FOR_REGISTERED_USER } = require('../../config');
+const { createMediaEndpoint } = require('../../utils/media-endpoint');

 const removeNullProperties = (obj) => (Object.keys(obj).forEach((key) => obj[key] === null && delete obj[key]), obj);
 const removeNulls = (req, res, next) => {
@@ -67,7 +68,7 @@ router.post('/',
      const {
        lookupAppBySid
      }  = srf.locals.dbHelpers;
-      const {getSBC, getFreeswitch} = srf.locals;
+      const {getSBC} = srf.locals;
      let sbcAddress = getSBC();
      if (!sbcAddress) throw new Error('no available SBCs for outbound call creation');
      const target = restDial.to;
@@ -170,9 +171,7 @@ router.post('/',
      }

      /* create endpoint for outdial */
-      const ms = getFreeswitch();
-      if (!ms) throw new Error('no available Freeswitch for outbound call creation');
-      const ep = await ms.createEndpoint();
+      const ep = await createMediaEndpoint(srf, logger);
      logger.debug(`createCall: successfully allocated endpoint, sending INVITE to ${sbcAddress}`);

      /* launch outdial */
@@ -181,10 +180,14 @@ router.post('/',
      let localSdp = ep.local.sdp;

      if (req.body.dual_streams) {
-        dualEp = await ms.createEndpoint();
+        dualEp = await createMediaEndpoint(srf, logger);
        localSdp = mergeSdpMedia(localSdp, dualEp.local.sdp);
      }
-
+      if (process.env.JAMBONES_VIDEO_CALLS_ENABLED_IN_FS) {
+        logger.debug('createCall: removing video sdp');
+        localSdp = removeVideoSdp(localSdp);
+        ep.modify(localSdp);
+      }
      const connectStream = async(remoteSdp) => {
        if (remoteSdp !== sdp) {
          sdp = remoteSdp;
@@ -235,6 +238,7 @@ router.post('/',
        if (app.call_hook.url === app.call_status_hook?.url  || !app.call_status_hook?.url) {
          logger.debug('reusing websocket for call status hook');
          app.notifier = app.requestor;
+          app.call_status_hook = app.call_hook;
        }
      }
      else {
--- a/lib/session/call-session.js
+++ b/lib/session/call-session.js
@@ -10,8 +10,9 @@ const {
  RecordState,
  AllowedSipRecVerbs,
  AllowedConfirmSessionVerbs,
-  TtsStreamingEvents
-} = require('../utils/constants');
+  TtsStreamingEvents,
+  ListenStatus
+} = require('../utils/constants.json');
 const moment = require('moment');
 const assert = require('assert');
 const sessionTracker = require('./session-tracker');
@@ -29,10 +30,6 @@ const {
  JAMBONES_INJECT_CONTENT,
  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
  AWS_REGION,
-  JAMBONES_USE_FREESWITCH_TIMER_FD,
-  JAMBONES_MEDIA_TIMEOUT_MS,
-  JAMBONES_MEDIA_HOLD_TIMEOUT_MS,
-  JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS
 } = require('../config');
 const bent = require('bent');
 const BackgroundTaskManager = require('../utils/background-task-manager');
@@ -40,7 +37,8 @@ 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 { sleepFor } = require('../utils/helpers');
+const { createMediaEndpoint } = require('../utils/media-endpoint');
+const SttLatencyCalculator = require('../utils/stt-latency-calculator');
 const sqlRetrieveQueueEventHook = `SELECT * FROM webhooks 
 WHERE webhook_sid = 
 (
@@ -150,6 +148,30 @@ class CallSession extends Emitter {
    this.conversationTurns = [];
    this.on('userSaid', this._onUserSaid.bind(this));
    this.on('botSaid', this._onBotSaid.bind(this));
+    /**
+     * Support STT latency
+     */
+    this.sttLatencyCalculator = new SttLatencyCalculator({
+      logger,
+      cs: this
+    });
+    this.on('transcribe-start', () => {
+      this.sttLatencyCalculator.resetTime();
+    });
+    this.on('on-transcription', () => {
+      this.sttLatencyCalculator.onTranscriptionReceived();
+    });
+    this.on('transcribe-stop', () => {
+      this.sttLatencyCalculator.onTranscribeStop();
+    });
+  }
+
+  get notifySttLatencyEnabled() {
+    return this._notifySttLatencyEnabled || false;
+  }
+
+  set notifySttLatencyEnabled(enabled) {
+    this._notifySttLatencyEnabled = enabled;
  }

  /**
@@ -720,10 +742,25 @@ class CallSession extends Emitter {
    return this._fillerNoise;
  }

+  async pauseOrResumeBackgroundListenIfRequired(action, silence = false) {
+    if ((action == 'pauseCallRecording' || action == 'resumeCallRecording') &&
+    this.backgroundTaskManager.isTaskRunning('record')) {
+      this.logger.debug({action, silence}, 'CallSession:pauseOrResumeBackgroundListenIfRequired');
+      const backgroundListenTask = this.backgroundTaskManager.getTask('record');
+      const status = action === 'pauseCallRecording' ? ListenStatus.Pause : ListenStatus.Resume;
+      backgroundListenTask.updateListen(
+        status,
+        silence
+      );
+    }
+  }
+
  async notifyRecordOptions(opts) {
-    const {action} = opts;
+    const {action, silence} = opts;
    this.logger.debug({opts}, 'CallSession:notifyRecordOptions');

+    this.pauseOrResumeBackgroundListenIfRequired(action, silence);
+
    /* if we have not answered yet, just save the details for later */
    if (!this.dlg) {
      if (action === 'startCallRecording') {
@@ -1097,6 +1134,7 @@ class CallSession extends Emitter {
          return {
            api_key: credential.api_key,
            model_id: credential.model_id,
+            stt_model_id: credential.stt_model_id,
            embedding: credential.embedding,
            options: credential.options
          };
@@ -1108,10 +1146,31 @@ class CallSession extends Emitter {
            options: credential.options
          };
        }
+        else if ('resemble' === vendor) {
+          return {
+            api_key: credential.api_key,
+            resemble_tts_use_tls: credential.resemble_tts_use_tls,
+            resemble_tts_uri: credential.resemble_tts_uri,
+          };
+        }
+        else if ('inworld' === vendor) {
+          return {
+            api_key: credential.api_key,
+            model_id: credential.model_id,
+            options: credential.options
+          };
+        }
        else if ('assemblyai' === vendor) {
          return {
            speech_credential_sid: credential.speech_credential_sid,
-            api_key: credential.api_key
+            api_key: credential.api_key,
+            service_version: credential.service_version
+          };
+        }
+        else if ('deepgramriver' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            api_key: credential.api_key,
          };
        }
        else if ('voxist' === vendor) {
@@ -1421,6 +1480,14 @@ class CallSession extends Emitter {
        method: 'POST'
      };
    }
+
+    /* if given a relative url then send to same base url as current session */
+    if (this.requestor._isRelativeUrl(opts.call_hook?.url)) {
+      opts.call_hook.url = `${this.requestor.baseUrl}${opts.call_hook?.url}`;
+    }
+    if (this.requestor._isRelativeUrl(opts.call_status_hook?.url)) {
+      opts.call_status_hook.url = `${this.requestor.baseUrl}${opts.call_status_hook?.url}`;
+    }
  }

  /**
@@ -1648,8 +1715,8 @@ Duration=${duration} `
  async _lccWhisper(opts, callSid) {
    const {whisper} = opts;
    let tasks;
-    const b3 = this.b3;
-    const httpHeaders = b3 && {b3};
+    //const b3 = this.b3;
+    //const httpHeaders = b3 && {b3};

    // this whole thing requires us to be in a Dial verb
    const task = this.currentTask;
@@ -1658,12 +1725,15 @@ Duration=${duration} `
    }

    // allow user to provide a url object, a url string, an array of tasks, or a single task
-    if (typeof whisper === 'string' || (typeof whisper === 'object' && whisper.url)) {
-      // retrieve a url
-      const json = await this.requestor(opts.call_hook, this.callInfo.toJSON(), httpHeaders);
-      tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
-    }
-    else if (Array.isArray(whisper)) {
+    // Disable passing a URL as not functional Sam Machin - 17/07/2025
+    //if (typeof whisper === 'string' || (typeof whisper === 'object' && whisper.url && !whisper.verb)) {
+    //  // retrieve a url
+    //  const json = await this.requestor(opts.call_hook, this.callInfo.toJSON(), httpHeaders);
+    //  tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
+    //}
+    //else if (Array.isArray(whisper)) {
+
+    if (Array.isArray(whisper)) {
      // an inline array of tasks
      tasks = normalizeJambones(this.logger, whisper).map((tdata) => makeTask(this.logger, tdata));
    }
@@ -1900,6 +1970,17 @@ Duration=${duration} `
      }
    } catch (err) {
      this.logger.info({err, opts, callSid}, 'CallSession:updateCall - error updating call');
+      const {writeAlerts} = this.srf.locals;
+      try {
+        writeAlerts({
+          alert_type: 'error-updating-call',
+          account_sid: this.accountSid,
+          message: err.message,
+          target_sid: callSid
+        });
+      } catch (err) {
+        this.logger.error({err}, 'Error writing error-updating-call alert');
+      }
    }
  }

@@ -2301,17 +2382,16 @@ Duration=${duration} `

    // need to allocate an endpoint
    try {
-      if (!this.ms) this.ms = this.getMS();
-      const ep = await this.ms.createEndpoint({
+      const ep = await this._createMediaEndpoint({
        headers: {
          'X-Jambones-Call-ID': this.callId,
+          'X-Call-Sid': this.callSid,
        },
        remoteSdp: this.req.body
      });
      //ep.cs = this;
      this.ep = ep;
      this.logger.info(`allocated endpoint ${ep.uuid}`);
-      this._configMsEndpoint();

      this.ep.on('destroy', () => {
        this.logger.debug(`endpoint was destroyed!! ${this.ep.uuid}`);
@@ -2382,9 +2462,6 @@ Duration=${duration} `
      this.logger.error('CallSession:replaceEndpoint cannot be called without stable dlg');
      return;
    }
-    // When this call kicked out from conference, session need to replace endpoint
-    // but this.ms might be undefined/null at this case.
-    this.ms = this.ms || this.getMS();
    // Destroy previous ep if it's still running.
    if (this.ep?.connected) this.ep.destroy();

@@ -2409,8 +2486,7 @@ Duration=${duration} `
     * This prevents call failures during media renegotiation.
     */

-    this.ep = await this.ms.createEndpoint();
-    this._configMsEndpoint();
+    this.ep = await this._createMediaEndpoint();

    const sdp = await this.dlg.modify(this.ep.local.sdp);
    await this.ep.modify(sdp);
@@ -2454,6 +2530,8 @@ Duration=${duration} `
    this.clearOrRestoreActionHookDelayProcessor().catch((err) => {});

    this.ttsStreamingBuffer?.stop();
+
+    this.sttLatencyCalculator?.stop();
  }

  /**
@@ -2670,15 +2748,8 @@ Duration=${duration} `
  async createOrRetrieveEpAndMs() {
    if (this.ms && this.ep) return {ms: this.ms, ep: this.ep};

-    // get a media server
-    if (!this.ms) {
-      const ms = this.srf.locals.getFreeswitch();
-      if (!ms) throw new Error('no available freeswitch');
-      this.ms = ms;
-    }
    if (!this.ep) {
-      this.ep = await this.ms.createEndpoint({remoteSdp: this.req.body});
-      this._configMsEndpoint();
+      this.ep = await this._createMediaEndpoint({remoteSdp: this.req.body});
    }
    return {ms: this.ms, ep: this.ep};
  }
@@ -2831,8 +2902,7 @@ Duration=${duration} `
  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();
+    this.ep = await this._createMediaEndpoint({remoteSdp: this.dlg.remote.sdp});
    await this.dlg.modify(this.ep.local.sdp, {
      headers: {
        'X-Reason': 'anchor-media'
@@ -2907,59 +2977,13 @@ 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_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);
-    }
-
-    const origDestroy = this.ep.destroy.bind(this.ep);
-    this.ep.destroy = async() => {
-      try {
-        if (this.currentTask?.name === TaskName.Transcribe && JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS) {
-          // transcribe task is being used, wait for some time before destroy
-          // if final transcription is received but endpoint is already closed,
-          // freeswitch module will not be able to send the transcription
-
-          this.logger.debug('callSession:_configMsEndpoint -' +
-            ' transcribe task, wait for some time before destroy');
-          await sleepFor(JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS);
-        }
-        await origDestroy();
-      } catch (err) {
-        this.logger.error(err, 'callSession:_configMsEndpoint - error destroying endpoint');
-      }
-    };
-  }
-
  async _handleMediaTimeout(evt) {
-    if (evt.reason === 'MEDIA_TIMEOUT' && !this.callGone) {
+    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
   */
@@ -3011,6 +3035,18 @@ Duration=${duration} `
    });
  }

+  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
+      ep.execute('start_dtmf').catch((err) => {
+        this.logger.info({err}, 'CallSession:_enableInbandDtmfIfRequired - error starting DTMF');
+      });
+      ep.inbandDtmfEnabled = true;
+    }
+  }
+
  _clearTasks(backgroundGather, evt) {
    if (this.requestor instanceof WsRequestor && !backgroundGather.cleared) {
      this.logger.debug({evt}, 'CallSession:_clearTasks on event from background gather');
@@ -3044,6 +3080,11 @@ Duration=${duration} `
    const task = this.currentTask;
    if (task && TaskName.Say === task.name) {
      task.notifyTtsStreamIsEmpty();
+    } else if (
+      // If Gather nested say task is streaming
+      TaskName.Gather === task.name && task.sayTask && task.sayTask.isStreamingTts) {
+      const sayTask = task.sayTask;
+      sayTask.notifyTtsStreamIsEmpty();
    }
  }

@@ -3118,6 +3159,16 @@ Duration=${duration} `
    }
  }

+  async _createMediaEndpoint(drachtioFsmrfOptions = {}) {
+    return await createMediaEndpoint(this.srf, this.logger, {
+      activeMs: this.getMS(),
+      drachtioFsmrfOptions,
+      onHoldMusic: this.onHoldMusic,
+      inbandDtmfEnabled: this.inbandDtmfEnabled,
+      mediaTimeoutHandler: this._handleMediaTimeout.bind(this),
+    });
+  }
+
  getFormattedConversation(numTurns) {
    const turns = this.conversationTurns.slice(-numTurns);
    if (turns.length === 0) return null;
@@ -3128,6 +3179,20 @@ Duration=${duration} `
      return `assistant: ${t.text}`;
    }).join('\n');
  }
+
+  startSttLatencyVad() {
+    if (this.notifySttLatencyEnabled) {
+      this.sttLatencyCalculator.start();
+    }
+  }
+
+  stopSttLatencyVad() {
+    this.sttLatencyCalculator.stop();
+  }
+
+  calculateSttLatency() {
+    return this.sttLatencyCalculator.calculateLatency();
+  }
 }

 module.exports = CallSession;
--- a/lib/session/confirm-call-session.js
+++ b/lib/session/confirm-call-session.js
@@ -8,7 +8,8 @@ const CallSession = require('./call-session');

 */
 class ConfirmCallSession extends CallSession {
-  constructor({logger, application, dlg, ep, tasks, callInfo, accountInfo, memberId, confName, rootSpan, req}) {
+  // eslint-disable-next-line max-len
+  constructor({logger, application, dlg, ep, tasks, callInfo, accountInfo, memberId, confName, rootSpan, req, tmpFiles}) {
    super({
      logger,
      application,
@@ -24,6 +25,7 @@ class ConfirmCallSession extends CallSession {
    this.dlg = dlg;
    this.ep = ep;
    this.req = req;
+    this.tmpFiles = tmpFiles;
  }

  /**
--- a/lib/session/rest-call-session.js
+++ b/lib/session/rest-call-session.js
@@ -27,10 +27,13 @@ class RestCallSession extends CallSession {
    }

    this.on('callStatusChange', this._notifyCallStatusChange.bind(this));
-    this._notifyCallStatusChange({
-      callStatus: CallStatus.Trying,
-      sipStatus: 100,
-      sipReason: 'Trying'
+
+    setImmediate(() => {
+      this._notifyCallStatusChange({
+        callStatus: CallStatus.Trying,
+        sipStatus: 100,
+        sipReason: 'Trying'
+      });
    });
  }

--- a/lib/session/siprec-call-session.js
+++ b/lib/session/siprec-call-session.js
@@ -45,12 +45,11 @@ class SipRecCallSession extends InboundCallSession {

  async answerSipRecCall() {
    try {
-      this.ms = this.getMS();
      let remoteSdp = this.sdp1.replace(/sendonly/, 'sendrecv');
-      this.ep = await this.ms.createEndpoint({remoteSdp});
+      this.ep = await this._createMediaEndpoint({remoteSdp});
      //this.logger.debug({remoteSdp, localSdp: this.ep.local.sdp}, 'SipRecCallSession - allocated first endpoint');
      remoteSdp = this.sdp2.replace(/sendonly/, 'sendrecv');
-      this.ep2 = await this.ms.createEndpoint({remoteSdp});
+      this.ep2 = await this._createMediaEndpoint({remoteSdp});
      //this.logger.debug({remoteSdp, localSdp: this.ep2.local.sdp}, 'SipRecCallSession - allocated second endpoint');
      await this.ep.bridge(this.ep2);
      const combinedSdp = await createSipRecPayload(this.ep.local.sdp, this.ep2.local.sdp, this.logger);
--- a/lib/tasks/alert.js
+++ b/lib/tasks/alert.js
@@ -0,0 +1,31 @@
+const Task = require('./task');
+const {TaskName} = require('../utils/constants');
+
+class TaskAlert extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts);
+    this.message = this.data.message;
+  }
+
+  get name() { return TaskName.Alert; }
+
+  async exec(cs) {
+    const {srf, accountSid:account_sid, callSid:target_sid, applicationSid:application_sid} = cs;
+    const {writeAlerts, AlertType} = srf.locals;
+    await super.exec(cs);
+    writeAlerts({
+      account_sid,
+      alert_type: AlertType.APPLICATION,
+      detail: `Application SID ${application_sid}`,
+      message: this.message,
+      target_sid
+    }).catch((err) => this.logger.info({err}, 'Error generating alert application'));
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+    this.notifyTaskDone();
+  }
+}
+
+module.exports = TaskAlert;
--- a/lib/tasks/conference.js
+++ b/lib/tasks/conference.js
@@ -673,7 +673,8 @@ class Conference extends Task {
        confName: this.confName,
        tasks,
        rootSpan: cs.rootSpan,
-        req: cs.req
+        req: cs.req,
+        tmpFiles: cs.tmpFiles,
      });
      await this._playSession.exec();
      this._playSession = null;
--- a/lib/tasks/config.js
+++ b/lib/tasks/config.js
@@ -24,6 +24,9 @@ class TaskConfig extends Task {
    if ('notifyEvents' in this.data) {
      this.notifyEvents = !!this.data.notifyEvents;
    }
+    if (this.hasNotifySttLatency) {
+      this.notifySttLatency = !!this.data.notifySttLatency;
+    }

    if (this.bargeIn.enable) {
      this.gatherOpts = {
@@ -83,6 +86,7 @@ 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 hasNotifySttLatency() { return Object.keys(this.data).includes('notifySttLatency'); }
  get hasTtsStream() { return Object.keys(this.ttsStream).length; }

  get summary() {
@@ -112,6 +116,8 @@ class TaskConfig extends Task {
    if (this.hasFillerNoise) phrase.push(`fillerNoise ${this.fillerNoise.enable ? 'on' : 'off'}`);
    if (this.data.amd) phrase.push('enable amd');
    if (this.notifyEvents) phrase.push(`event notification ${this.notifyEvents ? 'on' : 'off'}`);
+    if (this.hasNotifySttLatency) phrase.push(
+      `notifySttLatency ${this.notifySttLatency ? 'on' : 'off'}`);
    if (this.onHoldMusic) phrase.push(`onHoldMusic: ${this.onHoldMusic}`);
    if ('boostAudioSignal' in this.data) phrase.push(`setGain ${this.data.boostAudioSignal}`);
    if (this.hasReferHook) phrase.push('set referHook');
@@ -130,6 +136,11 @@ class TaskConfig extends Task {
      cs.notifyEvents = !!this.data.notifyEvents;
    }

+    if (this.hasNotifySttLatency) {
+      this.logger.debug(`turning notifySttLatency ${this.notifySttLatency ? 'on' : 'off'}`);
+      cs.notifySttLatencyEnabled = this.notifySttLatency;
+    }
+
    if (this.onHoldMusic) {
      cs.onHoldMusic = this.onHoldMusic;
    }
@@ -318,7 +329,10 @@ class TaskConfig extends Task {
        voiceMs: this.vad.voiceMs || 250,
        silenceMs: this.vad.silenceMs || 150,
        strategy: this.vad.strategy || 'one-shot',
-        mode: (this.vad.mode !== undefined && this.vad.mode !== null) ? this.vad.mode : 2
+        mode: (this.vad.mode !== undefined && this.vad.mode !== null) ? this.vad.mode : 2,
+        vendor: this.vad.vendor || 'silero',
+        threshold: this.vad.threshold || 0.5,
+        speechPadMs: this.vad.speechPadMs || 30,
      };
    }

--- a/lib/tasks/dial.js
+++ b/lib/tasks/dial.js
@@ -19,7 +19,7 @@ const parseDecibels = require('../utils/parse-decibels');
 const debug = require('debug')('jambonz:feature-server');
 const {parseUri} = require('drachtio-srf');
 const {ANCHOR_MEDIA_ALWAYS,
-  JAMBONZ_DISABLE_DIAL_PAI_HEADER,
+  JAMBONZ_DIAL_PAI_HEADER,
  JAMBONES_DIAL_SBC_FOR_REGISTERED_USER} = require('../config');
 const { isOnhold, isOpusFirst } = require('../utils/sdp-utils');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
@@ -109,6 +109,7 @@ class TaskDial extends Task {
    this.tag = this.data.tag;
    this.boostAudioSignal = this.data.boostAudioSignal;
    this._mediaPath = MediaPath.FullMedia;
+    this.forwardPAI = this.data.forwardPAI;

    if (this.dtmfHook) {
      const {parentDtmfCollector, childDtmfCollector} = parseDtmfOptions(logger, this.data.dtmfCapture || {});
@@ -397,7 +398,11 @@ class TaskDial extends Task {
            ...customHeaders
          }
        }, httpHeaders);
-        if (json && Array.isArray(json)) {
+        res.send(202);
+        this.logger.info('DialTask:handleRefer - sent 202 Accepted');
+
+        const returnedInstructions = !!json && Array.isArray(json);
+        if (returnedInstructions) {
          try {
            const logger = isChild ? this.logger : this.sd.logger;
            const tasks = normalizeJambones(logger, json).map((tdata) => makeTask(this.logger, tdata));
@@ -415,16 +420,21 @@ class TaskDial extends Task {
                /* need to update the callSid of the child with its own (new) AdultingCallSession */
                sessionTracker.add(adultingSession.callSid, adultingSession);
              }
+              if (this.ep) this.ep.unbridge();
+
+              /* if we got the REFER on the parent leg, end the dial task after completing the refer */
+              if (!isChild) {
+                logger.info('DialTask:handleRefer - killing dial task after processing REFER on parent leg');
+                cs.currentTask?.kill(cs, KillReason.ReferComplete);
+              }
            }
          } catch (err) {
            this.logger.info(err, 'Dial:handleRefer - error setting new application after receiving REFER');
          }
        }
-        //caller and callee legs are briged together, accept refer with 202 will release callee leg endpoint
-        //that makes freeswitch release endpoint for caller leg.
-        if (this.ep) this.ep.unbridge();
-        res.send(202);
-        this.logger.info('DialTask:handleRefer - sent 202 Accepted');
+        else {
+          this.logger.info('DialTask:handleRefer - no tasks returned from referHook, not setting new application');
+        }
      } catch (err) {
        this.logger.info({err}, 'DialTask:handleRefer - error processing incoming REFER');
        res.send(err.statusCode || 501);
@@ -539,13 +549,14 @@ class TaskDial extends Task {
    let sbcAddress = this.proxy || getSBC();
    const teamsInfo = {};
    let fqdn;
-
+    const forwardPAI = this.forwardPAI ?? JAMBONZ_DIAL_PAI_HEADER; // dial verb overides env var
+    this.logger.debug(forwardPAI, 'forwardPAI value');
    if (!sbcAddress) throw new Error('no SBC found for outbound call');
    this.headers = {
      'X-Account-Sid': cs.accountSid,
      ...(req && req.has('X-CID') && {'X-CID': req.get('X-CID')}),
      ...(direction === 'outbound' && callInfo.sbcCallid && {'X-CID': callInfo.sbcCallid}),
-      ...(!JAMBONZ_DISABLE_DIAL_PAI_HEADER && req && {
+      ...(req && forwardPAI && {
        ...(req.has('P-Asserted-Identity') && {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
        ...(req.has('Privacy') && {'Privacy': req.get('Privacy')}),
      }),
@@ -986,7 +997,7 @@ class TaskDial extends Task {
      this.epOther = null;
      this._mediaPath = releaseEntirely ? MediaPath.NoMedia : MediaPath.PartialMedia;
      this.logger.info(
-        `Dial:_releaseMedia - successfully released media from freewitch, media path is now ${this._mediaPath}`);
+        `Dial:_releaseMedia - successfully released media from freeswitch, media path is now ${this._mediaPath}`);
    } catch (err) {
      this.logger.info({err}, 'Dial:_releaseMedia error');
    }
@@ -995,7 +1006,7 @@ class TaskDial extends Task {
  async reAnchorMedia(cs, sd) {
    if (cs.ep && sd.ep) return;

-    this.logger.info('Dial:reAnchorMedia - re-anchoring media to freewitch');
+    this.logger.info('Dial:reAnchorMedia - re-anchoring media to freeswitch');
    await Promise.all([sd.reAnchorMedia(this._mediaPath), cs.reAnchorMedia(this._mediaPath)]);
    this.epOther = cs.ep;

@@ -1003,7 +1014,7 @@ class TaskDial extends Task {

    this._mediaPath = MediaPath.FullMedia;
    this.logger.info(
-      `Dial:_releaseMedia - successfully re-anchored media to freewitch, media path is now ${this._mediaPath}`);
+      `Dial:_releaseMedia - successfully re-anchored media to freeswitch, media path is now ${this._mediaPath}`);
  }

  // Handle RE-INVITE hold from caller leg.
@@ -1087,7 +1098,8 @@ class TaskDial extends Task {
              accountInfo: this.cs.accountInfo,
              tasks,
              rootSpan: this.cs.rootSpan,
-              req: this.cs.req
+              req: this.cs.req,
+              tmpFiles: this.cs.tmpFiles,
            });
            await this._onHoldSession.exec();
            this._onHoldSession = null;
--- a/lib/tasks/dub.js
+++ b/lib/tasks/dub.js
@@ -83,7 +83,8 @@ class TaskDub extends TtsTask {
      action: 'playOnTrack',
      track: this.track,
      play: this.play,
-      loop: this.loop ? 'loop' : 'once',
+      // drachtio-fsmrf will convert loop from boolean to 'loop' or 'once'
+      loop: this.loop,
      gain: this.gain
    });
  }
--- a/lib/tasks/enqueue.js
+++ b/lib/tasks/enqueue.js
@@ -370,7 +370,8 @@ class TaskEnqueue extends Task {
        accountInfo: cs.accountInfo,
        tasks: tasksToRun,
        rootSpan: cs.rootSpan,
-        req: cs.req
+        req: cs.req,
+        tmpFiles: cs.tmpFiles,
      });
      await this._playSession.exec();
      this._playSession = null;
--- a/lib/tasks/gather.js
+++ b/lib/tasks/gather.js
@@ -11,7 +11,9 @@ const {
  NvidiaTranscriptionEvents,
  JambonzTranscriptionEvents,
  AssemblyAiTranscriptionEvents,
+  DeepgramRiverTranscriptionEvents,
  VoxistTranscriptionEvents,
+  CartesiaTranscriptionEvents,
  OpenAITranscriptionEvents,
  VadDetection,
  VerbioTranscriptionEvents,
@@ -138,7 +140,11 @@ class TaskGather extends SttTask {
    try {
      await this.handling(cs, obj);
    } catch (error) {
-      if (error instanceof SpeechCredentialError) {
+      if (
+        // avoid bargein task with sticky will restart forever
+        // throw exception to stop the loop.
+        !this.sticky &&
+        error instanceof SpeechCredentialError) {
        this.logger.info('Gather failed due to SpeechCredentialError, finished!');
        this.notifyTaskDone();
        return;
@@ -262,11 +268,16 @@ class TaskGather extends SttTask {
          .catch((err) => {
            process();
          });
-        this.sayTask.on('playDone', (err) => {
-          span.end();
-          if (err) this.logger.error({err}, 'Gather:exec Error playing tts');
+        if (this.sayTask.isStreamingTts && !this.sayTask.closeOnStreamEmpty) {
+          // if streaming tts, we do not wait for it to complete if it is not closing the stream automatically
          process();
-        });
+        } else {
+          this.sayTask.on('playDone', (err) => {
+            span.end();
+            if (err) this.logger.error({err}, 'Gather:exec Error playing tts');
+            process();
+          });
+        }
      }
      else if (this.playTask) {
        const {span, ctx} = this.startChildSpan(`nested:${this.playTask.summary}`);
@@ -341,6 +352,7 @@ class TaskGather extends SttTask {
    this.sayTask?.span.end();
    this._stopVad();
    this._resolve('killed');
+    cs.stopSttLatencyVad();
  }

  updateTaskInProgress(opts) {
@@ -356,6 +368,9 @@ class TaskGather extends SttTask {

  _onDtmf(cs, ep, evt) {
    this.logger.debug(evt, 'TaskGather:_onDtmf');
+    if (!this._timeoutTimer && this.timeout > 0) {
+      this._startTimer();
+    }
    clearTimeout(this.interDigitTimer);
    let resolved = false;
    if (this.dtmfBargein) {
@@ -380,12 +395,7 @@ class TaskGather extends SttTask {
      if (this.digitBuffer.length === 0 && this.needsStt) {
        // DTMF is higher priority than STT.
        this.removeCustomEventListeners();
-        ep.stopTranscription({
-          vendor: this.vendor,
-          bugname: this.bugname,
-        })
-          .catch((err) => this.logger.error({err},
-            ` Received DTMF, Error stopping transcription for vendor ${this.vendor}`));
+        this._stopTranscribing(ep);
      }
      this.digitBuffer += evt.dtmf;
      const len = this.digitBuffer.length;
@@ -458,6 +468,16 @@ class TaskGather extends SttTask {
          this._onVendorConnectFailure.bind(this, cs, ep));
        break;

+      case 'deepgramriver':
+        this.bugname = `${this.bugname_prefix}deepgramriver_transcribe`;
+        this.addCustomEventListener(
+          ep, DeepgramRiverTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, DeepgramRiverTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, DeepgramRiverTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        break;
+
      case 'soniox':
        this.bugname = `${this.bugname_prefix}soniox_transcribe`;
        this.addCustomEventListener(
@@ -546,6 +566,17 @@ class TaskGather extends SttTask {
          this._onVendorConnectFailure.bind(this, cs, ep));
        break;

+      case 'cartesia':
+        this.bugname = `${this.bugname_prefix}cartesia_transcribe`;
+        this.addCustomEventListener(ep, CartesiaTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, CartesiaTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, CartesiaTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, CartesiaTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        break;
+
      case 'speechmatics':
        this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
        this.addCustomEventListener(
@@ -661,10 +692,14 @@ class TaskGather extends SttTask {
        target_sid: this.cs.callSid
      });
    }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
+
+    // Some vendor use single connection, that we cannot use onConnect event to track transcription start
+    this.cs.emit('transcribe-start');
  }

  _startTimer() {
    if (0 === this.timeout) return;
+    this.logger.debug(`Starting timoutTimer of ${this.timeout}ms`);
    this._clearTimer();
    this._timeoutTimer = setTimeout(() => {
      if (this.isContinuousAsr) this._startAsrTimer();
@@ -828,6 +863,10 @@ class TaskGather extends SttTask {
  }

  _onTranscription(cs, ep, evt, fsEvent) {
+    // check if we are in graceful shutdown mode
+    if (ep.gracefulShutdownResolver) {
+      ep.gracefulShutdownResolver();
+    }
    // make sure this is not a transcript from answering machine detection
    const bugname = fsEvent.getHeader('media-bugname');
    const finished = fsEvent.getHeader('transcription-session-finished');
@@ -840,6 +879,10 @@ class TaskGather extends SttTask {
    if (finished === 'true') return;

    if (this.vendor === 'ibm' && evt?.state === 'listening') return;
+
+    // emit an event to the call session to track the time transcription is received
+    cs.emit('on-transcription');
+
    if (this.vendor === 'deepgram' && evt.type === 'UtteranceEnd') {
      /* we will only get this when we have set utterance_end_ms */
      if (this._bufferedTranscripts.length === 0) {
@@ -923,6 +966,12 @@ class TaskGather extends SttTask {
        }
      }

+      // receive a final transcript, calculate the stt latency for this transcript
+      const sttLatency = this.cs.calculateSttLatency();
+      if (!emptyTranscript && sttLatency) {
+        this.stt_latency_ms += `${sttLatency.stt_latency_ms},`;
+      }
+
      if (this.isContinuousAsr) {
        /* append the transcript and start listening again for asrTimeout */
        const t = evt.alternatives[0].transcript;
@@ -1074,11 +1123,7 @@ class TaskGather extends SttTask {

  async _startFallback(cs, ep, evt) {
    if (this.canFallback) {
-      ep.stopTranscription({
-        vendor: this.vendor,
-        bugname: this.bugname
-      })
-        .catch((err) => this.logger.error({err}, `Error stopping transcription for primary vendor ${this.vendor}`));
+      this._stopTranscribing(ep);
      try {
        this.logger.debug('gather:_startFallback');
        this.notifyError({ msg: 'ASR error',
@@ -1207,17 +1252,26 @@ class TaskGather extends SttTask {
    }
  }

+  async _stopTranscribing(ep) {
+    // Fix for https://github.com/jambonz/jambonz-feature-server/issues/1281
+    // We should immediately call stop transcription from gather
+    // so that next gather can start transcription immediately
+    ep.stopTranscription({
+      vendor: this.vendor,
+      bugname: this.bugname,
+      gracefulShutdown: false
+    })
+      .catch((err) => {
+        if (this.resolved) return;
+        this.logger.error({err}, 'Error stopping transcription');
+      });
+    this.cs.emit('transcribe-stop');
+  }
+
  async _resolve(reason, evt) {
    this.logger.info({evt}, `TaskGather:resolve with reason ${reason}`);
    if (this.needsStt && this.ep && this.ep.connected) {
-      this.ep.stopTranscription({
-        vendor: this.vendor,
-        bugname: this.bugname
-      })
-        .catch((err) => {
-          if (this.resolved) return;
-          this.logger.error({err}, 'Error stopping transcription');
-        });
+      this._stopTranscribing(this.ep);
    }
    if (this.resolved) {
      this.logger.debug('TaskGather:_resolve - already resolved');
@@ -1236,11 +1290,28 @@ class TaskGather extends SttTask {
    this._clearAsrTimer();
    this._clearFinalAsrTimer();

+    let sttLatencyMetrics = {};
+    if (this.needsStt) {
+      const sttLatency = this.cs.calculateSttLatency();
+      if (sttLatency) {
+        this.stt_latency_ms = this.stt_latency_ms.endsWith(',') ?
+          this.stt_latency_ms.slice(0, -1) : this.stt_latency_ms;
+        sttLatencyMetrics = {
+          'stt.latency_ms': this.stt_latency_ms,
+          'stt.talkspurts': JSON.stringify(sttLatency.talkspurts),
+          'stt.start_time': sttLatency.stt_start_time,
+          'stt.stop_time': sttLatency.stt_stop_time,
+          'stt.usage': sttLatency.stt_usage,
+        };
+      }
+    }
+
    this.span.setAttributes({
      channel: 1,
      'stt.label': this.label || 'None',
      'stt.resolve': reason,
-      'stt.result': JSON.stringify(evt)
+      'stt.result': JSON.stringify(evt),
+      ...sttLatencyMetrics
    });

    if (this.callSession && this.callSession.callGone) {
@@ -1268,6 +1339,9 @@ class TaskGather extends SttTask {

    let returnedVerbs = false;
    try {
+      const latencies = Object.fromEntries(
+        Object.entries(sttLatencyMetrics).map(([key, value]) => [key.replace('stt.', 'stt_'), value])
+      );
      if (reason.startsWith('dtmf')) {
        if (this.parentTask) this.parentTask.emit('dtmf', evt);
        else {
@@ -1281,7 +1355,7 @@ class TaskGather extends SttTask {
        else {
          this.emit('transcription', evt);
          this.logger.debug('TaskGather:_resolve - invoking performAction');
-          returnedVerbs = await this.performAction({speech: evt, reason: 'speechDetected'});
+          returnedVerbs = await this.performAction({speech: evt, reason: 'speechDetected', ...latencies});
          this.logger.debug({returnedVerbs}, 'TaskGather:_resolve - back from performAction');
        }
      }
@@ -1289,20 +1363,20 @@ class TaskGather extends SttTask {
        if (this.parentTask) this.parentTask.emit('timeout', evt);
        else {
          this.emit('timeout', evt);
-          returnedVerbs = await this.performAction({reason: 'timeout'});
+          returnedVerbs = await this.performAction({reason: 'timeout', ...latencies});
        }
      }
      else if (reason.startsWith('stt-error')) {
        if (this.parentTask) this.parentTask.emit('stt-error', evt);
        else {
          this.emit('stt-error', evt);
-          returnedVerbs = await this.performAction({reason: 'error', details: evt.error});
+          returnedVerbs = await this.performAction({reason: 'error', details: evt.error, ...latencies});
        }
      } else if (reason.startsWith('stt-low-confidence')) {
        if (this.parentTask) this.parentTask.emit('stt-low-confidence', evt);
        else {
          this.emit('stt-low-confidence', evt);
-          returnedVerbs = await this.performAction({speech:evt, reason: 'stt-low-confidence'});
+          returnedVerbs = await this.performAction({speech:evt, reason: 'stt-low-confidence', ...latencies});
        }
      }
    } catch (err) {  /*already logged error*/ }
--- a/lib/tasks/listen.js
+++ b/lib/tasks/listen.js
@@ -1,5 +1,5 @@
 const Task = require('./task');
-const {TaskName, TaskPreconditions, ListenEvents, ListenStatus} = require('../utils/constants');
+const {TaskName, TaskPreconditions, ListenEvents, ListenStatus} = require('../utils/constants.json');
 const makeTask = require('./make_task');
 const moment = require('moment');
 const MAX_PLAY_AUDIO_QUEUE_SIZE = 10;
@@ -72,7 +72,7 @@ class TaskListen extends Task {
    } catch (err) {
      this.logger.info(err, `TaskListen:exec - error ${this.url}`);
    }
-    if (this.transcribeTask) this.transcribeTask.kill();
+    if (this.transcribeTask) this.transcribeTask.kill(cs);
    this._removeListeners(ep);
  }

@@ -103,9 +103,12 @@ class TaskListen extends Task {
    this.notifyTaskDone();
  }

-  async updateListen(status) {
+  async updateListen(status, silence = false) {
    if (!this.killed && this.ep && this.ep.connected) {
-      const args = this._bugname ? [this._bugname] : [];
+      const args = [
+        ...(this._bugname ? [this._bugname] : []),
+        ...(status === ListenStatus.Pause ? ([silence]) : []),
+      ];
      this.logger.info(`TaskListen:updateListen status ${status}`);
      switch (status) {
        case ListenStatus.Pause:
--- a/lib/tasks/llm/llms/ultravox_s2s.js
+++ b/lib/tasks/llm/llms/ultravox_s2s.js
@@ -32,12 +32,14 @@ class TaskLlmUltravox_S2S extends Task {
    this.auth = this.parent.auth;
    this.connectionOptions = this.parent.connectOptions;

-    const {apiKey} = this.auth || {};
+    const {apiKey, agent_id} = this.auth || {};
    if (!apiKey) throw new Error('auth.apiKey is required for Vendor: Ultravox');
    this.apiKey = apiKey;
+    this.agentId = agent_id;
    this.actionHook = this.data.actionHook;
    this.eventHook = this.data.eventHook;
    this.toolHook = this.data.toolHook;
+    this.llmOptions = this.data.llmOptions || {};

    this.results = {
      completionReason: 'normal conversation end'
@@ -105,24 +107,29 @@ class TaskLlmUltravox_S2S extends Task {
      }
      );
      // merge with any existing tools
-      this.data.llmOptions.selectedTools = [
+      this.llmOptions.selectedTools = [
        ...convertedTools,
-        ...(this.data.llmOptions.selectedTools || [])
+        ...(this.llmOptions.selectedTools || [])
      ];
    }

    const payload = {
-      ...this.data.llmOptions,
-      model: this.model,
+      ...this.llmOptions,
+      ...(!this.agentId && {
+        model: this.model,
+      }),
      medium: {
-        ...(this.data.llmOptions.medium || {}),
+        ...(this.llmOptions.medium || {}),
        serverWebSocket: {
          inputSampleRate: 8000,
          outputSampleRate: 8000,
        }
      }
    };
-    const {statusCode, body} = await request('https://api.ultravox.ai/api/calls', {
+    const baseUrl = 'https://api.ultravox.ai';
+    const url = this.agentId ?
+      `${baseUrl}/api/agents/${this.agentId}/calls` : `${baseUrl}/api/calls`;
+    const {statusCode, body} = await request(url, {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
@@ -211,7 +218,7 @@ class TaskLlmUltravox_S2S extends Task {
  async _onServerEvent(_ep, evt) {
    let endConversation = false;
    const type = evt.type;
-    this.logger.debug({evt}, 'TaskLlmUltravox_S2S:_onServerEvent');
+    //this.logger.debug({evt}, 'TaskLlmUltravox_S2S:_onServerEvent');

    /* server errors of some sort */
    if (type === 'error') {
--- a/lib/tasks/make_task.js
+++ b/lib/tasks/make_task.js
@@ -96,6 +96,9 @@ function makeTask(logger, obj, parent) {
    case TaskName.Tag:
      const TaskTag = require('./tag');
      return new TaskTag(logger, data, parent);
+    case TaskName.Alert:
+      const TaskAlert = require('./alert');
+      return new TaskAlert(logger, data, parent);
  }

  // should never reach
--- a/lib/tasks/play.js
+++ b/lib/tasks/play.js
@@ -6,7 +6,9 @@ class TaskPlay extends Task {
    super(logger, opts);
    this.preconditions = TaskPreconditions.Endpoint;

-    this.url = this.data.url;
+    this.url = this.data.url.includes('?')
+      ? this.data.url.split('?')[0] + '?' + this.data.url.split('?')[1].replaceAll('.', '%2E')
+      : this.data.url;
    this.seekOffset = this.data.seekOffset || -1;
    this.timeoutSecs = this.data.timeoutSecs || -1;
    this.loop = this.data.loop || 1;
--- a/lib/tasks/redirect.js
+++ b/lib/tasks/redirect.js
@@ -17,28 +17,33 @@ 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}`);
-      }
-    } else if (cs.application.requestor._isAbsoluteUrl(this.actionHook)) {
-      const baseUrl =  this.cs.application.requestor.baseUrl;
-      const newUrl = URL.parse(this.actionHook);
-      const newBaseUrl = newUrl.protocol + '//' + newUrl.host;
-      if (baseUrl != newBaseUrl) {
+    const isAbsoluteUrl = cs.application?.requestor?._isAbsoluteUrl(this.actionHook);
+
+    if (isAbsoluteUrl) {
+      this.logger.info(`TaskRedirect redirecting to new absolute URL ${this.actionHook}, requires new requestor`);
+
+      if (cs.requestor instanceof WsRequestor) {
        try {
-          this.logger.info(`Task:redirect updating base url to ${newBaseUrl}`);
-          const newRequestor = new HttpRequestor(this.logger, cs.accountSid, {url: this.actionHook},
-            cs.accountInfo.account.webhook_secret);
-          this.cs.requestor.removeAllListeners();
-          this.cs.application.requestor = newRequestor;
+          const requestor = new WsRequestor(this.logger, cs.accountSid, {url: this.actionHook},
+            cs.accountInfo.account.webhook_secret) ;
+          cs.requestor.emit('handover', requestor);
        } catch (err) {
-          this.logger.info(err, `Task:redirect error updating base url to ${this.actionHook}`);
+          this.logger.info(err, `TaskRedirect error redirecting to ${this.actionHook}`);
+        }
+      }
+      else {
+        const baseUrl =  this.cs.application.requestor.baseUrl;
+        const newUrl = URL.parse(this.actionHook);
+        const newBaseUrl = newUrl.protocol + '//' + newUrl.host;
+        if (baseUrl != newBaseUrl) {
+          try {
+            this.logger.info(`Task:redirect updating base url to ${newBaseUrl}`);
+            const newRequestor = new HttpRequestor(this.logger, cs.accountSid, {url: this.actionHook},
+              cs.accountInfo.account.webhook_secret);
+            cs.requestor.emit('handover', newRequestor);
+          } catch (err) {
+            this.logger.info(err, `TaskRedirect error updating base url to ${this.actionHook}`);
+          }
        }
      }
    }
--- a/lib/tasks/say.js
+++ b/lib/tasks/say.js
@@ -247,6 +247,7 @@ class TaskSay extends TtsTask {
        throw new SpeechCredentialError(error.message);
      }
    };
+
    let filepath;
    try {
      filepath = await this._synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label});
@@ -266,62 +267,125 @@ class TaskSay extends TtsTask {
          const isStreaming = filepath[segment].startsWith('say:{');
          if (isStreaming) {
            const arr = /^say:\{.*\}\s*(.*)$/.exec(filepath[segment]);
-            if (arr) this.logger.debug(`Say:exec sending streaming tts request: ${arr[1].substring(0, 64)}..`);
+            if (arr) this.logger.debug(`Say:exec sending streaming tts request ${arr[1].substring(0, 64)}..`);
+            else this.logger.debug(`Say:exec sending ${filepath[segment].substring(0, 64)}`);
          }
-          else this.logger.debug(`Say:exec sending ${filepath[segment].substring(0, 64)}`);
-          ep.once('playback-start', (evt) => {
-            this.logger.debug({evt}, 'Say got playback-start');
-            if (this.otelSpan) {
-              this._addStreamingTtsAttributes(this.otelSpan, evt, vendor);
-              this.otelSpan.end();
-              this.otelSpan = null;
-              if (evt.variable_tts_cache_filename) {
-                cs.trackTmpFile(evt.variable_tts_cache_filename);
-              }
-            }
-          });
-          ep.once('playback-stop', (evt) => {
-            this.logger.debug({evt}, 'Say got playback-stop');
-            this.notifyStatus({event: 'stop-playback'});
-            this.notifiedPlayBackStop = true;
-            const tts_error = evt.variable_tts_error;
-            let response_code = 200;
-            // Check if any property ends with _response_code
-            for (const [key, value] of Object.entries(evt)) {
-              if (key.endsWith('_response_code')) {
-                response_code = parseInt(value, 10) || 200;
-                break;
-              }
-            }

-            if (tts_error) {
-              writeAlerts({
-                account_sid,
-                alert_type: AlertType.TTS_FAILURE,
-                vendor,
-                detail: evt.variable_tts_error,
-                target_sid
-              }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-            }
-            if (!tts_error && response_code < 300 && evt.variable_tts_cache_filename && !this.killed) {
-              const text = parseTextFromSayString(this.text[segment]);
-              addFileToCache(evt.variable_tts_cache_filename, {
-                account_sid,
-                vendor,
-                language,
-                voice,
-                engine,
-                model: this.model || this.model_id,
-                text,
-                instructions: this.instructions
-              }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
-            }
+          const onPlaybackStop = (evt) => {
+            try {
+              this.logger.debug({evt},
+                `Say got playback-stop ${evt.variable_tts_playback_id ? evt.variable_tts_playback_id : ''}`);

-            if (this._playResolve) {
-              (tts_error || response_code >= 300) ? this._playReject(new Error(evt.variable_tts_error)) :
-                this._playResolve();
+              /**
+               * If we got a playback id on both the start and stop events, and they don't match,
+               * then we must have received a playback-stop event for an earlier play request.
+               */
+              const playbackId = this.getPlaybackId(segment);
+              // eslint-disable-next-line max-len
+              const unmatchedResponse = (!!playbackId && !!evt.variable_tts_playback_id) && evt.variable_tts_playback_id !== playbackId;
+              if (unmatchedResponse) {
+                this.logger.info({currentPlaybackId: playbackId, stopPlaybackId: evt.variable_tts_playback_id},
+                  'Say:exec discarding playback-stop for earlier play');
+                ep.once('playback-stop',  this._boundOnPlaybackStop);
+
+                return;
+              }
+
+              this.notifyStatus({event: 'stop-playback'});
+              this.notifiedPlayBackStop = true;
+              const tts_error = evt.variable_tts_error;
+              // some tts vendor may not provide response code, so we default to 200
+              let response_code = 200;
+              // Check if any property ends with _response_code
+              for (const [key, value] of Object.entries(evt)) {
+                if (key.endsWith('_response_code')) {
+                  response_code = parseInt(value, 10);
+                  if (isNaN(response_code)) {
+                    this.logger.info(`Say:exec playback-stop - Invalid response code: ${value}`);
+                    response_code = 0;
+                  }
+                  break;
+                }
+              }
+
+              if (tts_error ||
+                // error response codes indicate failure
+                response_code <= 199 || response_code >= 300) {
+                writeAlerts({
+                  account_sid,
+                  alert_type: AlertType.TTS_FAILURE,
+                  vendor,
+                  detail: evt.variable_tts_error || `TTS playback failed with response code ${response_code}`,
+                  target_sid
+                }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
+              }
+              if (
+                !tts_error &&
+                //2xx response codes indicate success
+                199 < response_code && response_code < 300 &&
+                evt.variable_tts_cache_filename &&
+                !this.killed &&
+                // if tts cache is not disabled, add the file to cache
+                !this.disableTtsCache
+              ) {
+                const text = parseTextFromSayString(this.text[segment]);
+                addFileToCache(evt.variable_tts_cache_filename, {
+                  account_sid,
+                  vendor,
+                  language,
+                  voice,
+                  engine,
+                  model: this.model || this.model_id,
+                  text,
+                  instructions: this.instructions
+                }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
+              }
+
+              if (this._playResolve) {
+                (tts_error ||
+                  // error response codes indicate failure
+                  response_code <= 199 || response_code >= 300
+                ) ?
+                  this._playReject(
+                    new Error(evt.variable_tts_error || `TTS playback failed with response code ${response_code}`)
+                  ) : this._playResolve();
+              }
+            } catch (err) {
+              this.logger.info({err}, 'Error handling playback-stop event');
            }
-          });
+          };
+          this._boundOnPlaybackStop = onPlaybackStop.bind(this);
+
+          const onPlaybackStart = (evt) => {
+            try {
+              const playbackId = this.getPlaybackId(segment);
+              // eslint-disable-next-line max-len
+              const unmatchedResponse = (!!playbackId && !!evt.variable_tts_playback_id) && evt.variable_tts_playback_id !== playbackId;
+              if (unmatchedResponse) {
+                this.logger.info({currentPlaybackId: playbackId, stopPlaybackId: evt.variable_tts_playback_id},
+                  'Say:exec playback-start - unmatched playback_id');
+                ep.once('playback-start',  this._boundOnPlaybackStart);
+                return;
+              }
+              this.logger.debug({evt},
+                `Say got playback-start ${evt.variable_tts_playback_id ? evt.variable_tts_playback_id : ''}`);
+              if (this.otelSpan) {
+                this._addStreamingTtsAttributes(this.otelSpan, evt, vendor);
+                this.otelSpan.end();
+                this.otelSpan = null;
+                if (evt.variable_tts_cache_filename) {
+                  cs.trackTmpFile(evt.variable_tts_cache_filename);
+                }
+              }
+            } catch (err) {
+              this.logger.info({err}, 'Error handling playback-start event');
+            }
+          };
+          this._boundOnPlaybackStart = onPlaybackStart.bind(this);
+
+          ep.once('playback-stop',  this._boundOnPlaybackStop);
+          ep.once('playback-start',  this._boundOnPlaybackStart);
+
          // wait for playback-stop event received to confirm if the playback is successful
          this._playPromise = new Promise((resolve, reject) => {
            this._playResolve = resolve;
@@ -397,6 +461,8 @@ class TaskSay extends TtsTask {
          .replace('playht_', 'playht.')
          .replace('cartesia_', 'cartesia.')
          .replace('rimelabs_', 'rimelabs.')
+          .replace('resemble_', 'resemble.')
+          .replace('inworld_', 'inworld.')
          .replace('verbio_', 'verbio.')
          .replace('elevenlabs_', 'elevenlabs.');
        if (spanMapping[newKey]) newKey = spanMapping[newKey];
@@ -461,6 +527,14 @@ const spanMapping = {
  'rimelabs.name_lookup_time_ms': 'name_lookup_ms',
  'rimelabs.connect_time_ms': 'connect_ms',
  'rimelabs.final_response_time_ms': 'final_response_ms',
+  // Resemble
+  'resemble.connect_time_ms': 'connect_ms',
+  'resemble.final_response_time_ms': 'final_response_ms',
+  // inworld
+  'inworld.name_lookup_time_ms': 'name_lookup_ms',
+  'inworld.connect_time_ms': 'connect_ms',
+  'inworld.final_response_time_ms': 'final_response_ms',
+  'inworld.x_envoy_upstream_service_time': 'upstream_service_time',
  // verbio
  'verbio.name_lookup_time_ms': 'name_lookup_ms',
  'verbio.connect_time_ms': 'connect_ms',
--- a/lib/tasks/sip_refer.js
+++ b/lib/tasks/sip_refer.js
@@ -111,7 +111,12 @@ class TaskSipRefer extends Task {
    /* get IP address of the SBC to use as hostname if needed */
    const {host} = parseUri(dlg.remote.uri);

-    if (!referTo.startsWith('<') && !referTo.startsWith('sip') && !referTo.startsWith('"')) {
+    if (
+      !referTo.startsWith('<') &&
+      !referTo.startsWith('sip:') &&
+      !referTo.startsWith('"') &&
+      !referTo.startsWith('tel:')
+    ) {
      /* they may have only provided a phone number/user */
      referTo = `sip:${referTo}@${host}`;
    }
@@ -124,7 +129,12 @@ class TaskSipRefer extends Task {
    if (!referredByDisplayName) {
      referredByDisplayName = cs.req?.callingName;
    }
-    if (!referredBy.startsWith('<') && !referredBy.startsWith('sip') && !referredBy.startsWith('"')) {
+    if (
+      !referredBy.startsWith('<') &&
+      !referredBy.startsWith('sip:') &&
+      !referredBy.startsWith('"') &&
+      !referredBy.startsWith('tel:')
+    ) {
      /* they may have only provided a phone number/user */
      referredBy = `${referredByDisplayName ? `"${referredByDisplayName}"` : ''}<sip:${referredBy}@${host}>`;
    }
--- a/lib/tasks/stt-task.js
+++ b/lib/tasks/stt-task.js
@@ -4,6 +4,7 @@ const crypto = require('crypto');
 const { TaskPreconditions, CobaltTranscriptionEvents } = require('../utils/constants');
 const { SpeechCredentialError } = require('../utils/error');
 const {JAMBONES_AWS_TRANSCRIBE_USE_GRPC} = require('../config');
+const {TaskName} = require('../utils/constants.json');

 /**
 * "Please insert turns here: {{turns:4}}"
@@ -84,6 +85,9 @@ class SttTask extends Task {
    /*bug name prefix */
    this.bugname_prefix = '';

+    // stt latency calculator
+    this.stt_latency_ms = '';
+
  }

  async exec(cs, {ep, ep2}) {
@@ -91,6 +95,12 @@ class SttTask extends Task {
    this.ep = ep;
    this.ep2 = ep2;

+    // start vad from stt latency calculator
+    if (this.name !== TaskName.Gather ||
+      this.name === TaskName.Gather && this.needsStt) {
+      cs.startSttLatencyVad();
+    }
+
    // use session preferences if we don't have specific verb-level settings.
    if (cs.recognizer) {
      for (const k in cs.recognizer) {
@@ -400,7 +410,7 @@ class SttTask extends Task {
    dgOptions.utteranceEndMs = dgOptions.utteranceEndMs || asrTimeout;
  }

-  _onVendorConnect(_cs, _ep) {
+  _onVendorConnect(cs, _ep) {
    this.logger.debug(`TaskGather:_on${this.vendor}Connect`);
  }

--- a/lib/tasks/transcribe.js
+++ b/lib/tasks/transcribe.js
@@ -6,6 +6,7 @@ const {
  AwsTranscriptionEvents,
  AzureTranscriptionEvents,
  DeepgramTranscriptionEvents,
+  DeepgramRiverTranscriptionEvents,
  SonioxTranscriptionEvents,
  CobaltTranscriptionEvents,
  IbmTranscriptionEvents,
@@ -14,6 +15,7 @@ const {
  TranscribeStatus,
  AssemblyAiTranscriptionEvents,
  VoxistTranscriptionEvents,
+  CartesiaTranscriptionEvents,
  OpenAITranscriptionEvents,
  VerbioTranscriptionEvents,
  SpeechmaticsTranscriptionEvents
@@ -136,22 +138,30 @@ class TaskTranscribe extends SttTask {
      stopTranscription = true;
      this.ep.stopTranscription({
        vendor: this.vendor,
-        bugname: this.bugname
+        bugname: this.bugname,
+        gracefulShutdown: this.paused ? false : true
      })
        .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
    }
    if (this.transcribing2 && this.ep2?.connected) {
      stopTranscription = true;
-      this.ep2.stopTranscription({vendor: this.vendor, bugname: this.bugname})
+      this.ep2.stopTranscription({
+        vendor: this.vendor,
+        bugname: this.bugname,
+        gracefulShutdown: this.paused ? false : true
+      })
        .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
    }

+    this.cs.emit('transcribe-stop');
+
    return stopTranscription;
  }

  async kill(cs) {
    super.kill(cs);
    const stopTranscription = this._stopTranscription();
+    cs.stopSttLatencyVad();
    // hangup after 1 sec if we don't get a final transcription
    if (stopTranscription) this._timer = setTimeout(() => this.notifyTaskDone(), 1500);
    else this.notifyTaskDone();
@@ -232,6 +242,15 @@ class TaskTranscribe extends SttTask {
        //if (opts.DEEPGRAM_SPEECH_UTTERANCE_END_MS) this.isContinuousAsr = true;

        break;
+      case 'deepgramriver':
+        this.bugname = `${this.bugname_prefix}deepgramriver_transcribe`;
+        this.addCustomEventListener(ep, DeepgramRiverTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep, DeepgramRiverTranscriptionEvents.Connect,
+          this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, DeepgramRiverTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
+        break;
      case 'soniox':
        this.bugname = `${this.bugname_prefix}soniox_transcribe`;
        this.addCustomEventListener(ep, SonioxTranscriptionEvents.Transcription,
@@ -312,6 +331,17 @@ class TaskTranscribe extends SttTask {
          this._onVendorConnectFailure.bind(this, cs, ep, channel));
        break;

+      case 'cartesia':
+        this.bugname = `${this.bugname_prefix}cartesia_transcribe`;
+        this.addCustomEventListener(ep, CartesiaTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep,
+          CartesiaTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, CartesiaTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, CartesiaTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
+        break;
+
      case 'speechmatics':
        this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
        this.addCustomEventListener(
@@ -406,9 +436,16 @@ class TaskTranscribe extends SttTask {
      bugname: this.bugname,
      hostport: this.hostport
    });
+
+    // Some vendor use single connection, that we cannot use onConnect event to track transcription start
+    this.cs.emit('transcribe-start');
  }

  async _onTranscription(cs, ep, channel, evt, fsEvent) {
+    // check if we are in graceful shutdown mode
+    if (ep.gracefulShutdownResolver) {
+      ep.gracefulShutdownResolver();
+    }
    // make sure this is not a transcript from answering machine detection
    const bugname = fsEvent.getHeader('media-bugname');
    const finished = fsEvent.getHeader('transcription-session-finished');
@@ -420,6 +457,9 @@ class TaskTranscribe extends SttTask {

    if (this.vendor === 'ibm' && evt?.state === 'listening') return;

+    // emit an event to the call session to track the time transcription is received
+    cs.emit('on-transcription');
+
    if (this.vendor === 'deepgram' && evt.type === 'UtteranceEnd') {
      /* we will only get this when we have set utterance_end_ms */

@@ -581,14 +621,28 @@ class TaskTranscribe extends SttTask {
  }

  async _resolve(channel, evt) {
+    let sttLatencyMetrics = {};
    if (evt.is_final) {
+      const sttLatency = this.cs.calculateSttLatency();
+      if (sttLatency) {
+        sttLatencyMetrics = {
+          'stt.latency_ms': `${sttLatency.stt_latency_ms}`,
+          'stt.talkspurts': JSON.stringify(sttLatency.talkspurts),
+          'stt.start_time': sttLatency.stt_start_time,
+          'stt.stop_time': sttLatency.stt_stop_time,
+          'stt.usage': sttLatency.stt_usage,
+        };
+      }
+      // time to reset the stt latency
+      this.cs.emit('transcribe-start');
      /* we've got a final transcript, so end the otel child span for this channel */
      if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
        this.childSpan[channel - 1].span.setAttributes({
          channel,
          'stt.label': this.label || 'None',
          'stt.resolve': 'transcript',
-          'stt.result': JSON.stringify(evt)
+          'stt.result': JSON.stringify(evt),
+          ...sttLatencyMetrics
        });
        this.childSpan[channel - 1].span.end();
      }
@@ -597,9 +651,13 @@ class TaskTranscribe extends SttTask {
    if (this.transcriptionHook) {
      const b3 = this.getTracingPropagation();
      const httpHeaders = b3 && {b3};
+      const latencies = Object.fromEntries(
+        Object.entries(sttLatencyMetrics).map(([key, value]) => [key.replace('stt.', 'stt_'), value])
+      );
      const payload = {
        ...this.cs.callInfo,
        ...httpHeaders,
+        ...latencies,
        ...(evt.alternatives && {speech: evt}),
        ...(evt.type && {speechEvent: evt})
      };
@@ -691,7 +749,8 @@ class TaskTranscribe extends SttTask {
    if (this.canFallback) {
      _ep.stopTranscription({
        vendor: this.vendor,
-        bugname: this.bugname
+        bugname: this.bugname,
+        gracefulShutdown: false
      })
        .catch((err) => this.logger.error({err}, `Error stopping transcription for primary vendor ${this.vendor}`));
      try {
--- a/lib/tasks/tts-task.js
+++ b/lib/tasks/tts-task.js
@@ -3,6 +3,16 @@ const { TaskPreconditions } = require('../utils/constants');
 const { SpeechCredentialError } = require('../utils/error');
 const dbUtils = require('../utils/db-utils');

+const extractPlaybackId = (str) => {
+  // Match say:{...} and capture the content inside braces
+  const match = str.match(/say:\{([^}]*)\}/);
+  if (!match) return null;
+
+  // Look for playback_id=value within the captured content
+  const playbackMatch = match[1].match(/playback_id=([^,]*)/);
+  return playbackMatch ? playbackMatch[1] : null;
+};
+
 class TtsTask extends Task {

  constructor(logger, data, parentTask) {
@@ -22,6 +32,11 @@ class TtsTask extends Task {
    this.disableTtsCache = this.data.disableTtsCache;
    this.options = this.synthesizer.options || {};
    this.instructions = this.data.instructions;
+    this.playbackIds = [];
+  }
+
+  getPlaybackId(offset) {
+    return this.playbackIds[offset];
  }

  async exec(cs) {
@@ -185,6 +200,9 @@ class TtsTask extends Task {
    } else if (vendor === 'rimelabs') {
      credentials = credentials || {};
      credentials.model_id = this.options.model_id || credentials.model_id;
+    } else if (vendor === 'inworld') {
+      credentials = credentials || {};
+      credentials.model_id = this.options.model_id || credentials.model_id;
    } else if (vendor === 'whisper') {
      credentials = credentials || {};
      credentials.model_id = this.options.model_id || credentials.model_id;
@@ -277,6 +295,7 @@ class TtsTask extends Task {
            renderForCaching: preCache
          });
          if (!filePath.startsWith('say:')) {
+            this.playbackIds.push(null);
            this.logger.debug(`Say: file ${filePath}, served from cache ${servedFromCache}`);
            if (filePath) cs.trackTmpFile(filePath);
            if (this.otelSpan) {
@@ -290,12 +309,24 @@ class TtsTask extends Task {
                vendor,
                language,
                characters: text.length,
-                elapsedTime: rtt
+                elapsedTime: rtt,
+                servedFromCache,
+                'id': this.id
+              });
+            }
+            if (servedFromCache) {
+              this.notifyStatus({
+                event: 'synthesized-audio',
+                vendor,
+                language,
+                servedFromCache,
+                'id': this.id
              });
            }
          }
          else {
-            this.logger.debug('Say: a streaming tts api will be used');
+            this.playbackIds.push(extractPlaybackId(filePath));
+            this.logger.debug({playbackIds: this.playbackIds}, 'Say: a streaming tts api will be used');
            const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${ep.uuid},`);
            return modifiedPath;
          }
--- a/lib/utils/amd-utils.js
+++ b/lib/utils/amd-utils.js
@@ -281,13 +281,17 @@ module.exports = (logger) => {

    /* set stt options */
    logger.info(`starting amd for vendor ${vendor} and language ${language}`);
-    const sttOpts = amd.setChannelVarsForStt({name: TaskName.Gather}, sttCredentials, language, {
-      vendor,
-      hints,
-      enhancedModel: true,
-      altLanguages: opts.recognizer?.altLanguages || [],
-      initialSpeechTimeoutMs: opts.resolveTimeoutMs,
-    });
+    /* if opts contains recognizer object use that config for stt, otherwise use defaults */
+    const rOpts = opts.recognizer ?
+      opts.recognizer :
+      {
+        vendor,
+        hints,
+        enhancedModel: true,
+        altLanguages: opts.recognizer?.altLanguages || [],
+        initialSpeechTimeoutMs: opts.resolveTimeoutMs,
+      };
+    const sttOpts = amd.setChannelVarsForStt({name: TaskName.Gather}, sttCredentials, language, rOpts);

    await ep.set(sttOpts).catch((err) => logger.info(err, 'Error setting channel variables'));

@@ -418,7 +422,11 @@ module.exports = (logger) => {
    }

    if (ep.connected) {
-      ep.stopTranscription({vendor, bugname})
+      ep.stopTranscription({
+        vendor,
+        bugname,
+        gracefulShutdown: false
+      })
        .catch((err) => logger.info(err, 'stopAmd: Error stopping transcription'));
      task.emit('amd', {type: AmdEvents.Stopped});
      ep.execute('avmd_stop').catch((err) => this.logger.info(err, 'Error stopping avmd'));
--- a/lib/utils/background-task-manager.js
+++ b/lib/utils/background-task-manager.js
@@ -65,7 +65,7 @@ class BackgroundTaskManager extends Emitter {
      this.logger.info(`stopping background task: ${type}`);
      task.removeAllListeners();
      task.span.end();
-      task.kill();
+      task.kill(this.cs);
      // Remove task from managed List
      this.tasks.delete(type);
    }
--- a/lib/utils/constants.json
+++ b/lib/utils/constants.json
@@ -1,5 +1,6 @@
 {
  "TaskName": {
+    "Alert": "alert",
    "Answer": "answer",
    "Conference": "conference",
    "Config": "config",
@@ -95,6 +96,11 @@
    "ConnectFailure": "deepgram_transcribe::connect_failed",
    "Connect": "deepgram_transcribe::connect"
  },
+  "DeepgramRiverTranscriptionEvents": {
+    "Transcription": "deepgramriver_transcribe::transcription",
+    "ConnectFailure": "deepgramriver_transcribe::connect_failed",
+    "Connect": "deepgramriver_transcribe::connect"
+  },
  "SonioxTranscriptionEvents": {
    "Transcription": "soniox_transcribe::transcription",
    "Error": "soniox_transcribe::error"
@@ -167,9 +173,18 @@
    "ConnectFailure": "voxist_transcribe::connect_failed",
    "Connect": "voxist_transcribe::connect"
  },
+  "CartesiaTranscriptionEvents": {
+    "Transcription": "cartesia_transcribe::transcription",
+    "Error": "cartesia_transcribe::error",
+    "ConnectFailure": "cartesia_transcribe::connect_failed",
+    "Connect": "cartesia_transcribe::connect"
+  },
  "VadDetection": {
    "Detection": "vad_detect:detection"
  },
+  "SileroVadDetection": {
+    "Detection": "vad_silero:detect"
+  },
  "ListenEvents": {
    "Connect": "mod_audio_fork::connect",
    "ConnectFailure": "mod_audio_fork::connect_failed",
@@ -237,6 +252,7 @@
  "KillReason": {
    "Hangup": "hangup",
    "Replaced": "replaced",
+    "ReferComplete": "refer-complete",
    "MediaTimeout": "media_timeout"
  },
  "HookMsgTypes": [
--- a/lib/utils/db-utils.js
+++ b/lib/utils/db-utils.js
@@ -81,6 +81,10 @@ const speechMapper = (cred) => {
      obj.deepgram_tts_uri = o.deepgram_tts_uri;
      obj.deepgram_stt_use_tls = o.deepgram_stt_use_tls;
    }
+    else if ('deepgramriver' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+    }
    else if ('soniox' === obj.vendor) {
      const o = JSON.parse(decrypt(credential));
      obj.api_key = o.api_key;
@@ -110,6 +114,7 @@ const speechMapper = (cred) => {
      const o = JSON.parse(decrypt(credential));
      obj.api_key = o.api_key;
      obj.model_id = o.model_id;
+      obj.stt_model_id = o.stt_model_id;
      obj.embedding = o.embedding;
      obj.options = o.options;
    }
@@ -119,9 +124,22 @@ const speechMapper = (cred) => {
      obj.model_id = o.model_id;
      obj.options = o.options;
    }
+    else if ('resemble' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.resemble_tts_use_tls = o.resemble_tts_use_tls;
+      obj.resemble_tts_uri = o.resemble_tts_uri;
+    }
+    else if ('inworld' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.model_id = o.model_id;
+      obj.options = o.options;
+    }
    else if ('assemblyai' === obj.vendor) {
      const o = JSON.parse(decrypt(credential));
      obj.api_key = o.api_key;
+      obj.service_version = o.service_version;
    }
    else if ('voxist' === obj.vendor) {
      const o = JSON.parse(decrypt(credential));
--- a/lib/utils/http-requestor.js
+++ b/lib/utils/http-requestor.js
@@ -41,7 +41,7 @@ class HttpRequestor extends BaseRequestor {
  constructor(logger, account_sid, hook, secret) {
    super(logger, account_sid, hook, secret);

-    this.method = hook.method || 'POST';
+    this.method = hook.method?.toUpperCase() || 'POST';
    this.authHeader = basicAuth(hook.username, hook.password);
    this.backoffMs = 500;

@@ -111,7 +111,7 @@ class HttpRequestor extends BaseRequestor {

    const payload = params ? snakeCaseKeys(params, ['customerData', 'sip', 'env_vars', 'args']) : null;
    const url = hook.url || hook;
-    const method = hook.method || 'POST';
+    const method = hook.method?.toUpperCase() || 'POST';
    let buf = '';
    httpHeaders = {
      ...httpHeaders,
@@ -119,7 +119,7 @@ class HttpRequestor extends BaseRequestor {
    };

    assert.ok(url, 'HttpRequestor:request url was not provided');
-    assert.ok, (['GET', 'POST'].includes(method), `HttpRequestor:request method must be 'GET' or 'POST' not ${method}`);
+    assert.ok(['GET', 'POST'].includes(method), `HttpRequestor:request method must be 'GET' or 'POST' not ${method}`);
    const startAt = process.hrtime();

    /* if we have an absolute url, and it is ws then do a websocket connection */
--- a/lib/utils/media-endpoint.js
+++ b/lib/utils/media-endpoint.js
@@ -0,0 +1,115 @@
+const {
+  JAMBONES_USE_FREESWITCH_TIMER_FD,
+  JAMBONES_MEDIA_TIMEOUT_MS,
+  JAMBONES_MEDIA_HOLD_TIMEOUT_MS,
+  JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS,
+} = require('../config');
+const { sleepFor } = require('./helpers');
+
+const createMediaEndpoint = async(srf, logger, {
+  activeMs,
+  drachtioFsmrfOptions = {},
+  onHoldMusic,
+  inbandDtmfEnabled,
+  mediaTimeoutHandler,
+} = {}) => {
+  const { getFreeswitch } = srf.locals;
+  const ms = activeMs || getFreeswitch();
+  if (!ms)
+    throw new Error('no available Freeswitch for creating media endpoint');
+
+  const ep = await ms.createEndpoint(drachtioFsmrfOptions);
+
+  // Configure the endpoint
+  const opts = {
+    ...(onHoldMusic && {holdMusic: `shout://${onHoldMusic.replace(/^https?:\/\//, '')}`}),
+    ...(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) {
+    ep.set(opts);
+  }
+  // inbandDtmfEnabled
+  if (inbandDtmfEnabled) {
+    // https://developer.signalwire.com/freeswitch/FreeSWITCH-Explained/Modules/mod-dptools/6587132/#0-about
+    ep.execute('start_dtmf').catch((err) => {
+      logger.error('Error starting inband DTMF', { error: err });
+    });
+    ep.inbandDtmfEnabled = true;
+  }
+  // Handle Media Timeout
+  if (mediaTimeoutHandler) {
+    ep.once('destroy', (evt) => {
+      mediaTimeoutHandler(evt, ep);
+    });
+  }
+  // Handle graceful shutdown for endpoint if required
+  if (JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS > 0) {
+    const getEpGracefulShutdownPromise = () => {
+      if (!ep.gracefulShutdownPromise) {
+        ep.gracefulShutdownPromise = new Promise((resolve) => {
+          // this resolver will be called when stt task received transcription.
+          ep.gracefulShutdownResolver = () => {
+            resolve();
+            ep.gracefulShutdownPromise = null;
+          };
+        });
+      }
+      return ep.gracefulShutdownPromise;
+    };
+
+    const gracefulShutdownHandler = async() => {
+      // resolve when one of the following happens:
+      // 1. stt task received transcription
+      // 2. JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS passed
+      await Promise.race([
+        getEpGracefulShutdownPromise(),
+        sleepFor(JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS)
+      ]);
+    };
+
+    const origStartTranscription = ep.startTranscription.bind(ep);
+    ep.startTranscription = async(...args) => {
+      try {
+        const result = await origStartTranscription(...args);
+        ep.isTranscribeActive = true;
+        return result;
+      } catch (err) {
+        ep.isTranscribeActive = false;
+        throw err;
+      }
+    };
+
+    const origStopTranscription = ep.stopTranscription.bind(ep);
+    ep.stopTranscription = async(opts = {}, ...args) => {
+      const { gracefulShutdown = true, ...others } = opts;
+      if (ep.isTranscribeActive && gracefulShutdown) {
+        // only wait for graceful shutdown if transcription is active
+        await gracefulShutdownHandler();
+      }
+      try {
+        const result = await origStopTranscription({...others}, ...args);
+        ep.isTranscribeActive = false;
+        return result;
+      } catch (err) {
+        ep.isTranscribeActive = false;
+        throw err;
+      }
+    };
+
+    const origDestroy = ep.destroy.bind(ep);
+    ep.destroy = async() => {
+      if (ep.isTranscribeActive) {
+        await gracefulShutdownHandler();
+      }
+      return await origDestroy();
+    };
+  }
+
+  return ep;
+};
+
+module.exports = {
+  createMediaEndpoint,
+};
--- a/lib/utils/place-outdial.js
+++ b/lib/utils/place-outdial.js
@@ -16,13 +16,7 @@ const crypto = require('crypto');
 const HttpRequestor = require('./http-requestor');
 const WsRequestor = require('./ws-requestor');
 const {makeOpusFirst, removeVideoSdp} = require('./sdp-utils');
-const {
-  JAMBONES_USE_FREESWITCH_TIMER_FD,
-  JAMBONES_MEDIA_TIMEOUT_MS,
-  JAMBONES_MEDIA_HOLD_TIMEOUT_MS,
-  JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS
-} = require('../config');
-const { sleepFor } = require('./helpers');
+const { createMediaEndpoint } = require('./media-endpoint');

 class SingleDialer extends Emitter {
  constructor({logger, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask,
@@ -98,6 +92,7 @@ class SingleDialer extends Emitter {
      };
    }
    this.ms = ms;
+    this.srf = srf;
    let uri, to, inviteSpan;
    try {
      switch (this.target.type) {
@@ -139,8 +134,7 @@ class SingleDialer extends Emitter {
      this.updateCallStatus = srf.locals.dbHelpers.updateCallStatus;
      this.serviceUrl = srf.locals.serviceUrl;

-      this.ep = await ms.createEndpoint();
-      this._configMsEndpoint();
+      this.ep = await this._createMediaEndpoint();
      this.logger.debug(`SingleDialer:exec - created endpoint ${this.ep.uuid}`);

      /**
@@ -352,43 +346,19 @@ class SingleDialer extends Emitter {
    }
  }

-  _configMsEndpoint() {
-    const opts = {
-      ...(this.onHoldMusic && {holdMusic: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`}),
-      ...(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');
-      }
-    }
+  async _handleMediaTimeout(evt, ep) {
+    this.logger.info({evt}, 'SingleDialer:_handleMediaTimeout - media timeout event received');
+    this.dialTask.kill(this.dialTask.cs, 'media-timeout');
+  }

-    const origDestroy = this.ep.destroy.bind(this.ep);
-    this.ep.destroy = async() => {
-      try {
-        if (this.dialTask.transcribeTask && JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS) {
-          // transcribe task is being used, wait for some time before destroy
-          // if final transcription is received but endpoint is already closed,
-          // freeswitch module will not be able to send the transcription
-
-          this.logger.info('SingleDialer:_configMsEndpoint -' +
-            ' Dial with transcribe task, wait for some time before destroy');
-          await sleepFor(JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS);
-        }
-        await origDestroy();
-      } catch (err) {
-        this.logger.error(err, 'SingleDialer:_configMsEndpoint - error destroying endpoint');
-      }
-    };
+  async _createMediaEndpoint(drachtioFsmrfOptions = {}) {
+    return await createMediaEndpoint(this.srf, this.logger, {
+      acactiveMs: this.ms,
+      drachtioFsmrfOptions,
+      onHoldMusic: this.onHoldMusic,
+      inbandDtmfEnabled: this.dialTask?.inbandDtmfEnabled,
+      mediaTimeoutHandler: this._handleMediaTimeout.bind(this),
+    });
  }

  /**
@@ -431,7 +401,8 @@ class SingleDialer extends Emitter {
        accountInfo: this.accountInfo,
        tasks,
        rootSpan: this.rootSpan,
-        req: this.req
+        req: this.req,
+        tmpFiles: cs.tmpFiles,
      });
      await cs.exec();

@@ -528,8 +499,7 @@ class SingleDialer extends Emitter {
    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();
+    this.ep = await this._createMediaEndpoint({remoteSdp: this.dlg.remote.sdp});
    await this.dlg.modify(this.ep.local.sdp, {
      headers: {
        'X-Reason': 'anchor-media'
--- a/lib/utils/process-monitor.js
+++ b/lib/utils/process-monitor.js
--- a/lib/utils/stt-latency-calculator.js
+++ b/lib/utils/stt-latency-calculator.js
@@ -0,0 +1,197 @@
+const { assert } = require('console');
+const Emitter = require('events');
+const {
+  VadDetection,
+  SileroVadDetection
+} = require('../utils/constants.json');
+
+class SttLatencyCalculator extends Emitter {
+  constructor({ logger, cs}) {
+    super();
+    this.logger = logger;
+    this.cs = cs;
+    this.isRunning = false;
+    this.isInTalkSpurt = false;
+    this.start_talking_time = 0;
+    this.talkspurts = [];
+    this.vendor = this.cs.vad?.vendor || 'silero';
+    this.stt_start_time = 0;
+    this.stt_stop_time = 0;
+    this.stt_on_transcription_time = 0;
+  }
+
+  set sttStartTime(time) {
+    this.stt_start_time = time;
+  }
+
+  get sttStartTime() {
+    return this.stt_start_time || 0;
+  }
+
+  set sttStopTime(time) {
+    this.stt_stop_time = time;
+  }
+
+  get sttStopTime() {
+    return this.stt_stop_time || 0;
+  }
+
+  set sttOnTranscriptionTime(time) {
+    this.stt_on_transcription_time = time;
+  }
+
+  get sttOnTranscriptionTime() {
+    return this.stt_on_transcription_time || 0;
+  }
+
+  _onVadDetected(_ep, _evt, fsEvent) {
+    if (fsEvent.getHeader('detected-event') === 'stop_talking') {
+      if (this.isInTalkSpurt) {
+        this.talkspurts.push({
+          start: this.start_talking_time,
+          stop: Date.now()
+        });
+      }
+
+      this.start_talking_time = 0;
+      this.isInTalkSpurt = false;
+    } else if (fsEvent.getHeader('detected-event') === 'start_talking') {
+      this.start_talking_time = Date.now();
+      this.isInTalkSpurt = true;
+    }
+  }
+
+  _startVad() {
+    assert(!this.isRunning, 'Latency calculator is already running');
+    assert(this.cs.ep, 'Callsession has no endpoint to start the latency calculator');
+    const ep = this.cs.ep;
+    if (!ep.sttLatencyVadHandler) {
+      ep.sttLatencyVadHandler = this._onVadDetected.bind(this, ep);
+      if (this.vendor === 'silero') {
+        ep.addCustomEventListener(SileroVadDetection.Detection, ep.sttLatencyVadHandler);
+      } else {
+        ep.addCustomEventListener(VadDetection.Detection, ep.sttLatencyVadHandler);
+      }
+    }
+    this.stop_talking_time = 0;
+    this.start_talking_time = 0;
+    this.vad = {
+      ...(this.cs.vad || {}),
+      strategy: 'continuous',
+      bugname: 'stt-latency-calculator-vad',
+      vendor: this.vendor
+    };
+
+    ep.startVadDetection(this.vad);
+    this.isRunning = true;
+  }
+
+  _stopVad() {
+    if (this.isRunning) {
+      this.logger.warn('Latency calculator is still running, stopping VAD detection');
+      const ep = this.cs.ep;
+      ep.stopVadDetection(this.vad);
+      if (ep.sttLatencyVadHandler) {
+        if (this.vendor === 'silero') {
+          this.ep?.removeCustomEventListener(SileroVadDetection.Detection, ep.sttLatencyVadHandler);
+        } else {
+          this.ep?.removeCustomEventListener(VadDetection.Detection, ep.sttLatencyVadHandler);
+        }
+        ep.sttLatencyVadHandler = null;
+      }
+      this.isRunning = false;
+      this.logger.info('STT Latency Calculator stopped');
+    }
+  }
+
+  start() {
+    if (this.isRunning) {
+      this.logger.warn('Latency calculator is already running');
+      return;
+    }
+    if (!this.cs.ep) {
+      this.logger.error('Callsession has no endpoint to start the latency calculator');
+      return;
+    }
+    this._startVad();
+    this.logger.debug('STT Latency Calculator started');
+  }
+
+  stop() {
+    this._stopVad();
+  }
+
+  toUnixTimestamp(date) {
+    return Math.floor(date / 1000);
+  }
+
+  calculateLatency() {
+    if (!this.isRunning) {
+      this.logger.debug('Latency calculator is not running, cannot calculate latency, returning default values');
+      return null;
+    }
+
+    const stt_stop_time = this.stt_stop_time || Date.now();
+    if (this.isInTalkSpurt) {
+      this.talkspurts.push({
+        start: this.start_talking_time,
+        stop: stt_stop_time
+      });
+      this.isInTalkSpurt = false;
+      this.start_talking_time = 0;
+    }
+    const stt_on_transcription_time = this.stt_on_transcription_time || stt_stop_time;
+    const start_talking_time = this.talkspurts[0]?.start;
+    let lastIdx = this.talkspurts.length - 1;
+    lastIdx = lastIdx < 0 ? 0 : lastIdx;
+    const stop_talking_time = this.talkspurts[lastIdx]?.stop || stt_stop_time;
+
+    return {
+      stt_start_time: this.toUnixTimestamp(this.stt_start_time),
+      stt_stop_time: this.toUnixTimestamp(stt_stop_time),
+      start_talking_time: this.toUnixTimestamp(start_talking_time),
+      stop_talking_time: this.toUnixTimestamp(stop_talking_time),
+      stt_latency: parseFloat((Math.abs(stt_on_transcription_time - stop_talking_time)) / 1000).toFixed(2),
+      stt_latency_ms: Math.abs(stt_on_transcription_time - stop_talking_time),
+      stt_usage: parseFloat((stt_stop_time - this.stt_start_time) / 1000).toFixed(2),
+      talkspurts: this.talkspurts.map((ts) =>
+        ([this.toUnixTimestamp(ts.start || 0), this.toUnixTimestamp(ts.stop || 0)]))
+    };
+  }
+
+  resetTime() {
+    if (!this.isRunning) {
+      return;
+    }
+    this.stt_start_time = Date.now();
+    this.stt_stop_time = 0;
+    this.stt_on_transcription_time = 0;
+    this.clearTalkspurts();
+    this.logger.info('STT Latency Calculator reset');
+  }
+
+  onTranscriptionReceived() {
+    if (!this.isRunning) {
+      return;
+    }
+    this.stt_on_transcription_time = Date.now();
+    this.logger.debug(`CallSession:on-transcription set to ${this.stt_on_transcription_time}`);
+  }
+
+  onTranscribeStop() {
+    if (!this.isRunning) {
+      return;
+    }
+    this.stt_stop_time = Date.now();
+    this.logger.debug(`CallSession:transcribe-stop set to ${this.stt_stop_time}`);
+  }
+
+  clearTalkspurts() {
+    this.talkspurts = [];
+    if (!this.isInTalkSpurt) {
+      this.start_talking_time = 0;
+    }
+  }
+}
+
+module.exports = SttLatencyCalculator;
--- a/lib/utils/transcription-utils.js
+++ b/lib/utils/transcription-utils.js
@@ -110,6 +110,10 @@ const stickyVars = {
  voxist: [
    'VOXIST_API_KEY',
  ],
+  cartesia: [
+    'CARTESIA_API_KEY',
+    'CARTESIA_MODEL_ID'
+  ],
  speechmatics: [
    'SPEECHMATICS_API_KEY',
    'SPEECHMATICS_HOST',
@@ -335,6 +339,25 @@ const normalizeDeepgram = (evt, channel, language, shortUtterance) => {
  };
 };

+const normalizeDeepgramRiver = (evt, channel, language, shortUtterance) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.event === 'EndOfTurn',
+    alternatives: [
+      {
+        confidence: evt.end_of_turn_confidence,
+        transcript: evt.transcript,
+      }
+    ],
+    vendor: {
+      name: 'deepgramriver',
+      evt: copy
+    }
+  };
+};
+
 const normalizeNvidia = (evt, channel, language) => {
  const copy = JSON.parse(JSON.stringify(evt));
  const alternatives = (evt.alternatives || [])
@@ -519,16 +542,27 @@ const normalizeAws = (evt, channel, language) => {

 const normalizeAssemblyAi = (evt, channel, language) => {
  const copy = JSON.parse(JSON.stringify(evt));
+  const alternatives = [];
+  let is_final = false;
+  if (evt.type && evt.type === 'Turn') {
+    // v3 is here
+    alternatives.push({
+      confidence: evt.end_of_turn_confidence,
+      transcript: evt.transcript,
+    });
+    is_final = evt.end_of_turn;
+  } else {
+    alternatives.push({
+      confidence: evt.confidence,
+      transcript: evt.text,
+    });
+    is_final = evt.message_type === 'FinalTranscript';
+  }
  return {
    language_code: language,
    channel_tag: channel,
-    is_final: evt.message_type === 'FinalTranscript',
-    alternatives: [
-      {
-        confidence: evt.confidence,
-        transcript: evt.text,
-      }
-    ],
+    is_final,
+    alternatives,
    vendor: {
      name: 'assemblyai',
      evt: copy
@@ -555,6 +589,25 @@ const normalizeVoxist = (evt, channel, language) => {
  };
 };

+const normalizeCartesia = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.is_final,
+    alternatives: [
+      {
+        confidence: 1.00,
+        transcript: evt.text,
+      }
+    ],
+    vendor: {
+      name: 'cartesia',
+      evt: copy
+    }
+  };
+};
+
 const normalizeSpeechmatics = (evt, channel, language) => {
  const copy = JSON.parse(JSON.stringify(evt));
  const is_final = evt.message === 'AddTranscript';
@@ -616,6 +669,8 @@ module.exports = (logger) => {
    switch (vendor) {
      case 'deepgram':
        return normalizeDeepgram(evt, channel, language, shortUtterance);
+      case 'deepgramriver':
+        return normalizeDeepgramRiver(evt, channel, language, shortUtterance);
      case 'microsoft':
        return normalizeMicrosoft(evt, channel, language, punctuation);
      case 'google':
@@ -636,6 +691,8 @@ module.exports = (logger) => {
        return normalizeAssemblyAi(evt, channel, language, shortUtterance);
      case 'voxist':
        return normalizeVoxist(evt, channel, language);
+      case 'cartesia':
+        return normalizeCartesia(evt, channel, language);
      case 'verbio':
        return normalizeVerbio(evt, channel, language);
      case 'speechmatics':
@@ -728,12 +785,15 @@ module.exports = (logger) => {
          AWS_ACCESS_KEY_ID: sttCredentials.accessKeyId,
          AWS_SECRET_ACCESS_KEY: sttCredentials.secretAccessKey,
          AWS_REGION: sttCredentials.region,
-          AWS_SECURITY_TOKEN: sttCredentials.securityToken
+          AWS_SECURITY_TOKEN: sttCredentials.securityToken,
+          AWS_SESSION_TOKEN: sttCredentials.sessionToken ? sttCredentials.sessionToken : sttCredentials.securityToken
        }),
        ...(awsOptions.accessKey && {AWS_ACCESS_KEY_ID: awsOptions.accessKey}),
        ...(awsOptions.secretKey && {AWS_SECRET_ACCESS_KEY: awsOptions.secretKey}),
        ...(awsOptions.region && {AWS_REGION: awsOptions.region}),
        ...(awsOptions.securityToken && {AWS_SECURITY_TOKEN: awsOptions.securityToken}),
+        ...(awsOptions.sessionToken && {AWS_SESSION_TOKEN: awsOptions.sessionToken ?
+          awsOptions.sessionToken : awsOptions.securityToken}),
        ...(awsOptions.languageModelName && {AWS_LANGUAGE_MODEL_NAME: awsOptions.languageModelName}),
        ...(awsOptions.piiEntityTypes?.length && {AWS_PII_ENTITY_TYPES: awsOptions.piiEntityTypes.join(',')}),
        ...(awsOptions.piiIdentifyEntities && {AWS_PII_IDENTIFY_ENTITIES: true}),
@@ -834,6 +894,14 @@ module.exports = (logger) => {
      const deepgramUri = deepgramOptions.deepgramSttUri || sttCredentials.deepgram_stt_uri;
      const useTls = deepgramOptions.deepgramSttUseTls || sttCredentials.deepgram_stt_use_tls;

+      // DH (2025-08-11) entity_prompt is currently limited to 100 words
+      const entityPrompt = deepgramOptions.entityPrompt ?
+        deepgramOptions.entityPrompt
+          .split(/\s+/)
+          .slice(0, 100)
+          .join(' ')
+        : undefined;
+
      /* default to a sensible model if not supplied */
      if (!model) {
        model = selectDefaultDeepgramModel(task, language);
@@ -892,7 +960,24 @@ module.exports = (logger) => {
        ...(deepgramOptions.fillerWords) &&
          {DEEPGRAM_SPEECH_FILLER_WORDS: deepgramOptions.fillerWords},
        ...((Array.isArray(deepgramOptions.keyterms) && deepgramOptions.keyterms.length > 0) &&
-          {DEEPGRAM_SPEECH_KEYTERMS: deepgramOptions.keyterms.join(',')})
+          {DEEPGRAM_SPEECH_KEYTERMS: deepgramOptions.keyterms.join(',')}),
+        ...(deepgramOptions.mipOptOut && {DEEPGRAM_SPEECH_MIP_OPT_OUT: deepgramOptions.mipOptOut}),
+        ...(entityPrompt && {DEEPGRAM_SPEECH_ENTITY_PROMPT: entityPrompt}),
+      };
+    }
+    else if ('deepgramriver' === vendor) {
+      const {
+        preflightThreshold,
+        eotThreshold,
+        eotTimeoutMs,
+        mipOptOut
+      } = rOpts.deepgramOptions || {};
+      opts = {
+        DEEPGRAM_API_KEY: sttCredentials.api_key,
+        ...(preflightThreshold && {DEEPGRAM_SPEECH_PRELIGHT_THRESHOLD: preflightThreshold}),
+        ...(eotThreshold && {DEEPGRAM_SPEECH_EOT_THRESHOLD: eotThreshold}),
+        ...(eotTimeoutMs && {DEEPGRAM_SPEECH_EOT_TIMEOUT_MS: eotTimeoutMs}),
+        ...(mipOptOut && {DEEPGRAM_SPEECH_MIP_OPT_OUT: mipOptOut}),
      };
    }
    else if ('soniox' === vendor) {
@@ -993,8 +1078,28 @@ module.exports = (logger) => {
      };
    }
    else if ('assemblyai' === vendor) {
+      const serviceVersion = rOpts.assemblyAiOptions?.serviceVersion || sttCredentials.service_version || 'v2';
+      const {
+        formatTurns,
+        endOfTurnConfidenceThreshold,
+        minEndOfTurnSilenceWhenConfident,
+        maxTurnSilence
+      } = rOpts.assemblyAiOptions || {};
      opts = {
        ...opts,
+        ASSEMBLYAI_API_VERSION: serviceVersion,
+        ...(serviceVersion === 'v3' && {
+          ...(formatTurns && {
+            ASSEMBLYAI_FORMAT_TURNS: formatTurns
+          }),
+          ...(endOfTurnConfidenceThreshold && {
+            ASSEMBLYAI_END_OF_TURN_CONFIDENCE_THRESHOLD: endOfTurnConfidenceThreshold
+          }),
+          ASSEMBLYAI_MIN_END_OF_TURN_SILENCE_WHEN_CONFIDENT: minEndOfTurnSilenceWhenConfident || 500,
+          ...(maxTurnSilence && {
+            ASSEMBLYAI_MAX_TURN_SILENCE: maxTurnSilence
+          }),
+        }),
        ...(sttCredentials.api_key) &&
          {ASSEMBLYAI_API_KEY: sttCredentials.api_key},
        ...(rOpts.hints?.length > 0 &&
@@ -1008,6 +1113,16 @@ module.exports = (logger) => {
          {VOXIST_API_KEY: sttCredentials.api_key},
      };
    }
+    else if ('cartesia' === vendor) {
+      opts = {
+        ...opts,
+        ...(sttCredentials.api_key &&
+          {CARTESIA_API_KEY: sttCredentials.api_key}),
+        ...(sttCredentials.stt_model_id && {
+          CARTESIA_MODEL_ID: sttCredentials.stt_model_id
+        })
+      };
+    }
    else if ('openai' === vendor) {
      const {openaiOptions = {}} = rOpts;
      const model = openaiOptions.model || rOpts.model || sttCredentials.model_id || 'whisper-1';
--- a/lib/utils/ws-requestor.js
+++ b/lib/utils/ws-requestor.js
@@ -293,7 +293,7 @@ class WsRequestor extends BaseRequestor {

      /* send the message */
      this.ws.send(JSON.stringify(obj), async() => {
-        this.logger.debug({obj}, `WsRequestor:request websocket: sent (${url})`);
+        if (obj.type !== 'llm:event') this.logger.debug({obj}, `WsRequestor:request websocket: sent (${url})`);
        // If session:reconnect is waiting for ack, hold here until ack to send queuedMsgs
        if (this._reconnectPromise) {
          try {
--- a/package-lock.json
+++ b/package-lock.json
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
  "name": "jambonz-feature-server",
-  "version": "0.9.4",
+  "version": "0.9.5",
  "main": "app.js",
  "engines": {
    "node": ">= 18.x"
@@ -27,14 +27,14 @@
  "dependencies": {
    "@aws-sdk/client-auto-scaling": "^3.549.0",
    "@aws-sdk/client-sns": "^3.549.0",
-    "@jambonz/db-helpers": "^0.9.12",
+    "@jambonz/db-helpers": "^0.9.16",
    "@jambonz/http-health-check": "^0.0.1",
    "@jambonz/mw-registrar": "^0.2.7",
-    "@jambonz/realtimedb-helpers": "^0.8.13",
-    "@jambonz/speech-utils": "^0.2.12",
+    "@jambonz/realtimedb-helpers": "^0.8.15",
+    "@jambonz/speech-utils": "^0.2.22",
    "@jambonz/stats-collector": "^0.1.10",
-    "@jambonz/time-series": "^0.2.13",
-    "@jambonz/verb-specifications": "^0.0.105",
+    "@jambonz/time-series": "^0.2.14",
+    "@jambonz/verb-specifications": "^0.0.113",
    "@modelcontextprotocol/sdk": "^1.9.0",
    "@opentelemetry/api": "^1.8.0",
    "@opentelemetry/exporter-jaeger": "^1.23.0",
@@ -48,7 +48,7 @@
    "bent": "^7.3.12",
    "debug": "^4.3.4",
    "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^4.0.3",
+    "drachtio-fsmrf": "^4.1.2",
    "drachtio-srf": "^5.0.5",
    "express": "^4.19.2",
    "express-validator": "^7.0.1",
--- a/test/index.js
+++ b/test/index.js
@@ -16,6 +16,7 @@ require('./sip-request-tests');
 require('./create-call-test');
 require('./play-tests');
 require('./sip-refer-tests');
+require('./sip-refer-handler-tests');
 require('./listen-tests');
 require('./config-test');
 require('./queue-test');
--- a/test/scenarios/uas-dial-refer.xml
+++ b/test/scenarios/uas-dial-refer.xml
@@ -0,0 +1,117 @@
+<?xml version="1.0" encoding="ISO-8859-1" ?>
+<!DOCTYPE scenario SYSTEM "sipp.dtd">
+
+<scenario name="UAS that accepts call and sends REFER">
+  <!-- Receive incoming INVITE -->
+  <recv request="INVITE" crlf="true">
+    <action>
+      <ereg regexp=".*" search_in="hdr" header="Subject:" assign_to="1" />
+      <ereg regexp=".*" search_in="hdr" header="From:" assign_to="2" />
+    </action>
+  </recv>
+
+  <!-- Send 180 Ringing -->
+  <send>
+    <![CDATA[
+      SIP/2.0 180 Ringing
+      [last_Via:]
+      [last_From:]
+      [last_To:];tag=[pid]SIPpTag01[call_number]
+      [last_Call-ID:]
+      [last_CSeq:]
+      [last_Record-Route:]
+      Subject:[$1]
+      Content-Length: 0
+    ]]>
+  </send>
+
+  <!-- Send 200 OK with SDP -->
+  <send>
+    <![CDATA[
+      SIP/2.0 200 OK
+      [last_Via:]
+      [last_From:]
+      [last_To:];tag=[pid]SIPpTag01[call_number]
+      [last_Call-ID:]
+      [last_CSeq:]
+      [last_Record-Route:]
+      Subject:[$1]
+      Contact: <sip:[local_ip]:[local_port];transport=[transport]>
+      Content-Type: application/sdp
+      Content-Length: [len]
+
+      v=0
+      o=user1 53655765 2353687637 IN IP[local_ip_type] [local_ip]
+      s=-
+      c=IN IP[media_ip_type] [media_ip]
+      t=0 0
+      m=audio [media_port] RTP/AVP 0
+      a=rtpmap:0 PCMU/8000
+    ]]>
+  </send>
+
+<recv request="ACK" rtd="true" crlf="true">
+  <action>
+    <!-- Check if this is NOT the first call (tag ends with 012 or higher) -->
+    <ereg regexp="tag=1SIPpTag01[2-9]" search_in="hdr" header="To:" assign_to="3" />
+    <log message="Not first call check result: [$3]"/>
+  </action>
+</recv>
+
+<!-- Skip REFER if we found a non-first call tag -->
+<nop next="skip_refer" test="3" value="" compare="not_equal">
+  <action>
+    <log message="Found non-first call tag [$3], skipping REFER"/>
+  </action>
+</nop>
+
+<!-- Wait a moment, then send REFER (only on first call) -->
+<pause milliseconds="1000"/>
+
+<nop>
+  <action>
+    <log message="Sending REFER for first call"/>
+  </action>
+</nop>
+
+<!-- Send REFER (only on first iteration) -->
+  <send retrans="500">
+    <![CDATA[
+      REFER sip:service@[remote_ip]:[remote_port] SIP/2.0
+      Via: SIP/2.0/[transport] [local_ip]:[local_port];branch=[branch]
+      From: <sip:[local_ip]:[local_port]>;tag=[pid]SIPpTag01[call_number]
+      To: [$2]
+      [last_Call-ID:]
+      CSeq: 2 REFER
+      Contact: <sip:[local_ip]:[local_port];transport=[transport]>
+      Max-Forwards: 70
+      X-Call-Number: [call_number]
+      Refer-To: <sip:+15551234567@example.com>
+      Referred-By: <sip:[local_ip]:[local_port]>
+      Content-Length: 0
+    ]]>
+  </send>
+
+  <!-- Expect 202 Accepted (only on first iteration) -->
+  <recv response="202"/>
+  
+  <label id="skip_refer"/>
+
+  <!-- Wait for BYE from feature server -->
+  <recv request="BYE"/>
+
+  <!-- Send 200 OK to BYE -->
+  <send>
+    <![CDATA[
+      SIP/2.0 200 OK
+      [last_Via:]
+      [last_From:]
+      [last_To:]
+      [last_Call-ID:]
+      [last_CSeq:]
+      Contact: <sip:[local_ip]:[local_port];transport=[transport]>
+      Content-Length: 0
+    ]]>
+  </send>
+
+</scenario>
--- a/test/sip-refer-handler-tests.js
+++ b/test/sip-refer-handler-tests.js
@@ -0,0 +1,90 @@
+const test = require('tape');
+const { sippUac } = require('./sipp')('test_fs');
+const bent = require('bent');
+const getJSON = bent('json')
+const clearModule = require('clear-module');
+const {provisionCallHook} = require('./utils');
+const { sleepFor } = require('../lib/utils/helpers');
+
+process.on('unhandledRejection', (reason, p) => {
+  console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
+});
+  
+function connect(connectable) {
+  return new Promise((resolve, reject) => {
+    connectable.on('connect', () => {
+      return resolve();
+    });
+  });
+}
+
+test('when parent leg recvs REFER it should end the dial after adulting child leg', async(t) => {
+    clearModule.all();
+    const {srf, disconnect} = require('../app');
+    try {
+      await connect(srf);
+      // wait for fs connected to drachtio server.
+      await sleepFor(1000);
+      
+      // GIVEN
+      const from = "dial_refer_handler";
+      let verbs = [
+        {
+          "verb": "dial",
+          "callerId": from,
+          "actionHook": "/actionHook",
+          "referHook": "/referHook",
+          "anchorMedia": true,
+          "target": [
+              {
+                  "type": "phone",
+                  "number": "15083084809"
+              }
+          ]
+        }
+      ];
+  
+      await provisionCallHook(from, verbs);
+  
+      // THEN
+      //const p = sippUac('uas-dial.xml', '172.38.0.10', undefined, undefined, 2);
+      const p = sippUac('uas-dial-refer.xml', '172.38.0.10', undefined, undefined, 2);
+      await sleepFor(1000);
+  
+      let account_sid = '622f62e4-303a-49f2-bbe0-eb1e1714e37a';
+  
+      let 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",
+        },
+        "from": from,
+        "to": {
+          "type": "phone",
+          "number": "15583084808"
+      }});
+  
+      await p;
+  
+      // Verify that the referHook was called
+      const obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_referHook`);
+      t.ok(obj.body.from === from,
+        'dial-refer-handler: referHook was called with correct from');
+      t.ok(obj.body.refer_details && obj.body.refer_details.sip_refer_to,
+        'dial-refer-handler: refer_details included in referHook');
+      t.ok(obj.body.refer_details.refer_to_user === '+15551234567',
+        'dial-refer-handler: refer_to_user correctly parsed');
+      t.ok(obj.body.refer_details.referring_call_sid,
+        'dial-refer-handler: referring_call_sid included');
+      t.ok(obj.body.refer_details.referred_call_sid,
+        'dial-refer-handler: referred_call_sid included');
+  
+      disconnect();
+    } catch (err) {
+      console.log(`error received: ${err}`);
+      disconnect();
+      t.error(err);
+    }
+});
--- a/test/sip-refer-tests.js
+++ b/test/sip-refer-tests.js
@@ -58,6 +58,46 @@ test('\'refer\' tests w/202 and NOTIFY', {timeout: 25000}, async(t) => {
  }
 });

+test('\'refer\' tests tel:', {timeout: 25000}, async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const verbs = [
+      {
+        verb: 'say',
+        text: 'silence_stream://100'
+      },
+      {
+        verb: 'sip:refer',
+        referTo: 'tel:+1234567890',
+        actionHook: '/actionHook'
+      }
+    ];
+    const noVerbs = [];
+
+    const from = 'refer_with_tel';
+    await provisionCallHook(from, verbs);
+    await provisionActionHook(from, noVerbs)
+
+    // THEN
+    await sippUac('uac-refer-with-notify.xml', '172.38.0.10', from);
+    t.pass('refer: successfully received 202 Accepted');
+    await sleepFor(1000);
+    const obj  = await getJSON(`http:127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    t.ok(obj.body.final_referred_call_status === 200, 'refer: successfully received NOTIFY with 200 OK');
+    // console.log(`obj: ${JSON.stringify(obj)}`);
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
 test('\'refer\' tests w/202 but no NOTIFY', {timeout: 25000}, async(t) => {
  clearModule.all();
  const {srf, disconnect} = require('../app');
--- a/test/webhook/app.js
+++ b/test/webhook/app.js
@@ -99,6 +99,24 @@ app.post('/actionHook', (req, res) => {
  return res.sendStatus(200);
 });

+/*
+ * referHook
+ */
+app.post('/referHook', (req, res) => {
+  console.log({payload: req.body}, 'POST /referHook');
+  let key = req.body.from + "_referHook"
+  addRequestToMap(key, req, hook_mapping);
+  return res.json([{"verb": "pause", "length": 2}]);
+});
+
+/*
+ * adultingHook
+ */
+app.post('/adulting', (req, res) => {
+  console.log({payload: req.body}, 'POST /adulting');
+  return res.sendStatus(200);
+});
+
 /*
 * customHook
 * For the hook to return
Author	SHA1	Message	Date
Dave Horton	a4726cbf1d	add startup logging	2025-08-21 13:15:36 -04:00
Dave Horton	2800c4cc4c	turn down some logging	2025-08-21 13:00:23 -04:00
Dave Horton	b4b096d07c	wip	2025-08-20 18:16:36 -04:00
Dave Horton	10b5ceeef1	wip	2025-08-20 15:15:56 -04:00
Dave Horton	ea954eca0b	bump speech utils	2025-08-20 11:02:03 -04:00
Dave Horton	d93da88bff	modify tts and say task to track current playback id and match against start and stop events	2025-08-20 10:29:11 -04:00
Dave Horton	10fb6b1e67	update to speech-utils that generates playback id	2025-08-20 10:13:47 -04:00
Sam Machin	654ccd9d9d	start timeout on bargein digits (#1312 )	2025-08-19 08:34:33 -04:00
Dave Horton	ea27b20ac5	update speech-utils (#1324 )	2025-08-17 10:09:55 -04:00
Hoan Luu Huu	96aa705378	update speech util verbsion (#1323 )	2025-08-14 08:39:12 -04:00
rammohan-y	5e51849839	Sending synthesized-audio notification for servedFromCache as false (#1320 ) * Sending synthesized-audio notification for servedFromCache as well https://github.com/jambonz/jambonz-feature-server/issues/1319 * Sending back the id that was set, to track the synthesized-audio e.g if we send a say verb having 100, it's synthesized-audio event will return 100 in the data to correleate the say verb and synthesized-audio event	2025-08-13 20:56:56 -04:00
Hoan Luu Huu	44f69fa76d	Support resemble tts (#1322 ) * support resemble tts * update speech utils version	2025-08-13 08:15:29 -04:00
Dave Horton	73c77bea71	add support for deepgram entity_prompt	2025-08-11 20:58:14 -04:00
rammohan-y	babc0d0dbb	Fix for issue https://github.com/jambonz/jambonz-feature-server/issues/1317 . (#1318 ) Unable to use mod_aws_transcribe module due to security error as sessionId is not populated	2025-08-11 09:17:32 -04:00
Hoan Luu Huu	6b8d0fe1a8	update db-helper 0.9.16 (#1316 )	2025-08-08 10:43:59 -04:00
Dave Horton	66bb466297	fix bug where task.kill is not passed cs (#1315 )	2025-08-07 16:01:45 -04:00
Dave Horton	1933f4ec0b	Feat/freeswitch logging (#1309 ) * include callSid on INVITEs to freeeswitch * remove unnecessary warning	2025-08-04 09:19:47 -04:00
Sam Machin	b1089a1ae9	pass recogniser opts in amd to stt (#1308 )	2025-08-01 22:26:51 -04:00
sathish kumar pasham	93e06d887e	Fix security vulnerabilities by upgrading @jambonz/realtimedb-helpers (#1305 )	2025-07-30 13:24:19 -04:00
Sam Machin	b478e0ecd2	fix assert, and force methods to upper case (#1304 ) * fix assert, and force methods to upper case * add alert for updateCall errors * lint * handle missing method	2025-07-30 08:32:15 -04:00
Sam Machin	94d43d4b70	use tmpFiles list of parent call-session (#1301 ) fixes #1299	2025-07-29 22:08:00 -04:00
Hoan Luu Huu	eb449e9169	support deepgram river (#1273 ) * support deepgram river * wip * rebase * fix review comment --------- Co-authored-by: Dave Horton <daveh@beachdognet.com>	2025-07-29 13:49:43 -04:00
Hoan Luu Huu	158d9d7d25	support stt latency metrics (#1252 ) * support stt latency metrics * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * wip * enable stt latency calculator by config verb * wip * wip * wip * fix jslint * fixed gather timeout does not have latency calculation * upadte verb specification to use notifySttLatency * move stt latency metric from call session to stt-latency calculator * wip	2025-07-29 09:56:37 -04:00
Hoan Luu Huu	5886d1d945	allow pause/resume background listen with silence/blank (#1300 ) * allow pause/resume background listen with silence/blank * wip * wip * wip * wip * update drachtio-fsmrf version	2025-07-28 07:58:30 -04:00
Hoan Luu Huu	352106ec0c	support referTo with tel: prefix (#1297 ) * support referTo with tel: prefix * fix review comment	2025-07-23 10:52:10 -04:00
Sam Machin	05a6bf51a7	corrected typos (#1295 ) 🧙‍♀️	2025-07-21 07:22:51 -04:00
Sam Machin	bd1c763e72	urlencode the . in a url querystring on play (#1293 ) * urlencode the . in a url querystring on play * lint * fix for non querystring urls too * lint	2025-07-20 14:03:21 -04:00
Sam Machin	d831a4ca7f	don't fetch if whisper is an object with a single verb in it (#1290 ) * don't fetch if whisper is an object with a single verb in it * disable URL fetching of verbs on whisper * fixes for lint * lint	2025-07-20 13:47:31 -04:00
Sam Machin	0cc6ea987f	Fix/1277 (#1278 ) * wip * forwardPAI control in verb * update deps * package-loxck * Update package-lock.json	2025-07-20 13:26:33 -04:00
rammohan-y	6e7521c91c	Fix for issue where we should not delay from gather when JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS is set (#1283 ) * Fix for issue https://github.com/jambonz/jambonz-feature-server/issues/1281 * Commented the reason for the change	2025-07-17 10:36:22 -04:00
Hoan Luu Huu	e0e2ade289	fixed gather cannot timeout if listenDuringPrompt is true (#1276 ) * fixed gather cannot timeout if listenDuringPrompt is true * wip * wip * wip	2025-07-15 22:53:22 -04:00
Dave Horton	ee895b4046	bump version	2025-07-15 11:40:14 -04:00
Dave Horton	2a42ccb0e1	fix for #1285 (#1286 ) * fix for #1285 * typo	2025-07-15 11:38:48 -04:00
Dave Horton	62b6a814b7	fixes for LCC dial where a relative url is given as actionhook (#1282 ) * fixes for LCC dial where a relative url is given as actionhook * update to speech-utils 0.2.15 with configurable tmp folder location	2025-07-13 11:08:15 -04:00
Dave Horton	e415420150	Fix/cached audio race (#1279 ) * fix for race condition when play-stop event from earlier command received * wip * say verb should not cache if disableTtsCache = true * logging * modify race condition logic to validate playback id in playback-stopped matches that from playback-start * logging --------- Co-authored-by: Quan HL <quan.luuhoang8@gmail.com>	2025-07-13 10:56:41 -04:00
Sam Machin	e6e039e0f2	add alert verb (#1270 ) * add alert verb * update dependencies * Update package-lock.json * remove await taskDone	2025-07-10 07:39:42 -04:00
Dave Horton	657e2d4a49	update speech-utils (#1275 )	2025-07-09 16:10:08 -04:00
Hoan Luu Huu	337c1cded0	fixed transcription is not received when call is terminated (#1259 ) * fixed transcription is not received when call is terminated * wip * fixed failing testcases * wip * wip * wip * wip * should not do gracefulshutdown on stopAmd	2025-07-09 10:20:09 -04:00
Dave Horton	444abcb036	(fd 1056) if no payload returned from referHook in dial then dont do anything (#1271 )	2025-07-07 12:50:46 -04:00
Hoan Luu Huu	c82a835e70	FD_1079: tts modules response_code = 0 should make say fail to exec. (#1269 ) * FD_1079: tts modules response_code = 0 should make say fail to exec. * fixed tts azure does not have response code * fixed tts error does not raise alarm * wip * fixed	2025-07-07 07:59:29 -04:00
Hoan Luu Huu	3c185d4bd2	consistent assemblyAiOptions name (#1267 ) * consistent assemblyAiOptions name * update verb specification version	2025-07-03 08:05:56 -04:00
Hoan Luu Huu	ba2049b705	support assemblyai v3 (#1265 ) * support assemblyai v3 * wip * wip * wip * wip * wip * wip	2025-07-01 15:46:19 -04:00
Dave Horton	7691af30de	Fix/dial refer (#1264 ) * Revert "Update dial.js (#1243)" This reverts commit `259dedcded`. * add to .gitignore * when we receive a REFER on the parent leg, after adulting the child the dial task in the parent session should end	2025-06-28 15:01:09 -04:00
Hoan Luu Huu	ab83b21979	support inworld tts (#1262 ) * support inworld tts * wip * wip	2025-06-27 10:05:18 -04:00
Hoan Luu Huu	f18b62e165	support ultravox agent id (#1254 ) * support ultravox agent id * wip	2025-06-23 09:37:10 -04:00
Dave Horton	f98bf2a1f8	update to drachtio-fsmrf@4.0.4 (#1255 )	2025-06-22 15:18:30 +02:00
Hoan Luu Huu	8c67c05d87	fixed bargin task loop forever (#1253 )	2025-06-22 06:20:24 +02:00
Hoan Luu Huu	3f11ee58a7	fixed dub playOnTrack loop forever (#1247 )	2025-06-18 21:37:32 +02:00
rammohan-y	c8d94026ff	Removing video sdp when making an outbound call (#1242 )	2025-06-18 21:31:00 +02:00
Hoan Luu Huu	5be6c54339	support mod_cartesia_transcribe (#1245 )	2025-06-17 20:54:26 +02:00
Sam Machin	259dedcded	Update dial.js (#1243 )	2025-06-13 18:23:40 +02:00