bump speech-utils

* add ws msgid to telemetry span if it's ws requestor

* wip

.github/workflows/build.yml

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

.gitignore

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

Dockerfile

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

README.md

@@ -13,7 +13,7 @@ Configuration is provided via environment variables:
 |AWS_ACCESS_KEY_ID| aws access key id, used for TTS/STT as well SNS notifications|no|
 |AWS_REGION| aws region| no|
 |AWS_SECRET_ACCESS_KEY| aws secret access key, used per above|no|
-|AWS_SNS_TOPIC_ARM| aws sns topic arn that scale-in lifecycle notifications will be published to|no|
+|AWS_SNS_TOPIC_ARN| aws sns topic arn that scale-in lifecycle notifications will be published to|no|
 |DRACHTIO_HOST| ip address of drachtio server (typically '127.0.0.1')|yes|
 |DRACHTIO_PORT| listening port of drachtio server for control connections (typically 9022)|yes|
 |DRACHTIO_SECRET| shared secret|yes|
@@ -21,6 +21,7 @@ Configuration is provided via environment variables:
 |ENCRYPTION_SECRET| secret for credential encryption(JWT_SECRET is deprecated) |yes|
 |GOOGLE_APPLICATION_CREDENTIALS| path to gcp service key file|yes|
 |HTTP_PORT| tcp port to listen on for API requests from jambonz-api-server|yes|
+|HTTP_IP| IP Address for API requests from jambonz-api-server |no|
 |JAMBONES_GATHER_EARLY_HINTS_MATCH| if true and hints are provided, gather will opportunistically review interim transcripts if possible to reduce ASR latency |no|
 |JAMBONES_FREESWITCH| IP:port:secret for Freeswitch server (e.g. '127.0.0.1:8021:JambonzR0ck$'|yes|
 |JAMBONES_LOGLEVEL| log level for application, 'info' or 'debug'|no|
@@ -71,7 +72,7 @@ module.exports = {
       STATS_PORT: 8125,
       STATS_PROTOCOL: 'tcp',
       STATS_TELEGRAF: 1,
-      AWS_SNS_TOPIC_ARM: 'arn:aws:sns:us-west-1:xxxxxxxxxxx:terraform-20201107200347128600000002',
+      AWS_SNS_TOPIC_ARN: 'arn:aws:sns:us-west-1:xxxxxxxxxxx:terraform-20201107200347128600000002',
       JAMBONES_NETWORK_CIDR: '172.31.0.0/16',
       JAMBONES_MYSQL_HOST: 'aurora-cluster-jambonz.cluster-yyyyyyyyyyy.us-west-1.rds.amazonaws.com',
       JAMBONES_MYSQL_USER: 'admin',

app.js

@@ -25,9 +25,74 @@ const opts = {
 };
 const pino = require('pino');
 const logger = pino(opts, pino.destination({sync: false}));
-const {LifeCycleEvents, FS_UUID_SET_NAME} = require('./lib/utils/constants');
+const {LifeCycleEvents, FS_UUID_SET_NAME, SystemState, FEATURE_SERVER} = require('./lib/utils/constants');
 const installSrfLocals = require('./lib/utils/install-srf-locals');
-installSrfLocals(srf, logger);
+const createHttpListener = require('./lib/utils/http-listener');
+const healthCheck = require('@jambonz/http-health-check');
+
+logger.on('level-change', (lvl, _val, prevLvl, _prevVal, instance) => {
+  if (logger !== instance) {
+    return;
+  }
+  logger.info('system log level %s was changed to %s', prevLvl, lvl);
+});
+
+// Install the srf locals
+installSrfLocals(srf, logger, {
+  onFreeswitchConnect: (wraper) => {
+    // Only connect to drachtio if freeswitch is connected
+    logger.info(`connected to freeswitch at ${wraper.ms.address}, start drachtio server`);
+    if (DRACHTIO_HOST) {
+      srf.connect({host: DRACHTIO_HOST, port: DRACHTIO_PORT, secret: DRACHTIO_SECRET });
+      srf.on('connect', (err, hp) => {
+        const arr = /^(.*)\/(.*)$/.exec(hp.split(',').pop());
+        srf.locals.localSipAddress = `${arr[2]}`;
+        logger.info(`connected to drachtio listening on ${hp}, local sip address is ${srf.locals.localSipAddress}`);
+      });
+    }
+    else {
+      logger.info(`listening for drachtio requests on port ${DRACHTIO_PORT}`);
+      srf.listen({port: DRACHTIO_PORT, secret: DRACHTIO_SECRET});
+    }
+    // Start Http server
+    createHttpListener(logger, srf)
+      .then(({server, app}) => {
+        httpServer = server;
+        healthCheck({app, logger, path: '/', fn: getCount});
+        return {server, app};
+      })
+      .catch((err) => {
+        logger.error(err, 'Error creating http listener');
+      });
+  },
+  onFreeswitchDisconnect: (wraper) => {
+    // check if all freeswitch connections are lost, disconnect drachtio server
+    logger.info(`lost connection to freeswitch at ${wraper.ms.address}`);
+    const ms = srf.locals.getFreeswitch();
+    if (!ms) {
+      logger.info('no freeswitch connections, stopping drachtio server');
+      disconnect();
+    }
+  }
+});
+if (NODE_ENV === 'test') {
+  srf.on('error', (err) => {
+    logger.info(err, 'Error connecting to drachtio');
+  });
+}
+
+// Init services
+const writeSystemAlerts = srf.locals?.writeSystemAlerts;
+if (writeSystemAlerts) {
+  writeSystemAlerts({
+    system_component: FEATURE_SERVER,
+    state : SystemState.Online,
+    fields : {
+      detail: `feature-server with process_id ${process.pid} started`,
+      host: srf.locals?.ipv4
+    }
+  });
+}
 
 const {
   initLocals,
@@ -42,24 +107,6 @@ const {
 const InboundCallSession = require('./lib/session/inbound-call-session');
 const SipRecCallSession = require('./lib/session/siprec-call-session');
 
-if (DRACHTIO_HOST) {
-  srf.connect({host: DRACHTIO_HOST, port: DRACHTIO_PORT, secret: DRACHTIO_SECRET });
-  srf.on('connect', (err, hp) => {
-    const arr = /^(.*)\/(.*)$/.exec(hp.split(',').pop());
-    srf.locals.localSipAddress = `${arr[2]}`;
-    logger.info(`connected to drachtio listening on ${hp}, local sip address is ${srf.locals.localSipAddress}`);
-  });
-}
-else {
-  logger.info(`listening for drachtio requests on port ${DRACHTIO_PORT}`);
-  srf.listen({port: DRACHTIO_PORT, secret: DRACHTIO_SECRET});
-}
-if (NODE_ENV === 'test') {
-  srf.on('error', (err) => {
-    logger.info(err, 'Error connecting to drachtio');
-  });
-}
-
 srf.use('invite', [
   initLocals,
   createRootSpan,
@@ -85,23 +132,28 @@ sessionTracker.on('idle', () => {
   }
 });
 const getCount = () => sessionTracker.count;
-const healthCheck = require('@jambonz/http-health-check');
+
 let httpServer;
 
-const createHttpListener = require('./lib/utils/http-listener');
-createHttpListener(logger, srf)
-  .then(({server, app}) => {
-    httpServer = server;
-    healthCheck({app, logger, path: '/', fn: getCount});
-    return {server, app};
-  })
-  .catch((err) => {
-    logger.error(err, 'Error creating http listener');
-  });
-
-
-setInterval(() => {
+const monInterval = setInterval(async() => {
   srf.locals.stats.gauge('fs.sip.calls.count', sessionTracker.count);
+  try {
+    const systemInformation = await srf.locals.dbHelpers.lookupSystemInformation();
+    if (systemInformation && systemInformation.log_level) {
+      const envLogLevel = logger.levels.values[JAMBONES_LOGLEVEL.toLowerCase()];
+      const dbLogLevel = logger.levels.values[systemInformation.log_level];
+      const appliedLogLevel = Math.min(envLogLevel, dbLogLevel);
+      if (logger.levelVal !== appliedLogLevel) {
+        logger.level = logger.levels.labels[Math.min(envLogLevel, dbLogLevel)];
+      }
+    }
+  } catch (err) {
+    if (process.env.NODE_ENV === 'test') {
+      clearInterval(monInterval);
+      logger.error('all tests complete');
+    }
+    else logger.error({err}, 'Error checking system log level in database');
+  }
 }, 20000);
 
 const disconnect = () => {
@@ -109,16 +161,29 @@ const disconnect = () => {
     httpServer?.on('close', resolve);
     httpServer?.close();
     srf.disconnect();
+    srf.removeAllListeners();
     srf.locals.mediaservers?.forEach((ms) => ms.disconnect());
   });
 };
-
 process.on('SIGTERM', handle);
+process.on('SIGINT', handle);
 
-function handle(signal) {
+async function handle(signal) {
   const {removeFromSet} = srf.locals.dbHelpers;
   srf.locals.disabled = true;
   logger.info(`got signal ${signal}`);
+  const writeSystemAlerts = srf.locals?.writeSystemAlerts;
+  if (writeSystemAlerts) {
+    // it has to be synchronous call, or else by the time system saves the app terminates
+    await writeSystemAlerts({
+      system_component: FEATURE_SERVER,
+      state : SystemState.Offline,
+      fields : {
+        detail: `feature-server with process_id ${process.pid} stopped, signal ${signal}`,
+        host: srf.locals?.ipv4
+      }
+    });
+  }
   const setName = `${(JAMBONES_CLUSTER_ID || 'default')}:active-fs`;
   const fsServiceUrlSetName = `${(JAMBONES_CLUSTER_ID || 'default')}:fs-service-url`;
   if (setName && srf.locals.localSipAddress) {

lib/config.js

@@ -73,6 +73,7 @@ const JAMBONES_LOGLEVEL = process.env.JAMBONES_LOGLEVEL || 'info';
 const JAMBONES_INJECT_CONTENT = process.env.JAMBONES_INJECT_CONTENT;
 
 const PORT = parseInt(process.env.HTTP_PORT, 10) || 3000;
+const HTTP_IP = process.env.HTTP_IP;
 const HTTP_PORT_MAX = parseInt(process.env.HTTP_PORT_MAX, 10);
 
 const K8S = process.env.K8S;
@@ -92,7 +93,7 @@ const AWS_REGION = process.env.AWS_REGION;
 const AWS_ACCESS_KEY_ID = process.env.AWS_ACCESS_KEY_ID;
 const AWS_SECRET_ACCESS_KEY = process.env.AWS_SECRET_ACCESS_KEY;
 const AWS_SNS_PORT = parseInt(process.env.AWS_SNS_PORT, 10) || 3001;
-const AWS_SNS_TOPIC_ARM = process.env.AWS_SNS_TOPIC_ARM;
+const AWS_SNS_TOPIC_ARN = process.env.AWS_SNS_TOPIC_ARN;
 const AWS_SNS_PORT_MAX = parseInt(process.env.AWS_SNS_PORT_MAX, 10) || 3005;
 
 const GCP_JSON_KEY = process.env.GCP_JSON_KEY;
@@ -107,6 +108,8 @@ const DEEPGRAM_API_KEY = process.env.DEEPGRAM_API_KEY;
 const ANCHOR_MEDIA_ALWAYS = process.env.ANCHOR_MEDIA_ALWAYS;
 const VMD_HINTS_FILE = process.env.VMD_HINTS_FILE;
 
+const JAMBONES_AWS_TRANSCRIBE_USE_GRPC = process.env.JAMBONES_AWS_TRANSCRIBE_USE_GRPC;
+
 /* security, secrets */
 const LEGACY_CRYPTO = !!process.env.LEGACY_CRYPTO;
 const JWT_SECRET = process.env.JWT_SECRET;
@@ -133,6 +136,12 @@ const JAMBONES_DISABLE_DIRECT_P2P_CALL = process.env.JAMBONES_DISABLE_DIRECT_P2P
 const JAMBONES_EAGERLY_PRE_CACHE_AUDIO = parseInt(process.env.JAMBONES_EAGERLY_PRE_CACHE_AUDIO, 10) || 0;
 
 const JAMBONES_USE_FREESWITCH_TIMER_FD = process.env.JAMBONES_USE_FREESWITCH_TIMER_FD;
+const JAMBONES_DIAL_SBC_FOR_REGISTERED_USER = process.env.JAMBONES_DIAL_SBC_FOR_REGISTERED_USER || false;
+const JAMBONES_MEDIA_TIMEOUT_MS = process.env.JAMBONES_MEDIA_TIMEOUT_MS || 0;
+const JAMBONES_MEDIA_HOLD_TIMEOUT_MS = process.env.JAMBONES_MEDIA_HOLD_TIMEOUT_MS || 0;
+// jambonz
+const JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS =
+  process.env.JAMBONES_TRANSCRIBE_EP_DESTROY_DELAY_MS;
 
 module.exports = {
   JAMBONES_MYSQL_HOST,
@@ -170,6 +179,7 @@ module.exports = {
   JAMBONES_CLUSTER_ID,
   PORT,
   HTTP_PORT_MAX,
+  HTTP_IP,
   K8S,
   K8S_SBC_SIP_SERVICE_NAME,
   JAMBONES_SUBNET,
@@ -182,12 +192,13 @@ module.exports = {
   AWS_ACCESS_KEY_ID,
   AWS_SECRET_ACCESS_KEY,
   AWS_SNS_PORT,
-  AWS_SNS_TOPIC_ARM,
+  AWS_SNS_TOPIC_ARN,
   AWS_SNS_PORT_MAX,
 
   ANCHOR_MEDIA_ALWAYS,
   VMD_HINTS_FILE,
   JAMBONES_FREESWITCH_MAX_CALL_DURATION_MINS,
+  JAMBONES_AWS_TRANSCRIBE_USE_GRPC,
 
   LEGACY_CRYPTO,
   JWT_SECRET,
@@ -216,5 +227,9 @@ module.exports = {
   JAMBONZ_RECORD_WS_PASSWORD,
   JAMBONZ_DISABLE_DIAL_PAI_HEADER,
   JAMBONES_DISABLE_DIRECT_P2P_CALL,
-  JAMBONES_USE_FREESWITCH_TIMER_FD
+  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
 };

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

@@ -3,7 +3,7 @@ const makeTask = require('../../tasks/make_task');
 const RestCallSession = require('../../session/rest-call-session');
 const CallInfo = require('../../session/call-info');
 const {CallDirection, CallStatus} = require('../../utils/constants');
-const uuidv4 = require('uuid-random');
+const crypto = require('crypto');
 const SipError = require('drachtio-srf').SipError;
 const { validationResult, body } = require('express-validator');
 const { validate } = require('@jambonz/verb-specifications');
@@ -14,6 +14,8 @@ const RootSpan = require('../../utils/call-tracer');
 const dbUtils = require('../../utils/db-utils');
 const { mergeSdpMedia, extractSdpMedia } = 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 removeNullProperties = (obj) => (Object.keys(obj).forEach((key) => obj[key] === null && delete obj[key]), obj);
 const removeNulls = (req, res, next) => {
@@ -65,7 +67,7 @@ router.post('/',
         lookupAppBySid
       }  = srf.locals.dbHelpers;
       const {getSBC, getFreeswitch} = srf.locals;
-      const sbcAddress = getSBC();
+      let sbcAddress = getSBC();
       if (!sbcAddress) throw new Error('no available SBCs for outbound call creation');
       const target = restDial.to;
       const opts = {
@@ -78,7 +80,7 @@ router.post('/',
       const {lookupTeamsByAccount, lookupAccountBySid} = srf.locals.dbHelpers;
       const account = await lookupAccountBySid(req.body.account_sid);
       const accountInfo = await lookupAccountDetails(req.body.account_sid);
-      const callSid = uuidv4();
+      const callSid = crypto.randomUUID();
       const application = req.body.application_sid ? await lookupAppBySid(req.body.application_sid) : null;
       const record_all_calls = account.record_all_calls || (application && application.record_all_calls);
       const recordOutputFormat = account.record_format || 'mp3';
@@ -97,7 +99,9 @@ router.post('/',
         'X-Trace-ID': rootSpan.traceId,
         ...(req.body?.application_sid && {'X-Application-Sid': req.body.application_sid}),
         ...(restDial.fromHost && {'X-Preferred-From-Host': restDial.fromHost}),
-        ...(record_all_calls && {'X-Record-All-Calls': recordOutputFormat})
+        ...(record_all_calls && {'X-Record-All-Calls': recordOutputFormat}),
+        ...(target.proxy && {'X-SIP-Proxy': target.proxy}),
+        ...target.headers
       };
 
       switch (target.type) {
@@ -139,6 +143,16 @@ router.post('/',
         }
       }
 
+      // find handling sbc sip for called user
+      if (JAMBONES_DIAL_SBC_FOR_REGISTERED_USER && target.type === 'user') {
+        const { registrar }  = srf.locals.dbHelpers;
+        const reg = await registrar.query(target.name);
+        if (reg) {
+          sbcAddress = selectHostPort(logger, reg.sbcAddress, 'tcp')[1];
+        }
+        //sbc outbound return 404 Notfound to handle case called user is not reigstered.
+      }
+
       /**
      * trunk isn't specified,
      * check if from-number matches any existing numbers on Jambonz
@@ -195,10 +209,13 @@ router.post('/',
 
       /**
      * create our application object -
-     * not from the database as per an inbound call,
-     * but from the provided params in the request
+     * we merge the inbound call application,
+     * with the provided app params from the request body
      */
-      const app = req.body;
+      const app = {
+        ...application,
+        ...req.body
+      };
 
       /**
      * attach our requestor and notifier objects
@@ -218,7 +235,7 @@ router.post('/',
       }
       if (!app.notifier && app.call_status_hook) {
         app.notifier = new HttpRequestor(logger, account.account_sid, app.call_status_hook, account.webhook_secret);
-        logger.debug({call_hook: app.call_hook}, 'creating http client for call status hook');
+        logger.debug({call_status_hook: app.call_status_hook}, 'creating http client for call status hook');
       }
       else if (!app.notifier) {
         logger.debug('creating null call status hook');
@@ -257,6 +274,8 @@ router.post('/',
               callId: inviteReq.get('Call-ID'),
               accountSid,
               traceId: rootSpan.traceId
+            }, {
+              ...(account.enable_debug_log && {level: 'debug'})
             });
             app.requestor.logger = app.notifier.logger = sipLogger;
             const callInfo = new CallInfo({
@@ -290,6 +309,8 @@ router.post('/',
           },
           cbProvisional: (prov) => {
             const callStatus = prov.body ? CallStatus.EarlyMedia : CallStatus.Ringing;
+            // Update call-id for sbc outbound INVITE
+            cs.callInfo.sbcCallid = prov.get('X-CID');
             if ([180, 183].includes(prov.status) && prov.body) connectStream(prov.body);
             restDial.emit('callStatus', prov.status, !!prov.body);
             cs.emit('callStatusChange', {callStatus, sipStatus: prov.status});

lib/middleware.js

@@ -1,5 +1,5 @@
-const uuidv4 = require('uuid-random');
-const {CallDirection, AllowedSipRecVerbs} = require('./utils/constants');
+const crypto = require('crypto');
+const {CallDirection, AllowedSipRecVerbs, WS_CLOSE_CODES} = require('./utils/constants');
 const {parseSiprecPayload} = require('./utils/siprec-utils');
 const CallInfo = require('./session/call-info');
 const HttpRequestor = require('./utils/http-requestor');
@@ -15,6 +15,7 @@ const {
   JAMBONES_DISABLE_DIRECT_P2P_CALL
 } = require('./config');
 const { createJambonzApp } = require('./dynamic-apps');
+const { decrypt } = require('./utils/encrypt-decrypt');
 
 module.exports = function(srf, logger) {
   const {
@@ -45,7 +46,7 @@ module.exports = function(srf, logger) {
       logger.info('getAccountDetails - rejecting call due to missing X-Account-Sid header');
       return res.send(500);
     }
-    const callSid = req.has('X-Retain-Call-Sid') ? req.get('X-Retain-Call-Sid') : uuidv4();
+    const callSid = req.has('X-Retain-Call-Sid') ? req.get('X-Retain-Call-Sid') : crypto.randomUUID();
     const account_sid = req.get('X-Account-Sid');
     req.locals = {callSid, account_sid, callId};
 
@@ -187,14 +188,20 @@ module.exports = function(srf, logger) {
 
     const {span} = rootSpan.startChildSpan('lookupAccountDetails');
     try {
-      req.locals.accountInfo = await lookupAccountDetails(account_sid);
-      req.locals.service_provider_sid = req.locals.accountInfo?.account?.service_provider_sid;
+      const accountDetail = await lookupAccountDetails(account_sid);
+      const account = accountDetail?.account;
+      req.locals.accountInfo = accountDetail;
+      req.locals.service_provider_sid = account?.service_provider_sid;
       span.end();
-      if (!req.locals.accountInfo.account.is_active) {
+      if (!account?.is_active) {
         logger.info(`Account is inactive or suspended ${account_sid}`);
         // TODO: alert
         return res.send(503, {headers: {'X-Reason': 'Account exists but is inactive'}});
       }
+      // Change the default log level to debug
+      if (account?.enable_debug_log) {
+        req.locals.logger.level = 'debug';
+      }
       logger.debug({accountInfo: req.locals?.accountInfo?.account}, `retrieved account info for ${account_sid}`);
       next();
     } catch (err) {
@@ -330,7 +337,9 @@ module.exports = function(srf, logger) {
         if (arr) {
           const google_custom_voice_sid = arr[1];
           const [custom_voice] = await lookupGoogleCustomVoice(google_custom_voice_sid);
-          if (custom_voice) {
+          //google voice cloning key has size 200kb, jambonz should not resolve the voice here that the app's calling
+          //webhook will receive big payload, tts-task should resolve the voice later.
+          if (!custom_voice.use_voice_cloning_key) {
             app2.speech_synthesis_voice = {
               reportedUsage: custom_voice.reported_usage,
               model: custom_voice.model
@@ -340,11 +349,10 @@ module.exports = function(srf, logger) {
       }
 
       req.locals.application = app2;
-
       // eslint-disable-next-line no-unused-vars
       const {call_hook, call_status_hook, ...appInfo} = app;  // mask sensitive data like user/pass on webhook
       // eslint-disable-next-line no-unused-vars
-      const {requestor, notifier, ...loggable} = appInfo;
+      const {requestor, notifier, env_vars, ...loggable} = appInfo;
       logger.info({app: loggable}, `retrieved application for incoming call to ${req.locals.calledNumber}`);
       req.locals.callInfo = new CallInfo({
         req,
@@ -354,11 +362,13 @@ module.exports = function(srf, logger) {
       });
       // if transferred call contains callInfo, let update original data to newly created callInfo in this instance.
       if (app.transferredCall && app.callInfo) {
-        req.locals.callInfo.callerName = app.callInfo.callerName;
-        req.locals.callInfo.from = app.callInfo.from;
-        req.locals.callInfo.to = app.callInfo.to;
-        req.locals.callInfo.originatingSipIp = app.callInfo.originatingSipIp;
-        req.locals.callInfo.originatingSipTrunkName = app.callInfo.originatingSipTrunkName;
+        const {direction, callerName, from, to, originatingSipIp, originatingSipTrunkName} = app.callInfo;
+        req.locals.callInfo.direction = direction;
+        req.locals.callInfo.callerName = callerName;
+        req.locals.callInfo.from = from;
+        req.locals.callInfo.to = to;
+        req.locals.callInfo.originatingSipIp = originatingSipIp;
+        req.locals.callInfo.originatingSipTrunkName = originatingSipTrunkName;
         delete app.callInfo;
       }
       next();
@@ -377,7 +387,7 @@ module.exports = function(srf, logger) {
     const {rootSpan, siprec, application:app} = req.locals;
     let span;
     try {
-      if (app.tasks && !JAMBONES_MYSQL_REFRESH_TTL) {
+      if (app.tasks && app.tasks?.length > 0 && !JAMBONES_MYSQL_REFRESH_TTL) {
         app.tasks = normalizeJambones(logger, app.tasks).map((tdata) => makeTask(logger, tdata));
         if (0 === app.tasks.length) throw new Error('no application provided');
         return next();
@@ -407,16 +417,28 @@ module.exports = function(srf, logger) {
             ...(app.fallback_speech_recognizer_language && {fallback_language: app.fallback_speech_recognizer_language})
           }
         };
+        let env_vars;
+        try {
+          if (app.env_vars) {
+            const d_env_vars = JSON.parse(decrypt(app.env_vars));
+            logger.info(`Setting env_vars: ${Object.keys(d_env_vars)}`); // Only log the keys not the values
+            env_vars = d_env_vars;
+          }
+        } catch (error) {
+          logger.info('Unable to set env_vars', error);
+        }
         const params = Object.assign(['POST', 'WS'].includes(app.call_hook.method) ? { sip: req.msg } : {},
           req.locals.callInfo,
           { service_provider_sid: req.locals.service_provider_sid },
-          { defaults });
+          { defaults },
+          { env_vars }
+        );
         logger.debug({ params }, 'sending initial webhook');
         const obj = rootSpan.startChildSpan('performAppWebhook');
         span = obj.span;
         const b3 = rootSpan.getTracingPropagation();
         const httpHeaders = b3 && { b3 };
-        json = await app.requestor.request('session:new', app.call_hook, params, httpHeaders);
+        json = await app.requestor.request('session:new', app.call_hook, params, httpHeaders, span);
       }
 
       app.tasks = normalizeJambones(logger, json).map((tdata) => makeTask(logger, tdata));
@@ -450,7 +472,7 @@ module.exports = function(srf, logger) {
       }).catch((err) => this.logger.info({err}, 'Error generating alert for parsing application'));
       logger.info({err}, `Error retrieving or parsing application: ${err?.message}`);
       res.send(480, {headers: {'X-Reason': err?.message || 'unknown'}});
-      app.requestor.close();
+      app.requestor.close(WS_CLOSE_CODES.GoingAway);
     }
   }
 

lib/session/adulting-call-session.js

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

lib/session/call-info.js

@@ -1,6 +1,6 @@
 const {CallDirection, CallStatus} = require('../utils/constants');
 const parseUri = require('drachtio-srf').parseUri;
-const uuidv4 = require('uuid-random');
+const crypto = require('crypto');
 const {JAMBONES_API_BASE_URL} = require('../config');
 /**
  * @classdesc Represents the common information for all calls
@@ -32,6 +32,7 @@ class CallInfo {
       this.sipStatus = 100;
       this.sipReason = 'Trying';
       this.callStatus = CallStatus.Trying;
+      this.sbcCallid = req.get('X-CID');
       this.originatingSipIp = req.get('X-Forwarded-For');
       this.originatingSipTrunkName = req.get('X-Originating-Carrier');
       const {siprec} = req.locals;
@@ -56,7 +57,7 @@ class CallInfo {
       // outbound call that is a child of an existing call
       const {req, parentCallInfo, to, callSid} = opts;
       srf = req.srf;
-      this.callSid = callSid || uuidv4();
+      this.callSid = callSid || crypto.randomUUID();
       this.parentCallSid = parentCallInfo.callSid;
       this.accountSid = parentCallInfo.accountSid;
       this.applicationSid = parentCallInfo.applicationSid;
@@ -129,6 +130,7 @@ class CallInfo {
       from: this.from,
       to: this.to,
       callId: this.callId,
+      sbcCallid: this.sbcCallid,
       sipStatus: this.sipStatus,
       sipReason: this.sipReason,
       callStatus: this.callStatus,

lib/session/confirm-call-session.js

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

lib/session/inbound-call-session.js

@@ -70,8 +70,14 @@ class InboundCallSession extends CallSession {
     this._hangup('caller');
   }
 
-  _jambonzHangup() {
-    this.dlg?.destroy();
+  _jambonzHangup(reason) {
+    this.dlg?.destroy({
+      headers: {
+        ...(reason && {'X-Reason': reason})
+      }
+    });
+    // kill current task or wakeup the call session.
+    this._callReleased();
   }
 
   _hangup(terminatedBy = 'jambonz') {

lib/session/rest-call-session.js

@@ -1,10 +1,6 @@
 const CallSession = require('./call-session');
 const {CallStatus} = require('../utils/constants');
 const moment = require('moment');
-const {parseUri} = require('drachtio-srf');
-const { normalizeJambones } = require('@jambonz/verb-specifications');
-const makeTask = require('../tasks/make_task');
-
 /**
  * @classdesc Subclass of CallSession.  This represents a CallSession that is
  * created for an outbound call that is initiated via the REST API.
@@ -46,61 +42,9 @@ class RestCallSession extends CallSession {
     this.dlg = dlg;
     dlg.on('destroy', this._callerHungup.bind(this));
     dlg.on('refer', this._onRefer.bind(this));
+    dlg.on('modify', this._onReinvite.bind(this));
     this.wrapDialog(dlg);
   }
-
-  /**
-   * global referHook
-   */
-
-  set referHook(hook) {
-    this._referHook = hook;
-  }
-
-  /**
-   * This is invoked when the called party sends REFER to Jambonz.
-   */
-  async _onRefer(req, res) {
-    if (this._referHook) {
-      try {
-        const to = parseUri(req.getParsedHeader('Refer-To').uri);
-        const by = parseUri(req.getParsedHeader('Referred-By').uri);
-        const b3 = this.b3;
-        const httpHeaders = b3 && {b3};
-        const json = await this.requestor.request('verb:hook', this._referHook, {
-          ...(this.callInfo.toJSON()),
-          refer_details: {
-            sip_refer_to: req.get('Refer-To'),
-            sip_referred_by: req.get('Referred-By'),
-            sip_user_agent: req.get('User-Agent'),
-            refer_to_user: to.scheme === 'tel' ? to.number : to.user,
-            referred_by_user: by.scheme === 'tel' ? by.number : by.user,
-            referring_call_sid: this.callSid,
-            referred_call_sid: null,
-          }
-        }, httpHeaders);
-
-        if (json && Array.isArray(json)) {
-          const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
-          if (tasks && tasks.length > 0) {
-            this.logger.info('RestCallSession:handleRefer received REFER, get new tasks');
-            this.replaceApplication(tasks);
-            if (this.wakeupResolver) {
-              this.wakeupResolver({reason: 'RestCallSession: referHook new taks'});
-              this.wakeupResolver = null;
-            }
-          }
-        }
-        res.send(202);
-        this.logger.info('RestCallSession:handleRefer - sent 202 Accepted');
-      } catch (err) {
-        this.logger.error({err}, 'RestCallSession:handleRefer - error while asking referHook');
-        res.send(err.statusCode || 501);
-      }
-    } else {
-      res.send(501);
-    }
-  }
   /**
    * This is invoked when the called party hangs up, in order to calculate the call duration.
    */
@@ -119,7 +63,7 @@ class RestCallSession extends CallSession {
     this.callInfo.callTerminationBy = terminatedBy;
     const duration = moment().diff(this.dlg.connectTime, 'seconds');
     this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});
-    this.logger.debug(`RestCallSession: called party hung up by ${terminatedBy}`);
+    this.logger.info(`RestCallSession: called party hung up by ${terminatedBy}`);
     this._callReleased();
   }
 

lib/tasks/conference.js

@@ -83,7 +83,11 @@ class Conference extends Task {
     // reset answer time if we were transferred from another feature server
     if (this.connectTime) dlg.connectTime = this.connectTime;
 
-
+    if (cs.sipRequestWithinDialogHook) {
+      /* remove any existing listener to escape from duplicating events */
+      this._removeSipIndialogRequestListener(this.dlg);
+      this._initSipIndialogRequestListener(cs, dlg);
+    }
     this.ep.on('destroy', this._kicked.bind(this, cs, dlg));
 
     try {
@@ -103,6 +107,7 @@ class Conference extends Task {
 
       this.logger.debug(`Conference:exec - conference ${this.confName} is over`);
       if (this.callMoved !== false) await this.performAction(this.results);
+      this._removeSipIndialogRequestListener(dlg);
     } catch (err) {
       this.logger.info(err, `TaskConference:exec - error in conference ${this.confName}`);
     }
@@ -118,7 +123,9 @@ class Conference extends Task {
     this.emitter.emit('kill');
     await this._doFinalMemberCheck(cs);
     if (this.ep && this.ep.connected) {
-      this.ep.conn.removeAllListeners('esl::event::CUSTOM::*');
+      // drachtio-fsmrf override esl::event::CUSTOM to conference join listerner, After finish the conference
+      // the application need to reset the esl::event::CUSTOM for another use on the same endpoint
+      this.ep.resetEslCustomEvent();
       this.ep.api(`conference ${this.confName} kick ${this.memberId}`)
         .catch((err) => this.logger.info({err}, 'Error kicking participant'));
     }
@@ -135,15 +142,10 @@ class Conference extends Task {
    * @param {SipDialog} dlg
    */
   async _init(cs, dlg) {
-    const friendlyName = this.confName;
     const {createHash, retrieveHash} = cs.srf.locals.dbHelpers;
+    this.friendlyName = this.confName;
     this.confName = `conf:${cs.accountSid}:${this.confName}`;
 
-    this.statusParams = Object.assign({
-      conferenceSid: this.confName,
-      friendlyName
-    }, cs.callInfo);
-
     // check if conference is in progress
     const obj = await retrieveHash(this.confName);
     if (obj) {
@@ -419,6 +421,20 @@ class Conference extends Task {
     }
   }
 
+  _initSipIndialogRequestListener(cs, dlg) {
+    dlg.on('info', this._onRequestWithinDialog.bind(this, cs));
+    dlg.on('message', this._onRequestWithinDialog.bind(this, cs));
+  }
+
+  _removeSipIndialogRequestListener(dlg) {
+    dlg && dlg.removeAllListeners('message');
+    dlg && dlg.removeAllListeners('info');
+  }
+
+  _onRequestWithinDialog(cs, req, res) {
+    cs._onRequestWithinDialog(req, res);
+  }
+
   /**
    * The conference we have been waiting for has started.
    * It may be on this server or a different one, and we are
@@ -493,7 +509,7 @@ class Conference extends Task {
   }
 
   async doConferenceParticipantAction(cs, opts) {
-    const {action, tag} = opts;
+    const {action, tag, wait_hook } = opts;
 
     switch (action) {
       case 'tag':
@@ -509,7 +525,10 @@ class Conference extends Task {
         await this.clearCoachMode();
         break;
       case 'hold':
-        this.doConferenceHold(cs, {conf_hold_status: 'hold'});
+        this.doConferenceHold(cs, {
+          conf_hold_status: 'hold',
+          ...(wait_hook && {wait_hook})
+        });
         break;
       case 'unhold':
         this.doConferenceHold(cs, {conf_hold_status: 'unhold'});
@@ -546,6 +565,13 @@ class Conference extends Task {
     } while (!this.killed && this.conf_hold_status === 'hold');
   }
 
+  /**
+   * mute or unmute side of the call
+   */
+  mute(callSid, doMute) {
+    this.doConferenceMute(this.callSession, {conf_mute_status: doMute});
+  }
+
   /**
    * Add ourselves to the waitlist of sessions to be notified once
    * the conference starts
@@ -604,7 +630,7 @@ class Conference extends Task {
      * when we hang up as the last member, the current member count = 1
      * when we are kicked out of the call when the moderator leaves, the member count = 0
      */
-    if (this.participantCount === 0) {
+    if (this.participantCount === 0 || this.endConferenceOnExit) {
       const {deleteKey} = cs.srf.locals.dbHelpers;
       try {
         this._notifyConferenceEvent(cs, 'end');
@@ -612,7 +638,8 @@ class Conference extends Task {
         this.logger.info(`conf ${this.confName} deprovisioned: ${removed ? 'success' : 'failure'}`);
       }
       catch (err) {
-        this.logger.error(err, `Error deprovisioning conference ${this.confName}`);
+        this.logger.error(err, `Error deprovisioning conference ${this.confName},
+          might be the conference already cleaned by another moderator`);
       }
     }
   }
@@ -645,7 +672,8 @@ class Conference extends Task {
         memberId: this.memberId,
         confName: this.confName,
         tasks,
-        rootSpan: cs.rootSpan
+        rootSpan: cs.rootSpan,
+        req: cs.req
       });
       await this._playSession.exec();
       this._playSession = null;
@@ -689,8 +717,24 @@ class Conference extends Task {
       if (!params.time) params.time = (new Date()).toISOString();
       if (!params.members && typeof this.participantCount === 'number') params.members = this.participantCount;
       cs.application.requestor
-        .request('verb:hook', this.statusHook, Object.assign(params, this.statusParams, httpHeaders))
-        .catch((err) => this.logger.info(err, 'Conference:notifyConferenceEvent - error'));
+        .request(
+          'verb:hook',
+          this.statusHook,
+          Object.assign(
+            params,
+            Object.assign(
+              {
+                conferenceSid: this.confName,
+                friendlyName: this.friendlyName,
+              },
+              cs.callInfo.toJSON()
+            ),
+            httpHeaders
+          )
+        )
+        .catch((err) =>
+          this.logger.info(err, 'Conference:notifyConferenceEvent - error')
+        );
     }
   }
 

lib/tasks/config.js

@@ -16,7 +16,8 @@ class TaskConfig extends Task {
       'fillerNoise',
       'actionHookDelayAction',
       'boostAudioSignal',
-      'vad'
+      'vad',
+      'ttsStream'
     ].forEach((k) => this[k] = this.data[k] || {});
 
     if ('notifyEvents' in this.data) {
@@ -34,7 +35,8 @@ class TaskConfig extends Task {
         'finishOnKey', 'input', 'numDigits', 'minDigits', 'maxDigits',
         'interDigitTimeout', 'bargein', 'dtmfBargein', 'minBargeinWordCount', 'actionHook'
       ].forEach((k) => {
-        if (this.bargeIn[k]) this.gatherOpts[k] = this.bargeIn[k];
+        const val = this.bargeIn[k];
+        if (val !== undefined && val !== null) this.gatherOpts[k] = val;
       });
     }
     if (this.transcribe?.enable) {
@@ -44,6 +46,12 @@ class TaskConfig extends Task {
       };
       delete this.transcribeOpts.enable;
     }
+    if (this.ttsStream.enable) {
+      this.sayOpts = {
+        verb: 'say',
+        stream: true
+      };
+    }
 
     if (this.data.reset) {
       if (typeof this.data.reset === 'string') this.data.reset = [this.data.reset];
@@ -73,6 +81,8 @@ class TaskConfig extends Task {
   get hasDub() { return Object.keys(this.dub).length; }
   get hasVad() { return Object.keys(this.vad).length; }
   get hasFillerNoise() { return Object.keys(this.fillerNoise).length; }
+  get hasReferHook() { return Object.keys(this.data).includes('referHook'); }
+  get hasTtsStream() { return Object.keys(this.ttsStream).length; }
 
   get summary() {
     const phrase = [];
@@ -82,13 +92,13 @@ class TaskConfig extends Task {
 
     if (this.bargeIn.enable) phrase.push('enable barge-in');
     if (this.hasSynthesizer) {
-      const {vendor:v, language:l, voice} = this.synthesizer;
-      const s = `{${v},${l},${voice}}`;
+      const {vendor:v, language:l, voice, label} = this.synthesizer;
+      const s = `{${v},${l},${voice},${label || 'None'}}`;
       phrase.push(`set synthesizer${s}`);
     }
     if (this.hasRecognizer) {
-      const {vendor:v, language:l} = this.recognizer;
-      const s = `{${v},${l}}`;
+      const {vendor:v, language:l, label} = this.recognizer;
+      const s = `{${v},${l},${label || 'None'}}`;
       phrase.push(`set recognizer${s}`);
     }
     if (this.hasRecording) phrase.push(this.record.action);
@@ -103,6 +113,10 @@ class TaskConfig extends Task {
     if (this.notifyEvents) phrase.push(`event notification ${this.notifyEvents ? 'on' : 'off'}`);
     if (this.onHoldMusic) phrase.push(`onHoldMusic: ${this.onHoldMusic}`);
     if ('boostAudioSignal' in this.data) phrase.push(`setGain ${this.data.boostAudioSignal}`);
+    if (this.hasReferHook) phrase.push('set referHook');
+    if (this.hasTtsStream) {
+      phrase.push(`${this.ttsStream.enable ? 'enable' : 'disable'} ttsStream`);
+    }
     return `${this.name}{${phrase.join(',')}}`;
   }
 
@@ -125,7 +139,7 @@ class TaskConfig extends Task {
 
       try {
         this.ep = ep;
-        this.startAmd(cs, ep, this, this.data.amd);
+        await this.startAmd(cs, ep, this, this.data.amd);
       } catch (err) {
         this.logger.info({err}, 'Config:exec - Error calling startAmd');
       }
@@ -173,18 +187,20 @@ class TaskConfig extends Task {
         : cs.speechRecognizerVendor;
       cs.speechRecognizerLabel = this.recognizer.label === 'default'
         ? cs.speechRecognizerLabel : this.recognizer.label;
-      cs.speechRecognizerLanguage = this.recognizer.language !== 'default'
+      cs.speechRecognizerLanguage = this.recognizer.language !== undefined && this.recognizer.language !== 'default'
         ? this.recognizer.language
         : cs.speechRecognizerLanguage;
 
       //fallback
-      cs.fallbackSpeechRecognizerVendor = this.recognizer.fallbackVendor !== 'default'
+      cs.fallbackSpeechRecognizerVendor = this.recognizer.fallbackVendor !== undefined &&
+        this.recognizer.fallbackVendor !== 'default'
         ? this.recognizer.fallbackVendor
         : cs.fallbackSpeechRecognizerVendor;
       cs.fallbackSpeechRecognizerLabel = this.recognizer.fallbackLabel === 'default' ?
         cs.fallbackSpeechRecognizerLabel :
         this.recognizer.fallbackLabel;
-      cs.fallbackSpeechRecognizerLanguage = this.recognizer.fallbackLanguage !== 'default'
+      cs.fallbackSpeechRecognizerLanguage = this.recognizer.fallbackLanguage !== undefined &&
+        this.recognizer.fallbackLanguage !== 'default'
         ? this.recognizer.fallbackLanguage
         : cs.fallbackSpeechRecognizerLanguage;
 
@@ -295,9 +311,29 @@ class TaskConfig extends Task {
         voiceMs: this.vad.voiceMs || 250,
         silenceMs: this.vad.silenceMs || 150,
         strategy: this.vad.strategy || 'one-shot',
-        mode: this.vad.mod || 2
+        mode: (this.vad.mode !== undefined && this.vad.mode !== null) ? this.vad.mode : 2
       };
     }
+
+    if (this.hasReferHook) {
+      cs.referHook = this.data.referHook;
+    }
+
+    if (this.ttsStream.enable && this.sayOpts) {
+      this.sayOpts.synthesizer = this.hasSynthesizer ? this.synthesizer : {
+        vendor: cs.speechSynthesisVendor,
+        language: cs.speechSynthesisLanguage,
+        voice: cs.speechSynthesisVoice,
+        ...(cs.speechSynthesisLabel && {
+          label: cs.speechSynthesisLabel
+        })
+      };
+      this.logger.info({opts: this.gatherOpts}, 'Config: enabling ttsStream');
+      cs.enableBackgroundTtsStream(this.sayOpts);
+    } else if (!this.ttsStream.enable) {
+      this.logger.info('Config: disabling ttsStream');
+      cs.disableTtsStream();
+    }
   }
 
   async kill(cs) {

lib/tasks/dequeue.js

@@ -3,8 +3,7 @@ const {TaskName, TaskPreconditions, DequeueResults, BONG_TONE} = require('../uti
 const Emitter = require('events');
 const bent = require('bent');
 const assert = require('assert');
-
-const sleepFor = (ms) => new Promise((resolve) => setTimeout(() => resolve(), ms));
+const { sleepFor } = require('../utils/helpers');
 
 const getUrl = (cs) => `${cs.srf.locals.serviceUrl}/v1/dequeue/${cs.callSid}`;
 

lib/tasks/dial.js

@@ -6,6 +6,7 @@ const {
   TaskName,
   TaskPreconditions,
   MAX_SIMRINGS,
+  MediaPath,
   KillReason
 } = require('../utils/constants');
 const assert = require('assert');
@@ -17,9 +18,13 @@ const dbUtils = require('../utils/db-utils');
 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} = require('../config');
+const {ANCHOR_MEDIA_ALWAYS,
+  JAMBONZ_DISABLE_DIAL_PAI_HEADER,
+  JAMBONES_DIAL_SBC_FOR_REGISTERED_USER} = require('../config');
 const { isOnhold, isOpusFirst } = require('../utils/sdp-utils');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
+const { selectHostPort } = require('../utils/network');
+const { sleepFor } = require('../utils/helpers');
 
 function parseDtmfOptions(logger, dtmfCapture) {
   let parentDtmfCollector, childDtmfCollector;
@@ -103,6 +108,7 @@ class TaskDial extends Task {
     this.proxy = this.data.proxy;
     this.tag = this.data.tag;
     this.boostAudioSignal = this.data.boostAudioSignal;
+    this._mediaPath = MediaPath.FullMedia;
 
     if (this.dtmfHook) {
       const {parentDtmfCollector, childDtmfCollector} = parseDtmfOptions(logger, this.data.dtmfCapture || {});
@@ -114,8 +120,9 @@ class TaskDial extends Task {
       }
     }
 
-    if (this.data.listen) {
-      this.listenTask = makeTask(logger, {'listen': this.data.listen}, this);
+    const listenData = this.data.listen || this.data.stream;
+    if (listenData) {
+      this.listenTask = makeTask(logger, {'listen': listenData }, this);
     }
     if (this.data.transcribe) {
       this.transcribeTask = makeTask(logger, {'transcribe' : this.data.transcribe}, this);
@@ -150,17 +157,21 @@ class TaskDial extends Task {
 
   get canReleaseMedia() {
     const keepAnchor = this.data.anchorMedia ||
-    this.cs.isBackGroundListen ||
-    this.cs.onHoldMusic ||
-    ANCHOR_MEDIA_ALWAYS ||
-    this.listenTask ||
-    this.dubTasks ||
-    this.transcribeTask ||
-    this.startAmd;
+      this.cs.isBackGroundListen ||
+      this.cs.onHoldMusic ||
+      ANCHOR_MEDIA_ALWAYS ||
+      this.listenTask ||
+      this.dubTasks ||
+      this.transcribeTask ||
+      this.startAmd;
 
     return !keepAnchor;
   }
 
+  get shouldExitMediaPathEntirely() {
+    return this.data.exitMediaPath;
+  }
+
   get summary() {
     if (this.target.length === 1) {
       const target = this.target[0];
@@ -181,6 +192,16 @@ class TaskDial extends Task {
 
   async exec(cs) {
     await super.exec(cs);
+
+    if (this.data.anchorMedia && this.data.exitMediaPath) {
+      this.logger.info('Dial:exec - incompatible anchorMedia and exitMediaPath are both set, will obey anchorMedia');
+      delete this.data.exitMediaPath;
+    }
+    if (!this.canReleaseMedia && this.data.exitMediaPath) {
+      this.logger.info(
+        'Dial:exec - exitMediaPath is set so features such as transcribe and record will not work on this call');
+    }
+
     try {
       if (this.listenTask) {
         const {span, ctx} = this.startChildSpan(`nested:${this.listenTask.summary}`);
@@ -203,7 +224,16 @@ class TaskDial extends Task {
       else {
         this.epOther = cs.ep;
         if (this.dialMusic && this.epOther && this.epOther.connected) {
-          this.epOther.play(this.dialMusic).catch((err) => {});
+          (async() => {
+            do {
+              try {
+                await this.epOther.play(this.dialMusic);
+              } catch (err) {
+                this.logger.error(err, `TaskDial:exec error playing dialMusic ${this.dialMusic}`);
+                await sleepFor(1000);
+              }
+            } while (!this.killed && !this.bridged && this._mediaPath === MediaPath.FullMedia);
+          })();
         }
       }
       if (!this.killed) await this._attemptCalls(cs);
@@ -243,7 +273,9 @@ class TaskDial extends Task {
     this._removeDtmfDetection(this.dlg);
     await this._killOutdials();
     if (this.sd) {
-      this.sd.kill();
+      const byeReasonHeader = this.killReason === KillReason.MediaTimeout ? 'Media Timeout' : undefined;
+      this.sd.kill(byeReasonHeader);
+      this.sd.ep?.removeListener('destroy', this._handleMediaTimeout.bind(this));
       this.sd.removeAllListeners();
       this.sd = null;
     }
@@ -289,7 +321,7 @@ class TaskDial extends Task {
       if (!cs.callGone && this.epOther) {
 
         /* if we can release the media back to the SBC, do so now */
-        if (this.canReleaseMedia) this._releaseMedia(cs, this.sd);
+        if (this.canReleaseMedia) this._releaseMedia(cs, this.sd, this.shouldExitMediaPathEntirely);
         else this.epOther.bridge(this.ep);
       }
     } catch (err) {
@@ -329,17 +361,26 @@ class TaskDial extends Task {
 
         const to = parseUri(req.getParsedHeader('Refer-To').uri);
         const by = parseUri(req.getParsedHeader('Referred-By').uri);
+        const referredBy = req.get('Referred-By');
+        const userAgent = req.get('User-Agent');
+        const customHeaders = Object.keys(req.headers)
+          .filter((h) => h.toLowerCase().startsWith('x-'))
+          .reduce((acc, h) => {
+            acc[h] = req.get(h);
+            return acc;
+          }, {});
         this.logger.info({to}, 'refer to parsed');
         const json = await cs.requestor.request('verb:hook', this.referHook, {
           ...(callInfo.toJSON()),
           refer_details: {
             sip_refer_to: req.get('Refer-To'),
-            sip_referred_by: req.get('Referred-By'),
-            sip_user_agent: req.get('User-Agent'),
             refer_to_user: to.scheme === 'tel' ? to.number : to.user,
-            referred_by_user: by.scheme === 'tel' ? by.number : by.user,
+            ...(referredBy && {sip_referred_by: referredBy}),
+            ...(userAgent && {sip_user_agent: userAgent}),
+            ...(by && {referred_by_user: by.scheme === 'tel' ? by.number : by.user}),
             referring_call_sid,
-            referred_call_sid
+            referred_call_sid,
+            ...customHeaders
           }
         }, httpHeaders);
         if (json && Array.isArray(json)) {
@@ -365,9 +406,13 @@ class TaskDial extends Task {
             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');
       } catch (err) {
+        this.logger.info({err}, 'DialTask:handleRefer - error processing incoming REFER');
         res.send(err.statusCode || 501);
       }
     }
@@ -453,7 +498,7 @@ class TaskDial extends Task {
     dlg && dlg.removeAllListeners('info');
   }
 
-  async _onRequestWithinDialog(cs, req, res) {
+  _onRequestWithinDialog(cs, req, res) {
     cs._onRequestWithinDialog(req, res);
   }
 
@@ -473,11 +518,11 @@ class TaskDial extends Task {
   }
 
   async _attemptCalls(cs) {
-    const {req, srf} = cs;
+    const {req, callInfo, direction, srf} = cs;
     const {getSBC} = srf.locals;
     const {lookupTeamsByAccount, lookupAccountBySid} = srf.locals.dbHelpers;
-    const {lookupCarrier, lookupCarrierByPhoneNumber} = dbUtils(this.logger, cs.srf);
-    const sbcAddress = this.proxy || getSBC();
+    const {lookupCarrier, lookupCarrierByPhoneNumber, lookupVoipCarrierBySid} = dbUtils(this.logger, cs.srf);
+    let sbcAddress = this.proxy || getSBC();
     const teamsInfo = {};
     let fqdn;
 
@@ -485,19 +530,28 @@ class TaskDial extends Task {
     this.headers = {
       'X-Account-Sid': cs.accountSid,
       ...(req && req.has('X-CID') && {'X-CID': req.get('X-CID')}),
-      ...(req && req.has('P-Asserted-Identity') && !JAMBONZ_DISABLE_DIAL_PAI_HEADER &&
-        {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
+      ...(direction === 'outbound' && callInfo.sbcCallid && {'X-CID': callInfo.sbcCallid}),
+      ...(!JAMBONZ_DISABLE_DIAL_PAI_HEADER && req && {
+        ...(req.has('P-Asserted-Identity') && {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
+        ...(req.has('Privacy') && {'Privacy': req.get('Privacy')}),
+      }),
       ...(req && req.has('X-Voip-Carrier-Sid') && {'X-Voip-Carrier-Sid': req.get('X-Voip-Carrier-Sid')}),
       // Put headers at the end to make sure opt.headers override all default behavior.
       ...this.headers
     };
 
+    // default to inband dtmf if not specified
+    this.inbandDtmfEnabled = cs.inbandDtmfEnabled;
+    // get calling user from From header
+    const parsedFrom = req.getParsedHeader('from');
+    const fromUri = parseUri(parsedFrom.uri);
     const opts = {
       headers: this.headers,
       proxy: `sip:${sbcAddress}`,
-      callingNumber: this.callerId || req.callingNumber,
+      callingNumber: this.callerId || fromUri.user,
       ...(this.callerName && {callingName: this.callerName}),
-      opusFirst: isOpusFirst(this.cs.ep.remote.sdp)
+      opusFirst: isOpusFirst(this.cs.ep.remote.sdp),
+      isVideoCall: this.cs.ep.remote.sdp.includes('m=video')
     };
 
     const t = this.target.find((t) => t.type === 'teams');
@@ -541,6 +595,15 @@ class TaskDial extends Task {
             this.logger.error({err}, 'Error looking up account by sid');
           }
         }
+        // find handling sbc sip for called user
+        if (JAMBONES_DIAL_SBC_FOR_REGISTERED_USER && t.type === 'user') {
+          const { registrar }  = srf.locals.dbHelpers;
+          const reg = await registrar.query(t.name);
+          if (reg) {
+            sbcAddress = selectHostPort(this.logger, reg.sbcAddress, 'tcp')[1];
+          }
+          //sbc outbound return 404 Notfound to handle case called user is not reigstered.
+        }
         if (t.type === 'phone' && t.trunk) {
           const voip_carrier_sid = await lookupCarrier(cs.accountSid, t.trunk);
           this.logger.info(`Dial:_attemptCalls: selected ${voip_carrier_sid} for requested carrier: ${t.trunk}`);
@@ -557,10 +620,17 @@ class TaskDial extends Task {
           const str = this.callerId || req.callingNumber || '';
           const callingNumber = str.startsWith('+') ? str.substring(1) : str;
           const voip_carrier_sid = await lookupCarrierByPhoneNumber(cs.accountSid, callingNumber);
+          const req_voip_carrier_sid = req.has('X-Voip-Carrier-Sid') ? req.get('X-Voip-Carrier-Sid') : null;
           if (voip_carrier_sid) {
             this.logger.info(
               `Dial:_attemptCalls: selected voip_carrier_sid ${voip_carrier_sid} for callingNumber: ${callingNumber}`);
             opts.headers['X-Requested-Carrier-Sid'] = voip_carrier_sid;
+
+            // Checking if outbound carrier is different from inbound carrier and has dtmf type tones
+            if (voip_carrier_sid !== req_voip_carrier_sid) {
+              const [voipCarrier] = await lookupVoipCarrierBySid(voip_carrier_sid);
+              this.inbandDtmfEnabled = voipCarrier?.dtmf_type === 'tones';
+            }
           }
         }
 
@@ -601,6 +671,7 @@ class TaskDial extends Task {
                 dialCallStatus: obj.callStatus,
                 dialSipStatus: obj.sipStatus,
                 dialCallSid: sd.callSid,
+                dialSbcCallid: sd.callInfo.sbcCallid
               });
             }
             switch (obj.callStatus) {
@@ -716,7 +787,7 @@ class TaskDial extends Task {
             // Offhold, time to release media
             const newSdp = await this.ep.modify(req.body);
             await res.send(200, {body: newSdp});
-            await this._releaseMedia(this.cs, this.sd);
+            await this._releaseMedia(this.cs, this.sd, this.shouldExitMediaPathEntirely);
             this.isOutgoingLegHold = false;
           } else {
             this.logger.debug('Dial: _onReinvite receive unhold Request, update media server');
@@ -800,10 +871,14 @@ class TaskDial extends Task {
 
     if (this.parentDtmfCollector) this._installDtmfDetection(cs, cs.dlg);
     if (this.childDtmfCollector) this._installDtmfDetection(cs, this.dlg);
-    if (cs.sipRequestWithinDialogHook) this._initSipIndialogRequestListener(cs, this.dlg);
+    if (cs.sipRequestWithinDialogHook) {
+      /* remove any existing listener to escape from duplicating events */
+      this._removeSipIndialogRequestListener(this.dlg);
+      this._initSipIndialogRequestListener(cs, this.dlg);
+    }
 
     if (this.transcribeTask) this.transcribeTask.exec(cs, {ep: this.epOther, ep2:this.ep});
-    if (this.listenTask) this.listenTask.exec(cs, {ep: this.epOther});
+    if (this.listenTask) this.listenTask.exec(cs, {ep: this.listenTask.channel === 2 ? this.ep : this.epOther});
     if (this.startAmd) {
       try {
         this.startAmd(cs, this.ep, this, this.data.amd);
@@ -825,7 +900,17 @@ class TaskDial extends Task {
     }
 
     /* if we can release the media back to the SBC, do so now */
-    if (this.canReleaseMedia) setTimeout(this._releaseMedia.bind(this, cs, sd), 200);
+    if (this.canReleaseMedia || this.shouldExitMediaPathEntirely) {
+      setTimeout(this._releaseMedia.bind(this, cs, sd, this.shouldExitMediaPathEntirely), 200);
+    }
+
+    this.sd.ep.once('destroy', this._handleMediaTimeout.bind(this));
+  }
+
+  _handleMediaTimeout(evt) {
+    if (evt?.reason === 'MEDIA_TIMEOUT' && this.sd && this.bridged) {
+      this.kill(this.cs, KillReason.MediaTimeout);
+    }
   }
 
   _bridgeEarlyMedia(sd) {
@@ -837,22 +922,57 @@ class TaskDial extends Task {
     }
   }
 
+  /* public api */
+  async updateMediaPath(desiredPath) {
+    this.logger.info(`Dial:updateMediaPath - ${this._mediaPath} => ${desiredPath}`);
+    switch (desiredPath) {
+      case MediaPath.NoMedia:
+        assert(this._mediaPath !== MediaPath.NoMedia, 'updateMediaPath: already no-media');
+        await this._releaseMedia(this.cs, this.sd, true);
+        break;
+
+      case MediaPath.PartialMedia:
+        assert(this._mediaPath !== MediaPath.PartialMedia, 'updateMediaPath: already partial-media');
+        if (this._mediaPath === MediaPath.FullMedia) {
+          await this._releaseMedia(this.cs, this.sd, false);
+        }
+        else {
+          // to go from no-media to partial-media we need to go through full-media first
+          await this.reAnchorMedia(this.cs, this.sd);
+          await this._releaseMedia(this.cs, this.sd, false);
+        }
+        assert(!this.epOther, 'updateMediaPath: epOther should be null');
+        assert(!this.ep, 'updateMediaPath: ep should be null');
+
+        break;
+      case MediaPath.FullMedia:
+        assert(this._mediaPath !== MediaPath.FullMedia, 'updateMediaPath: already full-media');
+        await this.reAnchorMedia(this.cs, this.sd);
+        break;
+
+      default:
+        assert(false, `updateMediaPath: invalid path request ${desiredPath}`);
+    }
+  }
+
   /**
    * Release the media from freeswitch
    * @param {*} cs
    * @param {*} sd
    */
-  async _releaseMedia(cs, sd) {
+  async _releaseMedia(cs, sd, releaseEntirely = false) {
     assert(cs.ep && sd.ep);
 
     try {
       // Wait until we got new SDP from B leg to ofter to A Leg
       const aLegSdp = cs.ep.remote.sdp;
-      await sd.releaseMediaToSBC(aLegSdp, cs.ep.local.sdp);
+      await sd.releaseMediaToSBC(aLegSdp, cs.ep.local.sdp, releaseEntirely);
       const bLegSdp = sd.dlg.remote.sdp;
-      await cs.releaseMediaToSBC(bLegSdp);
+      await cs.releaseMediaToSBC(bLegSdp, releaseEntirely);
       this.epOther = null;
-      this.logger.info('Dial:_releaseMedia - successfully released media from freewitch');
+      this._mediaPath = releaseEntirely ? MediaPath.NoMedia : MediaPath.PartialMedia;
+      this.logger.info(
+        `Dial:_releaseMedia - successfully released media from freewitch, media path is now ${this._mediaPath}`);
     } catch (err) {
       this.logger.info({err}, 'Dial:_releaseMedia error');
     }
@@ -862,8 +982,14 @@ class TaskDial extends Task {
     if (cs.ep && sd.ep) return;
 
     this.logger.info('Dial:reAnchorMedia - re-anchoring media to freewitch');
-    await Promise.all([sd.reAnchorMedia(), cs.reAnchorMedia()]);
+    await Promise.all([sd.reAnchorMedia(this._mediaPath), cs.reAnchorMedia(this._mediaPath)]);
     this.epOther = cs.ep;
+
+    this.epOther.bridge(this.ep);
+
+    this._mediaPath = MediaPath.FullMedia;
+    this.logger.info(
+      `Dial:_releaseMedia - successfully re-anchored media to freewitch, media path is now ${this._mediaPath}`);
   }
 
   // Handle RE-INVITE hold from caller leg.
@@ -882,11 +1008,12 @@ class TaskDial extends Task {
         }
         this._onHoldHook(req);
       } else if (!isOnhold(req.body)) {
-        if (this.epOther && this.ep && this.isIncomingLegHold && this.canReleaseMedia) {
+        if (this.epOther && this.ep && this.isIncomingLegHold &&
+          (this.canReleaseMedia || this.shouldExitMediaPathEntirely)) {
           // Offhold, time to release media
           const newSdp = await this.epOther.modify(req.body);
           await res.send(200, {body: newSdp});
-          await this._releaseMedia(this.cs, this.sd);
+          await this._releaseMedia(this.cs, this.sd, this.shouldExitMediaPathEntirely);
           isHandled = true;
         }
         this.isIncomingLegHold = false;
@@ -945,7 +1072,8 @@ class TaskDial extends Task {
               callInfo: this.cs.callInfo,
               accountInfo: this.cs.accountInfo,
               tasks,
-              rootSpan: this.cs.rootSpan
+              rootSpan: this.cs.rootSpan,
+              req: this.cs.req
             });
             await this._onHoldSession.exec();
             this._onHoldSession = null;

lib/tasks/enqueue.js

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

lib/tasks/gather.js

@@ -11,8 +11,11 @@ const {
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
   AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
+  OpenAITranscriptionEvents,
   VadDetection,
-  VerbioTranscriptionEvents
+  VerbioTranscriptionEvents,
+  SpeechmaticsTranscriptionEvents
 } = require('../utils/constants.json');
 const {
   JAMBONES_GATHER_EARLY_HINTS_MATCH,
@@ -22,6 +25,8 @@ const {
 const makeTask = require('./make_task');
 const assert = require('assert');
 const SttTask = require('./stt-task');
+const { SpeechCredentialError } = require('../utils/error');
+const SPEECHMATICS_DEFAULT_ASR_TIMEOUT = 1200;
 
 class TaskGather extends SttTask {
   constructor(logger, opts, parentTask) {
@@ -79,6 +84,7 @@ class TaskGather extends SttTask {
     this._bufferedTranscripts = [];
     this.partialTranscriptsCount = 0;
     this.bugname_prefix = 'gather_';
+
   }
 
   get name() { return TaskName.Gather; }
@@ -106,6 +112,12 @@ class TaskGather extends SttTask {
     return this.fillerNoise.startDelaySecs;
   }
 
+  get isStreamingTts() { return this.sayTask && this.sayTask.isStreamingTts; }
+
+  getTtsVendorData() {
+    if (this.sayTask) return this.sayTask.getTtsVendorData(this.cs);
+  }
+
   get summary() {
     let s = `${this.name}{`;
     if (this.input.length === 2) s += 'inputs=[speech,digits],';
@@ -122,10 +134,22 @@ class TaskGather extends SttTask {
     return s;
   }
 
-  async exec(cs, {ep}) {
+  async exec(cs, obj) {
+    try {
+      await this.handling(cs, obj);
+    } catch (error) {
+      if (error instanceof SpeechCredentialError) {
+        this.logger.info('Gather failed due to SpeechCredentialError, finished!');
+        this.notifyTaskDone();
+        return;
+      }
+      throw error;
+    }
+  }
+
+  async handling(cs, {ep}) {
     this.logger.debug({options: this.data}, 'Gather:exec');
     await super.exec(cs, {ep});
-    const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
 
     this.fillerNoise = {
       ...(cs.fillerNoise || {}),
@@ -141,12 +165,23 @@ class TaskGather extends SttTask {
       const {hints, hintsBoost} = cs.globalSttHints;
       const setOfHints = new Set((this.data.recognizer.hints || [])
         .concat(hints)
-        .filter((h) => typeof h === 'string' && h.length > 0));
+        // allow for hints to be an array of object
+        .filter((h) => (typeof h === 'string' && h.length > 0) || (typeof h === 'object')));
       this.data.recognizer.hints = [...setOfHints];
       if (!this.data.recognizer.hintsBoost && hintsBoost) this.data.recognizer.hintsBoost = hintsBoost;
       this.logger.debug({hints: this.data.recognizer.hints, hintsBoost: this.data.recognizer.hintsBoost},
         'Gather:exec - applying global sttHints');
     }
+
+    // specials case for speechmatics: they dont do endpointing so we need to enable continuous ASR
+    if (this.vendor === 'speechmatics' && !this.isContinuousAsr) {
+      const maxDelay = this.recognizer?.speechmaticsOptions?.transcription_config?.max_delay;
+      if (maxDelay) this.asrTimeout = Math.min(SPEECHMATICS_DEFAULT_ASR_TIMEOUT, maxDelay * 1000);
+      else this.asrTimeout = SPEECHMATICS_DEFAULT_ASR_TIMEOUT;
+      this.isContinuousAsr = true;
+      this.logger.debug(`Gather:exec - auto-enabling continuous ASR for speechmatics w/ timeout ${this.asrTimeout}`);
+    }
+
     if (!this.isContinuousAsr && cs.isContinuousAsr) {
       this.isContinuousAsr = true;
       this.asrTimeout = cs.asrTimeout * 1000;
@@ -177,8 +212,10 @@ class TaskGather extends SttTask {
     this._startVad();
 
     const startDtmfListener = () => {
+      assert(!this._dtmfListenerStarted);
       if (this.input.includes('digits') || this.dtmfBargein || this.asrDtmfTerminationDigit) {
         ep.on('dtmf', this._onDtmf.bind(this, cs, ep));
+        this._dtmfListenerStarted = true;
       }
     };
 
@@ -193,7 +230,6 @@ class TaskGather extends SttTask {
             return;
           }
           this._startTranscribing(ep);
-          return updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
         } catch (e) {
           await this._startFallback(cs, ep, {error: e});
         }
@@ -205,6 +241,7 @@ class TaskGather extends SttTask {
         const {span, ctx} = this.startChildSpan(`nested:${this.sayTask.summary}`);
         const process = () => {
           this.logger.debug('Gather: nested say task completed');
+          this.playComplete = true;
           if (!this.listenDuringPrompt) {
             startDtmfListener();
           }
@@ -235,6 +272,7 @@ class TaskGather extends SttTask {
         const {span, ctx} = this.startChildSpan(`nested:${this.playTask.summary}`);
         const process = () => {
           this.logger.debug('Gather: nested play task completed');
+          this.playComplete = true;
           if (!this.listenDuringPrompt) {
             startDtmfListener();
           }
@@ -273,15 +311,14 @@ class TaskGather extends SttTask {
         await this._setSpeechHandlers(cs, ep);
         if (!this.resolved && !this.killed) {
           this._startTranscribing(ep);
-          updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid)
-            .catch(() => {/*already logged error */});
         }
         else {
           this.logger.info('Gather:exec - task was killed or resolved quickly, not starting transcription');
         }
       }
 
-      if (this.listenDuringPrompt) {
+      // https://github.com/jambonz/jambonz-feature-server/issues/913
+      if (this.listenDuringPrompt || (!this.sayTask && !this.playTask)) {
         startDtmfListener();
       }
 
@@ -322,9 +359,19 @@ class TaskGather extends SttTask {
     clearTimeout(this.interDigitTimer);
     let resolved = false;
     if (this.dtmfBargein) {
+      if (!this.playComplete) {
+        this.notifyStatus({event: 'dtmf-bargein-detected', ...evt});
+      }
       this._killAudio(cs);
       this.emit('dtmf', evt);
     }
+    if (this.isContinuousAsr && evt.dtmf === this.asrDtmfTerminationDigit && this._bufferedTranscripts.length > 0) {
+      this.logger.info(`continuousAsr triggered with dtmf ${this.asrDtmfTerminationDigit}`);
+      this._clearAsrTimer();
+      this._clearTimer();
+      this._startFinalAsrTimer();
+      return;
+    }
     if (evt.dtmf === this.finishOnKey && this.input.includes('digits')) {
       resolved = true;
       this._resolve('dtmf-terminator-key');
@@ -347,13 +394,6 @@ class TaskGather extends SttTask {
         this._resolve('dtmf-num-digits');
       }
     }
-    else if (this.isContinuousAsr && evt.dtmf === this.asrDtmfTerminationDigit) {
-      this.logger.info(`continuousAsr triggered with dtmf ${this.asrDtmfTerminationDigit}`);
-      this._clearAsrTimer();
-      this._clearTimer();
-      this._startFinalAsrTimer();
-      return;
-    }
     if (!resolved && this.interDigitTimeout > 0 && this.digitBuffer.length >= this.minDigits) {
       /* start interDigitTimer */
       const ms = this.interDigitTimeout * 1000;
@@ -494,6 +534,60 @@ class TaskGather extends SttTask {
         this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.ConnectFailure,
           this._onVendorConnectFailure.bind(this, cs, ep));
         break;
+
+      case 'voxist':
+        this.bugname = `${this.bugname_prefix}voxist_transcribe`;
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, VoxistTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        break;
+
+      case 'speechmatics':
+        this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
+        this.addCustomEventListener(
+          ep, SpeechmaticsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.Info,
+          this._onSpeechmaticsInfo.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.RecognitionStarted,
+          this._onSpeechmaticsRecognitionStarted.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.Connect,
+          this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.Error,
+          this._onSpeechmaticsErrror.bind(this, cs, ep));
+
+        break;
+
+      case 'openai':
+        this.bugname = `${this.bugname_prefix}openai_transcribe`;
+        this.addCustomEventListener(
+          ep, OpenAITranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, OpenAITranscriptionEvents.SpeechStarted, this._onOpenAISpeechStarted.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, OpenAITranscriptionEvents.SpeechStopped, this._onOpenAISpeechStopped.bind(this, cs, ep));
+        this.addCustomEventListener(ep, OpenAITranscriptionEvents.Connect,
+          this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, OpenAITranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        this.addCustomEventListener(ep, OpenAITranscriptionEvents.Error,
+          this._onOpenAIErrror.bind(this, cs, ep));
+
+        /* openai delta transcripts are useful only for minBargeinWordCount eval */
+        if (this.minBargeinWordCount > 1) {
+          this.openaiPartials = [];
+          opts.OPENAI_WANT_PARTIALS = 1;
+          this.addCustomEventListener(
+            ep, OpenAITranscriptionEvents.PartialTranscript, this._onOpenAIPartialTranscript.bind(this, cs, ep));
+        }
+        this.modelSupportsConversationTracking = opts.OPENAI_MODEL !== 'whisper-1';
+        break;
+
       default:
         if (this.vendor.startsWith('custom:')) {
           this.bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
@@ -518,13 +612,32 @@ class TaskGather extends SttTask {
   }
 
   _startTranscribing(ep) {
-    this.logger.debug({
+    this.logger.info({
       vendor: this.vendor,
       locale: this.language,
       interim: this.interim,
       bugname: this.bugname
     }, 'Gather:_startTranscribing');
 
+
+    /* special feature for openai: we can provide a prompt that includes recent conversation history */
+    let prompt;
+    if (this.vendor === 'openai') {
+      if (this.modelSupportsConversationTracking) {
+        prompt = this.formatOpenAIPrompt(this.cs, {
+          prompt: this.data.recognizer?.openaiOptions?.prompt,
+          hintsTemplate: this.data.recognizer?.openaiOptions?.promptTemplates?.hintsTemplate,
+          // eslint-disable-next-line max-len
+          conversationHistoryTemplate: this.data.recognizer?.openaiOptions?.promptTemplates?.conversationHistoryTemplate,
+          hints: this.data.recognizer?.hints,
+        });
+        this.logger.debug({prompt}, 'Gather:_startTranscribing - created an openai prompt');
+      }
+      else if (this.data.recognizer?.hints?.length > 0) {
+        prompt = this.data.recognizer?.hints.join(', ');
+      }
+    }
+
     /**
      * Note: we don't need to ask deepgram for interim results, because they
      * already send us words as they are finalized (is_final=true) even before
@@ -536,6 +649,7 @@ class TaskGather extends SttTask {
       interim: this.interim,
       bugname: this.bugname,
       hostport: this.hostport,
+      prompt
     }).catch((err) => {
       const {writeAlerts, AlertType} = this.cs.srf.locals;
       this.logger.error(err, 'TaskGather:_startTranscribing error');
@@ -543,7 +657,8 @@ class TaskGather extends SttTask {
         account_sid: this.cs.accountSid,
         alert_type: AlertType.STT_FAILURE,
         vendor: this.vendor,
-        detail: err.message
+        detail: err.message,
+        target_sid: this.cs.callSid
       });
     }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
   }
@@ -570,15 +685,24 @@ class TaskGather extends SttTask {
   }
 
   _startAsrTimer() {
-    if (this.vendor === 'deepgram') return; // no need
+    // Deepgram has a case that UtteranceEnd is not sent to cover the last word end time.
+    // So we need to wait for the asrTimeout to be sure that the last word is sent.
+    // if (this.vendor === 'deepgram') return; // no need
     assert(this.isContinuousAsr);
     this._clearAsrTimer();
     this._asrTimer = setTimeout(() => {
-      this.logger.debug('_startAsrTimer - asr timer went off');
+      this.logger.info('_startAsrTimer - asr timer went off');
       const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
+
+      /* special case for speechmatics - keep listening if we dont have any transcripts */
+      if (this.vendor === 'speechmatics' && this._bufferedTranscripts.length === 0) {
+        this.logger.debug('Gather:_startAsrTimer - speechmatics, no transcripts yet, keep listening');
+        this._startAsrTimer();
+        return;
+      }
       this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
     }, this.asrTimeout);
-    this.logger.debug(`_startAsrTimer: set for ${this.asrTimeout}ms`);
+    this.logger.info(`_startAsrTimer: set for ${this.asrTimeout}ms`);
   }
 
   _clearAsrTimer() {
@@ -590,7 +714,7 @@ class TaskGather extends SttTask {
   }
 
   _hangupCall() {
-    this.logger.debug('_hangupCall');
+    this.logger.debug('TaskGather:_hangupCall');
     this.cs.hangup();
   }
 
@@ -613,7 +737,7 @@ class TaskGather extends SttTask {
   _startFinalAsrTimer() {
     this._clearFinalAsrTimer();
     this._finalAsrTimer = setTimeout(() => {
-      this.logger.debug('_startFinalAsrTimer - final asr timer went off');
+      this.logger.info('_startFinalAsrTimer - final asr timer went off');
       const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
       this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
     }, 1000);
@@ -686,6 +810,7 @@ class TaskGather extends SttTask {
         this._fillerNoiseOn = false;  // in a race, if we just started audio it may sneak through here
         this.ep.api('uuid_break', this.ep.uuid)
           .catch((err) => this.logger.info(err, 'Error killing audio'));
+        cs.clearTtsStream();
       }
       return;
     }
@@ -699,6 +824,7 @@ class TaskGather extends SttTask {
       this.playTask.kill(cs);
       this.playTask = null;
     }
+    this.playComplete = true;
   }
 
   _onTranscription(cs, ep, evt, fsEvent) {
@@ -706,20 +832,31 @@ class TaskGather extends SttTask {
     const bugname = fsEvent.getHeader('media-bugname');
     const finished = fsEvent.getHeader('transcription-session-finished');
     this.logger.debug({evt, bugname, finished, vendor: this.vendor}, 'Gather:_onTranscription raw transcript');
-    if (bugname && this.bugname !== bugname) return;
+    if (bugname && this.bugname !== bugname) {
+      this.logger.info(
+        `Gather:_onTranscription - ignoring transcript from ${bugname} because our bug is ${this.bugname}`);
+      return;
+    }
     if (finished === 'true') return;
 
     if (this.vendor === 'ibm' && evt?.state === 'listening') return;
     if (this.vendor === 'deepgram' && evt.type === 'UtteranceEnd') {
       /* we will only get this when we have set utterance_end_ms */
       if (this._bufferedTranscripts.length === 0) {
-        this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram but no buffered transcripts');
+        this.logger.info('Gather:_onTranscription - got UtteranceEnd event from deepgram but no buffered transcripts');
       }
       else {
-        this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram, return buffered transcript');
-        evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
-        this._bufferedTranscripts = [];
-        this._resolve('speech', evt);
+        const utteranceTime = evt.last_word_end;
+        // eslint-disable-next-line max-len
+        if (utteranceTime && this._dgTimeOfLastUnprocessedWord && utteranceTime < this._dgTimeOfLastUnprocessedWord && utteranceTime != -1) {
+          this.logger.info('Gather:_onTranscription - got UtteranceEnd with unprocessed words, continue listening');
+        }
+        else {
+          this.logger.info('Gather:_onTranscription - got UtteranceEnd from deepgram, return buffered transcript');
+          evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language, this.vendor);
+          this._bufferedTranscripts = [];
+          this._resolve('speech', evt);
+        }
       }
       return;
     }
@@ -738,8 +875,6 @@ class TaskGather extends SttTask {
     }
     const confidence = evt.alternatives[0].confidence;
     const minConfidence = this.data.recognizer?.minConfidence;
-    this.logger.debug({evt},
-      `TaskGather:_onTranscription - confidence (${confidence}), minConfidence (${minConfidence})`);
     if (confidence && minConfidence && confidence < minConfidence) {
       this.logger.info({evt},
         'TaskGather:_onTranscription - Transcript confidence ' +
@@ -793,7 +928,7 @@ class TaskGather extends SttTask {
         const t = evt.alternatives[0].transcript;
         if (t) {
           /* remove trailing punctuation */
-          if (/[,;:\.!\?]$/.test(t)) {
+          if (this.vendor !== 'speechmatics' && /[,;:\.!\?]$/.test(t)) {
             this.logger.debug('TaskGather:_onTranscription - removing trailing punctuation');
             evt.alternatives[0].transcript = t.slice(0, -1);
           }
@@ -809,7 +944,10 @@ class TaskGather extends SttTask {
         this._startAsrTimer();
 
         /* some STT engines will keep listening after a final response, so no need to restart */
-        if (!['soniox', 'aws', 'microsoft', 'deepgram'].includes(this.vendor)) this._startTranscribing(ep);
+        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'speechmatics'].includes(this.vendor) &&
+          !this.vendor.startsWith('custom')) {
+          this._startTranscribing(ep);
+        }
       }
       else {
         /* this was removed to fix https://github.com/jambonz/jambonz-feature-server/issues/783 */
@@ -852,8 +990,21 @@ class TaskGather extends SttTask {
         if (originalEvent.is_final && evt.alternatives[0].transcript !== '') {
           this.logger.debug({evt}, 'Gather:_onTranscription - buffering a completed (partial) deepgram transcript');
           this._bufferedTranscripts.push(evt);
+          this._dgTimeOfLastUnprocessedWord = null;
+        }
+        if (evt.alternatives[0].transcript === '') {
+          emptyTranscript = true;
+        }
+        else if (!originalEvent.is_final) {
+          /* Deepgram: we have unprocessed words-save last word end time so we can later compare to UtteranceEnd */
+          const words = originalEvent.channel.alternatives[0].words;
+          if (words?.length > 0) {
+            this._dgTimeOfLastUnprocessedWord = words.slice(-1)[0].end;
+            this.logger.debug(
+              `TaskGather:_onTranscription - saving word end time: ${this._dgTimeOfLastUnprocessedWord}`);
+          }
+
         }
-        if (evt.alternatives[0].transcript === '') emptyTranscript = true;
       }
 
       if (!emptyTranscript) {
@@ -862,6 +1013,7 @@ class TaskGather extends SttTask {
           if (!this.playComplete) {
             this.logger.debug({transcript: evt.alternatives[0].transcript}, 'killing audio due to speech');
             this.emit('vad');
+            this.notifyStatus({event: 'speech-bargein-detected', ...evt});
           }
           this._killAudio(cs);
         }
@@ -876,15 +1028,18 @@ class TaskGather extends SttTask {
           this.cs.callInfo, httpHeaders));
       }
       if (this.vendor === 'soniox') {
-        this._clearTimer();
         if (evt.vendor.finalWords.length) {
           this.logger.debug({evt}, 'TaskGather:_onTranscription - buffering soniox transcript');
           this._sonioxTranscripts.push(evt.vendor.finalWords);
         }
       }
-
-      /* restart asr timer if we get a partial transcript */
-      if (this.isContinuousAsr) this._startAsrTimer();
+      // If transcription received, reset timeout timer.
+      if (this._timeoutTimer && !emptyTranscript) {
+        this._startTimer();
+      }
+      /* restart asr timer if we get a partial transcript (only if the asr timer is already running) */
+      /* note: https://github.com/jambonz/jambonz-feature-server/issues/866 */
+      if (this.isContinuousAsr && this._asrTimer) this._startAsrTimer();
     }
   }
   _onEndOfUtterance(cs, ep) {
@@ -924,7 +1079,6 @@ class TaskGather extends SttTask {
         bugname: this.bugname
       })
         .catch((err) => this.logger.error({err}, `Error stopping transcription for primary vendor ${this.vendor}`));
-      const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
       try {
         this.logger.debug('gather:_startFallback');
         this.notifyError({ msg: 'ASR error',
@@ -933,7 +1087,6 @@ class TaskGather extends SttTask {
         this._speechHandlersSet = false;
         await this._setSpeechHandlers(cs, ep);
         this._startTranscribing(ep);
-        updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
         return true;
       } catch (error) {
         this.logger.info({error}, `There is error while falling back to ${this.fallbackVendor}`);
@@ -961,11 +1114,13 @@ class TaskGather extends SttTask {
       if (code === 413 && error === 'Too much speech') return this._resolve('timeout');
     }
     this.logger.info({evt}, 'TaskGather:_onJambonzError');
+    const errMessage = evt.error || evt.Message;
     writeAlerts({
       account_sid: cs.accountSid,
       alert_type: AlertType.STT_FAILURE,
-      message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
+      message: `Custom speech vendor ${this.vendor} error: ${errMessage}`,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
     if (!(await this._startFallback(cs, ep, evt))) {
       this.notifyTaskDone();
@@ -979,12 +1134,52 @@ class TaskGather extends SttTask {
     }
   }
 
+  async _onSpeechmaticsErrror(cs, _ep, evt) {
+    // eslint-disable-next-line no-unused-vars
+    const {message, ...e} = evt;
+    this._onVendorError(cs, _ep, {error: JSON.stringify(e)});
+  }
+
+  async _onOpenAIErrror(cs, _ep, evt) {
+    // eslint-disable-next-line no-unused-vars
+    const {message, ...e} = evt;
+    this._onVendorError(cs, _ep, {error: JSON.stringify(e)});
+  }
+
+  async _onOpenAISpeechStarted(cs, _ep, evt) {
+    this.logger.debug({evt}, 'TaskGather:_onOpenAISpeechStarted');
+  }
+
+  async _onOpenAISpeechStopped(cs, _ep, evt) {
+    this.logger.debug({evt}, 'TaskGather:_onOpenAISpeechStopped');
+  }
+
+  async _onOpenAIPartialTranscript(cs, _ep, evt) {
+    if (!this.playComplete) {
+      const words = evt.delta.split(' ').filter((w) => /[A-Za-z0-0]/.test(w));
+      this.openaiPartials.push(...words);
+      this.logger.debug({words, partials: this.openaiPartials, evt}, 'TaskGather:_onOpenAIPartialTranscript - words');
+      if (this.openaiPartials.length >= this.minBargeinWordCount) {
+        this.logger.debug({partials: this.openaiPartials}, 'killing audio due to speech (openai)');
+        this._killAudio(cs);
+        this.notifyStatus({event: 'speech-bargein-detected', words: this.openaiPartials});
+      }
+    }
+  }
+
   async _onVendorError(cs, _ep, evt) {
     super._onVendorError(cs, _ep, evt);
     if (!(await this._startFallback(cs, _ep, evt))) {
       this._resolve('stt-error', evt);
     }
   }
+  async _onSpeechmaticsRecognitionStarted(_cs, _ep, evt) {
+    this.logger.debug({evt}, 'TaskGather:_onSpeechmaticsRecognitionStarted');
+  }
+
+  async _onSpeechmaticsInfo(_cs, _ep, evt) {
+    this.logger.debug({evt}, 'TaskGather:_onSpeechmaticsInfo');
+  }
 
   _onVadDetected(cs, ep) {
     if (this.bargein && this.minBargeinWordCount === 0) {
@@ -1013,7 +1208,7 @@ class TaskGather extends SttTask {
   }
 
   async _resolve(reason, evt) {
-    this.logger.debug(`TaskGather:resolve with reason ${reason}`);
+    this.logger.info({evt}, `TaskGather:resolve with reason ${reason}`);
     if (this.needsStt && this.ep && this.ep.connected) {
       this.ep.stopTranscription({
         vendor: this.vendor,
@@ -1043,12 +1238,13 @@ class TaskGather extends SttTask {
 
     this.span.setAttributes({
       channel: 1,
+      'stt.label': this.label || 'None',
       'stt.resolve': reason,
       'stt.result': JSON.stringify(evt)
     });
 
     if (this.callSession && this.callSession.callGone) {
-      this.logger.debug('TaskGather:_resolve - call is gone, not invoking web callback');
+      this.logger.info('TaskGather:_resolve - call is gone, not invoking web callback');
       this.notifyTaskDone();
       return;
     }
@@ -1080,6 +1276,7 @@ class TaskGather extends SttTask {
         }
       }
       else if (reason.startsWith('speech')) {
+        this.cs.emit('userSaid', evt.alternatives[0].transcript);
         if (this.parentTask) this.parentTask.emit('transcription', evt);
         else {
           this.emit('transcription', evt);
@@ -1105,13 +1302,12 @@ class TaskGather extends SttTask {
         if (this.parentTask) this.parentTask.emit('stt-low-confidence', evt);
         else {
           this.emit('stt-low-confidence', evt);
-          returnedVerbs = await this.performAction({reason: 'stt-low-confidence'});
+          returnedVerbs = await this.performAction({speech:evt, reason: 'stt-low-confidence'});
         }
       }
     } catch (err) {  /*already logged error*/ }
 
     // Gather got response from hook, cancel actionHookDelay processing
-    this.logger.debug('TaskGather:_resolve - checking ahd');
     if (this.cs.actionHookDelayProcessor) {
       if (returnedVerbs) {
         this.logger.debug('TaskGather:_resolve - got response from action hook, cancelling actionHookDelay');

lib/tasks/listen.js

@@ -17,7 +17,7 @@ class TaskListen extends Task {
 
     [
       'action', 'auth', 'method', 'url', 'finishOnKey', 'maxLength', 'metadata', 'mixType', 'passDtmf', 'playBeep',
-      'sampleRate', 'timeout', 'transcribe', 'wsAuth', 'disableBidirectionalAudio'
+      'sampleRate', 'timeout', 'transcribe', 'wsAuth', 'disableBidirectionalAudio', 'channel'
     ].forEach((k) => this[k] = this.data[k]);
 
     this.mixType = this.mixType || 'mono';
@@ -221,7 +221,7 @@ class TaskListen extends Task {
     }
   }
   _onConnect(ep) {
-    this.logger.debug('TaskListen:_onConnect');
+    this.logger.info('TaskListen:_onConnect');
   }
   _onConnectFailure(ep, evt) {
     this.logger.info(evt, 'TaskListen:_onConnectFailure');

lib/tasks/llm/index.js

@@ -0,0 +1,144 @@
+const Task = require('../task');
+const {TaskPreconditions} = require('../../utils/constants');
+const TaskLlmOpenAI_S2S = require('./llms/openai_s2s');
+const TaskLlmVoiceAgent_S2S = require('./llms/voice_agent_s2s');
+const TaskLlmUltravox_S2S = require('./llms/ultravox_s2s');
+const TaskLlmElevenlabs_S2S = require('./llms/elevenlabs_s2s');
+const TaskLlmGoogle_S2S = require('./llms/google_s2s');
+const LlmMcpService = require('../../utils/llm-mcp');
+
+class TaskLlm extends Task {
+  constructor(logger, opts) {
+    super(logger, opts);
+    this.preconditions = TaskPreconditions.Endpoint;
+
+    ['vendor', 'model', 'auth', 'connectOptions'].forEach((prop) => {
+      this[prop] = this.data[prop];
+    });
+
+    this.eventHandlers = [];
+
+    // delegate to the specific llm model
+    this.llm = this.createSpecificLlm();
+    // MCP
+    this.mcpServers = this.data.mcpServers || [];
+  }
+
+  get name() { return this.llm.name ; }
+
+  get toolHook() { return this.llm?.toolHook; }
+
+  get eventHook() { return this.llm?.eventHook; }
+
+  get ep() { return this.cs.ep; }
+
+  get mcpService() {
+    return this.llmMcpService;
+  }
+
+  get isMcpEnabled() {
+    return this.mcpServers.length > 0;
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs, {ep});
+
+    // create the MCP service if we have MCP servers
+    if (this.isMcpEnabled) {
+      this.llmMcpService = new LlmMcpService(this.logger, this.mcpServers);
+      await this.llmMcpService.init();
+    }
+    await this.llm.exec(cs, {ep});
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+    await this.llm.kill(cs);
+    // clean up MCP clients
+    if (this.isMcpEnabled) {
+      await this.mcpService.close();
+    }
+  }
+
+  createSpecificLlm() {
+    let llm;
+    switch (this.vendor) {
+      case 'openai':
+      case 'microsoft':
+        llm = new TaskLlmOpenAI_S2S(this.logger, this.data, this);
+        break;
+
+      case 'voiceagent':
+      case 'deepgram':
+        llm = new TaskLlmVoiceAgent_S2S(this.logger, this.data, this);
+        break;
+
+      case 'ultravox':
+        llm = new TaskLlmUltravox_S2S(this.logger, this.data, this);
+        break;
+
+      case 'elevenlabs':
+        llm = new TaskLlmElevenlabs_S2S(this.logger, this.data, this);
+        break;
+
+      case 'google':
+        llm = new TaskLlmGoogle_S2S(this.logger, this.data, this);
+        break;
+
+      default:
+        throw new Error(`Unsupported vendor ${this.vendor} for LLM`);
+    }
+
+    if (!llm) {
+      throw new Error(`Unsupported vendor:model ${this.vendor}:${this.model}`);
+    }
+    return llm;
+  }
+
+  addCustomEventListener(ep, event, handler) {
+    this.eventHandlers.push({ep, event, handler});
+    ep.addCustomEventListener(event, handler);
+  }
+
+  removeCustomEventListeners() {
+    this.eventHandlers.forEach((h) => h.ep.removeCustomEventListener(h.event, h.handler));
+  }
+
+  async sendEventHook(data) {
+    await this.cs?.requestor.request('llm:event', this.eventHook, data);
+  }
+
+
+  async sendToolHook(tool_call_id, data) {
+    const tool_response = await this.cs?.requestor.request('llm:tool-call', this.toolHook, {tool_call_id, ...data});
+    // if the toolHook was a websocket it will return undefined, otherwise it should return an object
+    if (typeof tool_response != 'undefined') {
+      tool_response.type = 'client_tool_result';
+      tool_response.invocation_id = tool_call_id;
+      this.processToolOutput(tool_call_id, tool_response);
+    }
+  }
+
+  async processToolOutput(tool_call_id, data) {
+    if (!this.ep.connected) {
+      this.logger.info('TaskLlm:processToolOutput - no connected endpoint');
+      return;
+    }
+    this.llm.processToolOutput(this.ep, tool_call_id, data);
+  }
+
+  async processLlmUpdate(data, callSid) {
+    if (this.ep.connected) {
+      if (typeof this.llm.processLlmUpdate === 'function') {
+        this.llm.processLlmUpdate(this.ep, data, callSid);
+      }
+      else {
+        const {vendor, model} = this.llm;
+        this.logger.info({data, callSid},
+          `TaskLlm:_processLlmUpdate: LLM ${vendor}:${model} does not support llm:update`);
+      }
+    }
+  }
+}
+
+module.exports = TaskLlm;

lib/tasks/llm/llms/elevenlabs_s2s.js

@@ -0,0 +1,327 @@
+const Task = require('../../task');
+const TaskName = 'Llm_Elevenlabs_s2s';
+const {LlmEvents_Elevenlabs} = require('../../../utils/constants');
+const {request} = require('undici');
+const ClientEvent = 'client.event';
+const SessionDelete = 'session.delete';
+
+const elevenlabs_server_events = [
+  'conversation_initiation_metadata',
+  'user_transcript',
+  'agent_response',
+  'client_tool_call'
+];
+
+const expandWildcards = (events) => {
+  const expandedEvents = [];
+
+  events.forEach((evt) => {
+    if (evt.endsWith('.*')) {
+      const prefix = evt.slice(0, -2); // Remove the wildcard ".*"
+      const matchingEvents = elevenlabs_server_events.filter((e) => e.startsWith(prefix));
+      expandedEvents.push(...matchingEvents);
+    } else {
+      expandedEvents.push(evt);
+    }
+  });
+
+  return expandedEvents;
+};
+
+class TaskLlmElevenlabs_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.auth = this.parent.auth;
+
+    const {agent_id, api_key} = this.auth || {};
+    if (!agent_id) throw new Error('auth.agent_id is required for Elevenlabs S2S');
+
+    this.agent_id = agent_id;
+    this.api_key = api_key;
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+    const {
+      conversation_initiation_client_data,
+      input_sample_rate = 16000,
+      output_sample_rate = 16000
+    } = this.data.llmOptions;
+    this.conversation_initiation_client_data = conversation_initiation_client_data;
+    this.input_sample_rate = input_sample_rate;
+    this.output_sample_rate = output_sample_rate;
+    this.results = {
+      completionReason: 'normal conversation end'
+    };
+
+    /**
+     * only one of these will have items,
+     * if includeEvents, then these are the events to include
+     * if excludeEvents, then these are the events to exclude
+     */
+    this.includeEvents = [];
+    this.excludeEvents = [];
+
+    /* default to all events if user did not specify */
+    this._populateEvents(this.data.events || elevenlabs_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  async getSignedUrl() {
+    if (!this.api_key) {
+      return {
+        host: 'api.elevenlabs.io',
+        path: `/v1/convai/conversation?agent_id=${this.agent_id}`,
+      };
+    }
+
+    const {statusCode, body} = await request(
+      `https://api.elevenlabs.io/v1/convai/conversation/get_signed_url?agent_id=${this.agent_id}`, {
+        method: 'GET',
+        headers: {
+          'xi-api-key': this.api_key
+        },
+      }
+    );
+    const data = await body.json();
+    if (statusCode !== 200 || !data?.signed_url) {
+      this.logger.error({statusCode, data}, 'Elevenlabs Error registering call');
+      throw new Error(`Elevenlabs  Error registering call: ${data.message}`);
+    }
+
+    const url = new URL(data.signed_url);
+    return {
+      host: url.hostname,
+      path: url.pathname + url.search,
+    };
+  }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_elevenlabs_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error({args}, `Error calling uuid_elevenlabs_s2s: ${res.body}`);
+    }
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs);
+    await this._startListening(cs, ep);
+
+    await this.awaitTaskDone();
+
+    /* note: the parent llm verb started the span, which is why this is necessary */
+    await this.parent.performAction(this.results);
+
+    this._unregisterHandlers();
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+
+    this._api(cs.ep, [cs.ep.uuid, SessionDelete])
+      .catch((err) => this.logger.info({err}, 'TaskLlmElevenlabs_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  /**
+   * Send function call output to the Elevenlabs server in the form of conversation.item.create
+   * per https://elevenlabs.io/docs/conversational-ai/api-reference/conversational-ai/websocket
+   */
+  async processToolOutput(ep, tool_call_id, rawData) {
+    try {
+      const {data} = rawData;
+      this.logger.debug({tool_call_id, data}, 'TaskLlmElevenlabs_S2S:processToolOutput');
+
+      if (!data.type || data.type !== 'client_tool_result') {
+        this.logger.info({data},
+          'TaskLlmElevenlabs_S2S:processToolOutput - invalid tool output, must be client_tool_result');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmElevenlabs_S2S:processToolOutput');
+    }
+  }
+
+  /**
+   * Send a session.update to the Elevenlabs server
+   * Note: creating and deleting conversation items also supported as well as interrupting the assistant
+   */
+  async processLlmUpdate(ep, data, _callSid) {
+    this.logger.debug({data, _callSid}, 'TaskLlmElevenlabs_S2S:processLlmUpdate, ignored');
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    try {
+      const {host, path} = await this.getSignedUrl();
+      const args = this.conversation_initiation_client_data ?
+        [ep.uuid, 'session.create', this.input_sample_rate, this.output_sample_rate, host, path] :
+        [ep.uuid, 'session.create', this.input_sample_rate, this.output_sample_rate, host, path, 'no_initial_config'];
+      await this._api(ep, args);
+    } catch (err) {
+      this.logger.error({err}, 'TaskLlmElevenlabs_S2S:_startListening');
+      this.notifyTaskDone();
+    }
+  }
+
+  async _sendClientEvent(ep, obj) {
+    let ok = true;
+    this.logger.debug({obj}, 'TaskLlmElevenlabs_S2S:_sendClientEvent');
+    try {
+      const args = [ep.uuid, ClientEvent, JSON.stringify(obj)];
+      await this._api(ep, args);
+    } catch (err) {
+      ok = false;
+      this.logger.error({err}, 'TaskLlmElevenlabs_S2S:_sendClientEvent - Error');
+    }
+    return ok;
+  }
+
+  async _sendInitialMessage(ep) {
+    if (this.conversation_initiation_client_data) {
+      if (!await this._sendClientEvent(ep, {
+        type: 'conversation_initiation_client_data',
+        ...this.conversation_initiation_client_data
+      })) {
+        this.notifyTaskDone();
+      }
+    }
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Elevenlabs.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _onError(ep, evt) {
+    this.logger.info({evt}, 'TaskLlmElevenlabs_S2S:_onError');
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.debug('TaskLlmElevenlabs_S2S:_onConnect');
+    this._sendInitialMessage(ep);
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmElevenlabs_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmElevenlabs_S2S:_onConnectFailure');
+    this.results = {completionReason: 'disconnect from remote end'};
+    this.notifyTaskDone();
+  }
+  async _onServerEvent(ep, evt) {
+    let endConversation = false;
+    const type = evt.type;
+    this.logger.info({evt}, 'TaskLlmElevenlabs_S2S:_onServerEvent');
+
+    if (type === 'error') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server error',
+        error: evt.error
+      };
+    }
+
+    /* tool calls */
+    else if (type === 'client_tool_call') {
+      this.logger.debug({evt}, 'TaskLlmElevenlabs_S2S:_onServerEvent - function_call');
+      const {tool_name: name, tool_call_id: call_id, parameters: args} = evt.client_tool_call;
+
+      const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+      if (mcpTools.some((tool) => tool.name === name)) {
+        this.logger.debug({name, args}, 'TaskLlmElevenlabs_S2S:_onServerEvent - calling mcp tool');
+        try {
+          const res = await this.parent.mcpService.callMcpTool(name, args);
+          this.logger.debug({res}, 'TaskLlmElevenlabs_S2S:_onServerEvent - function_call - mcp result');
+          this.processToolOutput(ep, call_id, {
+            data: {
+              type: 'client_tool_result',
+              tool_call_id: call_id,
+              result: res.content?.length ? res.content[0] : res.content,
+              is_error: false
+            }
+          });
+          return;
+        }
+        catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmElevenlabs_S2S - error calling mcp tool');
+          this.results = {
+            completionReason: 'client error calling mcp function',
+            error: err
+          };
+          endConversation = true;
+        }
+      } else if (!this.toolHook) {
+        this.logger.warn({evt}, 'TaskLlmElevenlabs_S2S:_onServerEvent - no toolHook defined!');
+      }
+      else {
+        try {
+          await this.parent.sendToolHook(call_id, {name, args});
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmElevenlabs_S2S - error calling function');
+          this.results = {
+            completionReason: 'client error calling function',
+            error: err
+          };
+          endConversation = true;
+        }
+      }
+    }
+
+    /* check whether we should notify on this event */
+    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
+      this.parent.sendEventHook(evt)
+        .catch((err) => this.logger.info({err},
+          'TaskLlmElevenlabs_S2S:_onServerEvent - error sending event hook'));
+    }
+
+    if (endConversation) {
+      this.logger.info({results: this.results},
+        'TaskLlmElevenlabs_S2S:_onServerEvent - ending conversation due to error');
+      this.notifyTaskDone();
+    }
+  }
+
+  _populateEvents(events) {
+    if (events.includes('all')) {
+      /* work by excluding specific events */
+      const exclude = events
+        .filter((evt) => evt.startsWith('-'))
+        .map((evt) => evt.slice(1));
+      if (exclude.length === 0) this.includeEvents = elevenlabs_server_events;
+      else this.excludeEvents = expandWildcards(exclude);
+    }
+    else {
+      /* work by including specific events */
+      const include = events
+        .filter((evt) => !evt.startsWith('-'));
+      this.includeEvents = expandWildcards(include);
+    }
+
+    this.logger.debug({
+      includeEvents: this.includeEvents,
+      excludeEvents: this.excludeEvents
+    }, 'TaskLlmElevenlabs_S2S:_populateEvents');
+  }
+}
+
+module.exports = TaskLlmElevenlabs_S2S;

lib/tasks/llm/llms/google_s2s.js

@@ -0,0 +1,319 @@
+const Task = require('../../task');
+const TaskName = 'Llm_Google_s2s';
+const {LlmEvents_Google} = require('../../../utils/constants');
+const ClientEvent = 'client.event';
+const SessionDelete = 'session.delete';
+
+const google_server_events = [
+  'error',
+  'session.created',
+  'session.updated',
+];
+
+const expandWildcards = (events) => {
+  const expandedEvents = [];
+
+  events.forEach((evt) => {
+    if (evt.endsWith('.*')) {
+      const prefix = evt.slice(0, -2); // Remove the wildcard ".*"
+      const matchingEvents = google_server_events.filter((e) => e.startsWith(prefix));
+      expandedEvents.push(...matchingEvents);
+    } else {
+      expandedEvents.push(evt);
+    }
+  });
+
+  return expandedEvents;
+};
+
+class TaskLlmGoogle_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.vendor = this.parent.vendor;
+    this.model = this.parent.model || 'models/gemini-2.0-flash-live-001';
+    this.auth = this.parent.auth;
+    this.connectionOptions = this.parent.connectOptions;
+
+    const {apiKey} = this.auth || {};
+    if (!apiKey) throw new Error('auth.apiKey is required for Google S2S');
+
+    this.apiKey = apiKey;
+
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+
+    const {setup} = this.data.llmOptions;
+
+    if (typeof setup !== 'object') {
+      throw new Error('llmOptions with an initial setup is required for Google S2S');
+    }
+    this.setup = {
+      ...setup,
+      model: this.model,
+      // make sure output is always audio
+      generationConfig: {
+        ...(setup.generationConfig || {}),
+        responseModalities: 'audio'
+      }
+    };
+
+    this.results = {
+      completionReason: 'normal conversation end'
+    };
+
+    /**
+     * only one of these will have items,
+     * if includeEvents, then these are the events to include
+     * if excludeEvents, then these are the events to exclude
+     */
+    this.includeEvents = [];
+    this.excludeEvents = [];
+
+    /* default to all events if user did not specify */
+    this._populateEvents(this.data.events || google_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_google_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error({args}, `Error calling uuid_openai_s2s: ${res.body}`);
+    }
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs);
+
+    await this._startListening(cs, ep);
+
+    await this.awaitTaskDone();
+
+    /* note: the parent llm verb started the span, which is why this is necessary */
+    await this.parent.performAction(this.results);
+
+    this._unregisterHandlers();
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+
+    this._api(cs.ep, [cs.ep.uuid, SessionDelete])
+      .catch((err) => this.logger.info({err}, 'TaskLlmGoogle_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  _populateEvents(events) {
+    if (events.includes('all')) {
+      /* work by excluding specific events */
+      const exclude = events
+        .filter((evt) => evt.startsWith('-'))
+        .map((evt) => evt.slice(1));
+      if (exclude.length === 0) this.includeEvents = google_server_events;
+      else this.excludeEvents = expandWildcards(exclude);
+    }
+    else {
+      /* work by including specific events */
+      const include = events
+        .filter((evt) => !evt.startsWith('-'));
+      this.includeEvents = expandWildcards(include);
+    }
+
+    this.logger.debug({
+      includeEvents: this.includeEvents,
+      excludeEvents: this.excludeEvents
+    }, 'TaskLlmGoogle_S2S:_populateEvents');
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    try {
+      const args = [ep.uuid, 'session.create', this.apiKey];
+      await this._api(ep, args);
+    } catch (err) {
+      this.logger.error({err}, 'TaskLlmGoogle_S2S:_startListening');
+      this.notifyTaskDone();
+    }
+  }
+
+  async _sendClientEvent(ep, obj) {
+    let ok = true;
+    this.logger.debug({obj}, 'TaskLlmGoogle_S2S:_sendClientEvent');
+    try {
+      const args = [ep.uuid, ClientEvent, JSON.stringify(obj)];
+      await this._api(ep, args);
+    } catch (err) {
+      ok = false;
+      this.logger.error({err}, 'TaskLlmGoogle_S2S:_sendClientEvent - Error');
+    }
+    return ok;
+  }
+
+  async _sendInitialMessage(ep) {
+    const setup = this.setup;
+    const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+    if (mcpTools && mcpTools.length > 0) {
+      const convertedTools = [
+        {
+          functionDeclarations: mcpTools.map((tool) => {
+            if (tool.inputSchema) {
+              delete tool.inputSchema.additionalProperties;
+              delete tool.inputSchema['$schema'];
+            }
+            return {
+              name: tool.name,
+              description: tool.description,
+              parameters: tool.inputSchema,
+            };
+          })
+        }
+      ];
+      // merge with any existing tools
+      setup.tools = [...convertedTools, ...(this.setup.tools || [])];
+    }
+    if (!await this._sendClientEvent(ep, {
+      setup,
+    })) {
+      this.logger.debug(this.setup, 'TaskLlmGoogle_S2S:_sendInitialMessage - sending session.update');
+      this.notifyTaskDone();
+    }
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_Google.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Google.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Google.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Google.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _onError(ep, evt) {
+    this.logger.info({evt}, 'TaskLlmGoogle_S2S:_onError');
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.debug('TaskLlmGoogle_S2S:_onConnect');
+    this._sendInitialMessage(ep);
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmGoogle_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmGoogle_S2S:_onConnectFailure');
+    this.results = {completionReason: 'disconnect from remote end'};
+    this.notifyTaskDone();
+  }
+
+  async _onServerEvent(ep, evt) {
+    let endConversation = false;
+    this.logger.debug({evt}, 'TaskLlmGoogle_S2S:_onServerEvent');
+    const {toolCall /**toolCallCancellation*/}  = evt;
+
+    if (toolCall) {
+      this.logger.debug({toolCall}, 'TaskLlmGoogle_S2S:_onServerEvent - toolCall');
+      if (!this.toolHook) {
+        this.logger.info({evt}, 'TaskLlmGoogle_S2S:_onServerEvent - no toolHook defined!');
+      }
+      else {
+        const {functionCalls} = toolCall;
+        const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+        const functionResponses = [];
+        if (mcpTools && mcpTools.length > 0) {
+          for (const functionCall of functionCalls) {
+            const {name, args, id} = functionCall;
+            const tool = mcpTools.find((tool) => tool.name === name);
+            if (tool) {
+              const response = await this.parent.mcpService.callMcpTool(name, args);
+              functionResponses.push({
+                response: {
+                  output: response,
+                },
+                id
+              });
+            }
+          }
+        }
+
+        if (functionResponses && functionResponses.length > 0) {
+          this.logger.debug({functionResponses}, 'TaskLlmGoogle_S2S:_onServerEvent - function_call - mcp result');
+          this.processToolOutput(ep, 'tool_call_id', {
+            toolResponse: {
+              functionResponses
+            }
+          });
+        } else {
+          try {
+            await this.parent.sendToolHook('function_call_id', {type: 'toolCall', functionCalls});
+          } catch (err) {
+            this.logger.info({err, evt}, 'TaskLlmGoogle_S2S - error calling function');
+            this.results = {
+              completionReason: 'client error calling function',
+              error: err
+            };
+            endConversation = true;
+          }
+        }
+      }
+    }
+
+    this._sendLlmEvent('llm_event', evt);
+
+    if (endConversation) {
+      this.logger.info({results: this.results},
+        'TaskLlmGoogle_S2S:_onServerEvent - ending conversation due to error');
+      this.notifyTaskDone();
+    }
+  }
+
+  _sendLlmEvent(type, evt) {
+    /* check whether we should notify on this event */
+    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
+      this.parent.sendEventHook(evt)
+        .catch((err) => this.logger.info({err}, 'TaskLlmGoogle_S2S:_onServerEvent - error sending event hook'));
+    }
+  }
+
+  async processLlmUpdate(ep, data, _callSid) {
+    try {
+      this.logger.debug({data, _callSid}, 'TaskLlmGoogle_S2S:processLlmUpdate');
+
+      await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+    } catch (err) {
+      this.logger.info({err, data}, 'TaskLlmGoogle_S2S:processLlmUpdate - Error processing LLM update');
+    }
+  }
+
+  async processToolOutput(ep, tool_call_id, data) {
+    try {
+      this.logger.debug({tool_call_id, data}, 'TaskLlmGoogle_S2S:processToolOutput');
+      const {toolResponse} = data;
+
+      if (!toolResponse) {
+        this.logger.info({data},
+          'TaskLlmGoogle_S2S:processToolOutput - invalid tool output, must be functionResponses');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err, data}, 'TaskLlmGoogle_S2S:processToolOutput - Error processing tool output');
+    }
+  }
+}
+
+module.exports = TaskLlmGoogle_S2S;

lib/tasks/llm/llms/openai_s2s.js

@@ -0,0 +1,398 @@
+const Task = require('../../task');
+const TaskName = 'Llm_OpenAI_s2s';
+const {LlmEvents_OpenAI} = require('../../../utils/constants');
+const ClientEvent = 'client.event';
+const SessionDelete = 'session.delete';
+
+const openai_server_events = [
+  'error',
+  'session.created',
+  'session.updated',
+  'conversation.created',
+  'input_audio_buffer.committed',
+  'input_audio_buffer.cleared',
+  'input_audio_buffer.speech_started',
+  'input_audio_buffer.speech_stopped',
+  'conversation.item.created',
+  'conversation.item.input_audio_transcription.completed',
+  'conversation.item.input_audio_transcription.failed',
+  'conversation.item.truncated',
+  'conversation.item.deleted',
+  'response.created',
+  'response.done',
+  'response.output_item.added',
+  'response.output_item.done',
+  'response.content_part.added',
+  'response.content_part.done',
+  'response.text.delta',
+  'response.text.done',
+  'response.audio_transcript.delta',
+  'response.audio_transcript.done',
+  'response.audio.delta',
+  'response.audio.done',
+  'response.function_call_arguments.delta',
+  'response.function_call_arguments.done',
+  'rate_limits.updated',
+  'output_audio.playback_started',
+  'output_audio.playback_stopped',
+];
+
+const expandWildcards = (events) => {
+  const expandedEvents = [];
+
+  events.forEach((evt) => {
+    if (evt.endsWith('.*')) {
+      const prefix = evt.slice(0, -2); // Remove the wildcard ".*"
+      const matchingEvents = openai_server_events.filter((e) => e.startsWith(prefix));
+      expandedEvents.push(...matchingEvents);
+    } else {
+      expandedEvents.push(evt);
+    }
+  });
+
+  return expandedEvents;
+};
+
+class TaskLlmOpenAI_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.model = this.parent.model || 'gpt-4o-realtime-preview-2024-12-17';
+    this.auth = this.parent.auth;
+    this.connectionOptions = this.parent.connectOptions;
+
+    const {apiKey} = this.auth || {};
+    if (!apiKey) throw new Error('auth.apiKey is required for OpenAI S2S');
+
+    if (['openai', 'microsoft'].indexOf(this.vendor) === -1) {
+      throw new Error(`Invalid vendor ${this.vendor} for OpenAI S2S`);
+    }
+
+    if ('microsoft' === this.vendor && !this.connectionOptions?.host) {
+      throw new Error('connectionOptions.host is required for Microsoft OpenAI S2S');
+    }
+
+    this.apiKey = apiKey;
+    this.authType = 'microsoft' === this.vendor ? 'query' : 'bearer';
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+    const {response_create, session_update} = this.data.llmOptions;
+
+    if (typeof response_create !== 'object') {
+      throw new Error('llmOptions with an initial response.create is required for OpenAI S2S');
+    }
+
+    this.response_create = response_create;
+    this.session_update = session_update;
+
+    this.results = {
+      completionReason: 'normal conversation end'
+    };
+
+    /**
+     * only one of these will have items,
+     * if includeEvents, then these are the events to include
+     * if excludeEvents, then these are the events to exclude
+     */
+    this.includeEvents = [];
+    this.excludeEvents = [];
+
+    /* default to all events if user did not specify */
+    this._populateEvents(this.data.events || openai_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  get host() {
+    const {host} = this.connectionOptions || {};
+    return host || (this.vendor === 'openai' ? 'api.openai.com' : void 0);
+  }
+
+  get path() {
+    const {path} = this.connectionOptions || {};
+    if (path) return path;
+
+    switch (this.vendor) {
+      case 'openai':
+        return `v1/realtime?model=${this.model}`;
+      case 'microsoft':
+        return `openai/realtime?api-version=2024-10-01-preview&deployment=${this.model}`;
+    }
+  }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_openai_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error({args}, `Error calling uuid_openai_s2s: ${res.body}`);
+    }
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs);
+
+    await this._startListening(cs, ep);
+
+    await this.awaitTaskDone();
+
+    /* note: the parent llm verb started the span, which is why this is necessary */
+    await this.parent.performAction(this.results);
+
+    this._unregisterHandlers();
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+
+    this._api(cs.ep, [cs.ep.uuid, SessionDelete])
+      .catch((err) => this.logger.info({err}, 'TaskLlmOpenAI_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  /**
+   * Send function call output to the OpenAI server in the form of conversation.item.create
+   * per https://platform.openai.com/docs/guides/realtime/function-calls
+   */
+  async processToolOutput(ep, tool_call_id, data) {
+    try {
+      this.logger.debug({tool_call_id, data}, 'TaskLlmOpenAI_S2S:processToolOutput');
+
+      if (!data.type || data.type !== 'conversation.item.create') {
+        this.logger.info({data},
+          'TaskLlmOpenAI_S2S:processToolOutput - invalid tool output, must be conversation.item.create');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+
+        // spec also recommends to send immediate response.create
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify({type: 'response.create'})]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmOpenAI_S2S:processToolOutput');
+    }
+  }
+
+  /**
+   * Send a session.update to the OpenAI server
+   * Note: creating and deleting conversation items also supported as well as interrupting the assistant
+   */
+  async processLlmUpdate(ep, data, _callSid) {
+    try {
+      this.logger.debug({data, _callSid}, 'TaskLlmOpenAI_S2S:processLlmUpdate');
+
+      if (!data.type || ![
+        'session.update',
+        'conversation.item.create',
+        'conversation.item.delete',
+        'response.cancel'
+      ].includes(data.type)) {
+        this.logger.info({data}, 'TaskLlmOpenAI_S2S:processLlmUpdate - invalid mid-call request');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmOpenAI_S2S:processLlmUpdate');
+    }
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    try {
+      const args = [ep.uuid, 'session.create', this.host, this.path, this.authType, this.apiKey];
+      await this._api(ep, args);
+    } catch (err) {
+      this.logger.error({err}, 'TaskLlmOpenAI_S2S:_startListening');
+      this.notifyTaskDone();
+    }
+  }
+
+  async _sendClientEvent(ep, obj) {
+    let ok = true;
+    this.logger.debug({obj}, 'TaskLlmOpenAI_S2S:_sendClientEvent');
+    try {
+      const args = [ep.uuid, ClientEvent, JSON.stringify(obj)];
+      await this._api(ep, args);
+    } catch (err) {
+      ok = false;
+      this.logger.error({err}, 'TaskLlmOpenAI_S2S:_sendClientEvent - Error');
+    }
+    return ok;
+  }
+
+  async _sendInitialMessage(ep) {
+    let obj = {type: 'response.create', response: this.response_create};
+    if (!await this._sendClientEvent(ep, obj)) {
+      this.notifyTaskDone();
+    }
+
+    /* send immediate session.update if present */
+    else if (this.session_update) {
+      if (this.parent.isMcpEnabled) {
+        this.logger.debug('TaskLlmOpenAI_S2S:_sendInitialMessage - mcp enabled');
+        const tools = await this.parent.mcpService.getAvailableMcpTools();
+        if (tools && tools.length > 0 && this.session_update) {
+          const convertedTools = tools.map((tool) => ({
+            name: tool.name,
+            type: 'function',
+            description: tool.description,
+            parameters: tool.inputSchema
+          }));
+
+          this.session_update.tools = [
+            ...convertedTools,
+            ...(this.session_update.tools || [])
+          ];
+        }
+      }
+      obj = {type: 'session.update', session: this.session_update};
+      this.logger.debug({obj}, 'TaskLlmOpenAI_S2S:_sendInitialMessage - sending session.update');
+      if (!await this._sendClientEvent(ep, obj)) {
+        this.notifyTaskDone();
+      }
+    }
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_OpenAI.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_OpenAI.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_OpenAI.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_OpenAI.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _onError(ep, evt) {
+    this.logger.info({evt}, 'TaskLlmOpenAI_S2S:_onError');
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.debug('TaskLlmOpenAI_S2S:_onConnect');
+    this._sendInitialMessage(ep);
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmOpenAI_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmOpenAI_S2S:_onConnectFailure');
+    this.results = {completionReason: 'disconnect from remote end'};
+    this.notifyTaskDone();
+  }
+  async _onServerEvent(ep, evt) {
+    let endConversation = false;
+    const type = evt.type;
+    this.logger.info({evt}, 'TaskLlmOpenAI_S2S:_onServerEvent');
+
+    /* check for failures, such as rate limit exceeded, that should terminate the conversation */
+    if (type === 'response.done' && evt.response.status === 'failed') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server failure',
+        error: evt.response.status_details?.error
+      };
+    }
+
+    /* server errors of some sort */
+    else if (type === 'error') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server error',
+        error: evt.error
+      };
+    }
+
+    /* tool calls */
+    else if (type === 'response.output_item.done' && evt.item?.type === 'function_call') {
+      this.logger.debug({evt}, 'TaskLlmOpenAI_S2S:_onServerEvent - function_call');
+      const {name, call_id} = evt.item;
+      const args = JSON.parse(evt.item.arguments);
+
+      const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+      if (mcpTools.some((tool) => tool.name === name)) {
+        this.logger.debug({call_id, name, args}, 'TaskLlmOpenAI_S2S:_onServerEvent - calling mcp tool');
+        try {
+          const res = await this.parent.mcpService.callMcpTool(name, args);
+          this.logger.debug({res}, 'TaskLlmOpenAI_S2S:_onServerEvent - function_call - mcp result');
+          this.processToolOutput(ep, call_id, {
+            type: 'conversation.item.create',
+            item: {
+              type: 'function_call_output',
+              call_id,
+              output: res.content[0]?.text || 'There is no output from the function call',
+            }
+          });
+          return;
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmOpenAI_S2S - error calling function');
+          this.results = {
+            completionReason: 'client error calling mcp function',
+            error: err
+          };
+          endConversation = true;
+        }
+      }
+      else if (!this.toolHook) {
+        this.logger.warn({evt}, 'TaskLlmOpenAI_S2S:_onServerEvent - no toolHook defined!');
+      }
+      else {
+        try {
+          await this.parent.sendToolHook(call_id, {name, args});
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmOpenAI - error calling function');
+          this.results = {
+            completionReason: 'client error calling function',
+            error: err
+          };
+          endConversation = true;
+        }
+      }
+    }
+
+    /* check whether we should notify on this event */
+    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
+      this.parent.sendEventHook(evt)
+        .catch((err) => this.logger.info({err}, 'TaskLlmOpenAI_S2S:_onServerEvent - error sending event hook'));
+    }
+
+    if (endConversation) {
+      this.logger.info({results: this.results}, 'TaskLlmOpenAI_S2S:_onServerEvent - ending conversation due to error');
+      this.notifyTaskDone();
+    }
+  }
+
+  _populateEvents(events) {
+    if (events.includes('all')) {
+      /* work by excluding specific events */
+      const exclude = events
+        .filter((evt) => evt.startsWith('-'))
+        .map((evt) => evt.slice(1));
+      if (exclude.length === 0) this.includeEvents = openai_server_events;
+      else this.excludeEvents = expandWildcards(exclude);
+    }
+    else {
+      /* work by including specific events */
+      const include = events
+        .filter((evt) => !evt.startsWith('-'));
+      this.includeEvents = expandWildcards(include);
+    }
+
+    this.logger.debug({
+      includeEvents: this.includeEvents,
+      excludeEvents: this.excludeEvents
+    }, 'TaskLlmOpenAI_S2S:_populateEvents');
+  }
+}
+
+module.exports = TaskLlmOpenAI_S2S;

lib/tasks/llm/llms/ultravox_s2s.js

@@ -0,0 +1,344 @@
+const Task = require('../../task');
+const TaskName = 'Llm_Ultravox_s2s';
+const {request} = require('undici');
+const {LlmEvents_Ultravox} = require('../../../utils/constants');
+
+const ultravox_server_events = [
+  'createCall',
+  'pong',
+  'state',
+  'transcript',
+  'conversationText',
+  'clientToolInvocation',
+  'playbackClearBuffer',
+];
+
+const ClientEvent = 'client.event';
+
+const expandWildcards = (events) => {
+  // no-op for deepgram
+  return events;
+};
+
+const SessionDelete = 'session.delete';
+
+class TaskLlmUltravox_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.model = this.parent.model || 'fixie-ai/ultravox';
+    this.auth = this.parent.auth;
+    this.connectionOptions = this.parent.connectOptions;
+
+    const {apiKey} = this.auth || {};
+    if (!apiKey) throw new Error('auth.apiKey is required for Vendor: Ultravox');
+    this.apiKey = apiKey;
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+
+    this.results = {
+      completionReason: 'normal conversation end'
+    };
+
+    /**
+     * only one of these will have items,
+     * if includeEvents, then these are the events to include
+     * if excludeEvents, then these are the events to exclude
+     */
+    this.includeEvents = [];
+    this.excludeEvents = [];
+
+    /* default to all events if user did not specify */
+    this._populateEvents(this.data.events || ultravox_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_ultravox_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error(`Error calling uuid_ultravox_s2s: ${JSON.stringify(res.body)}`);
+    }
+  }
+
+  /**
+   * Converts a JSON Schema to the dynamic parameters format used in the Ultravox API
+   * @param {Object} jsonSchema - A JSON Schema object defining parameters
+   * @param {string} locationDefault - Default location value for parameters (default: 'PARAMETER_LOCATION_BODY')
+   * @returns {Array} Array of dynamic parameters objects
+   */
+  transformSchemaToParameters(jsonSchema, locationDefault = 'PARAMETER_LOCATION_BODY') {
+    if (jsonSchema.properties) {
+      const required = jsonSchema.required || [];
+
+      return Object.entries(jsonSchema.properties).map(([name]) => {
+        return {
+          name,
+          location: locationDefault,
+          required: required.includes(name)
+        };
+      });
+    }
+
+    return [];
+  }
+
+  async createCall() {
+    const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+    if (mcpTools && mcpTools.length > 0) {
+      const convertedTools = mcpTools.map((tool) => {
+        return {
+          temporaryTool: {
+            modelToolName: tool.name,
+            description: tool.description,
+            dynamicParameters: this.transformSchemaToParameters(tool.inputSchema),
+            // use client tool that ultravox call tool via freeswitch module.
+            client: {}
+          }
+        };
+      }
+      );
+      // merge with any existing tools
+      this.data.llmOptions.selectedTools = [
+        ...convertedTools,
+        ...(this.data.llmOptions.selectedTools || [])
+      ];
+    }
+
+    const payload = {
+      ...this.data.llmOptions,
+      model: this.model,
+      medium: {
+        ...(this.data.llmOptions.medium || {}),
+        serverWebSocket: {
+          inputSampleRate: 8000,
+          outputSampleRate: 8000,
+        }
+      }
+    };
+    const {statusCode, body} = await request('https://api.ultravox.ai/api/calls', {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+        'X-API-Key': this.apiKey
+      },
+      body: JSON.stringify(payload)
+    });
+    const data = await body.json();
+    if (statusCode !== 201 || !data?.joinUrl) {
+      this.logger.info({statusCode, data}, 'Ultravox Error registering call');
+      throw new Error(`Ultravox Error registering call:${statusCode} - ${data.detail}`);
+    }
+    this.logger.debug({joinUrl: data.joinUrl}, 'Ultravox Call registered');
+    return data;
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_Ultravox.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    try {
+      const data = await this.createCall();
+      const {joinUrl} = data;
+      // split the joinUrl into host and path
+      const {host, pathname, search} = new URL(joinUrl);
+      const args = [ep.uuid, 'session.create', host, pathname + search];
+      await this._api(ep, args);
+      // Notify the application that the session has been created with detail information
+      this._sendLlmEvent('createCall', {
+        type: 'createCall',
+        ...data
+      });
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmUltraVox_S2S:_startListening - Error sending createCall');
+      this.results = {completionReason: `connection failure - ${err}`};
+      this.notifyTaskDone();
+    }
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs);
+
+    await this._startListening(cs, ep);
+
+    await this.awaitTaskDone();
+
+    /* note: the parent llm verb started the span, which is why this is necessary */
+    await this.parent.performAction(this.results);
+
+    this._unregisterHandlers();
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+
+    this._api(cs.ep, [cs.ep.uuid, SessionDelete])
+      .catch((err) => this.logger.info({err}, 'TaskLlmUltravox_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.info('TaskLlmUltravox_S2S:_onConnect');
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmUltravox_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmUltravox_S2S:_onConnectFailure');
+    this.results = {completionReason: 'disconnect from remote end'};
+    this.notifyTaskDone();
+  }
+
+  async _onServerEvent(_ep, evt) {
+    let endConversation = false;
+    const type = evt.type;
+    this.logger.debug({evt}, 'TaskLlmUltravox_S2S:_onServerEvent');
+
+    /* server errors of some sort */
+    if (type === 'error') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server error',
+        error: evt.error
+      };
+    }
+
+    /* tool calls */
+    else if (type === 'client_tool_invocation') {
+      this.logger.debug({evt}, 'TaskLlmUltravox_S2S:_onServerEvent - function_call');
+      const {toolName: name, invocationId: call_id, parameters: args} = evt;
+
+      const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+      if (mcpTools.some((tool) => tool.name === name)) {
+        this.logger.debug({
+          name,
+          input: args
+        }, 'TaskLlmUltravox_S2S:_onServerEvent - function_call - mcp tool');
+        try {
+          const res = await this.parent.mcpService.callMcpTool(name, args);
+          this.logger.debug({res}, 'TaskLlmUltravox_S2S:_onServerEvent - function_call - mcp result');
+          this.processToolOutput(_ep, call_id, {
+            type: 'client_tool_result',
+            invocation_id: call_id,
+            result: res.content
+          });
+          return;
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmUltravox_S2S - error calling mcp tool');
+          this.results = {
+            completionReason: 'client error calling mcp function',
+            error: err
+          };
+          endConversation = true;
+        }
+      } else if (!this.toolHook) {
+        this.logger.info({evt}, 'TaskLlmUltravox_S2S:_onServerEvent - no toolHook defined!');
+      }
+      else {
+        try {
+          await this.parent.sendToolHook(call_id, {name, args});
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmUltravox_S2S - error calling function');
+          this.results = {
+            completionReason: 'client error calling function',
+            error: err
+          };
+          endConversation = true;
+        }
+      }
+    }
+
+    this._sendLlmEvent(type, evt);
+
+    if (endConversation) {
+      this.logger.info({results: this.results},
+        'TaskLlmUltravox_S2S:_onServerEvent - ending conversation due to error');
+      this.notifyTaskDone();
+    }
+  }
+
+  _sendLlmEvent(type, evt) {
+    /* check whether we should notify on this event */
+    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
+      this.parent.sendEventHook(evt)
+        .catch((err) => this.logger.info({err}, 'TaskLlmUltravox_S2S:_onServerEvent - error sending event hook'));
+    }
+  }
+
+  async processLlmUpdate(ep, data, _callSid) {
+    try {
+      this.logger.debug({data, _callSid}, 'TaskLlmUltravox_S2S:processLlmUpdate');
+
+      if (!data.type || ![
+        'input_text_message'
+      ].includes(data.type)) {
+        this.logger.info({data},
+          'TaskLlmUltravox_S2S:processLlmUpdate - invalid mid-call request, only input_text_message supported');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err, data}, 'TaskLlmUltravox_S2S:processLlmUpdate - Error processing LLM update');
+    }
+  }
+
+  async processToolOutput(ep, tool_call_id, data) {
+    try {
+      this.logger.debug({tool_call_id, data}, 'TaskLlmUltravox_S2S:processToolOutput');
+
+      if (!data.type || data.type !== 'client_tool_result') {
+        this.logger.info({data},
+          'TaskLlmUltravox_S2S:processToolOutput - invalid tool output, must be client_tool_result');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err, data}, 'TaskLlmUltravox_S2S:processToolOutput - Error processing tool output');
+    }
+  }
+
+  _populateEvents(events) {
+    if (events.includes('all')) {
+      /* work by excluding specific events */
+      const exclude = events
+        .filter((evt) => evt.startsWith('-'))
+        .map((evt) => evt.slice(1));
+      if (exclude.length === 0) this.includeEvents = ultravox_server_events;
+      else this.excludeEvents = expandWildcards(exclude);
+    }
+    else {
+      /* work by including specific events */
+      const include = events
+        .filter((evt) => !evt.startsWith('-'));
+      this.includeEvents = expandWildcards(include);
+    }
+
+    this.logger.debug({
+      includeEvents: this.includeEvents,
+      excludeEvents: this.excludeEvents
+    }, 'TaskLlmUltravox_S2S:_populateEvents');
+  }
+}
+
+module.exports = TaskLlmUltravox_S2S;

lib/tasks/llm/llms/voice_agent_s2s.js

@@ -0,0 +1,352 @@
+const Task = require('../../task');
+const TaskName = 'Llm_VoiceAgent_s2s';
+const {LlmEvents_VoiceAgent} = require('../../../utils/constants');
+const ClientEvent = 'client.event';
+const SessionDelete = 'session.delete';
+
+const va_server_events = [
+  'Error',
+  'Welcome',
+  'SettingsApplied',
+  'ConversationText',
+  'UserStartedSpeaking',
+  'EndOfThought',
+  'AgentThinking',
+  'FunctionCallRequest',
+  'FunctionCalling',
+  'AgentStartedSpeaking',
+  'AgentAudioDone',
+];
+
+const expandWildcards = (events) => {
+  // no-op for deepgram
+  return events;
+};
+
+class TaskLlmVoiceAgent_S2S extends Task {
+  constructor(logger, opts, parentTask) {
+    super(logger, opts, parentTask);
+    this.parent = parentTask;
+
+    this.vendor = this.parent.vendor;
+    this.model = this.parent.model || 'voice-agent';
+    this.auth = this.parent.auth;
+    this.connectionOptions = this.parent.connectOptions;
+
+    const {apiKey} = this.auth || {};
+    if (!apiKey) throw new Error('auth.apiKey is required for VoiceAgent S2S');
+
+    this.apiKey = apiKey;
+    this.authType = 'bearer';
+    this.actionHook = this.data.actionHook;
+    this.eventHook = this.data.eventHook;
+    this.toolHook = this.data.toolHook;
+    const {Settings} = this.data.llmOptions;
+
+    if (typeof Settings !== 'object') {
+      throw new Error('llmOptions with an initial Settings is required for VoiceAgent S2S');
+    }
+
+    // eslint-disable-next-line no-unused-vars
+    const {audio, ...rest} = Settings;
+    const cfg = this.Settings = rest;
+
+    if (!cfg.agent) throw new Error('llmOptions.Settings.agent is required for VoiceAgent S2S');
+    if (!cfg.agent.think) {
+      throw new Error('llmOptions.Settings.agent.think is required for VoiceAgent S2S');
+    }
+    if (!cfg.agent.think.provider?.model) {
+      throw new Error('llmOptions.Settings.agent.think.provider.model is required for VoiceAgent S2S');
+    }
+    if (!cfg.agent.think.provider?.type) {
+      throw new Error('llmOptions.Settings.agent.think.provider.type is required for VoiceAgent S2S');
+    }
+
+    this.results = {
+      completionReason: 'normal conversation end'
+    };
+
+    /**
+     * only one of these will have items,
+     * if includeEvents, then these are the events to include
+     * if excludeEvents, then these are the events to exclude
+     */
+    this.includeEvents = [];
+    this.excludeEvents = [];
+
+    /* default to all events if user did not specify */
+    this._populateEvents(this.data.events || va_server_events);
+
+    this.addCustomEventListener = parentTask.addCustomEventListener.bind(parentTask);
+    this.removeCustomEventListeners = parentTask.removeCustomEventListeners.bind(parentTask);
+  }
+
+  get name() { return TaskName; }
+
+  get host() {
+    const {host} = this.connectionOptions || {};
+    return host || 'agent.deepgram.com';
+  }
+
+  get path() {
+    const {path} = this.connectionOptions || {};
+    if (path) return path;
+
+    return '/v1/agent/converse';
+  }
+
+  async _api(ep, args) {
+    const res = await ep.api('uuid_voice_agent_s2s', `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      throw new Error(`Error calling uuid_voice_agent_s2s: ${JSON.stringify(res.body)}`);
+    }
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs);
+
+    await this._startListening(cs, ep);
+
+    await this.awaitTaskDone();
+
+    /* note: the parent llm verb started the span, which is why this is necessary */
+    await this.parent.performAction(this.results);
+
+    this._unregisterHandlers();
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+
+    this._api(cs.ep, [cs.ep.uuid, SessionDelete])
+      .catch((err) => this.logger.info({err}, 'TaskLlmVoiceAgent_S2S:kill - error deleting session'));
+
+    this.notifyTaskDone();
+  }
+
+  /**
+   * Send function call response to the VoiceAgent server
+   */
+  async processToolOutput(ep, tool_call_id, data) {
+    try {
+      const {data:response} = data;
+      this.logger.debug({tool_call_id, response}, 'TaskLlmVoiceAgent_S2S:processToolOutput');
+
+      if (!response.type || response.type !== 'FunctionCallResponse') {
+        this.logger.info({response},
+          'TaskLlmVoiceAgent_S2S:processToolOutput - invalid tool output, must be FunctionCallResponse');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(response)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmVoiceAgent_S2S:processToolOutput');
+    }
+  }
+
+  /**
+   * Send a session.update to the VoiceAgent server
+   * Note: creating and deleting conversation items also supported as well as interrupting the assistant
+   */
+  async processLlmUpdate(ep, data, _callSid) {
+    try {
+      this.logger.debug({data, _callSid}, 'TaskLlmVoiceAgent_S2S:processLlmUpdate');
+
+      if (!data.type || ![
+        'UpdateInstructions',
+        'UpdateSpeak',
+        'InjectAgentMessage',
+      ].includes(data.type)) {
+        this.logger.info({data}, 'TaskLlmVoiceAgent_S2S:processLlmUpdate - invalid mid-call request');
+      }
+      else {
+        await this._api(ep, [ep.uuid, ClientEvent, JSON.stringify(data)]);
+      }
+    } catch (err) {
+      this.logger.info({err}, 'TaskLlmVoiceAgent_S2S:processLlmUpdate');
+    }
+  }
+
+  async _startListening(cs, ep) {
+    this._registerHandlers(ep);
+
+    try {
+      const args = [ep.uuid, 'session.create', this.host, this.path, this.authType, this.apiKey];
+      await this._api(ep, args);
+    } catch (err) {
+      this.logger.error({err}, `TaskLlmVoiceAgent_S2S:_startListening: ${JSON.stringify(err)}`);
+      this.notifyTaskDone();
+    }
+  }
+
+  async _sendClientEvent(ep, obj) {
+    let ok = true;
+    this.logger.debug({obj}, 'TaskLlmVoiceAgent_S2S:_sendClientEvent');
+    try {
+      const args = [ep.uuid, ClientEvent, JSON.stringify(obj)];
+      await this._api(ep, args);
+    } catch (err) {
+      ok = false;
+      this.logger.error({err}, 'TaskLlmVoiceAgent_S2S:_sendClientEvent - Error');
+    }
+    return ok;
+  }
+
+  async _sendInitialMessage(ep) {
+    const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+    if (mcpTools && mcpTools.length > 0 && this.Settings.agent?.think) {
+      const convertedTools = mcpTools.map((tool) => ({
+        name: tool.name,
+        description: tool.description,
+        parameters: tool.inputSchema
+      }));
+
+      this.Settings.agent.think.functions = [
+        ...convertedTools,
+        ...(this.Settings.agent.think?.functions || [])
+      ];
+    }
+    if (!await this._sendClientEvent(ep, this.Settings)) {
+      this.notifyTaskDone();
+    }
+  }
+
+  _registerHandlers(ep) {
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.Connect, this._onConnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.ConnectFailure, this._onConnectFailure.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.Disconnect, this._onDisconnect.bind(this, ep));
+    this.addCustomEventListener(ep, LlmEvents_VoiceAgent.ServerEvent, this._onServerEvent.bind(this, ep));
+  }
+
+  _unregisterHandlers() {
+    this.removeCustomEventListeners();
+  }
+
+  _onError(_ep, evt) {
+    this.logger.info({evt}, 'TaskLlmVoiceAgent_S2S:_onError');
+    this.notifyTaskDone();
+  }
+
+  _onConnect(ep) {
+    this.logger.debug('TaskLlmVoiceAgent_S2S:_onConnect');
+    this._sendInitialMessage(ep);
+  }
+  _onConnectFailure(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmVoiceAgent_S2S:_onConnectFailure');
+    this.results = {completionReason: 'connection failure'};
+    this.notifyTaskDone();
+  }
+  _onDisconnect(_ep, evt) {
+    this.logger.info(evt, 'TaskLlmVoiceAgent_S2S:_onConnectFailure');
+    this.results = {completionReason: 'disconnect from remote end'};
+    this.notifyTaskDone();
+  }
+  async _onServerEvent(_ep, evt) {
+    let endConversation = false;
+    const type = evt.type;
+    this.logger.info({evt}, 'TaskLlmVoiceAgent_S2S:_onServerEvent');
+
+    /* check for failures, such as rate limit exceeded, that should terminate the conversation */
+    if (type === 'response.done' && evt.response.status === 'failed') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server failure',
+        error: evt.response.status_details?.error
+      };
+    }
+
+    /* server errors of some sort */
+    else if (type === 'error') {
+      endConversation = true;
+      this.results = {
+        completionReason: 'server error',
+        error: evt.error
+      };
+    }
+
+    /* tool calls */
+    else if (type === 'FunctionCallRequest') {
+      this.logger.debug({evt}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - function_call');
+
+      const mcpTools = this.parent.isMcpEnabled ? await this.parent.mcpService.getAvailableMcpTools() : [];
+      if (!this.toolHook && mcpTools.length === 0) {
+        this.logger.warn({evt}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - no toolHook defined!');
+      } else {
+        const {functions} = evt;
+        const handledFunctions = [];
+
+        try {
+          if (mcpTools && mcpTools.length > 0) {
+            for (const func of functions) {
+              const {name, arguments: args, id} = func;
+              const tool = mcpTools.find((tool) => tool.name === name);
+              if (tool) {
+                handledFunctions.push(name);
+                const response = await this.parent.mcpService.callMcpTool(name, JSON.parse(args));
+                this.logger.debug({response}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - function_call - mcp result');
+                this.processToolOutput(_ep, id, {
+                  data: {
+                    type: 'FunctionCallResponse',
+                    id,
+                    name,
+                    content: response.length > 0 ? response[0].text : 'There is no output from the function call'
+                  }
+                });
+              }
+            }
+          }
+
+          for (const func of functions) {
+            const {name, arguments: args, id} = func;
+            if (!handledFunctions.includes(name)) {
+              await this.parent.sendToolHook(id, {name, args: JSON.parse(args)});
+            }
+          }
+        } catch (err) {
+          this.logger.info({err, evt}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - error calling function');
+          this.results = {
+            completionReason: 'client error calling function',
+            error: err
+          };
+          endConversation = true;
+        }
+      }
+    }
+
+    /* check whether we should notify on this event */
+    if (this.includeEvents.length > 0 ? this.includeEvents.includes(type) : !this.excludeEvents.includes(type)) {
+      this.parent.sendEventHook(evt)
+        .catch((err) => this.logger.info({err}, 'TaskLlmVoiceAgent_S2S:_onServerEvent - error sending event hook'));
+    }
+
+    if (endConversation) {
+      this.logger.info({results: this.results},
+        'TaskLlmVoiceAgent_S2S:_onServerEvent - ending conversation due to error');
+      this.notifyTaskDone();
+    }
+  }
+
+  _populateEvents(events) {
+    if (events.includes('all')) {
+      /* work by excluding specific events */
+      const exclude = events
+        .filter((evt) => evt.startsWith('-'))
+        .map((evt) => evt.slice(1));
+      if (exclude.length === 0) this.includeEvents = va_server_events;
+      else this.excludeEvents = expandWildcards(exclude);
+    }
+    else {
+      /* work by including specific events */
+      const include = events
+        .filter((evt) => !evt.startsWith('-'));
+      this.includeEvents = expandWildcards(include);
+    }
+
+    this.logger.debug({
+      includeEvents: this.includeEvents,
+      excludeEvents: this.excludeEvents
+    }, 'TaskLlmVoiceAgent_S2S:_populateEvents');
+  }
+}
+
+module.exports = TaskLlmVoiceAgent_S2S;

lib/tasks/make_task.js

@@ -62,6 +62,9 @@ function makeTask(logger, obj, parent) {
     case TaskName.Message:
       const TaskMessage = require('./message');
       return new TaskMessage(logger, data, parent);
+    case TaskName.Llm:
+      const TaskLlm = require('./llm');
+      return new TaskLlm(logger, data, parent);
     case TaskName.Rasa:
       const TaskRasa = require('./rasa');
       return new TaskRasa(logger, data, parent);
@@ -81,6 +84,7 @@ function makeTask(logger, obj, parent) {
       const TaskTranscribe = require('./transcribe');
       return new TaskTranscribe(logger, data, parent);
     case TaskName.Listen:
+    case TaskName.Stream:
       const TaskListen = require('./listen');
       return new TaskListen(logger, data, parent);
     case TaskName.Redirect:

lib/tasks/message.js

@@ -1,7 +1,7 @@
 const Task = require('./task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
 const bent = require('bent');
-const uuidv4 = require('uuid-random');
+const crypto = require('crypto');
 const {K8S} = require('../config');
 class TaskMessage extends Task {
   constructor(logger, opts) {
@@ -9,7 +9,7 @@ class TaskMessage extends Task {
     this.preconditions = TaskPreconditions.None;
 
     this.payload = {
-      message_sid: this.data.message_sid || uuidv4(),
+      message_sid: this.data.message_sid || crypto.randomUUID(),
       carrier: this.data.carrier,
       to: this.data.to,
       from: this.data.from,

lib/tasks/play.js

@@ -1,6 +1,6 @@
 const Task = require('./task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
-
+const { PlayFileNotFoundError } = require('../utils/error');
 class TaskPlay extends Task {
   constructor(logger, opts) {
     super(logger, opts);
@@ -26,6 +26,7 @@ class TaskPlay extends Task {
     let playbackSeconds = 0;
     let playbackMilliseconds = 0;
     let completed = !(this.timeoutSecs > 0 || this.loop);
+    cs.playingAudio = true;
     if (this.timeoutSecs > 0) {
       timeout = setTimeout(async() => {
         completed = true;
@@ -39,6 +40,22 @@ class TaskPlay extends Task {
     try {
       this.notifyStatus({event: 'start-playback'});
       while (!this.killed && (this.loop === 'forever' || this.loop--) && this.ep.connected) {
+        /* Listen for playback-start event and set up a one-time listener for uuid_break
+         * that will kill the audio playback if the taskIds match. This ensures that
+         * we only kill the currently playing audio and not audio from other tasks.
+         * As we are using stickyEventEmitter, even if the event is emitted before the listener is registered,
+         * the listener will receive the most recent event.
+         */
+        ep.once('playback-start', (evt) => {
+          this.logger.debug({evt}, 'Play got playback-start');
+          this.cs.stickyEventEmitter.once('uuid_break', (t) => {
+            if (t?.taskId === this.taskId) {
+              this.logger.debug(`Play got kill-playback, executing uuid_break, taskId: ${t?.taskId}`);
+              this.ep.api('uuid_break', this.ep.uuid).catch((err) => this.logger.info(err, 'Error killing audio'));
+              this.notifyStatus({event: 'kill-playback'});
+            }
+          });
+        });
         if (cs.isInConference) {
           const {memberId, confName, confUuid} = cs;
           if (Array.isArray(this.url)) {
@@ -66,23 +83,35 @@ class TaskPlay extends Task {
         }
       }
     } catch (err) {
-      if (timeout) clearTimeout(timeout);
-      this.logger.info(err, `TaskPlay:exec - error playing ${this.url}`);
+      this.logger.info(`TaskPlay:exec - error playing ${this.url}: ${err.message}`);
+      this.playComplete = true;
+      if (err.message === 'File Not Found') {
+        const {writeAlerts, AlertType} = cs.srf.locals;
+        await this.performAction({status: 'fail', reason: 'playFailed'}, !(this.parentTask || cs.isConfirmCallSession));
+        this.emit('playDone');
+        writeAlerts({
+          account_sid: cs.accountSid,
+          alert_type: AlertType.PLAY_FILENOTFOUND,
+          url: this.url,
+          target_sid: cs.callSid
+        });
+        throw new PlayFileNotFoundError(this.url);
+      }
     }
     this.emit('playDone');
   }
 
   async kill(cs) {
     super.kill(cs);
-    if (this.ep.connected && !this.playComplete) {
+    if (this.ep?.connected && !this.playComplete) {
       this.logger.debug('TaskPlay:kill - killing audio');
       if (cs.isInConference) {
         const {memberId, confName} = cs;
         this.killPlayToConfMember(this.ep, memberId, confName);
       }
       else {
-        this.notifyStatus({event: 'kill-playback'});
-        this.ep.api('uuid_break', this.ep.uuid).catch((err) => this.logger.info(err, 'Error killing audio'));
+        //this.ep.api('uuid_break', this.ep.uuid).catch((err) => this.logger.info(err, 'Error killing audio'));
+        cs.stickyEventEmitter.emit('uuid_break', this);
       }
     }
   }

lib/tasks/redirect.js

@@ -1,5 +1,8 @@
 const Task = require('./task');
 const {TaskName} = require('../utils/constants');
+const WsRequestor = require('../utils/ws-requestor');
+const URL = require('url');
+const HttpRequestor = require('../utils/http-requestor');
 
 /**
  * Redirects to a new application
@@ -13,6 +16,32 @@ 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) {
+        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;
+        } catch (err) {
+          this.logger.info(err, `Task:redirect error updating base url to ${this.actionHook}`);
+        }
+      }
+    }
     await this.performAction();
   }
 }

lib/tasks/rest_dial.js

@@ -12,6 +12,7 @@ class TaskRestDial extends Task {
 
     this.from = this.data.from;
     this.callerName = this.data.callerName;
+    this.timeLimit = this.data.timeLimit;
     this.fromHost = this.data.fromHost;
     this.to = this.data.to;
     this.call_hook = this.data.call_hook;
@@ -39,9 +40,9 @@ class TaskRestDial extends Task {
 
     if (this.data.amd) {
       this.startAmd = cs.startAmd;
-      this.stopAmd = cs.stopAmd;
       this.on('amd', this._onAmdEvent.bind(this, cs));
     }
+    this.stopAmd = cs.stopAmd;
 
     this._setCallTimer();
     await this.awaitTaskDone();
@@ -66,6 +67,9 @@ class TaskRestDial extends Task {
     const cs = this.callSession;
     cs.setDialog(dlg);
     cs.referHook = this.referHook;
+    if (this.timeLimit) {
+      cs.startMaxCallDurationTimer(this.timeLimit);
+    }
     this.logger.debug('TaskRestDial:_onConnect - call connected');
     if (this.sipRequestWithinDialogHook) this._initSipRequestWithinDialogHandler(cs, dlg);
     try {
@@ -77,11 +81,13 @@ class TaskRestDial extends Task {
           synthesizer: {
             vendor: cs.speechSynthesisVendor,
             language: cs.speechSynthesisLanguage,
-            voice: cs.speechSynthesisVoice
+            voice: cs.speechSynthesisVoice,
+            label: cs.speechSynthesisLabel,
           },
           recognizer: {
             vendor:  cs.speechRecognizerVendor,
-            language: cs.speechRecognizerLanguage
+            language: cs.speechRecognizerLanguage,
+            label: cs.speechRecognizerLabel,
           }
         }
       };

lib/tasks/say.js

@@ -1,6 +1,8 @@
+const assert = require('assert');
 const TtsTask = require('./tts-task');
 const {TaskName, TaskPreconditions} = require('../utils/constants');
 const pollySSMLSplit = require('polly-ssml-split');
+const { SpeechCredentialError } = require('../utils/error');
 
 const breakLengthyTextIfNeeded = (logger,  text) => {
   const chunkSize = 1000;
@@ -34,24 +36,40 @@ class TaskSay extends TtsTask {
     super(logger, opts, parentTask);
     this.preconditions = TaskPreconditions.Endpoint;
 
-    this.text = (Array.isArray(this.data.text) ? this.data.text : [this.data.text])
-      .map((t) => breakLengthyTextIfNeeded(this.logger, t))
-      .flat();
+    assert.ok((typeof this.data.text === 'string' || Array.isArray(this.data.text)) || this.data.stream === true,
+      'Say: either text or stream:true is required');
 
-    this.loop = this.data.loop || 1;
-    this.isHandledByPrimaryProvider = true;
+
+    if (this.data.stream === true) {
+      this._isStreamingTts = true;
+      this.closeOnStreamEmpty = this.data.closeOnStreamEmpty !== false;
+    }
+    else {
+      this._isStreamingTts = false;
+      this.text = (Array.isArray(this.data.text) ? this.data.text : [this.data.text])
+        .map((t) => breakLengthyTextIfNeeded(this.logger, t))
+        .flat();
+
+      this.loop = this.data.loop || 1;
+      this.isHandledByPrimaryProvider = true;
+    }
   }
 
   get name() { return TaskName.Say; }
 
   get summary() {
-    for (let i = 0; i < this.text.length; i++) {
-      if (this.text[i].startsWith('silence_stream')) continue;
-      return `${this.name}{text=${this.text[i].slice(0, 15)}${this.text[i].length > 15 ? '...' : ''}}`;
+    if (this.isStreamingTts) return `${this.name} streaming`;
+    else {
+      for (let i = 0; i < this.text.length; i++) {
+        if (this.text[i].startsWith('silence_stream')) continue;
+        return `${this.name}{text=${this.text[i].slice(0, 15)}${this.text[i].length > 15 ? '...' : ''}}`;
+      }
+      return `${this.name}{${this.text[0]}}`;
     }
-    return `${this.name}{${this.text[0]}}`;
   }
 
+  get isStreamingTts() { return this._isStreamingTts; }
+
   _validateURL(urlString) {
     try {
       new URL(urlString);
@@ -61,147 +79,58 @@ class TaskSay extends TtsTask {
     }
   }
 
-  async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
-    const {srf, accountSid:account_sid} = cs;
-    const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
-    const {writeAlerts, AlertType, stats} = srf.locals;
-    const {synthAudio} = srf.locals.dbHelpers;
-    const engine = this.synthesizer.engine || cs.synthesizer?.engine || 'neural';
-    const salt = cs.callSid;
-
-    let credentials = cs.getSpeechCredentials(vendor, 'tts', label);
-    /* parse Nuance voices into name and model */
-    let model;
-    if (vendor === 'nuance' && voice) {
-      const arr = /([A-Za-z-]*)\s+-\s+(enhanced|standard)/.exec(voice);
-      if (arr) {
-        voice = arr[1];
-        model = arr[2];
-      }
-    } else if (vendor === 'deepgram') {
-      model = voice;
+  async exec(cs, obj) {
+    if (this.isStreamingTts && !cs.appIsUsingWebsockets) {
+      throw new Error('Say: streaming say verb requires applications to use the websocket API');
     }
 
-    /* allow for microsoft custom region voice and api_key to be specified as an override */
-    if (vendor === 'microsoft' && this.options.deploymentId) {
-      credentials = credentials || {};
-      credentials.use_custom_tts = true;
-      credentials.custom_tts_endpoint = this.options.deploymentId;
-      credentials.api_key = this.options.apiKey || credentials.apiKey;
-      credentials.region = this.options.region || credentials.region;
-      voice = this.options.voice || voice;
-    } else if (vendor === 'elevenlabs') {
-      credentials = credentials || {};
-      credentials.model_id = this.options.model_id || credentials.model_id;
-      credentials.voice_settings = this.options.voice_settings || {};
-      credentials.optimize_streaming_latency = this.options.optimize_streaming_latency
-      || credentials.optimize_streaming_latency;
-      voice = this.options.voice_id || voice;
-    }
-
-    ep.set({
-      tts_engine: vendor.startsWith('custom:') ? 'custom' : vendor,
-      tts_voice: voice,
-      cache_speech_handles: !cs.currentTtsVendor || cs.currentTtsVendor === vendor ? 1 : 0,
-    }).catch((err) => this.logger.info({err}, 'Error setting tts_engine on endpoint'));
-    // set the current vendor on the call session
-    // If vendor is changed from the previous one, then reset the cache_speech_handles flag
-    cs.currentTtsVendor = vendor;
-
-    if (!preCache && !this._disableTracing) this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
     try {
-      if (!credentials) {
-        writeAlerts({
-          account_sid,
-          alert_type: AlertType.TTS_NOT_PROVISIONED,
-          vendor
-        }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-        throw new Error('no provisioned speech credentials for TTS');
+      if (this.isStreamingTts) await this.handlingStreaming(cs, obj);
+      else await this.handling(cs, obj);
+      this.emit('playDone');
+    } catch (error) {
+      if (error instanceof SpeechCredentialError) {
+        // if say failed due to speech credentials, alarm is writtern and error notification is sent
+        // finished this say to move to next task.
+        this.logger.info({error}, 'Say failed due to SpeechCredentialError, finished!');
+        this.emit('playDone');
+        return;
       }
-      // synthesize all of the text elements
-      let lastUpdated = false;
-
-      /* produce an audio segment from the provided text */
-      const generateAudio = async(text) => {
-        if (this.killed) return;
-        if (text.startsWith('silence_stream://')) return text;
-
-        /* otel: trace time for tts */
-        if (!preCache && !this._disableTracing)  {
-          const {span} = this.startChildSpan('tts-generation', {
-            'tts.vendor': vendor,
-            'tts.language': language,
-            'tts.voice': voice
-          });
-          this.otelSpan = span;
-        }
-        try {
-          const {filePath, servedFromCache, rtt} = await synthAudio(stats, {
-            account_sid,
-            text,
-            vendor,
-            language,
-            voice,
-            engine,
-            model,
-            salt,
-            credentials,
-            options: this.options,
-            disableTtsCache : this.disableTtsCache,
-            preCache
-          });
-          if (!filePath.startsWith('say:')) {
-            this.logger.debug(`Say: file ${filePath}, served from cache ${servedFromCache}`);
-            if (filePath) cs.trackTmpFile(filePath);
-            if (this.otelSpan) {
-              this.otelSpan.setAttributes({'tts.cached': servedFromCache});
-              this.otelSpan.end();
-              this.otelSpan = null;
-            }
-            if (!servedFromCache && !lastUpdated) {
-              lastUpdated = true;
-              updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
-            }
-            if (!servedFromCache && rtt && !preCache  && !this._disableTracing) {
-              this.notifyStatus({
-                event: 'synthesized-audio',
-                vendor,
-                language,
-                characters: text.length,
-                elapsedTime: rtt
-              });
-            }
-          }
-          else {
-            this.logger.debug('Say: a streaming tts api will be used');
-            const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${ep.uuid},`);
-            return modifiedPath;
-          }
-          return filePath;
-        } catch (err) {
-          this.logger.info({err}, 'Error synthesizing tts');
-          if (this.otelSpan) this.otelSpan.end();
-          writeAlerts({
-            account_sid: cs.accountSid,
-            alert_type: AlertType.TTS_FAILURE,
-            vendor,
-            detail: err.message
-          }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
-          throw err;
-        }
-      };
-
-      const arr = this.text.map((t) => (this._validateURL(t) ? t : generateAudio(t)));
-      return (await Promise.all(arr)).filter((fp) => fp && fp.length);
-    } catch (err) {
-      this.logger.info(err, 'TaskSay:exec error');
-      throw err;
+      throw error;
     }
-
   }
 
-  async exec(cs, {ep}) {
-    const {srf, accountSid:account_sid} = cs;
+  async handlingStreaming(cs, {ep}) {
+    const {vendor, language, voice, label} = this.getTtsVendorData(cs);
+    const  credentials = cs.getSpeechCredentials(vendor, 'tts', label);
+    if (!credentials) {
+      throw new SpeechCredentialError(
+        `No text-to-speech service credentials for ${vendor} with labels: ${label} have been configured`);
+    }
+    this.ep = ep;
+    try {
+
+      await this.setTtsStreamingChannelVars(vendor, language, voice, credentials, ep);
+
+      await cs.startTtsStream();
+
+      cs.requestor?.request('tts:streaming-event', '/streaming-event', {event_type: 'stream_open'})
+        .catch((err) => this.logger.info({err}, 'TaskSay:handlingStreaming - Error sending'));
+    } catch (err) {
+      this.logger.info({err}, 'TaskSay:handlingStreaming - Error setting channel vars');
+      cs.requestor?.request('tts:streaming-event', '/streaming-event', {event_type: 'stream_closed'})
+        .catch((err) => this.logger.info({err}, 'TaskSay:handlingStreaming - Error sending'));
+
+      //TODO: send tts:streaming-event with error?
+      this.notifyTaskDone();
+    }
+
+    await this.awaitTaskDone();
+    this.logger.info('TaskSay:handlingStreaming - done');
+  }
+
+  async handling(cs, {ep}) {
+    const {srf, accountSid:account_sid, callSid:target_sid} = cs;
     const {writeAlerts, AlertType} = srf.locals;
     const {addFileToCache} = srf.locals.dbHelpers;
     const engine = this.synthesizer.engine || cs.synthesizer?.engine || 'neural';
@@ -218,10 +147,7 @@ class TaskSay extends TtsTask {
     let voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
       this.synthesizer.voice :
       cs.speechSynthesisVoice;
-    // label can be null/empty in synthesizer config, just use application level label if it's default
-    let label = this.synthesizer.label === 'default' ?
-      cs.speechSynthesisLabel :
-      this.synthesizer.label;
+    let label = this.taskIncludeSynthesizer ? this.synthesizer.label : cs.speechSynthesisLabel;
 
     const fallbackVendor = this.synthesizer.fallbackVendor && this.synthesizer.fallbackVendor !== 'default' ?
       this.synthesizer.fallbackVendor :
@@ -232,10 +158,8 @@ class TaskSay extends TtsTask {
     const fallbackVoice =  this.synthesizer.fallbackVoice && this.synthesizer.fallbackVoice !== 'default' ?
       this.synthesizer.fallbackVoice :
       cs.fallbackSpeechSynthesisVoice;
-    // label can be null/empty in synthesizer config, just use application level label if it's default
-    const fallbackLabel = this.synthesizer.fallbackLabel === 'default' ?
-      cs.fallbackSpeechSynthesisLabel :
-      this.synthesizer.fallbackLabel;
+    const fallbackLabel = this.taskIncludeSynthesizer ?
+      this.synthesizer.fallbackLabel : cs.fallbackSpeechSynthesisLabel;
 
     if (cs.hasFallbackTts) {
       vendor = fallbackVendor;
@@ -261,7 +185,7 @@ class TaskSay extends TtsTask {
       } else {
         this.notifyError(
           { msg: 'TTS error', details:`TTS vendor ${vendor} error: ${error}`, failover: 'not available'});
-        throw error;
+        throw new SpeechCredentialError(error.message);
       }
     };
     let filepath;
@@ -280,55 +204,63 @@ class TaskSay extends TtsTask {
           await this.playToConfMember(ep, memberId, confName, confUuid, filepath[segment]);
         }
         else {
-          let tts_cache_filename;
-          if (filepath[segment].startsWith('say:{')) {
+          const isStreaming = filepath[segment].startsWith('say:{');
+          if (isStreaming) {
             const arr = /^say:\{.*\}\s*(.*)$/.exec(filepath[segment]);
             if (arr) this.logger.debug(`Say:exec sending streaming tts request: ${arr[1].substring(0, 64)}..`);
           }
-          else this.logger.debug(`Say:exec sending  ${filepath[segment].substring(0, 64)}`);
+          else this.logger.debug(`Say:exec sending ${filepath[segment].substring(0, 64)}`);
           ep.once('playback-start', (evt) => {
             this.logger.debug({evt}, 'Say got playback-start');
             if (this.otelSpan) {
-              this._addStreamingTtsAttributes(this.otelSpan, evt);
+              this._addStreamingTtsAttributes(this.otelSpan, evt, vendor);
               this.otelSpan.end();
               this.otelSpan = null;
               if (evt.variable_tts_cache_filename) {
-                tts_cache_filename = evt.variable_tts_cache_filename;
                 cs.trackTmpFile(evt.variable_tts_cache_filename);
               }
-              else {
-                this.logger.info('No tts_cache_filename in playback-start event');
-              }
             }
           });
           ep.once('playback-stop', (evt) => {
-            if (!tts_cache_filename || evt.variable_tts_cache_filename !== tts_cache_filename) {
-              this.logger.info({evt}, 'Say: discarding playback-stop from other say verb');
-            }
-            else {
-              this.logger.debug({evt}, 'Say got playback-stop');
-              if (evt.variable_tts_error) {
-                writeAlerts({
-                  account_sid,
-                  alert_type: AlertType.TTS_FAILURE,
-                  vendor,
-                  detail: evt.variable_tts_error
-                }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-              }
-              if (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,
-                  text
-                }).catch((err) => this.logger.info({err}, 'Error adding file to cache'));
+            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'));
+            }
+
             if (this._playResolve) {
-              evt.variable_tts_error ? this._playReject(new Error(evt.variable_tts_error)) : this._playResolve();
+              (tts_error || response_code >= 300) ? this._playReject(new Error(evt.variable_tts_error)) :
+                this._playResolve();
             }
           });
           // wait for playback-stop event received to confirm if the playback is successful
@@ -347,6 +279,7 @@ class TaskSay extends TtsTask {
               continue;
             } catch (err) {
               this.logger.info({err}, 'Error waiting for playback-stop event');
+              throw err;
             }
           } finally {
             this._playPromise = null;
@@ -364,7 +297,6 @@ class TaskSay extends TtsTask {
         segment++;
       }
     }
-    this.emit('playDone');
   }
 
   async kill(cs) {
@@ -374,8 +306,13 @@ class TaskSay extends TtsTask {
       if (cs.isInConference) {
         const {memberId, confName} = cs;
         this.killPlayToConfMember(this.ep, memberId, confName);
-      }
-      else {
+      } else if (this.isStreamingTts) {
+        this.logger.debug('TaskSay:kill - clearing TTS stream for streaming audio');
+        cs.clearTtsStream();
+      } else {
+        if (!this.notifiedPlayBackStop) {
+          this.notifyStatus({event: 'stop-playback'});
+        }
         this.notifyStatus({event: 'kill-playback'});
         this.ep.api('uuid_break', this.ep.uuid);
       }
@@ -387,26 +324,39 @@ class TaskSay extends TtsTask {
         this._playResolve = null;
       }
     }
+    this.notifyTaskDone();
   }
 
-  _addStreamingTtsAttributes(span, evt) {
+  _addStreamingTtsAttributes(span, evt, vendor) {
     const attrs = {'tts.cached': false};
     for (const [key, value] of Object.entries(evt)) {
       if (key.startsWith('variable_tts_')) {
         let newKey = key.substring('variable_tts_'.length)
           .replace('whisper_', 'whisper.')
+          .replace('nvidia_', 'nvidia.')
           .replace('deepgram_', 'deepgram.')
           .replace('playht_', 'playht.')
+          .replace('cartesia_', 'cartesia.')
           .replace('rimelabs_', 'rimelabs.')
           .replace('verbio_', 'verbio.')
           .replace('elevenlabs_', 'elevenlabs.');
         if (spanMapping[newKey]) newKey = spanMapping[newKey];
         attrs[newKey] = value;
+        if (key === 'variable_tts_time_to_first_byte_ms' && value) {
+          this.cs.srf.locals.stats.histogram('tts.response_time', value, [`vendor:${vendor}`]);
+        }
       }
     }
     delete attrs['cache_filename']; //no value in adding this to the span
     span.setAttributes(attrs);
   }
+
+  notifyTtsStreamIsEmpty() {
+    if (this.isStreamingTts && this.closeOnStreamEmpty) {
+      this.logger.info('TaskSay:notifyTtsStreamIsEmpty - stream is empty, killing task');
+      this.notifyTaskDone();
+    }
+  }
 }
 
 const spanMapping = {
@@ -443,6 +393,11 @@ const spanMapping = {
   'playht.name_lookup_time_ms': 'name_lookup_ms',
   'playht.connect_time_ms': 'connect_ms',
   'playht.final_response_time_ms': 'final_response_ms',
+  // Cartesia
+  'cartesia.request_id': 'cartesia.req_id',
+  'cartesia.name_lookup_time_ms': 'name_lookup_ms',
+  'cartesia.connect_time_ms': 'connect_ms',
+  'cartesia.final_response_time_ms': 'final_response_ms',
   // Rimelabs
   'rimelabs.name_lookup_time_ms': 'name_lookup_ms',
   'rimelabs.connect_time_ms': 'connect_ms',

lib/tasks/sip_decline.js

@@ -18,6 +18,11 @@ class TaskSipDecline extends Task {
     super.exec(cs);
     res.send(this.data.status, this.data.reason, {
       headers: this.headers
+    }, (err) => {
+      if (!err) {
+        // Call was successfully declined
+        cs._callReleased();
+      }
     });
     cs.emit('callStatusChange', {
       callStatus: CallStatus.Failed,

lib/tasks/sip_refer.js

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

lib/tasks/stt-task.js

@@ -2,6 +2,32 @@ const Task = require('./task');
 const assert = require('assert');
 const crypto = require('crypto');
 const { TaskPreconditions, CobaltTranscriptionEvents } = require('../utils/constants');
+const { SpeechCredentialError } = require('../utils/error');
+const {JAMBONES_AWS_TRANSCRIBE_USE_GRPC} = require('../config');
+
+/**
+ * "Please insert turns here: {{turns:4}}"
+// -> { processed: 'Please insert turns here: {{turns}}', turns: 4 }
+
+processTurnString("Please insert turns here: {{turns}}"));
+// -> { processed: 'Please insert turns here: {{turns}}', turns: null }
+ */
+const processTurnString = (input) => {
+  const regex = /\{\{turns(?::(\d+))?\}\}/;
+  const match = input.match(regex);
+
+  if (!match) {
+    return {
+      processed: input,
+      turns: null
+    };
+  }
+
+  const turns = match[1] ? parseInt(match[1], 10) : null;
+  const processed = input.replace(regex, '{{turns}}');
+
+  return { processed, turns };
+};
 
 class SttTask extends Task {
 
@@ -16,14 +42,22 @@ class SttTask extends Task {
       normalizeTranscription,
       setSpeechCredentialsAtRuntime,
       compileSonioxTranscripts,
-      consolidateTranscripts
+      consolidateTranscripts,
+      updateSpeechmaticsPayload
     } = require('../utils/transcription-utils')(logger);
     this.setChannelVarsForStt = setChannelVarsForStt;
     this.normalizeTranscription = normalizeTranscription;
     this.compileSonioxTranscripts = compileSonioxTranscripts;
     this.consolidateTranscripts = consolidateTranscripts;
+    this.updateSpeechmaticsPayload = updateSpeechmaticsPayload;
     this.eventHandlers = [];
     this.isHandledByPrimaryProvider = true;
+    /**
+     * Task use taskIncludeRecognizer to identify
+     * if taskIncludeRecognizer === true, use label from verb.recognizer, even it's empty
+     * if taskIncludeRecognizer === false, use label from application.recognizer
+     */
+    this.taskIncludeRecognizer = !!this.data.recognizer;
     if (this.data.recognizer) {
       const recognizer = this.data.recognizer;
       this.vendor = recognizer.vendor;
@@ -33,7 +67,6 @@ class SttTask extends Task {
       //fallback
       this.fallbackVendor = recognizer.fallbackVendor || 'default';
       this.fallbackLanguage = recognizer.fallbackLanguage || 'default';
-      // label can be empty and should not have default value.
       this.fallbackLabel = recognizer.fallbackLabel;
 
       /* let credentials be supplied in the recognizer object at runtime */
@@ -82,8 +115,7 @@ class SttTask extends Task {
       this.language = cs.speechRecognizerLanguage;
       if (this.data.recognizer) this.data.recognizer.language = this.language;
     }
-    // label can be empty, should not assign application level label
-    if ('default' === this.label) {
+    if (!this.taskIncludeRecognizer) {
       this.label = cs.speechRecognizerLabel;
       if (this.data.recognizer) this.data.recognizer.label = this.label;
     }
@@ -96,17 +128,21 @@ class SttTask extends Task {
       this.fallbackLanguage = cs.fallbackSpeechRecognizerLanguage;
       if (this.data.recognizer) this.data.recognizer.fallbackLanguage = this.fallbackLanguage;
     }
-    // label can be empty, should not assign application level label
-    if ('default' === this.fallbackLabel) {
+    if (!this.taskIncludeRecognizer) {
       this.fallbackLabel = cs.fallbackSpeechRecognizerLabel;
       if (this.data.recognizer) this.data.recognizer.fallbackLabel = this.fallbackLabel;
     }
-    // If call is already fallback to 2nd ASR vendor
-    // use that.
+
     if (cs.hasFallbackAsr) {
-      this.vendor = this.fallbackVendor;
-      this.language = this.fallbackLanguage;
-      this.label = this.fallbackLabel;
+      if (this.taskIncludeRecognizer) {
+        // reset fallback ASR from previous run if this verb contains data.recognizer.
+        cs.hasFallbackAsr = false;
+      } else {
+        this.logger.debug('Call session has fallback to 2nd ASR, use 2nd recognizer configuration');
+        this.vendor = this.fallbackVendor;
+        this.language = this.fallbackLanguage;
+        this.label = this.fallbackLabel;
+      }
     }
     if (!this.data.recognizer.vendor) {
       this.data.recognizer.vendor = this.vendor;
@@ -178,10 +214,11 @@ class SttTask extends Task {
       writeAlerts({
         account_sid: cs.accountSid,
         alert_type: AlertType.STT_NOT_PROVISIONED,
-        vendor
+        vendor,
+        target_sid: cs.callSid
       }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
-      this.notifyTaskDone();
-      throw new Error(`No speech-to-text service credentials for ${vendor} have been configured`);
+      // the ASR might have fallback configuration, should not done task here.
+      throw new SpeechCredentialError(`No speech-to-text service credentials for ${vendor} have been configured`);
     }
 
     if (vendor === 'nuance' && credentials.client_id) {
@@ -205,14 +242,30 @@ class SttTask extends Task {
           region,
           roleArn
         });
-      this.logger.debug({roleArn}, `got aws access token ${servedFromCache ? 'from cache' : ''}`);
-      credentials = {...credentials, accessKeyId, secretAccessKey, sessionToken};
-    } else if (vendor === 'verbio' && credentials.client_id && credentials.client_secret) {
+      this.logger.debug({roleArn}, `(roleArn) got aws access token ${servedFromCache ? 'from cache' : ''}`);
+      // from role ARN, we will get SessionToken, but feature server use it as securityToken.
+      credentials = {...credentials, accessKeyId, secretAccessKey, securityToken: sessionToken};
+    }
+    else if (vendor === 'verbio' && credentials.client_id && credentials.client_secret) {
       const {access_token, servedFromCache} = await getVerbioAccessToken(credentials);
       this.logger.debug({client_id: credentials.client_id},
         `got verbio access token ${servedFromCache ? 'from cache' : ''}`);
       credentials.access_token = access_token;
     }
+    else if (vendor == 'aws' && !JAMBONES_AWS_TRANSCRIBE_USE_GRPC) {
+      /* get AWS access token */
+      const {speech_credential_sid, accessKeyId, secretAccessKey, securityToken, region } = credentials;
+      if (!securityToken) {
+        const { servedFromCache, ...newCredentials} = await getAwsAuthToken({
+          speech_credential_sid,
+          accessKeyId,
+          secretAccessKey,
+          region});
+        this.logger.debug({newCredentials}, `got aws security token ${servedFromCache ? 'from cache' : ''}`);
+        credentials = {...newCredentials, region};
+      }
+    }
+
     return credentials;
   }
 
@@ -222,12 +275,12 @@ class SttTask extends Task {
 
   async _initFallback() {
     assert(this.fallbackVendor, 'fallback failed without fallbackVendor configuration');
+    this.logger.info(`Failed to use primary STT provider, fallback to ${this.fallbackVendor}`);
     this.isHandledByPrimaryProvider = false;
     this.cs.hasFallbackAsr = true;
-    this.logger.info(`Failed to use primary STT provider, fallback to ${this.fallbackVendor}`);
-    this.vendor = this.fallbackVendor;
-    this.language = this.fallbackLanguage;
-    this.label = this.fallbackLabel;
+    this.vendor = this.cs.fallbackSpeechRecognizerVendor = this.fallbackVendor;
+    this.language = this.cs.fallbackSpeechRecognizerLanguage = this.fallbackLanguage;
+    this.label = this.cs.fallbackSpeechRecognizerLabel = this.fallbackLabel;
     this.data.recognizer.vendor = this.vendor;
     this.data.recognizer.language = this.language;
     this.data.recognizer.label = this.label;
@@ -261,6 +314,57 @@ class SttTask extends Task {
     });
   }
 
+  formatOpenAIPrompt(cs, {prompt, hintsTemplate, conversationHistoryTemplate, hints}) {
+    let conversationHistoryPrompt, hintsPrompt;
+
+    /* generate conversation history from template */
+    if (conversationHistoryTemplate) {
+      const {processed, turns} = processTurnString(conversationHistoryTemplate);
+      this.logger.debug({processed, turns}, 'SttTask: processed conversation history template');
+      conversationHistoryPrompt = cs.getFormattedConversation(turns || 4);
+      //this.logger.debug({conversationHistoryPrompt}, 'SttTask: conversation history');
+      if (conversationHistoryPrompt) {
+        conversationHistoryPrompt = processed.replace('{{turns}}', `\n${conversationHistoryPrompt}\nuser: `);
+      }
+    }
+
+    /* generate hints from template */
+    if (hintsTemplate && Array.isArray(hints) && hints.length > 0) {
+      hintsPrompt = hintsTemplate.replace('{{hints}}', hints);
+    }
+
+    /* combine into final prompt */
+    let finalPrompt = prompt || '';
+    if (hintsPrompt) {
+      finalPrompt = `${finalPrompt}\n${hintsPrompt}`;
+    }
+    if (conversationHistoryPrompt) {
+      finalPrompt = `${finalPrompt}\n${conversationHistoryPrompt}`;
+    }
+
+    this.logger.debug({
+      finalPrompt,
+      hints,
+      hintsPrompt,
+      conversationHistoryTemplate,
+      conversationHistoryPrompt
+    }, 'SttTask: formatted OpenAI prompt');
+    return finalPrompt?.trimStart();
+  }
+
+  /* some STT engines will keep listening after a final response, so no need to restart */
+  doesVendorContinueListeningAfterFinalTranscript(vendor) {
+    return (vendor.startsWith('custom:') || [
+      'soniox',
+      'aws',
+      'microsoft',
+      'deepgram',
+      'google',
+      'speechmatics',
+      'openai',
+    ].includes(vendor));
+  }
+
   _onCompileContext(ep, key, evt) {
     const {addKey} = this.cs.srf.locals.dbHelpers;
     this.logger.debug({evt}, `received cobalt compile context event, will cache under ${key}`);
@@ -309,6 +413,7 @@ class SttTask extends Task {
       message: 'STT failure reported by vendor',
       detail: evt.error,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
   }
 
@@ -321,6 +426,7 @@ class SttTask extends Task {
       alert_type: AlertType.STT_FAILURE,
       message: `Failed connecting to ${this.vendor} speech recognizer: ${reason}`,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
   }
 }

lib/tasks/task.js

@@ -1,5 +1,5 @@
 const Emitter = require('events');
-const uuidv4 = require('uuid-random');
+const crypto = require('crypto');
 const {TaskPreconditions} = require('../utils/constants');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 const WsRequestor = require('../utils/ws-requestor');
@@ -19,6 +19,7 @@ class Task extends Emitter {
     this.data = data;
     this.actionHook = this.data.actionHook;
     this.id = data.id;
+    this.taskId = crypto.randomUUID();
 
     this._killInProgress = false;
     this._completionPromise = new Promise((resolve) => this._completionResolver = resolve);
@@ -165,7 +166,7 @@ class Task extends Emitter {
       span.setAttributes({'http.body': JSON.stringify(params)});
       try {
         if (this.id) params.verb_id = this.id;
-        const json = await this.cs.requestor.request(type, this.actionHook, params, httpHeaders);
+        const json = await this.cs.requestor.request(type, this.actionHook, params, httpHeaders, span);
         span.setAttributes({'http.statusCode': 200});
         const isWsConnection = this.cs.requestor instanceof WsRequestor;
         if (!isWsConnection || (expectResponse && json && Array.isArray(json) && json.length)) {
@@ -209,7 +210,7 @@ class Task extends Emitter {
     const httpHeaders = b3 && {b3};
     span.setAttributes({'http.body': JSON.stringify(params)});
     try {
-      const json = await cs.requestor.request('verb:hook', hook, params, httpHeaders);
+      const json = await cs.requestor.request('verb:hook', hook, params, httpHeaders, span);
       span.setAttributes({'http.statusCode': 200});
       span.end();
       if (json && Array.isArray(json)) {
@@ -272,7 +273,7 @@ class Task extends Emitter {
   }
 
   async transferCallToFeatureServer(cs, sipAddress, opts) {
-    const uuid = uuidv4();
+    const uuid = crypto.randomUUID();
     const {addKey} = cs.srf.locals.dbHelpers;
     const obj = Object.assign({}, cs.application);
     delete obj.requestor;

lib/tasks/transcribe.js

@@ -12,10 +12,15 @@ const {
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
   TranscribeStatus,
-  AssemblyAiTranscriptionEvents
+  AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
+  OpenAITranscriptionEvents,
+  VerbioTranscriptionEvents,
+  SpeechmaticsTranscriptionEvents
 } = require('../utils/constants.json');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 const SttTask = require('./stt-task');
+const { SpeechCredentialError } = require('../utils/error');
 
 const STT_LISTEN_SPAN_NAME = 'stt-listen';
 
@@ -24,8 +29,8 @@ class TaskTranscribe extends SttTask {
     super(logger, opts, parentTask);
 
     this.transcriptionHook = this.data.transcriptionHook;
+    this.translationHook = this.data.translationHook;
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
-
     if (this.data.recognizer) {
       this.interim = !!this.data.recognizer.interim;
       this.separateRecognitionPerChannel = this.data.recognizer.separateRecognitionPerChannel;
@@ -73,7 +78,20 @@ class TaskTranscribe extends SttTask {
     return this.channel === 2 || this.separateRecognitionPerChannel && this.ep2;
   }
 
-  async exec(cs, {ep, ep2}) {
+  async exec(cs, obj) {
+    try {
+      await this.handling(cs, obj);
+    } catch (error) {
+      if (error instanceof SpeechCredentialError) {
+        this.logger.info('Transcribe failed due to SpeechCredentialError, finished!');
+        this.notifyTaskDone();
+        return;
+      }
+      throw error;
+    }
+  }
+
+  async handling(cs, {ep, ep2}) {
     await super.exec(cs, {ep, ep2});
 
     if (this.data.recognizer.vendor === 'nuance') {
@@ -84,11 +102,10 @@ class TaskTranscribe extends SttTask {
         ...this.data.recognizer.nuanceOptions
       };
     }
-    const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
 
     if (cs.hasGlobalSttHints) {
       const {hints, hintsBoost} = cs.globalSttHints;
-      this.data.recognizer.hints = this.data.recognizer.hints.concat(hints);
+      this.data.recognizer.hints = this.data.recognizer?.hints?.concat(hints);
       if (!this.data.recognizer.hintsBoost && hintsBoost) this.data.recognizer.hintsBoost = hintsBoost;
       this.logger.debug({hints: this.data.recognizer.hints, hintsBoost: this.data.recognizer.hintsBoost},
         'Transcribe:exec - applying global sttHints');
@@ -101,9 +118,6 @@ class TaskTranscribe extends SttTask {
       if (this.transcribing2) {
         await this._startTranscribing(cs, ep2, 2);
       }
-
-      updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid)
-        .catch(() => {/*already logged error */});
     } catch (err) {
       if (!(await this._startFallback(cs, ep, {error: err}))) {
         this.logger.info(err, 'TaskTranscribe:exec - error');
@@ -223,6 +237,13 @@ class TaskTranscribe extends SttTask {
         this.addCustomEventListener(ep, SonioxTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         break;
+
+      case 'verbio':
+        this.bugname = `${this.bugname_prefix}verbio_transcribe`;
+        this.addCustomEventListener(
+          ep, VerbioTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        break;
+
       case 'cobalt':
         this.bugname = `${this.bugname_prefix}cobalt_transcribe`;
         this.addCustomEventListener(ep, CobaltTranscriptionEvents.Transcription,
@@ -280,6 +301,49 @@ class TaskTranscribe extends SttTask {
           this._onVendorConnectFailure.bind(this, cs, ep, channel));
         break;
 
+      case 'voxist':
+        this.bugname = `${this.bugname_prefix}voxist_transcribe`;
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep,
+          VoxistTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
+        break;
+
+      case 'speechmatics':
+        this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
+        this.addCustomEventListener(
+          ep, SpeechmaticsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(
+          ep, SpeechmaticsTranscriptionEvents.Translation, this._onTranslation.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.Info,
+          this._onSpeechmaticsInfo.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.RecognitionStarted,
+          this._onSpeechmaticsRecognitionStarted.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.Connect,
+          this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        this.addCustomEventListener(ep, SpeechmaticsTranscriptionEvents.Error,
+          this._onSpeechmaticsError.bind(this, cs, ep));
+        break;
+
+      case 'openai':
+        this.bugname = `${this.bugname_prefix}openai_transcribe`;
+        this.addCustomEventListener(
+          ep, OpenAITranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep, OpenAITranscriptionEvents.Connect,
+          this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, OpenAITranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        this.addCustomEventListener(ep, OpenAITranscriptionEvents.Error,
+          this._onOpenAIErrror.bind(this, cs, ep));
+
+        this.modelSupportsConversationTracking = opts.OPENAI_MODEL !== 'whisper-1';
+        break;
+
       default:
         if (this.vendor.startsWith('custom:')) {
           this.bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
@@ -315,6 +379,25 @@ class TaskTranscribe extends SttTask {
   async _transcribe(ep) {
     this.logger.debug(
       `TaskTranscribe:_transcribe - starting transcription vendor ${this.vendor} bugname ${this.bugname}`);
+
+    /* special feature for openai: we can provide a prompt that includes recent conversation history */
+    let prompt;
+    if (this.vendor === 'openai') {
+      if (this.modelSupportsConversationTracking) {
+        prompt = this.formatOpenAIPrompt(this.cs, {
+          prompt: this.data.recognizer?.openaiOptions?.prompt,
+          hintsTemplate: this.data.recognizer?.openaiOptions?.promptTemplates?.hintsTemplate,
+          // eslint-disable-next-line max-len
+          conversationHistoryTemplate: this.data.recognizer?.openaiOptions?.promptTemplates?.conversationHistoryTemplate,
+          hints: this.data.recognizer?.hints,
+        });
+        this.logger.debug({prompt}, 'Gather:_startTranscribing - created an openai prompt');
+      }
+      else if (this.data.recognizer?.hints?.length > 0) {
+        prompt = this.data.recognizer?.hints.join(', ');
+      }
+    }
+
     await ep.startTranscription({
       vendor: this.vendor,
       interim: this.interim ? true : false,
@@ -406,8 +489,9 @@ class TaskTranscribe extends SttTask {
         this._startAsrTimer(channel);
 
         /* some STT engines will keep listening after a final response, so no need to restart */
-        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google']
-          .includes(this.vendor)) this._startTranscribing(cs, ep, channel);
+        if (!this.doesVendorContinueListeningAfterFinalTranscript(this.vendor)) {
+          this._startTranscribing(cs, ep, channel);
+        }
       }
       else {
         if (this.vendor === 'soniox') {
@@ -430,9 +514,7 @@ class TaskTranscribe extends SttTask {
         this.logger.debug({evt}, 'TaskTranscribe:_onTranscription - sending final transcript');
         this._resolve(channel, evt);
 
-        /* some STT engines will keep listening after a final response, so no need to restart */
-        if (!['soniox', 'aws', 'microsoft', 'deepgram', 'google'].includes(this.vendor) &&
-          !this.vendor.startsWith('custom:')) {
+        if (!this.doesVendorContinueListeningAfterFinalTranscript(this.vendor)) {
           this.logger.debug('TaskTranscribe:_onTranscription - restarting transcribe');
           this._startTranscribing(cs, ep, channel);
         }
@@ -457,12 +539,54 @@ class TaskTranscribe extends SttTask {
     }
   }
 
+  async _onTranslation(_cs, _ep, channel, evt, _fsEvent) {
+    this.logger.debug({evt}, 'TaskTranscribe:_onTranslation');
+    if (this.translationHook && evt.results?.length > 0) {
+      try {
+        const b3 = this.getTracingPropagation();
+        const httpHeaders = b3 && {b3};
+        const payload = {
+          ...this.cs.callInfo,
+          ...httpHeaders,
+          translation: {
+            channel,
+            language: evt.language,
+            translation: evt.results[0].content
+          }
+        };
+
+        this.logger.debug({payload}, 'sending translationHook');
+        const json = await this.cs.requestor.request('verb:hook', this.translationHook, payload);
+        this.logger.info({json}, 'completed translationHook');
+        if (json && Array.isArray(json) && !this.parentTask) {
+          const makeTask = require('./make_task');
+          const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
+          if (tasks && tasks.length > 0) {
+            this.logger.info({tasks: tasks}, `${this.name} replacing application with ${tasks.length} tasks`);
+            this.cs.replaceApplication(tasks);
+          }
+        }
+      } catch (err) {
+        this.logger.info(err, 'TranscribeTask:_onTranslation error');
+      }
+      if (this.parentTask) {
+        this.parentTask.emit('translation', evt);
+      }
+    }
+    if (this.killed) {
+      this.logger.debug('TaskTranscribe:_onTranslation exiting after receiving final transcription');
+      this._clearTimer();
+      this.notifyTaskDone();
+    }
+  }
+
   async _resolve(channel, evt) {
     if (evt.is_final) {
       /* we've got a final transcript, so end the otel child span for this channel */
       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)
         });
@@ -516,7 +640,8 @@ class TaskTranscribe extends SttTask {
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
-        'stt.resolve': 'timeout'
+        'stt.resolve': 'timeout',
+        'stt.label': this.label || 'None',
       });
       this.childSpan[channel - 1].span.end();
     }
@@ -528,12 +653,22 @@ class TaskTranscribe extends SttTask {
   }
 
   _onMaxDurationExceeded(cs, ep, channel) {
-    this.logger.debug(`TaskTranscribe:_onMaxDurationExceeded on channel ${channel}`);
+    this.restartDueToError(ep, channel, 'Max duration exceeded');
+  }
+
+  _onMaxBufferExceeded(cs, ep, channel) {
+    this.restartDueToError(ep, channel, 'Max buffer exceeded');
+  }
+
+  restartDueToError(ep, channel, reason) {
+    this.logger.debug(`TaskTranscribe:${reason} on channel ${channel}`);
     if (this.paused) return;
+
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
-        'stt.resolve': 'max duration exceeded'
+        'stt.resolve': reason,
+        'stt.label': this.label || 'None',
       });
       this.childSpan[channel - 1].span.end();
     }
@@ -559,7 +694,6 @@ class TaskTranscribe extends SttTask {
         bugname: this.bugname
       })
         .catch((err) => this.logger.error({err}, `Error stopping transcription for primary vendor ${this.vendor}`));
-      const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
       try {
         this.notifyError({ msg: 'ASR error',
           details:`STT Vendor ${this.vendor} error: ${evt.error || evt.reason}`, failover: 'in progress'});
@@ -570,7 +704,6 @@ class TaskTranscribe extends SttTask {
         }
         this[`_speechHandlersSet_${channel}`] = false;
         this._startTranscribing(cs, _ep, channel);
-        updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
         return true;
       } catch (error) {
         this.notifyError({ msg: 'ASR error',
@@ -591,6 +724,14 @@ class TaskTranscribe extends SttTask {
       return;
     }
     this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
+    if (this.vendor === 'microsoft' &&
+      evt.error?.includes('Due to service inactivity, the client buffer exceeded maximum size. Resetting the buffer')) {
+      let channel = 1;
+      if (this.ep !== _ep) {
+        channel = 2;
+      }
+      return this._onMaxBufferExceeded(cs, _ep, channel);
+    }
     if (this.paused) return;
     const {writeAlerts, AlertType} = cs.srf.locals;
 
@@ -605,6 +746,7 @@ class TaskTranscribe extends SttTask {
       alert_type: AlertType.STT_FAILURE,
       message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
       vendor: this.vendor,
+      target_sid: cs.callSid
     }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
     if (!(await this._startFallback(cs, _ep, evt))) {
       this.notifyTaskDone();
@@ -616,7 +758,8 @@ class TaskTranscribe extends SttTask {
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
-        'stt.resolve': 'connection failure'
+        'stt.resolve': 'connection failure',
+        'stt.label': this.label || 'None',
       });
       this.childSpan[channel - 1].span.end();
     }
@@ -625,6 +768,26 @@ class TaskTranscribe extends SttTask {
     }
   }
 
+  async _onSpeechmaticsRecognitionStarted(_cs, _ep, evt) {
+    this.logger.debug({evt}, 'TaskGather:_onSpeechmaticsRecognitionStarted');
+  }
+
+  async _onSpeechmaticsInfo(_cs, _ep, evt) {
+    this.logger.debug({evt}, 'TaskGather:_onSpeechmaticsInfo');
+  }
+
+  async _onSpeechmaticsError(cs, _ep, evt) {
+    // eslint-disable-next-line no-unused-vars
+    const {message, ...e} = evt;
+    this._onVendorError(cs, _ep, {error: JSON.stringify(e)});
+  }
+
+  async _onOpenAIErrror(cs, _ep, evt) {
+    // eslint-disable-next-line no-unused-vars
+    const {message, ...e} = evt;
+    this._onVendorError(cs, _ep, {error: JSON.stringify(e)});
+  }
+
   _startAsrTimer(channel) {
     if (this.vendor === 'deepgram') return; // no need
     assert(this.isContinuousAsr);

lib/tasks/tts-task.js

@@ -1,5 +1,7 @@
 const Task = require('./task');
 const { TaskPreconditions } = require('../utils/constants');
+const { SpeechCredentialError } = require('../utils/error');
+const dbUtils = require('../utils/db-utils');
 
 class TtsTask extends Task {
 
@@ -10,41 +12,159 @@ class TtsTask extends Task {
     this.preconditions = TaskPreconditions.Endpoint;
 
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
+    /**
+     * Task use taskIncludeSynthesizer to identify
+     * if taskIncludeSynthesizer === true, use label from verb.synthesizer, even it's empty
+     * if taskIncludeSynthesizer === false, use label from application.synthesizer
+     */
+    this.taskIncludeSynthesizer = !!this.data.synthesizer;
     this.synthesizer = this.data.synthesizer || {};
     this.disableTtsCache = this.data.disableTtsCache;
     this.options = this.synthesizer.options || {};
+    this.instructions = this.data.instructions;
   }
 
   async exec(cs) {
     super.exec(cs);
+    if (cs.synthesizer) {
+      this.options = {...cs.synthesizer.options, ...this.options};
+      this.data.synthesizer = this.data.synthesizer || {};
+      for (const k in cs.synthesizer) {
+        const newValue = this.data.synthesizer && this.data.synthesizer[k] !== undefined ?
+          this.data.synthesizer[k] :
+          cs.synthesizer[k];
+
+        if (Array.isArray(newValue)) {
+          this.data.synthesizer[k] = [...(this.data.synthesizer[k] || []), ...cs.synthesizer[k]];
+        } else if (typeof newValue === 'object' && newValue !== null) {
+          this.data.synthesizer[k] = { ...(this.data.synthesizer[k] || {}), ...cs.synthesizer[k] };
+        } else {
+          this.data.synthesizer[k] = newValue;
+        }
+      }
+    }
+
+    const fullText = Array.isArray(this.text) ? this.text.join(' ') : this.text;
+    // in case dub verb, text might not be set.
+    if (fullText?.length > 0) {
+      cs.emit('botSaid', fullText);
+    }
   }
 
-  async _synthesizeWithSpecificVendor(cs, ep, {
-    vendor,
-    language,
-    voice,
-    label,
-    disableTtsStreaming,
-    preCache
-  }) {
+  getTtsVendorData(cs) {
+    const vendor = this.synthesizer.vendor && this.synthesizer.vendor !== 'default' ?
+      this.synthesizer.vendor :
+      cs.speechSynthesisVendor;
+    const language = this.synthesizer.language && this.synthesizer.language !== 'default' ?
+      this.synthesizer.language :
+      cs.speechSynthesisLanguage ;
+    const voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
+      this.synthesizer.voice :
+      cs.speechSynthesisVoice;
+    const label = this.taskIncludeSynthesizer ? this.synthesizer.label : cs.speechSynthesisLabel;
+    return {vendor, language, voice, label};
+  }
+
+  async setTtsStreamingChannelVars(vendor, language, voice, credentials, ep) {
+    const {api_key, model_id, custom_tts_streaming_url, auth_token} = credentials;
+    let obj;
+
+    this.logger.debug({credentials},
+      `setTtsStreamingChannelVars: vendor: ${vendor}, language: ${language}, voice: ${voice}`);
+
+    switch (vendor) {
+      case 'deepgram':
+        obj = {
+          DEEPGRAM_API_KEY: api_key,
+          DEEPGRAM_TTS_STREAMING_MODEL: voice
+        };
+        break;
+      case 'cartesia':
+        obj = {
+          CARTESIA_API_KEY: api_key,
+          CARTESIA_TTS_STREAMING_MODEL_ID: model_id,
+          CARTESIA_TTS_STREAMING_VOICE_ID: voice,
+          CARTESIA_TTS_STREAMING_LANGUAGE: language || 'en',
+        };
+        break;
+      case 'elevenlabs':
+        const {stability, similarity_boost, use_speaker_boost, style, speed} = this.options.voice_settings || {};
+        obj = {
+          ELEVENLABS_API_KEY: api_key,
+          ELEVENLABS_TTS_STREAMING_MODEL_ID: model_id,
+          ELEVENLABS_TTS_STREAMING_VOICE_ID: voice,
+          // 20/12/2024 - only eleven_turbo_v2_5 support multiple language
+          ...(['eleven_turbo_v2_5'].includes(model_id) && {ELEVENLABS_TTS_STREAMING_LANGUAGE: language}),
+          ...(stability && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_STABILITY: stability}),
+          ...(similarity_boost && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_SIMILARITY_BOOST: similarity_boost}),
+          ...(use_speaker_boost && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_USE_SPEAKER_BOOST: use_speaker_boost}),
+          ...(style && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_STYLE: style}),
+          // speed has value 0.7 to 1.2, 1.0 is default, make sure we send the value event it's 0
+          ...(speed !== null && speed !== undefined && {ELEVENLABS_TTS_STREAMING_VOICE_SETTINGS_SPEED: `${speed}`}),
+          ...(this.options.pronunciation_dictionary_locators &&
+            Array.isArray(this.options.pronunciation_dictionary_locators) && {
+            ELEVENLABS_TTS_STREAMING_PRONUNCIATION_DICTIONARY_LOCATORS:
+            JSON.stringify(this.options.pronunciation_dictionary_locators)
+          }),
+        };
+        break;
+      case 'rimelabs':
+        const {
+          pauseBetweenBrackets, phonemizeBetweenBrackets, inlineSpeedAlpha, speedAlpha, reduceLatency
+        } = this.options;
+        obj = {
+          RIMELABS_API_KEY: api_key,
+          RIMELABS_TTS_STREAMING_MODEL_ID: model_id,
+          RIMELABS_TTS_STREAMING_VOICE_ID: voice,
+          RIMELABS_TTS_STREAMING_LANGUAGE: language || 'en',
+          ...(pauseBetweenBrackets && {RIMELABS_TTS_STREAMING_PAUSE_BETWEEN_BRACKETS: pauseBetweenBrackets}),
+          ...(phonemizeBetweenBrackets &&
+            {RIMELABS_TTS_STREAMING_PHONEMIZE_BETWEEN_BRACKETS: phonemizeBetweenBrackets}),
+          ...(inlineSpeedAlpha && {RIMELABS_TTS_STREAMING_INLINE_SPEED_ALPHA: inlineSpeedAlpha}),
+          ...(speedAlpha && {RIMELABS_TTS_STREAMING_SPEED_ALPHA: speedAlpha}),
+          ...(reduceLatency && {RIMELABS_TTS_STREAMING_REDUCE_LATENCY: reduceLatency})
+        };
+        break;
+      default:
+        if (vendor.startsWith('custom:')) {
+          const use_tls = custom_tts_streaming_url.startsWith('wss://');
+          obj = {
+            CUSTOM_TTS_STREAMING_HOST: custom_tts_streaming_url.replace(/^(ws|wss):\/\//, ''),
+            CUSTOM_TTS_STREAMING_API_KEY: auth_token,
+            CUSTOM_TTS_STREAMING_VOICE_ID: voice,
+            CUSTOM_TTS_STREAMING_LANGUAGE: language || 'en',
+            CUSTOM_TTS_STREAMING_USE_TLS: use_tls
+          };
+        } else {
+          throw new Error(`vendor ${vendor} is not supported for tts streaming yet`);
+        }
+    }
+    this.logger.debug({vendor, credentials, obj}, 'setTtsStreamingChannelVars');
+
+    await ep.set(obj);
+  }
+
+  async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
     const {srf, accountSid:account_sid} = cs;
-    const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
     const {writeAlerts, AlertType, stats} = srf.locals;
     const {synthAudio} = srf.locals.dbHelpers;
     const engine = this.synthesizer.engine || cs.synthesizer?.engine || 'neural';
     const salt = cs.callSid;
 
     let credentials = cs.getSpeechCredentials(vendor, 'tts', label);
+    if (!credentials) {
+      throw new SpeechCredentialError(
+        `No text-to-speech service credentials for ${vendor} with labels: ${label} have been configured`);
+    }
     /* parse Nuance voices into name and model */
-    let model;
     if (vendor === 'nuance' && voice) {
       const arr = /([A-Za-z-]*)\s+-\s+(enhanced|standard)/.exec(voice);
       if (arr) {
         voice = arr[1];
-        model = arr[2];
+        this.model = arr[2];
       }
     } else if (vendor === 'deepgram') {
-      model = voice;
+      this.model = voice;
     }
 
     /* allow for microsoft custom region voice and api_key to be specified as an override */
@@ -62,30 +182,68 @@ class TtsTask extends Task {
       credentials.optimize_streaming_latency = this.options.optimize_streaming_latency
       || credentials.optimize_streaming_latency;
       voice = this.options.voice_id || voice;
+    } else if (vendor === 'rimelabs') {
+      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;
+    } else if (vendor === 'verbio') {
+      credentials = credentials || {};
+      credentials.engine_version = this.options.engine_version || credentials.engine_version;
+    } else if (vendor === 'playht') {
+      credentials = credentials || {};
+      credentials.voice_engine = this.options.voice_engine || credentials.voice_engine;
+    } else if (vendor === 'google' && typeof voice === 'string' && voice.startsWith('custom_')) {
+      const {lookupGoogleCustomVoice} = dbUtils(this.logger, cs.srf);
+      const arr = /custom_(.*)/.exec(voice);
+      if (arr) {
+        const google_custom_voice_sid = arr[1];
+        const [custom_voice] = await lookupGoogleCustomVoice(google_custom_voice_sid);
+        if (custom_voice.use_voice_cloning_key) {
+          voice = {
+            voice_cloning_key: custom_voice.voice_cloning_key,
+          };
+        }
+      }
+    } else if (vendor === 'cartesia') {
+      credentials.model_id = this.options.model_id || credentials.model_id;
     }
 
-    ep.set({
-      tts_engine: vendor,
-      tts_voice: voice,
-      cache_speech_handles: 1,
-    }).catch((err) => this.logger.info({err}, `${this.name}: Error setting tts_engine on endpoint`));
+    this.model_id = credentials.model_id;
 
-    if (!preCache) this.logger.info({vendor, language, voice, model}, `${this.name}:exec`);
+    /**
+     * note on cache_speech_handles.  This was found to be risky.
+     * It can cause a crash in the following sequence on a single call:
+     * 1. Stream tts on vendor A with cache_speech_handles=1, then
+     * 2. Stream tts on vendor B with cache_speech_handles=1
+     *
+     * we previously tried to track when vendors were switched and manage the flag accordingly,
+     * but it difficult to track all the scenarios and the benefit (slightly faster start to tts playout)
+     * is probably minimal.  DH.
+     */
+    ep.set({
+      tts_engine: vendor.startsWith('custom:') ? 'custom' : vendor,
+      tts_voice: voice,
+      //cache_speech_handles: !cs.currentTtsVendor || cs.currentTtsVendor === vendor ? 1 : 0,
+      cache_speech_handles: 0,
+    }).catch((err) => this.logger.info({err}, 'Error setting tts_engine on endpoint'));
+    // set the current vendor on the call session
+    // If vendor is changed from the previous one, then reset the cache_speech_handles flag
+    //cs.currentTtsVendor = vendor;
+
+    if (!preCache && !this._disableTracing)
+      this.logger.debug({vendor, language, voice, model: this.model}, 'TaskSay:exec');
     try {
       if (!credentials) {
         writeAlerts({
           account_sid,
           alert_type: AlertType.TTS_NOT_PROVISIONED,
-          vendor
+          vendor,
+          target_sid: cs.callSid
         }).catch((err) => this.logger.info({err}, 'Error generating alert for no tts'));
-        this.notifyError({
-          msg: 'TTS error',
-          details:`No speech credentials provisioned for selected vendor ${vendor}`
-        });
-        throw new Error('no provisioned speech credentials for TTS');
+        throw new SpeechCredentialError('no provisioned speech credentials for TTS');
       }
-      // synthesize all of the text elements
-      let lastUpdated = false;
 
       /* produce an audio segment from the provided text */
       const generateAudio = async(text) => {
@@ -93,11 +251,12 @@ class TtsTask extends Task {
         if (text.startsWith('silence_stream://')) return text;
 
         /* otel: trace time for tts */
-        if (!preCache && !this.parentTask)  {
+        if (!preCache && !this._disableTracing)  {
           const {span} = this.startChildSpan('tts-generation', {
             'tts.vendor': vendor,
             'tts.language': language,
-            'tts.voice': voice
+            'tts.voice': voice,
+            'tts.label': label || 'None',
           });
           this.otelSpan = span;
         }
@@ -105,31 +264,27 @@ class TtsTask extends Task {
           const {filePath, servedFromCache, rtt} = await synthAudio(stats, {
             account_sid,
             text,
+            instructions: this.instructions,
             vendor,
             language,
             voice,
             engine,
-            model,
+            model: this.model,
             salt,
             credentials,
             options: this.options,
             disableTtsCache : this.disableTtsCache,
-            disableTtsStreaming,
-            preCache
+            renderForCaching: preCache
           });
           if (!filePath.startsWith('say:')) {
-            this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
+            this.logger.debug(`Say: file ${filePath}, served from cache ${servedFromCache}`);
             if (filePath) cs.trackTmpFile(filePath);
             if (this.otelSpan) {
               this.otelSpan.setAttributes({'tts.cached': servedFromCache});
               this.otelSpan.end();
               this.otelSpan = null;
             }
-            if (!servedFromCache && !lastUpdated) {
-              lastUpdated = true;
-              updateSpeechCredentialLastUsed(credentials.speech_credential_sid).catch(() => {/* logged error */});
-            }
-            if (!servedFromCache && rtt && !preCache) {
+            if (!servedFromCache && rtt && !preCache  && !this._disableTracing) {
               this.notifyStatus({
                 event: 'synthesized-audio',
                 vendor,
@@ -140,7 +295,7 @@ class TtsTask extends Task {
             }
           }
           else {
-            this.logger.debug('a streaming tts api will be used');
+            this.logger.debug('Say: a streaming tts api will be used');
             const modifiedPath = filePath.replace('say:{', `say:{session-uuid=${ep.uuid},`);
             return modifiedPath;
           }
@@ -152,9 +307,9 @@ class TtsTask extends Task {
             account_sid: cs.accountSid,
             alert_type: AlertType.TTS_FAILURE,
             vendor,
-            detail: err.message
+            detail: err.message,
+            target_sid: cs.callSid
           }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
-          this.notifyError({msg: 'TTS error', details: err.message || err});
           throw err;
         }
       };
@@ -165,6 +320,7 @@ class TtsTask extends Task {
       this.logger.info(err, 'TaskSay:exec error');
       throw err;
     }
+
   }
 
   _validateURL(urlString) {

lib/utils/action-hook-delay.js

@@ -2,7 +2,6 @@ const makeTask = require('../tasks/make_task');
 const Emitter = require('events');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 const {TaskName} = require('../utils/constants');
-const assert = require('assert');
 
 /**
  * ActionHookDelayProcessor
@@ -25,10 +24,12 @@ class ActionHookDelayProcessor extends Emitter {
     this._active = false;
 
     const enabled = this.init(opts);
-    if (enabled && (!this.actions || !Array.isArray(this.actions) || this.actions.length === 0)) {
+    if (enabled && this.noResponseTimeout &&
+          (!this.actions || !Array.isArray(this.actions) || this.actions.length === 0)) {
       throw new Error('ActionHookDelayProcessor: no actions specified');
     }
-    else if (enabled && this.actions.some((a) => !a.verb || ![TaskName.Say, TaskName.Play].includes(a.verb))) {
+    else if (enabled && this.actions &&
+          this.actions.some((a) => !a.verb || ![TaskName.Say, TaskName.Play].includes(a.verb))) {
       throw new Error(`ActionHookDelayProcessor: invalid actions specified: ${JSON.stringify(this.actions)}`);
     }
   }
@@ -51,8 +52,9 @@ class ActionHookDelayProcessor extends Emitter {
 
     this.actions = opts.actions;
     this.retries = opts.retries || 0;
-    this.noResponseTimeout = opts.noResponseTimeout || 0;
+    this.noResponseTimeout = opts.noResponseTimeout;
     this.noResponseGiveUpTimeout = opts.noResponseGiveUpTimeout;
+    this.giveUpActions = opts.giveUpActions;
 
     // return false if these options actually disable the ahdp
     return ('enable' in opts && opts.enable === true) ||
@@ -66,11 +68,16 @@ class ActionHookDelayProcessor extends Emitter {
       this.logger.debug('ActionHookDelayProcessor#start: already started due to prior gather which is continuing');
       return;
     }
-    assert(!this._noResponseTimer);
     this._active = true;
     this._retryCount = 0;
-    const timeoutMs =  this.noResponseTimeout === 0 ? 1 : this.noResponseTimeout * 1000;
-    this._noResponseTimer = setTimeout(this._onNoResponseTimer.bind(this), timeoutMs);
+    if (this.noResponseTimeout > 0) {
+      const timeoutMs =  this.noResponseTimeout * 1000;
+      this._noResponseTimer = setTimeout(this._onNoResponseTimer.bind(this), timeoutMs);
+    } else {
+      this.logger.debug(
+        'ActionHookDelayProcessor#start: noResponseTimeout is 0 or undefined hence not calling _onNoResponseTimer'
+      );
+    }
 
     if (this.noResponseGiveUpTimeout > 0) {
       const timeoutMs = this.noResponseGiveUpTimeout * 1000;
@@ -79,7 +86,6 @@ class ActionHookDelayProcessor extends Emitter {
   }
 
   async stop() {
-    this.logger.debug('ActionHookDelayProcessor#stop');
     this._active = false;
 
     if (this._noResponseTimer) {
@@ -91,25 +97,19 @@ class ActionHookDelayProcessor extends Emitter {
       this._noResponseGiveUpTimer = null;
     }
     if (this._taskInProgress) {
-      this.logger.debug(`ActionHookDelayProcessor#stop: killing task in progress: ${this._taskInProgress.name}`);
+      this.logger.debug(`ActionHookDelayProcessor#stop: stopping ${this._taskInProgress.name}`);
 
-      /** if we are doing a play, kill it immediately
-       * if we are doing a say, wait for it to finish
-       */
-      if (TaskName.Say === this._taskInProgress.name) {
-        this._sayResolver = () => {
-          this.logger.debug('ActionHookDelayProcessor#stop: say is done, continue on..');
-          this._taskInProgress.kill(this.cs);
-          this._taskInProgress = null;
-        };
-        this.logger.debug('ActionHookDelayProcessor#stop returning promise');
-        return new Promise((resolve) => this._sayResolver = resolve);
-      }
-      else {
-        /* play */
-        this._taskInProgress.kill(this.cs);
+      this._sayResolver = () => {
+        this.logger.debug('ActionHookDelayProcessor#stop: play/say is done, continue on..');
+        //this._taskInProgress.kill(this.cs);
         this._taskInProgress = null;
+      };
+
+      /* we let Say finish, but interrupt Play */
+      if (TaskName.Play === this._taskInProgress.name) {
+        await this._taskInProgress.kill(this.cs);
       }
+      return new Promise((resolve) => this._sayResolver = resolve);
     }
     this.logger.debug('ActionHookDelayProcessor#stop returning');
   }
@@ -126,7 +126,12 @@ class ActionHookDelayProcessor extends Emitter {
     try {
       this._taskInProgress = makeTask(this.logger, t[0]);
       this._taskInProgress.disableTracing = true;
-      this._taskInProgress.exec(this.cs, {ep: this.ep});
+      this._taskInProgress.exec(this.cs, {ep: this.ep}).catch((err) => {
+        this.logger.info(`ActionHookDelayProcessor#_onNoResponseTimer: error playing file: ${err.message}`);
+        this._taskInProgress = null;
+        this.ep.removeAllListeners('playback-start');
+        this.ep.removeAllListeners('playback-stop');
+      });
     } catch (err) {
       this.logger.info(err, 'ActionHookDelayProcessor#_onNoResponseTimer: error starting action');
       this._taskInProgress = null;
@@ -137,7 +142,9 @@ class ActionHookDelayProcessor extends Emitter {
       this.logger.debug({evt}, 'got playback-start');
       if (!this._active) {
         this.logger.info({evt}, 'ActionHookDelayProcessor#_onNoResponseTimer: killing audio immediately');
-        this.ep.api('uuid_break', this.ep.uuid)
+
+        /* note: in race condition we may have just hung up and cs.ep cleared */
+        this.ep?.api('uuid_break', this.ep?.uuid)
           .catch((err) => this.logger.info(err,
             'ActionHookDelayProcessor#_onNoResponseTimer Error killing audio'));
       }
@@ -147,7 +154,7 @@ class ActionHookDelayProcessor extends Emitter {
       this._taskInProgress = null;
       if (this._sayResolver) {
         /* we were waiting for the play to finish before continuing to next task */
-        this.logger.debug({evt}, 'got playback-stop');
+        this.logger.debug({evt}, 'ActionHookDelayProcessor#_onNoResponseTimer got playback-stop');
         this._sayResolver();
         this._sayResolver = null;
       }
@@ -166,9 +173,14 @@ class ActionHookDelayProcessor extends Emitter {
 
   _onNoResponseGiveUpTimer() {
     this._active = false;
-    this.logger.info('ActionHookDelayProcessor#_onNoResponseGiveUpTimer');
-    this.stop().catch((err) => {});
-    this.emit('giveup');
+    if (!this.giveUpActions) {
+      this.logger.info('ActionHookDelayProcessor#_onNoResponseGiveUpTimer');
+      this.stop().catch((err) => {});
+      this.emit('giveup');
+    } else {
+      this.logger.info('ActionHookDelayProcessor#_onNoResponseGiveUpTimer - giveUpActions');
+      this.emit('giveupWithTasks', this.giveUpActions);
+    }
   }
 }
 

lib/utils/amd-utils.js

@@ -45,6 +45,7 @@ if (VMD_HINTS_FILE) {
   });
 }
 
+
 class Amd extends Emitter {
   constructor(logger, cs, opts) {
     super();
@@ -68,6 +69,8 @@ class Amd extends Emitter {
     this.getIbmAccessToken = getIbmAccessToken;
     const {setChannelVarsForStt} = require('./transcription-utils')(logger);
     this.setChannelVarsForStt = setChannelVarsForStt;
+    this.digitCount = opts.digitCount || 0;
+    this.numberRegEx = RegExp(`[0-9]{${this.digitCount}}`);
 
     const {
       noSpeechTimeoutMs = 5000,
@@ -153,7 +156,7 @@ class Amd extends Emitter {
       const wordCount = t.alternatives[0].transcript.split(' ').length;
       const final = t.is_final;
 
-      const foundHint = hints.find((h) =>  t.alternatives[0].transcript.includes(h));
+      const foundHint = hints.find((h) =>  t.alternatives[0].transcript.toLowerCase().includes(h.toLowerCase()));
       if (foundHint) {
         /* we detected a common voice mail greeting */
         this.logger.debug(`Amd:evaluateTranscription: found hint ${foundHint}`);
@@ -163,6 +166,14 @@ class Amd extends Emitter {
           language: t.language_code
         });
       }
+      else if (this.digitCount != 0 && this.numberRegEx.test(t.alternatives[0].transcript)) {
+        /* a string of numbers is typically a machine */
+        this.emit(this.decision = AmdEvents.MachineDetected, {
+          reason: 'digit count',
+          greeting: t.alternatives[0].transcript,
+          language: t.language_code
+        });
+      }
       else if (final && wordCount < this.thresholdWordCount) {
         /* a short greeting is typically a human */
         this.emit(this.decision = AmdEvents.HumanDetected, {
@@ -210,7 +221,8 @@ module.exports = (logger) => {
         account_sid: cs.accountSid,
         alert_type: AlertType.STT_FAILURE,
         vendor: vendor,
-        detail: err.message
+        detail: err.message,
+        target_sid: cs.callSid
       });
     }).catch((err) => logger.info({err}, 'Error generating alert for tts failure'));
 
@@ -245,7 +257,10 @@ module.exports = (logger) => {
     const amd = ep.amd = new Amd(logger, cs, opts);
     const {vendor, language} = amd;
     let sttCredentials = amd.sttCredentials;
-    const hints = voicemailHints[language] || [];
+    // hints from configuration might be too long for specific language and vendor that make transcribe freeswitch
+    // modules cannot connect to the vendor. hints is used in next step to validate if the transcription
+    // matchs voice mail hints.
+    const hints = [];
 
     if (vendor === 'nuance' && sttCredentials.client_id) {
       /* get nuance access token */

lib/utils/aws-sns-lifecycle.js

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

lib/utils/background-task-manager.js

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

lib/utils/base-requestor.js

@@ -1,6 +1,7 @@
 const assert = require('assert');
 const Emitter = require('events');
 const crypto = require('crypto');
+const parseUrl = require('parse-url');
 const timeSeries = require('@jambonz/time-series');
 const {NODE_ENV, JAMBONES_TIME_SERIES_HOST} = require('../config');
 let alerter ;
@@ -21,6 +22,10 @@ class BaseRequestor extends Emitter {
     const {stats} = require('../../').srf.locals;
     this.stats = stats;
 
+    const u = this._parsedUrl = parseUrl(this.url);
+    if (u.port) this._baseUrl = `${u.protocol}://${u.resource}:${u.port}`;
+    else this._baseUrl = `${u.protocol}://${u.resource}`;
+
     if (!alerter) {
       alerter = timeSeries(logger, {
         host: JAMBONES_TIME_SERIES_HOST,
@@ -30,6 +35,10 @@ class BaseRequestor extends Emitter {
     }
   }
 
+  get baseUrl() {
+    return this._baseUrl;
+  }
+
   get Alerter() {
     return alerter;
   }

lib/utils/constants.json

@@ -14,6 +14,7 @@
     "Leave": "leave",
     "Lex": "lex",
     "Listen": "listen",
+    "Llm": "llm",
     "Message": "message",
     "Pause": "pause",
     "Play": "play",
@@ -27,10 +28,11 @@
     "SipRedirect": "sip:redirect",
     "Say": "say",
     "SayLegacy": "say:legacy",
+    "Stream": "stream",
     "Tag": "tag",
     "Transcribe": "transcribe"
   },
-  "AllowedSipRecVerbs": ["answer", "config", "gather", "transcribe", "listen", "tag"],
+  "AllowedSipRecVerbs": ["answer", "config", "gather", "transcribe", "listen", "tag", "hangup", "sip:decline"],
   "AllowedConfirmSessionVerbs": ["config", "gather", "plays", "say", "tag"],
   "CallStatus": {
     "Trying": "trying",
@@ -126,6 +128,27 @@
     "NoSpeechDetected": "azure_transcribe::no_speech_detected",
     "VadDetected": "azure_transcribe::vad_detected"
   },
+  "SpeechmaticsTranscriptionEvents": {
+    "Transcription": "speechmatics_transcribe::transcription",
+    "Translation": "speechmatics_transcribe::translation",
+    "Info": "speechmatics_transcribe::info",
+    "RecognitionStarted": "speechmatics_transcribe::recognition_started",
+    "ConnectFailure": "speechmatics_transcribe::connect_failed",
+    "Connect": "speechmatics_transcribe::connect",
+    "Error": "speechmatics_transcribe::error"
+  },
+  "OpenAITranscriptionEvents": {
+    "Transcription": "openai_transcribe::transcription",
+    "Translation": "openai_transcribe::translation",
+    "SpeechStarted": "openai_transcribe::speech_started",
+    "SpeechStopped": "openai_transcribe::speech_stopped",
+    "PartialTranscript": "openai_transcribe::partial_transcript",
+    "Info": "openai_transcribe::info",
+    "RecognitionStarted": "openai_transcribe::recognition_started",
+    "ConnectFailure": "openai_transcribe::connect_failed",
+    "Connect": "openai_transcribe::connect",
+    "Error": "openai_transcribe::error"
+  },
   "JambonzTranscriptionEvents": {
     "Transcription": "jambonz_transcribe::transcription",
     "ConnectFailure": "jambonz_transcribe::connect_failed",
@@ -138,6 +161,12 @@
     "ConnectFailure": "assemblyai_transcribe::connect_failed",
     "Connect": "assemblyai_transcribe::connect"
   },
+  "VoxistTranscriptionEvents": {
+    "Transcription": "voxist_transcribe::transcription",
+    "Error": "voxist_transcribe::error",
+    "ConnectFailure": "voxist_transcribe::connect_failed",
+    "Connect": "voxist_transcribe::connect"
+  },
   "VadDetection": {
     "Detection": "vad_detect:detection"
   },
@@ -158,6 +187,41 @@
     "StandbyEnter": "standby-enter",
     "StandbyExit": "standby-exit"
   },
+  "LlmEvents_OpenAI": {
+    "Error": "error",
+    "Connect": "openai_s2s::connect",
+    "ConnectFailure": "openai_s2s::connect_failed",
+    "Disconnect": "openai_s2s::disconnect",
+    "ServerEvent": "openai_s2s::server_event"
+  },
+  "LlmEvents_Google": {
+    "Error": "error",
+    "Connect": "google_s2s::connect",
+    "ConnectFailure": "google_s2s::connect_failed",
+    "Disconnect": "google_s2s::disconnect",
+    "ServerEvent": "google_s2s::server_event"
+  },
+  "LlmEvents_Elevenlabs": {
+    "Error": "error",
+    "Connect": "elevenlabs_s2s::connect",
+    "ConnectFailure": "elevenlabs_s2s::connect_failed",
+    "Disconnect": "elevenlabs_s2s::disconnect",
+    "ServerEvent": "elevenlabs_s2s::server_event"
+  },
+  "LlmEvents_VoiceAgent": {
+    "Error": "error",
+    "Connect": "voice_agent_s2s::connect",
+    "ConnectFailure": "voice_agent_s2s::connect_failed",
+    "Disconnect": "voice_agent_s2s::disconnect",
+    "ServerEvent": "voice_agent_s2s::server_event"
+  },
+  "LlmEvents_Ultravox": {
+    "Error": "error",
+    "Connect": "ultravox_s2s::connect",
+    "ConnectFailure": "ultravox_s2s::connect_failed",
+    "Disconnect": "ultravox_s2s::disconnect",
+    "ServerEvent": "ultravox_s2s::server_event"
+  },
   "QueueResults": {
     "Bridged": "bridged",
     "Error": "error",
@@ -172,7 +236,8 @@
   },
   "KillReason": {
     "Hangup": "hangup",
-    "Replaced": "replaced"
+    "Replaced": "replaced",
+    "MediaTimeout": "media_timeout"
   },
   "HookMsgTypes": [
     "session:new",
@@ -184,6 +249,10 @@
     "dial:confirm",
     "verb:hook",
     "verb:status",
+    "llm:event",
+    "llm:tool-call",
+    "tts:tokens-result",
+    "tts:streaming-event",
     "jambonz:error"
   ],
   "RecordState": {
@@ -202,7 +271,59 @@
     "ToneTimeout": "amd_tone_timeout",
     "Stopped": "amd_stopped"
   },
+  "MediaPath": {
+    "NoMedia": "no-media",
+    "PartialMedia": "partial-media",
+    "FullMedia": "full-media"
+  },
+  "DeepgramTtsStreamingEvents": {
+    "Empty": "deepgram_tts_streaming::empty",
+    "ConnectFailure": "deepgram_tts_streaming::connect_failed",
+    "Connect": "deepgram_tts_streaming::connect"
+  },
+  "CartesiaTtsStreamingEvents": {
+    "Empty": "cartesia_tts_streaming::empty",
+    "ConnectFailure": "cartesia_tts_streaming::connect_failed",
+    "Connect": "cartesia_tts_streaming::connect"
+  },
+  "ElevenlabsTtsStreamingEvents": {
+    "Empty": "elevenlabs_tts_streaming::empty",
+    "ConnectFailure": "elevenlabs_tts_streaming::connect_failed",
+    "Connect": "elevenlabs_tts_streaming::connect"
+  },
+  "RimelabsTtsStreamingEvents": {
+    "Empty": "rimelabs_tts_streaming::empty",
+    "ConnectFailure": "rimelabs_tts_streaming::connect_failed",
+    "Connect": "rimelabs_tts_streaming::connect"
+  },
+  "CustomTtsStreamingEvents": {
+    "Empty": "custom_tts_streaming::empty",
+    "ConnectFailure": "custom_tts_streaming::connect_failed",
+    "Connect": "custom_tts_streaming::connect"
+  },
+  "TtsStreamingEvents": {
+    "Empty": "tts_streaming::empty",
+    "Pause": "tts_streaming::pause",
+    "Resume": "tts_streaming::resume",
+    "ConnectFailure": "tts_streaming::connect_failed"
+  },
+  "TtsStreamingConnectionStatus": {
+    "NotConnected": "not_connected",
+    "Connected": "connected",
+    "Connecting": "connecting",
+    "Failed": "failed"
+  },
   "MAX_SIMRINGS": 10,
   "BONG_TONE": "tone_stream://v=-7;%(100,0,941.0,1477.0);v=-7;>=2;+=.1;%(1400,0,350,440)",
-  "FS_UUID_SET_NAME": "fsUUIDs"
+  "FS_UUID_SET_NAME": "fsUUIDs",
+  "SystemState" : {
+    "Online": "ONLINE",
+    "Offline": "OFFLINE",
+    "GracefulShutdownInProgress":"SHUTDOWN_IN_PROGRESS"
+  },
+  "FEATURE_SERVER" : "feature-server",
+  "WS_CLOSE_CODES": {
+    "NormalClosure": 1000,
+    "GoingAway": 1001
+  }
 }

lib/utils/db-utils.js

@@ -76,7 +76,9 @@ const speechMapper = (cred) => {
     else if ('deepgram' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
+      obj.model_id = o.model_id;
       obj.deepgram_stt_uri = o.deepgram_stt_uri;
+      obj.deepgram_tts_uri = o.deepgram_tts_uri;
       obj.deepgram_stt_use_tls = o.deepgram_stt_use_tls;
     }
     else if ('soniox' === obj.vendor) {
@@ -90,39 +92,68 @@ const speechMapper = (cred) => {
     else if ('cobalt' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.cobalt_server_uri = o.cobalt_server_uri;
-    } else if ('elevenlabs' === obj.vendor) {
+    }
+    else if ('elevenlabs' === 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 ('playht' === obj.vendor) {
+    }
+    else if ('playht' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
       obj.user_id = o.user_id;
       obj.voice_engine = o.voice_engine;
       obj.options = o.options;
-    } else if ('rimelabs' === obj.vendor) {
+    }
+    else if ('cartesia' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.model_id = o.model_id;
+      obj.embedding = o.embedding;
+      obj.options = o.options;
+    }
+    else if ('rimelabs' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
       obj.model_id = o.model_id;
       obj.options = o.options;
-    } else if ('assemblyai' === obj.vendor) {
+    }
+    else if ('assemblyai' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
-    } else if ('whisper' === obj.vendor) {
+    }
+    else if ('voxist' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+    }
+    else if ('whisper' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
       obj.model_id = o.model_id;
-    }  else if ('verbio' === obj.vendor) {
+    }
+    else if ('verbio' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.client_id = o.client_id;
       obj.client_secret = o.client_secret;
       obj.engine_version = o.engine_version;
-    } else if (obj.vendor.startsWith('custom:')) {
+    }
+    else if ('speechmatics' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.speechmatics_stt_uri = o.speechmatics_stt_uri;
+    }
+    else if ('openai' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+      obj.model_id = o.model_id;
+    }
+    else if (obj.vendor.startsWith('custom:')) {
       const o = JSON.parse(decrypt(credential));
       obj.auth_token = o.auth_token;
       obj.custom_stt_url = o.custom_stt_url;
       obj.custom_tts_url = o.custom_tts_url;
+      obj.custom_tts_streaming_url = o.custom_tts_streaming_url;
     }
   } catch (err) {
     console.log(err);
@@ -202,11 +233,23 @@ module.exports = (logger, srf) => {
     }
   };
 
+  const lookupVoipCarrierBySid = async(sid) => {
+    const pp = pool.promise();
+    try {
+      const [r] = await pp.query('SELECT * FROM voip_carriers WHERE voip_carrier_sid = ?', [sid]);
+      return r;
+
+    } catch (err) {
+      logger.error({err}, `lookupVoipCarrierBySid: Error ${sid}`);
+    }
+  };
+
   return {
     lookupAccountDetails,
     updateSpeechCredentialLastUsed,
     lookupCarrier,
     lookupCarrierByPhoneNumber,
-    lookupGoogleCustomVoice
+    lookupGoogleCustomVoice,
+    lookupVoipCarrierBySid
   };
 };

lib/utils/error.js

@@ -0,0 +1,33 @@
+class NonFatalTaskError extends Error {
+  constructor(msg) {
+    super(msg);
+  }
+}
+
+class SpeechCredentialError extends NonFatalTaskError {
+  constructor(msg) {
+    super(msg);
+  }
+}
+
+class PlayFileNotFoundError extends NonFatalTaskError {
+  constructor(url) {
+    super('File not found');
+    this.url = url;
+  }
+}
+
+class HTTPResponseError extends Error {
+  constructor(statusCode) {
+    super('Unexpected HTTP Response');
+    delete this.stack;
+    this.statusCode = statusCode;
+  }
+}
+
+module.exports = {
+  SpeechCredentialError,
+  NonFatalTaskError,
+  PlayFileNotFoundError,
+  HTTPResponseError
+};

lib/utils/helpers.js

@@ -0,0 +1,5 @@
+
+const sleepFor = (ms) => new Promise((resolve) => setTimeout(() => resolve(), ms));
+module.exports = {
+  sleepFor
+};

lib/utils/http-requestor.js

@@ -16,6 +16,7 @@ const {
   NODE_ENV,
   HTTP_USER_AGENT_HEADER,
 } = require('../config');
+const {HTTPResponseError} = require('./error');
 
 const toBase64 = (str) => Buffer.from(str || '', 'utf8').toString('base64');
 
@@ -47,8 +48,6 @@ class HttpRequestor extends BaseRequestor {
     assert(['GET', 'POST'].includes(this.method));
 
     const u = this._parsedUrl = parseUrl(this.url);
-    if (u.port) this._baseUrl = `${u.protocol}://${u.resource}:${u.port}`;
-    else this._baseUrl = `${u.protocol}://${u.resource}`;
     this._protocol = u.protocol;
     this._resource = u.resource;
     this._port = u.port;
@@ -56,18 +55,18 @@ class HttpRequestor extends BaseRequestor {
     this._usePools = HTTP_POOL && parseInt(HTTP_POOL);
 
     if (this._usePools) {
-      if (pools.has(this._baseUrl)) {
-        this.client = pools.get(this._baseUrl);
+      if (pools.has(this.baseUrl)) {
+        this.client = pools.get(this.baseUrl);
       }
       else {
         const connections = HTTP_POOLSIZE ? parseInt(HTTP_POOLSIZE) : 10;
         const pipelining = HTTP_PIPELINING ? parseInt(HTTP_PIPELINING) : 1;
-        const pool = this.client = new Pool(this._baseUrl, {
+        const pool = this.client = new Pool(this.baseUrl, {
           connections,
           pipelining
         });
-        pools.set(this._baseUrl, pool);
-        this.logger.debug(`HttpRequestor:created pool for ${this._baseUrl}`);
+        pools.set(this.baseUrl, pool);
+        this.logger.debug(`HttpRequestor:created pool for ${this.baseUrl}`);
       }
     }
     else {
@@ -88,10 +87,6 @@ class HttpRequestor extends BaseRequestor {
     }
   }
 
-  get baseUrl() {
-    return this._baseUrl;
-  }
-
   close() {
     if (!this._usePools && !this.client?.closed) this.client.close();
   }
@@ -107,13 +102,13 @@ class HttpRequestor extends BaseRequestor {
    * @param {string} [hook.password] - if basic auth is protecting the endpoint
    * @param {object} [params] - request parameters
    */
-  async request(type, hook, params, httpHeaders = {}) {
+  async request(type, hook, params, httpHeaders = {}, span) {
     /* jambonz:error only sent over ws */
     if (type === 'jambonz:error') return;
 
     assert(HookMsgTypes.includes(type));
 
-    const payload = params ? snakeCaseKeys(params, ['customerData', 'sip']) : null;
+    const payload = params ? snakeCaseKeys(params, ['customerData', 'sip', 'env_vars', 'args']) : null;
     const url = hook.url || hook;
     const method = hook.method || 'POST';
     let buf = '';
@@ -136,7 +131,7 @@ class HttpRequestor extends BaseRequestor {
         this.close();
         this.emit('handover', requestor);
       }
-      return requestor.request('session:new', hook, params, httpHeaders);
+      return requestor.request('session:new', hook, params, httpHeaders, span);
     }
 
     let newClient;
@@ -190,8 +185,7 @@ class HttpRequestor extends BaseRequestor {
         followRedirects: false
       });
       if (![200, 202, 204].includes(statusCode)) {
-        const err = new Error();
-        err.statusCode = statusCode;
+        const err = new HTTPResponseError(statusCode);
         throw err;
       }
       if (headers['content-type']?.includes('application/json')) {
@@ -225,7 +219,7 @@ class HttpRequestor extends BaseRequestor {
     const rtt = this._roundTrip(startAt);
     if (buf) this.stats.histogram('app.hook.response_time', rtt, ['hook_type:app']);
 
-    if (buf && Array.isArray(buf)) {
+    if (buf && (Array.isArray(buf) || type == 'llm:tool-call')) {
       this.logger.info({response: buf}, `HttpRequestor:request ${method} ${url} succeeded in ${rtt}ms`);
       return buf;
     }

lib/utils/install-srf-locals.js

@@ -1,5 +1,5 @@
 const Mrf = require('drachtio-fsmrf');
-const ip = require('ip');
+const os = require('os');
 const {
   JAMBONES_MYSQL_HOST,
   JAMBONES_MYSQL_USER,
@@ -12,23 +12,45 @@ const {
   JAMBONES_TIME_SERIES_HOST,
   JAMBONES_ESL_LISTEN_ADDRESS,
   PORT,
+  HTTP_IP,
   NODE_ENV,
 } = require('../config');
 const Registrar = require('@jambonz/mw-registrar');
 const assert = require('assert');
 
-function initMS(logger, wrapper, ms) {
+function getLocalIp() {
+  const interfaces = os.networkInterfaces();
+  for (const interfaceName in interfaces) {
+    const interface = interfaces[interfaceName];
+    for (const iface of interface) {
+      if (iface.family === 'IPv4' && !iface.internal) {
+        return iface.address;
+      }
+    }
+  }
+  return '127.0.0.1'; // Fallback to localhost if no suitable interface found
+}
+
+function initMS(logger, wrapper, ms, {
+  onFreeswitchConnect,
+  onFreeswitchDisconnect
+}) {
   Object.assign(wrapper, {ms, active: true, connects: 1});
   logger.info(`connected to freeswitch at ${ms.address}`);
 
+  onFreeswitchConnect(wrapper);
+
   ms.conn
     .on('esl::end', () => {
       wrapper.active = false;
+      wrapper.connects = 0;
       logger.info(`lost connection to freeswitch at ${ms.address}`);
+      onFreeswitchDisconnect(wrapper);
+      ms.removeAllListeners();
     })
     .on('esl::ready', () => {
       if (wrapper.connects > 0) {
-        logger.info(`connected to freeswitch at ${ms.address}`);
+        logger.info(`esl::ready connected to freeswitch at ${ms.address}`);
       }
       wrapper.connects = 1;
       wrapper.active = true;
@@ -42,7 +64,10 @@ function initMS(logger, wrapper, ms) {
   });
 }
 
-function installSrfLocals(srf, logger) {
+function installSrfLocals(srf, logger, {
+  onFreeswitchConnect = () => {},
+  onFreeswitchDisconnect = () => {}
+}) {
   logger.debug('installing srf locals');
   assert(!srf.locals.dbHelpers);
   const {tracer} = srf.locals.otel;
@@ -77,7 +102,10 @@ function installSrfLocals(srf, logger) {
       mediaservers.push(val);
       try {
         const ms = await mrf.connect(fs);
-        initMS(logger, val, ms);
+        initMS(logger, val, ms, {
+          onFreeswitchConnect,
+          onFreeswitchDisconnect
+        });
       }
       catch (err) {
         logger.info({err}, `failed connecting to freeswitch at ${fs.address}, will retry shortly: ${err.message}`);
@@ -88,9 +116,15 @@ function installSrfLocals(srf, logger) {
       for (const val of mediaservers) {
         if (val.connects === 0) {
           try {
+            // make sure all listeners are removed before reconnecting
+            val.ms?.disconnect();
+            val.ms = null;
             logger.info({mediaserver: val.opts}, 'Retrying initial connection to media server');
             const ms = await mrf.connect(val.opts);
-            initMS(logger, val, ms);
+            initMS(logger, val, ms, {
+              onFreeswitchConnect,
+              onFreeswitchDisconnect
+            });
           } catch (err) {
             logger.info({err}, `failed connecting to freeswitch at ${val.opts.address}, will retry shortly`);
           }
@@ -138,7 +172,8 @@ function installSrfLocals(srf, logger) {
     lookupAccountBySid,
     lookupAccountCapacitiesBySid,
     lookupSmppGateways,
-    lookupClientByAccountAndUsername
+    lookupClientByAccountAndUsername,
+    lookupSystemInformation
   } = require('@jambonz/db-helpers')({
     host: JAMBONES_MYSQL_HOST,
     user: JAMBONES_MYSQL_USER,
@@ -184,7 +219,8 @@ function installSrfLocals(srf, logger) {
   } = require('@jambonz/speech-utils')({}, logger);
   const {
     writeAlerts,
-    AlertType
+    AlertType,
+    writeSystemAlerts
   } = require('@jambonz/time-series')(logger, {
     host: JAMBONES_TIME_SERIES_HOST,
     commitSize: 50,
@@ -193,7 +229,8 @@ function installSrfLocals(srf, logger) {
 
   let localIp;
   try {
-    localIp = ip.address();
+    // Either use the configured IP address or discover it
+    localIp = HTTP_IP || getLocalIp();
   } catch (err) {
     logger.error({err}, 'installSrfLocals - error detecting local ipv4 address');
   }
@@ -213,6 +250,7 @@ function installSrfLocals(srf, logger) {
       lookupAccountCapacitiesBySid,
       lookupSmppGateways,
       lookupClientByAccountAndUsername,
+      lookupSystemInformation,
       updateCallStatus,
       retrieveCall,
       listCalls,
@@ -252,7 +290,8 @@ function installSrfLocals(srf, logger) {
     getFreeswitch,
     stats: stats,
     writeAlerts,
-    AlertType
+    AlertType,
+    writeSystemAlerts
   };
 
   if (localIp) {

lib/utils/llm-mcp.js

@@ -0,0 +1,103 @@
+const { Client } = require('@modelcontextprotocol/sdk/client/index.js');
+
+class LlmMcpService {
+
+  constructor(logger, mcpServers) {
+    this.logger = logger;
+    this.mcpServers = mcpServers || [];
+    this.mcpClients = [];
+  }
+
+  // make sure we call init() before using any of the mcp clients
+  // this is to ensure that we have a valid connection to the MCP server
+  // and that we have collected the available tools.
+  async init() {
+    if (this.mcpClients.length > 0) {
+      return;
+    }
+    const { SSEClientTransport } = await import('@modelcontextprotocol/sdk/client/sse.js');
+    for (const server of this.mcpServers) {
+      const { url } = server;
+      if (url) {
+        try {
+          const transport = new SSEClientTransport(new URL(url), {});
+          const client = new Client({ name: 'Jambonz MCP Client', version: '1.0.0' });
+          await client.connect(transport);
+          // collect available tools
+          const { tools } = await client.listTools();
+          this.mcpClients.push({
+            url,
+            client,
+            tools
+          });
+        } catch (err) {
+          this.logger.error(`LlmMcpService: Failed to connect to MCP server at ${url}: ${err.message}`);
+        }
+      }
+    }
+  }
+
+  async getAvailableMcpTools() {
+    // returns a list of available tools from all MCP clients
+    const tools = [];
+    for (const mcpClient of this.mcpClients) {
+      const {tools: availableTools} = mcpClient;
+      if (availableTools) {
+        tools.push(...availableTools);
+      }
+    }
+    return tools;
+  }
+
+  async getMcpClientByToolName(name) {
+    for (const mcpClient of this.mcpClients) {
+      const { tools } = mcpClient;
+      if (tools && tools.some((tool) => tool.name === name)) {
+        return mcpClient.client;
+      }
+    }
+    return null;
+  }
+
+  async getMcpClientByToolId(id) {
+    for (const mcpClient of this.mcpClients) {
+      const { tools } = mcpClient;
+      if (tools && tools.some((tool) => tool.id === id)) {
+        return mcpClient.client;
+      }
+    }
+    return null;
+  }
+
+  async callMcpTool(name, input) {
+    const client = await this.getMcpClientByToolName(name);
+    if (client) {
+      try {
+        const result = await client.callTool({
+          name,
+          arguments: input,
+        });
+        this.logger.debug({result}, 'LlmMcpService - result');
+        return result;
+      } catch (err) {
+        this.logger.error({err}, 'LlmMcpService - error calling tool');
+        throw err;
+      }
+    }
+  }
+
+  async close() {
+    for (const mcpClient of this.mcpClients) {
+      const { client } = mcpClient;
+      if (client) {
+        await client.close();
+        this.logger.debug({url: mcpClient.url}, 'LlmMcpService - mcp client closed');
+      }
+    }
+    this.mcpClients = [];
+  }
+
+}
+
+module.exports = LlmMcpService;
+

lib/utils/network.js

@@ -0,0 +1,32 @@
+
+/**
+ * Parses a list of hostport entries and selects the first one that matches the specified protocol,
+ * excluding any entries with the localhost IP address ('127.0.0.1').
+ *
+ * Each hostport entry should be in the format: 'protocol/ip:port'
+ *
+ * @param {Object} logger - A logging object with a 'debug' method for logging debug messages.
+ * @param {string} hostport - A comma-separated string containing hostport entries.
+ * @param {string} protocol - The protocol to match (e.g., 'udp', 'tcp').
+ * @returns {Array} An array containing:
+ * 0: protocol
+ * 1: ip address
+ * 2: port
+ */
+const selectHostPort = (logger, hostport, protocol) => {
+  logger.debug(`selectHostPort: ${hostport}, ${protocol}`);
+  const sel = hostport
+    .split(',')
+    .map((hp) => {
+      const arr = /(.*)\/(.*):(.*)/.exec(hp);
+      return [arr[1], arr[2], arr[3]];
+    })
+    .filter((hp) => {
+      return hp[0] === protocol && hp[1] !== '127.0.0.1';
+    });
+  return sel[0];
+};
+
+module.exports = {
+  selectHostPort
+};

lib/utils/place-outdial.js

@@ -1,5 +1,5 @@
 const Emitter = require('events');
-const {CallStatus} = require('./constants');
+const {CallStatus, MediaPath} = require('./constants');
 const SipError = require('drachtio-srf').SipError;
 const {TaskPreconditions, CallDirection} = require('../utils/constants');
 const CallInfo = require('../session/call-info');
@@ -12,13 +12,17 @@ const deepcopy = require('deepcopy');
 const moment = require('moment');
 const stripCodecs = require('./strip-ancillary-codecs');
 const RootSpan = require('./call-tracer');
-const uuidv4 = require('uuid-random');
+const crypto = require('crypto');
 const HttpRequestor = require('./http-requestor');
 const WsRequestor = require('./ws-requestor');
-const {makeOpusFirst} = require('./sdp-utils');
+const {makeOpusFirst, removeVideoSdp} = require('./sdp-utils');
 const {
-  JAMBONES_USE_FREESWITCH_TIMER_FD
+  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');
 
 class SingleDialer extends Emitter {
   constructor({logger, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask,
@@ -43,7 +47,7 @@ class SingleDialer extends Emitter {
 
     this.callGone = false;
 
-    this.callSid = uuidv4();
+    this.callSid = crypto.randomUUID();
     this.dialTask = dialTask;
     this.onHoldMusic = onHoldMusic;
 
@@ -150,15 +154,21 @@ class SingleDialer extends Emitter {
         return;
       }
       let lastSdp;
-      const connectStream = async(remoteSdp) => {
+      const connectStream = async(remoteSdp, isVideoCall) => {
         if (remoteSdp === lastSdp) return;
+        if (process.env.JAMBONES_VIDEO_CALLS_ENABLED_IN_FS && !isVideoCall) {
+          remoteSdp = removeVideoSdp(remoteSdp);
+        }
         lastSdp = remoteSdp;
         return this.ep.modify(remoteSdp);
       };
-
+      let localSdp = this.ep.local.sdp;
+      if (process.env.JAMBONES_VIDEO_CALLS_ENABLED_IN_FS && !opts.isVideoCall) {
+        localSdp = removeVideoSdp(localSdp);
+      }
       Object.assign(opts, {
         proxy: `sip:${this.sbcAddress}`,
-        localSdp: opts.opusFirst ? makeOpusFirst(this.ep.local.sdp) : this.ep.local.sdp
+        localSdp: opts.opusFirst ? makeOpusFirst(localSdp) : localSdp
       });
       if (this.target.auth) opts.auth = this.target.auth;
       inviteSpan = this.startSpan('invite', {
@@ -213,18 +223,20 @@ class SingleDialer extends Emitter {
         },
         cbProvisional: (prov) => {
           const status = {sipStatus: prov.status, sipReason: prov.reason};
+          // Update call-id for sbc outbound INVITE
+          this.callInfo.sbcCallid = prov.get('X-CID');
           if ([180, 183].includes(prov.status) && prov.body) {
             if (status.callStatus !== CallStatus.EarlyMedia) {
               status.callStatus = CallStatus.EarlyMedia;
               this.emit('earlyMedia');
             }
-            connectStream(prov.body);
+            connectStream(prov.body, opts.isVideoCall);
           }
           else status.callStatus = CallStatus.Ringing;
           this.emit('callStatusChange', status);
         }
       });
-      await connectStream(this.dlg.remote.sdp);
+      await connectStream(this.dlg.remote.sdp, opts.isVideoCall);
       this.dlg.callSid = this.callSid;
       this.inviteInProgress = null;
       this.emit('callStatusChange', {
@@ -267,7 +279,12 @@ class SingleDialer extends Emitter {
                 this.logger.info('dial is onhold, emit event');
                 this.emit('reinvite', req, res);
               } else {
-                const newSdp = await this.ep.modify(req.body);
+                let newSdp = await this.ep.modify(req.body);
+                // in case of reINVITE if video call is enabled in FS and the call is not a video call,
+                // remove video media from the SDP
+                if (process.env.JAMBONES_VIDEO_CALLS_ENABLED_IN_FS && !this.opts?.isVideoCall) {
+                  newSdp = removeVideoSdp(newSdp);
+                }
                 res.send(200, {body: newSdp});
                 this.logger.info({offer: req.body, answer: newSdp}, 'SingleDialer:exec: handling reINVITE');
               }
@@ -315,14 +332,19 @@ class SingleDialer extends Emitter {
   /**
    * kill the call in progress or the stable dialog, whichever we have
    */
-  async kill() {
+  async kill(Reason) {
     this.killed = true;
     if (this.inviteInProgress) await this.inviteInProgress.cancel();
     else if (this.dlg && this.dlg.connected) {
       const duration = moment().diff(this.dlg.connectTime, 'seconds');
       this.logger.debug('SingleDialer:kill hanging up called party');
       this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});
-      this.dlg.destroy();
+      const headers = {
+        ...(Reason && {'X-Reason': Reason})
+      };
+      this.dlg.destroy({
+        headers
+      });
     }
     if (this.ep) {
       this.logger.debug(`SingleDialer:kill - deleting endpoint ${this.ep.uuid}`);
@@ -333,11 +355,40 @@ class SingleDialer extends Emitter {
   _configMsEndpoint() {
     const opts = {
       ...(this.onHoldMusic && {holdMusic: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`}),
-      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'})
+      ...(JAMBONES_USE_FREESWITCH_TIMER_FD && {timer_name: 'timerfd'}),
+      ...(JAMBONES_MEDIA_TIMEOUT_MS && {media_timeout: JAMBONES_MEDIA_TIMEOUT_MS}),
+      ...(JAMBONES_MEDIA_HOLD_TIMEOUT_MS && {media_hold_timeout: JAMBONES_MEDIA_HOLD_TIMEOUT_MS})
     };
     if (Object.keys(opts).length > 0) {
       this.ep.set(opts);
     }
+    if (this.dialTask?.inbandDtmfEnabled && !this.ep.inbandDtmfEnabled) {
+      // https://developer.signalwire.com/freeswitch/FreeSWITCH-Explained/Modules/mod-dptools/6587132/#0-about
+      try {
+        this.ep.execute('start_dtmf');
+        this.ep.inbandDtmfEnabled = true;
+      } catch (err) {
+        this.logger.info(err, 'place-outdial:_configMsEndpoint - error enable inband DTMF');
+      }
+    }
+
+    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');
+      }
+    };
   }
 
   /**
@@ -379,7 +430,8 @@ class SingleDialer extends Emitter {
         callInfo: this.callInfo,
         accountInfo: this.accountInfo,
         tasks,
-        rootSpan: this.rootSpan
+        rootSpan: this.rootSpan,
+        req: this.req
       });
       await cs.exec();
 
@@ -388,7 +440,10 @@ class SingleDialer extends Emitter {
     } catch (err) {
       this.logger.debug(err, 'SingleDialer:_executeApp: error');
       this.emit('decline');
-      if (this.dlg.connected) this.dlg.destroy();
+      if (this.dlg.connected) {
+        this.dlg.destroy();
+        this.ep.destroy();
+      }
     }
   }
 
@@ -441,31 +496,38 @@ class SingleDialer extends Emitter {
     });
     app.requestor.request('session:adulting', '/adulting', {
       ...cs.callInfo.toJSON(),
-      parentCallInfo: this.parentCallInfo
+      parentCallInfo: this.parentCallInfo.toJSON()
     }).catch((err) => {
       newLogger.error({err}, 'doAdulting: error sending adulting request');
     });
 
     cs.req = this.req;
+    // fixed hangup an adulting session does not send status callback Completed
+    cs.wrapDialog(this.dlg);
     cs.exec().catch((err) => newLogger.error({err}, 'doAdulting: error executing session'));
     return cs;
   }
 
-  async releaseMediaToSBC(remoteSdp, localSdp) {
+  async releaseMediaToSBC(remoteSdp, localSdp, releaseMediaEntirely) {
     assert(this.dlg && this.dlg.connected && this.ep && typeof remoteSdp === 'string');
     const sdp = stripCodecs(this.logger, remoteSdp, localSdp) || remoteSdp;
     await this.dlg.modify(sdp, {
       headers: {
-        'X-Reason': 'release-media'
+        'X-Reason': releaseMediaEntirely ? 'release-media-entirely' : 'release-media'
       }
     });
-    this.ep.destroy()
-      .then(() => this.ep = null)
-      .catch((err) => this.logger.error({err}, 'SingleDialer:releaseMediaToSBC: Error destroying endpoint'));
+    try {
+      await this.ep.destroy();
+    } catch (err) {
+      this.logger.error({err}, 'SingleDialer:releaseMediaToSBC: Error destroying endpoint');
+    }
+    this.ep = null;
   }
 
-  async reAnchorMedia() {
+  async reAnchorMedia(currentMediaRoute = MediaPath.PartialMedia) {
     assert(this.dlg && this.dlg.connected && !this.ep);
+
+    this.logger.debug('SingleDialer:reAnchorMedia: re-anchoring media after partial media');
     this.ep = await this.ms.createEndpoint({remoteSdp: this.dlg.remote.sdp});
     this._configMsEndpoint();
     await this.dlg.modify(this.ep.local.sdp, {
@@ -473,6 +535,11 @@ class SingleDialer extends Emitter {
         'X-Reason': 'anchor-media'
       }
     });
+
+    if (currentMediaRoute === MediaPath.NoMedia) {
+      this.logger.debug('SingleDialer:reAnchorMedia: repoint endpoint after no media');
+      await this.ep.modify(this.dlg.remote.sdp);
+    }
   }
 
   _notifyCallStatusChange({callStatus, sipStatus, sipReason, duration}) {
@@ -503,7 +570,8 @@ function placeOutdial({
 }) {
   const myOpts = deepcopy(opts);
   const sd = new SingleDialer({
-    logger, sbcAddress, target, myOpts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask, onHoldMusic
+    logger, sbcAddress, target, opts: myOpts, application, callInfo,
+    accountInfo, rootSpan, startSpan, dialTask, onHoldMusic
   });
   sd.exec(srf, ms, myOpts);
   return sd;

lib/utils/sbc-pinger.js

@@ -1,5 +1,5 @@
 const assert = require('assert');
-const uuidv4 = require('uuid-random');
+const crypto = require('crypto');
 const {LifeCycleEvents, FS_UUID_SET_NAME} = require('./constants');
 const Emitter = require('events');
 const debug = require('debug')('jambonz:feature-server');
@@ -8,7 +8,7 @@ const {
   JAMBONES_SBCS,
   K8S,
   K8S_SBC_SIP_SERVICE_NAME,
-  AWS_SNS_TOPIC_ARM,
+  AWS_SNS_TOPIC_ARN,
   OPTIONS_PING_INTERVAL,
   AWS_REGION,
   NODE_ENV,
@@ -35,7 +35,7 @@ module.exports = (logger) => {
   // listen for SNS lifecycle changes
   let lifecycleEmitter = new Emitter();
   let dryUpCalls = false;
-  if (AWS_SNS_TOPIC_ARM && AWS_REGION) {
+  if (AWS_SNS_TOPIC_ARN && AWS_REGION) {
 
     (async function() {
       try {
@@ -46,12 +46,24 @@ module.exports = (logger) => {
             const {srf} = require('../..');
             srf.locals.publicIp = publicIp;
           })
-          .on(LifeCycleEvents.ScaleIn, () => {
+          .on(LifeCycleEvents.ScaleIn, async() => {
             logger.info('AWS scale-in notification: begin drying up calls');
             dryUpCalls = true;
             lifecycleEmitter.operationalState = LifeCycleEvents.ScaleIn;
 
             const {srf} = require('../..');
+            const {writeSystemAlerts} = srf.locals;
+            if (writeSystemAlerts) {
+              const {SystemState, FEATURE_SERVER} = require('./constants');
+              await writeSystemAlerts({
+                system_component: FEATURE_SERVER,
+                state : SystemState.GracefulShutdownInProgress,
+                fields : {
+                  detail: `feature-server with process_id ${process.pid} shutdown in progress`,
+                  host: srf.locals?.ipv4
+                }
+              });
+            }
             pingProxies(srf);
 
             // if we have zero calls, we can complete the scale-in right
@@ -118,7 +130,7 @@ module.exports = (logger) => {
       logger.info('disabling OPTIONS pings since we are running as a kubernetes service');
       const {srf} = require('../..');
       const {addToSet} = srf.locals.dbHelpers;
-      const uuid = srf.locals.fsUUID = uuidv4();
+      const uuid = srf.locals.fsUUID = crypto.randomUUID();
 
       /* in case redis is restarted, re-insert our key every so often */
       setInterval(() => {

lib/utils/sdp-utils.js

@@ -35,6 +35,12 @@ const makeOpusFirst = (sdp) => {
   }
   return sdpTransform.write(parsedSdp);
 };
+const removeVideoSdp = (sdp) => {
+  const parsedSdp = sdpTransform.parse(sdp);
+  // Filter out video media sections, keeping only non-video media
+  parsedSdp.media = parsedSdp.media.filter((media) => media.type !== 'video');
+  return sdpTransform.write(parsedSdp);
+};
 
 const extractSdpMedia = (sdp) => {
   const parsedSdp1 = sdpTransform.parse(sdp);
@@ -54,5 +60,6 @@ module.exports = {
   mergeSdpMedia,
   extractSdpMedia,
   isOpusFirst,
-  makeOpusFirst
+  makeOpusFirst,
+  removeVideoSdp
 };

lib/utils/siprec-utils.js

@@ -1,5 +1,5 @@
 const xmlParser = require('xml2js').parseString;
-const uuidv4 = require('uuid-random');
+const crypto = require('crypto');
 const parseUri = require('drachtio-srf').parseUri;
 const transform = require('sdp-transform');
 const debug = require('debug')('jambonz:feature-server');
@@ -52,7 +52,7 @@ const parseSiprecPayload = (req, logger) => {
       const arr = /^([^]+)(m=[^]+?)(m=[^]+?)$/.exec(sdp);
       opts.sdp1 = `${arr[1]}${arr[2]}`;
       opts.sdp2 = `${arr[1]}${arr[3]}\r\n`;
-      opts.sessionId = uuidv4();
+      opts.sessionId = crypto.randomUUID();
       logger.info({ payload: req.payload }, 'SIPREC payload with no metadata (e.g. Cisco NBR)');
       resolve(opts);
     } else if (!sdp || !meta) {
@@ -64,7 +64,7 @@ const parseSiprecPayload = (req, logger) => {
       if (err) { throw err; }
 
       opts.recordingData = result ;
-      opts.sessionId = uuidv4() ;
+      opts.sessionId = crypto.randomUUID();
 
       const arr = /^([^]+)(m=[^]+?)(m=[^]+?)$/.exec(sdp) ;
       opts.sdp1 = `${arr[1]}${arr[2]}` ;

lib/utils/sticky-event-emitter.js

@@ -0,0 +1,70 @@
+const EventEmitter = require('events');
+
+/**
+ * A specialized EventEmitter that caches the most recent event emissions.
+ * When new listeners are added, they immediately receive the most recent
+ * event if it was previously emitted. This is useful for handling state
+ * changes where late subscribers need to know the current state.
+ *
+ * Features:
+ * - Caches the most recent emission for each event type
+ * - New listeners immediately receive the cached event if available
+ * - Supports both regular (on) and one-time (once) listeners
+ * - Maintains compatibility with Node's EventEmitter interface
+ */
+class StickyEventEmitter extends EventEmitter {
+  constructor() {
+    super();
+    this._eventCache = new Map();
+    this._onceListeners = new Map(); // For storing once listeners if needed
+  }
+  destroy() {
+    this._eventCache.clear();
+    this._onceListeners.clear();
+    this.removeAllListeners();
+  }
+  emit(event, ...args) {
+    // Store the event and its args
+    this._eventCache.set(event, args);
+
+    // If there are any 'once' listeners waiting, call them
+    if (this._onceListeners.has(event)) {
+      const listeners = this._onceListeners.get(event);
+      for (const listener of listeners) {
+        listener(...args);
+      }
+      if (this.onSuccess) {
+        this.onSuccess();
+      }
+      this._onceListeners.delete(event);
+    }
+
+    return super.emit(event, ...args);
+  }
+
+  on(event, listener) {
+    if (this._eventCache.has(event)) {
+      listener(...this._eventCache.get(event));
+    }
+    return super.on(event, listener);
+  }
+
+  once(event, listener) {
+    if (this._eventCache.has(event)) {
+      listener(...this._eventCache.get(event));
+      if (this.onSuccess) {
+        this.onSuccess();
+      }
+    } else {
+      // Store listener in case emit comes before
+      if (!this._onceListeners.has(event)) {
+        this._onceListeners.set(event, []);
+      }
+      this._onceListeners.get(event).push(listener);
+      super.once(event, listener); // Also attach to native once
+    }
+    return this;
+  }
+}
+
+module.exports = StickyEventEmitter;

lib/utils/transcription-utils.js

@@ -1,7 +1,4 @@
-const {
-  TaskName,
-} = require('./constants.json');
-
+const {TaskName} = require('./constants.json');
 const stickyVars = {
   google: [
     'GOOGLE_SPEECH_HINTS',
@@ -33,6 +30,7 @@ const stickyVars = {
     'DEEPGRAM_SPEECH_TIER',
     'DEEPGRAM_SPEECH_MODEL',
     'DEEPGRAM_SPEECH_ENABLE_SMART_FORMAT',
+    'DEEPGRAM_SPEECH_ENABLE_NO_DELAY',
     'DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION',
     'DEEPGRAM_SPEECH_PROFANITY_FILTER',
     'DEEPGRAM_SPEECH_REDACT',
@@ -45,12 +43,21 @@ const stickyVars = {
     'DEEPGRAM_SPEECH_ENDPOINTING',
     'DEEPGRAM_SPEECH_UTTERANCE_END_MS',
     'DEEPGRAM_SPEECH_VAD_TURNOFF',
-    'DEEPGRAM_SPEECH_TAG'
+    'DEEPGRAM_SPEECH_TAG',
+    'DEEPGRAM_SPEECH_MODEL_VERSION',
+    'DEEPGRAM_SPEECH_FILLER_WORDS',
+    'DEEPGRAM_SPEECH_KEYTERMS',
   ],
   aws: [
     'AWS_VOCABULARY_NAME',
     'AWS_VOCABULARY_FILTER_METHOD',
-    'AWS_VOCABULARY_FILTER_NAME'
+    'AWS_VOCABULARY_FILTER_NAME',
+    'AWS_LANGUAGE_MODEL_NAME',
+    'AWS_ACCESS_KEY_ID',
+    'AWS_SECRET_ACCESS_KEY',
+    'AWS_REGION',
+    'AWS_SECURITY_TOKEN',
+    'AWS_PII_ENTITY_TYPES',
   ],
   nuance: [
     'NUANCE_ACCESS_TOKEN',
@@ -99,7 +106,27 @@ const stickyVars = {
   assemblyai: [
     'ASSEMBLYAI_API_KEY',
     'ASSEMBLYAI_WORD_BOOST'
-  ]
+  ],
+  voxist: [
+    'VOXIST_API_KEY',
+  ],
+  speechmatics: [
+    'SPEECHMATICS_API_KEY',
+    'SPEECHMATICS_HOST',
+    'SPEECHMATICS_PATH',
+    'SPEECHMATICS_SPEECH_HINTS',
+    'SPEECHMATICS_TRANSLATION_LANGUAGES',
+    'SPEECHMATICS_TRANSLATION_PARTIALS'
+  ],
+  openai: [
+    'OPENAI_API_KEY',
+    'OPENAI_MODEL',
+    'OPENAI_INPUT_AUDIO_NOISE_REDUCTION',
+    'OPENAI_TURN_DETECTION_TYPE',
+    'OPENAI_TURN_DETECTION_THRESHOLD',
+    'OPENAI_TURN_DETECTION_PREFIX_PADDING_MS',
+    'OPENAI_TURN_DETECTION_SILENCE_DURATION_MS',
+  ],
 };
 
 /**
@@ -141,7 +168,6 @@ const optimalDeepramModels = {
   tr: ['nova-2', 'nova-2'],
   uk: ['nova-2', 'nova-2']
 };
-
 const selectDefaultDeepgramModel = (task, language) => {
   if (language in optimalDeepramModels) {
     const [gather, transcribe] = optimalDeepramModels[language];
@@ -150,8 +176,34 @@ const selectDefaultDeepgramModel = (task, language) => {
   return 'base';
 };
 
+const optimalGoogleModels = {
+  'v1' : {
+    'en-IN':['telephony', 'telephony'],
+    'es-DO':['default', 'default'],
+    'es-MX':['default', 'default'],
+    'en-AU':['telephony', 'telephony'],
+    'en-GB':['telephony', 'telephony'],
+    'en-NZ':['telephony', 'telephony']
+  },
+  'v2' : {
+    'en-IN':['telephony', 'long']
+  }
+};
+const selectDefaultGoogleModel = (task, language, version) => {
+  const useV2 = version === 'v2';
+  if (language in optimalGoogleModels[version]) {
+    const [gather, transcribe] = optimalGoogleModels[version][language];
+    return task.name === TaskName.Gather ? gather : transcribe;
+  }
+  return task.name === TaskName.Gather ?
+    (useV2 ? 'telephony_short' : 'command_and_search') :
+    (useV2 ? 'long' : 'latest_long');
+};
 const consolidateTranscripts = (bufferedTranscripts, channel, language, vendor) => {
-  if (bufferedTranscripts.length === 1) return bufferedTranscripts[0];
+  if (bufferedTranscripts.length === 1) {
+    bufferedTranscripts[0].is_final = true;
+    return bufferedTranscripts[0];
+  }
   let totalConfidence = 0;
   const finalTranscript = bufferedTranscripts.reduce((acc, evt) => {
     totalConfidence += evt.alternatives[0].confidence;
@@ -171,7 +223,7 @@ const consolidateTranscripts = (bufferedTranscripts, channel, language, vendor)
     const lastChar = acc.alternatives[0].transcript.slice(-1);
     const firstChar = newTranscript.charAt(0);
 
-    if (lastChar.match(/\d/) && firstChar.match(/\d/)) {
+    if (vendor === 'speechmatics' || (lastChar.match(/\d/) && firstChar.match(/\d/))) {
       acc.alternatives[0].transcript += newTranscript;
     } else {
       acc.alternatives[0].transcript += ` ${newTranscript}`;
@@ -261,13 +313,18 @@ const normalizeDeepgram = (evt, channel, language, shortUtterance) => {
       confidence: alt.confidence,
       transcript: alt.transcript,
     }));
-
+  /**
+   * Some models (nova-2-general) return the detected language in the
+   * alternatives.languages array if the language is set as multi.
+   * If the language is detected, we use it as the language_code.
+   */
+  const detectedLanguage = evt.channel?.alternatives?.[0]?.languages?.[0];
   /**
    * note difference between is_final and speech_final in Deepgram:
    * https://developers.deepgram.com/docs/understand-endpointing-interim-results
    */
   return {
-    language_code: language,
+    language_code: detectedLanguage || language,
     channel_tag: channel,
     is_final: shortUtterance ? evt.is_final : evt.speech_final,
     alternatives: alternatives.length ? [alternatives[0]] : [],
@@ -316,8 +373,10 @@ const normalizeIbm = (evt, channel, language) => {
 
 const normalizeGoogle = (evt, channel, language) => {
   const copy = JSON.parse(JSON.stringify(evt));
+  const language_code = evt.language_code || language;
+
   return {
-    language_code: language,
+    language_code: language_code,
     channel_tag: channel,
     is_final: evt.is_final,
     alternatives: [evt.alternatives[0]],
@@ -421,16 +480,41 @@ const normalizeMicrosoft = (evt, channel, language, punctuation = true) => {
 
 const normalizeAws = (evt, channel, language) => {
   const copy = JSON.parse(JSON.stringify(evt));
-  return {
-    language_code: language,
-    channel_tag: channel,
-    is_final: evt[0].is_final,
-    alternatives: evt[0].alternatives,
-    vendor: {
-      name: 'aws',
-      evt: copy
-    }
-  };
+  const isGrpcPayload = Array.isArray(evt);
+  if (isGrpcPayload) {
+    /* legacy grpc api */
+    return {
+      language_code: language,
+      channel_tag: channel,
+      is_final: evt[0].is_final,
+      alternatives: evt[0].alternatives,
+      vendor: {
+        name: 'aws',
+        evt: copy
+      }
+    };
+  }
+  else {
+    /* websocket api */
+    const alternatives = evt.Transcript?.Results[0]?.Alternatives.map((alt) => {
+      const items = alt.Items.filter((item) => item.Type === 'pronunciation' && 'Confidence' in item);
+      const confidence = items.reduce((acc, item) => acc + item.Confidence, 0) / items.length;
+      return {
+        transcript: alt.Transcript,
+        confidence
+      };
+    });
+    return {
+      language_code: language,
+      channel_tag: channel,
+      is_final: evt.Transcript?.Results[0].IsPartial === false,
+      alternatives,
+      vendor: {
+        name: 'aws',
+        evt: copy
+      }
+    };
+  }
 };
 
 const normalizeAssemblyAi = (evt, channel, language) => {
@@ -446,12 +530,85 @@ const normalizeAssemblyAi = (evt, channel, language) => {
       }
     ],
     vendor: {
-      name: 'ASSEMBLYAI',
+      name: 'assemblyai',
       evt: copy
     }
   };
 };
 
+const normalizeVoxist = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.type === 'final',
+    alternatives: [
+      {
+        confidence: 1.00,
+        transcript: evt.text,
+      }
+    ],
+    vendor: {
+      name: 'voxist',
+      evt: copy
+    }
+  };
+};
+
+const normalizeSpeechmatics = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  const is_final = evt.message === 'AddTranscript';
+  const words = evt.results?.filter((r) => r.type === 'word') || [];
+  const confidence = words.length > 0 ?
+    words.reduce((acc, word) => acc + word.alternatives[0].confidence, 0) / words.length :
+    0;
+
+  const alternative = {
+    confidence,
+    transcript: evt.metadata?.transcript
+  };
+  const obj = {
+    language_code: language,
+    channel_tag: channel,
+    is_final,
+    alternatives: [alternative],
+    vendor: {
+      name: 'speechmatics',
+      evt: copy
+    }
+  };
+  return obj;
+};
+
+const calculateConfidence = (logprobsArray) => {
+  // Sum the individual log probabilities
+  const totalLogProb = logprobsArray.reduce((sum, tokenInfo) => sum + tokenInfo.logprob, 0);
+
+  // Convert the total log probability back to a regular probability
+  const confidence = Math.exp(totalLogProb);
+  return confidence;
+};
+
+const normalizeOpenAI = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  const obj = {
+    language_code: language,
+    channel_tag: channel,
+    is_final: true,
+    alternatives: [
+      {
+        transcript: evt.transcript,
+        confidence: evt.logprobs ? calculateConfidence(evt.logprobs) : 1.0,
+      }
+    ],
+    vendor: {
+      name: 'openai',
+      evt: copy
+    }
+  };
+  return obj;
+};
+
 module.exports = (logger) => {
   const normalizeTranscription = (evt, vendor, channel, language, shortUtterance, punctuation) => {
 
@@ -477,8 +634,14 @@ module.exports = (logger) => {
         return normalizeCobalt(evt, channel, language);
       case 'assemblyai':
         return normalizeAssemblyAi(evt, channel, language, shortUtterance);
+      case 'voxist':
+        return normalizeVoxist(evt, channel, language);
       case 'verbio':
         return normalizeVerbio(evt, channel, language);
+      case 'speechmatics':
+        return normalizeSpeechmatics(evt, channel, language);
+      case 'openai':
+        return normalizeOpenAI(evt, channel, language);
       default:
         if (vendor.startsWith('custom:')) {
           return normalizeCustom(evt, channel, language, vendor);
@@ -494,9 +657,9 @@ module.exports = (logger) => {
 
     if ('google' === vendor) {
       const useV2 = rOpts.googleOptions?.serviceVersion === 'v2';
-      const model = task.name === TaskName.Gather ?
-        (useV2 ? 'telephony_short' : 'command_and_search') :
-        (useV2 ? 'long' : 'latest_long');
+      const version = useV2 ? 'v2' : 'v1';
+      let {model} = rOpts;
+      model = model || selectDefaultGoogleModel(task, language, version);
       opts = {
         ...opts,
         ...(sttCredentials && {GOOGLE_APPLICATION_CREDENTIALS: JSON.stringify(sttCredentials.credentials)}),
@@ -552,17 +715,29 @@ module.exports = (logger) => {
       };
     }
     else if (['aws', 'polly'].includes(vendor)) {
+      const {awsOptions = {}} = rOpts;
+      const vocabularyName = awsOptions.vocabularyName || rOpts.vocabularyName;
+      const vocabularyFilterName = awsOptions.vocabularyFilterName || rOpts.vocabularyFilterName;
+      const filterMethod = awsOptions.vocabularyFilterMethod || rOpts.filterMethod;
       opts = {
         ...opts,
-        ...(rOpts.vocabularyName && {AWS_VOCABULARY_NAME: rOpts.vocabularyName}),
-        ...(rOpts.vocabularyFilterName && {AWS_VOCABULARY_FILTER_NAME: rOpts.vocabularyFilterName}),
-        ...(rOpts.filterMethod && {AWS_VOCABULARY_FILTER_METHOD: rOpts.filterMethod}),
+        ...(vocabularyName && {AWS_VOCABULARY_NAME: vocabularyName}),
+        ...(vocabularyFilterName && {AWS_VOCABULARY_FILTER_NAME: vocabularyFilterName}),
+        ...(filterMethod && {AWS_VOCABULARY_FILTER_METHOD: filterMethod}),
         ...(sttCredentials && {
-          ...(sttCredentials.accessKeyId && {AWS_ACCESS_KEY_ID: sttCredentials.accessKeyId}),
-          ...(sttCredentials.secretAccessKey && {AWS_SECRET_ACCESS_KEY: sttCredentials.secretAccessKey}),
+          AWS_ACCESS_KEY_ID: sttCredentials.accessKeyId,
+          AWS_SECRET_ACCESS_KEY: sttCredentials.secretAccessKey,
           AWS_REGION: sttCredentials.region,
-          ...(sttCredentials.sessionToken && {AWS_SESSION_TOKEN: sttCredentials.sessionToken}),
+          AWS_SECURITY_TOKEN: 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.languageModelName && {AWS_LANGUAGE_MODEL_NAME: awsOptions.languageModelName}),
+        ...(awsOptions.piiEntityTypes?.length && {AWS_PII_ENTITY_TYPES: awsOptions.piiEntityTypes.join(',')}),
+        ...(awsOptions.piiIdentifyEntities && {AWS_PII_IDENTIFY_ENTITIES: true}),
+        ...(awsOptions.languageModelName && {AWS_LANGUAGE_MODEL_NAME: awsOptions.languageModelName}),
       };
     }
     else if ('microsoft' === vendor) {
@@ -601,6 +776,8 @@ module.exports = (logger) => {
         //azureSttEndpointId overrides sttCredentials.custom_stt_endpoint
         ...(rOpts.azureSttEndpointId &&
           {AZURE_SERVICE_ENDPOINT_ID: rOpts.azureSttEndpointId}),
+        ...(azureOptions.speechRecognitionMode &&
+          {AZURE_RECOGNITION_MODE: azureOptions.speechRecognitionMode}),
       };
     }
     else if ('nuance' === vendor) {
@@ -652,7 +829,7 @@ module.exports = (logger) => {
       };
     }
     else if ('deepgram' === vendor) {
-      let {model} = rOpts;
+      let model = rOpts.deepgramOptions?.model || rOpts.model || sttCredentials.model_id;
       const {deepgramOptions = {}} = rOpts;
       const deepgramUri = deepgramOptions.deepgramSttUri || sttCredentials.deepgram_stt_uri;
       const useTls = deepgramOptions.deepgramSttUseTls || sttCredentials.deepgram_stt_use_tls;
@@ -674,6 +851,8 @@ module.exports = (logger) => {
           {DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION: 1},
         ...(deepgramOptions.smartFormatting) &&
           {DEEPGRAM_SPEECH_ENABLE_SMART_FORMAT: 1},
+        ...(deepgramOptions.noDelay) &&
+          {DEEPGRAM_SPEECH_ENABLE_NO_DELAY: 1},
         ...(deepgramOptions.profanityFilter) &&
           {DEEPGRAM_SPEECH_PROFANITY_FILTER: 1},
         ...(deepgramOptions.redact) &&
@@ -707,7 +886,13 @@ module.exports = (logger) => {
         ...(deepgramOptions.vadTurnoff) &&
           {DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.vadTurnoff},
         ...(deepgramOptions.tag) &&
-          {DEEPGRAM_SPEECH_TAG: deepgramOptions.tag}
+          {DEEPGRAM_SPEECH_TAG: deepgramOptions.tag},
+        ...(deepgramOptions.version) &&
+          {DEEPGRAM_SPEECH_MODEL_VERSION: deepgramOptions.version},
+        ...(deepgramOptions.fillerWords) &&
+          {DEEPGRAM_SPEECH_FILLER_WORDS: deepgramOptions.fillerWords},
+        ...((Array.isArray(deepgramOptions.keyterms) && deepgramOptions.keyterms.length > 0) &&
+          {DEEPGRAM_SPEECH_KEYTERMS: deepgramOptions.keyterms.join(',')})
       };
     }
     else if ('soniox' === vendor) {
@@ -806,7 +991,8 @@ module.exports = (logger) => {
         ...(cobaltOptions.enableConfusionNetwork && {COBALT_ENABLE_CONFUSION_NETWORK: 1}),
         ...(cobaltOptions.compiledContextData && {COBALT_COMPILED_CONTEXT_DATA: cobaltOptions.compiledContextData}),
       };
-    } else if ('assemblyai' === vendor) {
+    }
+    else if ('assemblyai' === vendor) {
       opts = {
         ...opts,
         ...(sttCredentials.api_key) &&
@@ -814,7 +1000,45 @@ module.exports = (logger) => {
         ...(rOpts.hints?.length > 0 &&
           {ASSEMBLYAI_WORD_BOOST: JSON.stringify(rOpts.hints)})
       };
-    } else if ('verbio' === vendor) {
+    }
+    else if ('voxist' === vendor) {
+      opts = {
+        ...opts,
+        ...(sttCredentials.api_key) &&
+          {VOXIST_API_KEY: sttCredentials.api_key},
+      };
+    }
+    else if ('openai' === vendor) {
+      const {openaiOptions = {}} = rOpts;
+      const model = openaiOptions.model || rOpts.model || sttCredentials.model_id || 'whisper-1';
+      const apiKey = openaiOptions.apiKey || sttCredentials.api_key;
+
+      opts = {
+        OPENAI_MODEL: model,
+        OPENAI_API_KEY: apiKey,
+        ...opts,
+        ...(openaiOptions.prompt && {OPENAI_PROMPT: openaiOptions.prompt}),
+        ...(openaiOptions.input_audio_noise_reduction &&
+          {OPENAI_INPUT_AUDIO_NOISE_REDUCTION: openaiOptions.input_audio_noise_reduction}),
+      };
+
+      if (openaiOptions.turn_detection) {
+        opts = {
+          ...opts,
+          OPENAI_TURN_DETECTION_TYPE: openaiOptions.turn_detection.type,
+          ...(openaiOptions.turn_detection.threshold && {
+            OPENAI_TURN_DETECTION_THRESHOLD: openaiOptions.turn_detection.threshold
+          }),
+          ...(openaiOptions.turn_detection.prefix_padding_ms && {
+            OPENAI_TURN_DETECTION_PREFIX_PADDING_MS: openaiOptions.turn_detection.prefix_padding_ms
+          }),
+          ...(openaiOptions.turn_detection.silence_duration_ms && {
+            OPENAI_TURN_DETECTION_SILENCE_DURATION_MS: openaiOptions.turn_detection.silence_duration_ms
+          }),
+        };
+      }
+    }
+    else if ('verbio' === vendor) {
       const {verbioOptions = {}} = rOpts;
       opts = {
         ...opts,
@@ -833,8 +1057,55 @@ module.exports = (logger) => {
         ...(verbioOptions.speech_incomplete_timeout &&
           {VERBIO_SPEECH_INCOMPLETE_TIMEOUT: verbioOptions.speech_incomplete_timeout}),
       };
-    } else if (vendor.startsWith('custom:')) {
-      let {options = {}} = rOpts;
+    }
+    else if ('speechmatics' === vendor) {
+      const {speechmaticsOptions = {}} = rOpts;
+      opts = {
+        ...opts,
+        ...(sttCredentials.api_key) && {SPEECHMATICS_API_KEY: sttCredentials.api_key},
+        ...(sttCredentials.speechmatics_stt_uri) && {SPEECHMATICS_HOST: sttCredentials.speechmatics_stt_uri},
+        ...(rOpts.hints?.length > 0 && {SPEECHMATICS_SPEECH_HINTS: rOpts.hints.join(',')}),
+        ...(speechmaticsOptions.translation_config &&
+          {
+            SPEECHMATICS_TRANSLATION_LANGUAGES: speechmaticsOptions.translation_config.target_languages.join(','),
+            SPEECHMATICS_TRANSLATION_PARTIALS: speechmaticsOptions.translation_config.enable_partials ? 1 : 0
+          }
+        ),
+        ...(speechmaticsOptions.transcription_config?.domain &&
+          {SPEECHMATICS_DOMAIN: speechmaticsOptions.transcription_config.domain}),
+        ...{SPEECHMATICS_MAX_DELAY: speechmaticsOptions.transcription_config?.max_delay || 0.7},
+        ...{SPEECHMATICS_MAX_DELAY_MODE: speechmaticsOptions.transcription_config?.max_delay_mode || 'flexible'},
+        ...(speechmaticsOptions.transcription_config?.diarization &&
+            {SPEECHMATICS_DIARIZATION: speechmaticsOptions.transcription_config.diarization}),
+        ...(speechmaticsOptions.transcription_config?.speaker_diarization_config?.speaker_sensitivity &&
+          {SPEECHMATICS_DIARIZATION_SPEAKER_SENSITIVITY:
+            speechmaticsOptions.transcription_config.speaker_diarization_config.speaker_sensitivity}),
+        ...(speechmaticsOptions.transcription_config?.speaker_diarization_config?.max_speakers &&
+          {SPEECHMATICS_DIARIZATION_MAX_SPEAKERS:
+            speechmaticsOptions.transcription_config.speaker_diarization_config.max_speakers}),
+        ...(speechmaticsOptions.transcription_config?.output_locale &&
+          {SPEECHMATICS_OUTPUT_LOCALE: speechmaticsOptions.transcription_config.output_locale}),
+        ...(speechmaticsOptions.transcription_config?.punctuation_overrides?.permitted_marks &&
+          {SPEECHMATICS_PUNCTUATION_ALLOWED:
+            speechmaticsOptions.transcription_config.punctuation_overrides.permitted_marks.join(',')}),
+        ...(speechmaticsOptions.transcription_config?.punctuation_overrides?.sensitivity &&
+          {SPEECHMATICS_PUNCTUATION_SENSITIVITY:
+            speechmaticsOptions.transcription_config?.punctuation_overrides?.sensitivity}),
+        ...(speechmaticsOptions.transcription_config?.operating_point &&
+          {SPEECHMATICS_OPERATING_POINT: speechmaticsOptions.transcription_config.operating_point}),
+        ...(speechmaticsOptions.transcription_config?.enable_entities &&
+          {SPEECHMATICS_ENABLE_ENTTIES: speechmaticsOptions.transcription_config.enable_entities}),
+        ...(speechmaticsOptions.transcription_config?.audio_filtering_config?.volume_threshold &&
+          {SPEECHMATICS_VOLUME_THRESHOLD:
+            speechmaticsOptions.transcription_config.audio_filtering_config.volume_threshold}),
+        ...(speechmaticsOptions.transcription_config?.transcript_filtering_config?.remove_disfluencies &&
+          {SPEECHMATICS_REMOVE_DISFLUENCIES:
+            speechmaticsOptions.transcription_config.transcript_filtering_config.remove_disfluencies})
+      };
+    }
+    else if (vendor.startsWith('custom:')) {
+      let {options = {}} = rOpts.customOptions || {};
+      const {sampleRate} = rOpts.customOptions || {};
       const {auth_token, custom_stt_url} = sttCredentials;
       options = {
         ...options,
@@ -842,14 +1113,15 @@ module.exports = (logger) => {
         {hints: rOpts.hints}),
         ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
         {hints: JSON.stringify(rOpts.hints)}),
-        ...(typeof rOpts.hintsBoost === 'number' && {hintsBoost: rOpts.hintsBoost})
+        ...(typeof rOpts.hintsBoost === 'number' && {hintsBoost: rOpts.hintsBoost}),
+        ...(task.cs?.callSid && {callSid: task.cs.callSid})
       };
-
       opts = {
         ...opts,
         ...(auth_token && {JAMBONZ_STT_API_KEY: auth_token}),
         JAMBONZ_STT_URL: custom_stt_url,
         ...(Object.keys(options).length > 0 && {JAMBONZ_STT_OPTIONS: JSON.stringify(options)}),
+        ...(sampleRate && {JAMBONZ_STT_SAMPLING: sampleRate})
       };
     }
 
@@ -899,6 +1171,6 @@ module.exports = (logger) => {
     setChannelVarsForStt,
     setSpeechCredentialsAtRuntime,
     compileSonioxTranscripts,
-    consolidateTranscripts
+    consolidateTranscripts,
   };
 };

lib/utils/tts-streaming-buffer.js

@@ -0,0 +1,452 @@
+const Emitter = require('events');
+const assert = require('assert');
+const {
+  TtsStreamingEvents,
+  TtsStreamingConnectionStatus
+} = require('../utils/constants');
+
+const MAX_CHUNK_SIZE = 1800;
+const HIGH_WATER_BUFFER_SIZE = 1000;
+const LOW_WATER_BUFFER_SIZE = 200;
+const TIMEOUT_RETRY_MSECS = 1000; // 1 second
+
+
+const isWhitespace = (str) => /^\s*$/.test(str);
+
+/**
+ * Each queue item is an object:
+ *   - { type: 'text', value: '…' } for text tokens.
+ *   - { type: 'flush' } for a flush command.
+ */
+class TtsStreamingBuffer extends Emitter {
+  constructor(cs) {
+    super();
+    this.cs = cs;
+    this.logger = cs.logger;
+
+    // Use an array to hold our structured items.
+    this.queue = [];
+    // Track total number of characters in text items.
+    this.bufferedLength = 0;
+    this.eventHandlers = [];
+    this._isFull = false;
+    this._connectionStatus = TtsStreamingConnectionStatus.NotConnected;
+    this.timer = null;
+    // Record the last time the text buffer was updated.
+    this.lastUpdateTime = 0;
+  }
+
+  get isEmpty() {
+    return this.queue.length === 0;
+  }
+
+  get size() {
+    return this.bufferedLength;
+  }
+
+  get isFull() {
+    return this._isFull;
+  }
+
+  get ep() {
+    return this.cs?.ep;
+  }
+
+  async start() {
+    assert.ok(
+      this._connectionStatus === TtsStreamingConnectionStatus.NotConnected,
+      'TtsStreamingBuffer:start already started, or has failed'
+    );
+
+    this.vendor = this.cs.getTsStreamingVendor();
+    if (!this.vendor) {
+      this.logger.info('TtsStreamingBuffer:start No TTS streaming vendor configured');
+      throw new Error('No TTS streaming vendor configured');
+    }
+
+    this.logger.info(`TtsStreamingBuffer:start Connecting to TTS streaming with vendor ${this.vendor}`);
+
+    this._connectionStatus = TtsStreamingConnectionStatus.Connecting;
+    try {
+      if (this.eventHandlers.length === 0) this._initHandlers(this.ep);
+      await this._api(this.ep, [this.ep.uuid, 'connect']);
+    } catch (err) {
+      this.logger.info({ err }, 'TtsStreamingBuffer:start Error connecting to TTS streaming');
+      this._connectionStatus = TtsStreamingConnectionStatus.Failed;
+    }
+  }
+
+  stop() {
+    clearTimeout(this.timer);
+    this.removeCustomEventListeners();
+    if (this.ep) {
+      this._api(this.ep, [this.ep.uuid, 'close'])
+        .catch((err) =>
+          this.logger.info({ err }, 'TtsStreamingBuffer:stop Error closing TTS streaming')
+        );
+    }
+    this.timer = null;
+    this.queue = [];
+    this.bufferedLength = 0;
+    this._connectionStatus = TtsStreamingConnectionStatus.NotConnected;
+  }
+
+  /**
+   * Buffer new text tokens.
+   */
+  async bufferTokens(tokens) {
+    if (this._connectionStatus === TtsStreamingConnectionStatus.Failed) {
+      this.logger.info('TtsStreamingBuffer:bufferTokens TTS streaming connection failed, rejecting request');
+      return { status: 'failed', reason: `connection to ${this.vendor} failed` };
+    }
+
+    if (0 === this.bufferedLength && isWhitespace(tokens)) {
+      this.logger.debug({tokens}, 'TtsStreamingBuffer:bufferTokens discarded whitespace tokens');
+      return { status: 'ok' };
+    }
+
+    const displayedTokens = tokens.length <= 40 ? tokens : tokens.substring(0, 40);
+    const totalLength = tokens.length;
+
+    if (this.bufferedLength + totalLength > HIGH_WATER_BUFFER_SIZE) {
+      this.logger.info(
+        `TtsStreamingBuffer throttling: buffer is full, rejecting request to buffer ${totalLength} tokens`
+      );
+      if (!this._isFull) {
+        this._isFull = true;
+        this.emit(TtsStreamingEvents.Pause);
+      }
+      return { status: 'failed', reason: 'full' };
+    }
+
+    this.logger.debug(
+      `TtsStreamingBuffer:bufferTokens "${displayedTokens}" (length: ${totalLength})`
+    );
+    this.queue.push({ type: 'text', value: tokens });
+    this.bufferedLength += totalLength;
+    // Update the last update time each time new text is buffered.
+    this.lastUpdateTime = Date.now();
+
+    await this._feedQueue();
+    return { status: 'ok' };
+  }
+
+  /**
+   * Insert a flush command. If no text is queued, flush immediately.
+   * Otherwise, append a flush marker so that all text preceding it will be sent
+   * (regardless of sentence boundaries) before the flush is issued.
+   */
+  flush() {
+    if (this._connectionStatus === TtsStreamingConnectionStatus.Connecting) {
+      this.logger.debug('TtsStreamingBuffer:flush TTS stream is not quite ready - wait for connect');
+      if (this.queue.length === 0 || this.queue[this.queue.length - 1].type !== 'flush') {
+        this.queue.push({ type: 'flush' });
+      }
+      return;
+    }
+    else if (this._connectionStatus === TtsStreamingConnectionStatus.Connected) {
+      if (this.isEmpty) {
+        this._doFlush();
+      }
+      else {
+        if (this.queue[this.queue.length - 1].type !== 'flush') {
+          this.queue.push({ type: 'flush' });
+          this.logger.debug('TtsStreamingBuffer:flush added flush marker to queue');
+        }
+      }
+    }
+    else {
+      this.logger.debug(
+        `TtsStreamingBuffer:flush TTS stream is not connected, status: ${this._connectionStatus}`
+      );
+    }
+  }
+
+  clear() {
+    this.logger.debug('TtsStreamingBuffer:clear');
+    if (this._connectionStatus !== TtsStreamingConnectionStatus.Connected) return;
+    clearTimeout(this.timer);
+    this._api(this.ep, [this.ep.uuid, 'clear']).catch((err) =>
+      this.logger.info({ err }, 'TtsStreamingBuffer:clear Error clearing TTS streaming')
+    );
+    this.queue = [];
+    this.bufferedLength = 0;
+    this.timer = null;
+    this._isFull = false;
+  }
+
+  /**
+   * Process the queue in two phases.
+   *
+   * Phase 1: Look for flush markers. When a flush marker is found (even if not at the very front),
+   *   send all text tokens that came before it immediately (ignoring sentence boundaries)
+   *   and then send the flush command. Repeat until there are no flush markers left.
+   *
+   * Phase 2: With the remaining queue (now containing only text items), accumulate text
+   *   up to MAX_CHUNK_SIZE and use sentence-boundary logic to determine a chunk.
+   *   Then, remove the exact tokens (or portions thereof) that were consumed.
+   */
+  async _feedQueue(handlingTimeout = false) {
+    this.logger.debug({ queue: this.queue }, 'TtsStreamingBuffer:_feedQueue');
+    try {
+      if (!this.cs.isTtsStreamOpen || !this.ep) {
+        this.logger.debug('TtsStreamingBuffer:_feedQueue TTS stream is not open or no endpoint available');
+        return;
+      }
+      if (
+        this._connectionStatus === TtsStreamingConnectionStatus.NotConnected ||
+        this._connectionStatus === TtsStreamingConnectionStatus.Failed
+      ) {
+        this.logger.debug('TtsStreamingBuffer:_feedQueue TTS stream is not connected');
+        return;
+      }
+
+      // --- Phase 1: Process flush markers ---
+      // Process any flush marker that isn’t in the very first position.
+      let flushIndex = this.queue.findIndex((item, idx) => item.type === 'flush' && idx > 0);
+      while (flushIndex !== -1) {
+        let flushText = '';
+        // Accumulate all text tokens preceding the flush marker.
+        for (let i = 0; i < flushIndex; i++) {
+          if (this.queue[i].type === 'text') {
+            flushText += this.queue[i].value;
+          }
+        }
+        // Remove those text items.
+        for (let i = 0; i < flushIndex; i++) {
+          const item = this.queue.shift();
+          if (item.type === 'text') {
+            this.bufferedLength -= item.value.length;
+          }
+        }
+        // Remove the flush marker (now at the front).
+        if (this.queue.length > 0 && this.queue[0].type === 'flush') {
+          this.queue.shift();
+        }
+        // Immediately send all accumulated text (ignoring sentence boundaries).
+        if (flushText.length > 0) {
+          const modifiedFlushText = flushText.replace(/\n\n/g, '\n \n');
+          try {
+            await this._api(this.ep, [this.ep.uuid, 'send', modifiedFlushText]);
+          } catch (err) {
+            this.logger.info({ err, flushText }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
+          }
+        }
+        // Send the flush command.
+        await this._doFlush();
+
+        flushIndex = this.queue.findIndex((item, idx) => item.type === 'flush' && idx > 0);
+      }
+
+      // If a flush marker is at the very front, process it.
+      while (this.queue.length > 0 && this.queue[0].type === 'flush') {
+        this.queue.shift();
+        await this._doFlush();
+      }
+
+      // --- Phase 2: Process remaining text tokens ---
+      if (this.queue.length === 0) {
+        this._removeTimer();
+        return;
+      }
+
+      // Accumulate contiguous text tokens (from the front) up to MAX_CHUNK_SIZE.
+      let combinedText = '';
+      for (const item of this.queue) {
+        if (item.type !== 'text') break;
+        combinedText += item.value;
+        if (combinedText.length >= MAX_CHUNK_SIZE) break;
+      }
+      if (combinedText.length === 0) {
+        this._removeTimer();
+        return;
+      }
+
+      const limit = Math.min(MAX_CHUNK_SIZE, combinedText.length);
+      let chunkEnd = findSentenceBoundary(combinedText, limit);
+      if (chunkEnd <= 0) {
+        if (handlingTimeout) {
+          chunkEnd = findWordBoundary(combinedText, limit);
+          if (chunkEnd <= 0) {
+            this._setTimerIfNeeded();
+            return;
+          }
+        } else {
+          this._setTimerIfNeeded();
+          return;
+        }
+      }
+      const chunk = combinedText.slice(0, chunkEnd);
+
+      // Now we iterate over the queue items
+      // and deduct their lengths until we've accounted for chunkEnd characters.
+      let remaining = chunkEnd;
+      let tokensProcessed = 0;
+      for (let i = 0; i < this.queue.length; i++) {
+        const token = this.queue[i];
+        if (token.type !== 'text') break;
+        if (remaining >= token.value.length) {
+          remaining -= token.value.length;
+          tokensProcessed = i + 1;
+        } else {
+          // Partially consumed token: update its value to remove the consumed part.
+          token.value = token.value.slice(remaining);
+          tokensProcessed = i;
+          remaining = 0;
+          break;
+        }
+      }
+      // Remove the fully consumed tokens from the front of the queue.
+      this.queue.splice(0, tokensProcessed);
+      this.bufferedLength -= chunkEnd;
+
+      const modifiedChunk = chunk.replace(/\n\n/g, '\n \n');
+      this.logger.debug(`TtsStreamingBuffer:_feedQueue sending chunk to tts: ${modifiedChunk}`);
+
+      try {
+        await this._api(this.ep, [this.ep.uuid, 'send', modifiedChunk]);
+      } catch (err) {
+        this.logger.info({ err, chunk }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
+      }
+
+      if (this._isFull && this.bufferedLength <= LOW_WATER_BUFFER_SIZE) {
+        this.logger.info('TtsStreamingBuffer throttling: buffer is no longer full - resuming');
+        this._isFull = false;
+        this.emit(TtsStreamingEvents.Resume);
+      }
+
+      return this._feedQueue();
+    } catch (err) {
+      this.logger.info({ err }, 'TtsStreamingBuffer:_feedQueue Error sending TTS chunk');
+      this.queue = [];
+      this.bufferedLength = 0;
+    }
+  }
+
+  async _api(ep, args) {
+    const apiCmd = `uuid_${this.vendor.startsWith('custom:') ? 'custom' : this.vendor}_tts_streaming`;
+    const res = await ep.api(apiCmd, `^^|${args.join('|')}`);
+    if (!res.body?.startsWith('+OK')) {
+      this.logger.info({ args }, `Error calling ${apiCmd}: ${res.body}`);
+      throw new Error(`Error calling ${apiCmd}: ${res.body}`);
+    }
+  }
+
+  _doFlush() {
+    return this._api(this.ep, [this.ep.uuid, 'flush'])
+      .then(() => this.logger.debug('TtsStreamingBuffer:_doFlush sent flush command'))
+      .catch((err) =>
+        this.logger.info(
+          { err },
+          `TtsStreamingBuffer:_doFlush Error flushing TTS streaming: ${JSON.stringify(err)}`
+        )
+      );
+  }
+
+  async _onConnect(vendor) {
+    this.logger.info(`TtsStreamingBuffer:_onConnect streaming tts connection made to ${vendor} successful`);
+    this._connectionStatus = TtsStreamingConnectionStatus.Connected;
+    if (this.queue.length > 0) {
+      await this._feedQueue();
+    }
+  }
+
+  _onConnectFailure(vendor) {
+    this.logger.info(`TtsStreamingBuffer:_onConnectFailure streaming tts connection failed to ${vendor}`);
+    this._connectionStatus = TtsStreamingConnectionStatus.Failed;
+    this.queue = [];
+    this.bufferedLength = 0;
+    this.emit(TtsStreamingEvents.ConnectFailure, { vendor });
+  }
+
+  _setTimerIfNeeded() {
+    if (this.bufferedLength > 0 && !this.timer) {
+      this.logger.debug({queue: this.queue},
+        `TtsStreamingBuffer:_setTimerIfNeeded setting timer because ${this.bufferedLength} buffered`);
+      this.timer = setTimeout(this._onTimeout.bind(this), TIMEOUT_RETRY_MSECS);
+    }
+  }
+
+  _removeTimer() {
+    if (this.timer) {
+      this.logger.debug('TtsStreamingBuffer:_removeTimer clearing timer');
+      clearTimeout(this.timer);
+      this.timer = null;
+    }
+  }
+
+  _onTimeout() {
+    this.logger.debug('TtsStreamingBuffer:_onTimeout Timeout waiting for sentence boundary');
+    this.timer = null;
+    // Check if new text has been added since the timer was set.
+    const now = Date.now();
+    if (now - this.lastUpdateTime < TIMEOUT_RETRY_MSECS) {
+      this.logger.debug('TtsStreamingBuffer:_onTimeout New text received recently; postponing flush.');
+      this._setTimerIfNeeded();
+      return;
+    }
+    this._feedQueue(true);
+  }
+
+  _onTtsEmpty(vendor) {
+    this.emit(TtsStreamingEvents.Empty, { vendor });
+  }
+
+  addCustomEventListener(ep, event, handler) {
+    this.eventHandlers.push({ ep, event, handler });
+    ep.addCustomEventListener(event, handler);
+  }
+
+  removeCustomEventListeners() {
+    this.eventHandlers.forEach((h) => h.ep.removeCustomEventListener(h.event, h.handler));
+  }
+
+  _initHandlers(ep) {
+    [
+      'deepgram',
+      'cartesia',
+      'elevenlabs',
+      'rimelabs',
+      'custom'
+    ].forEach((vendor) => {
+      const eventClassName = `${vendor.charAt(0).toUpperCase() + vendor.slice(1)}TtsStreamingEvents`;
+      const eventClass = require('../utils/constants')[eventClassName];
+      if (!eventClass) throw new Error(`Event class for vendor ${vendor} not found`);
+
+      this.addCustomEventListener(ep, eventClass.Connect, this._onConnect.bind(this, vendor));
+      this.addCustomEventListener(ep, eventClass.ConnectFailure, this._onConnectFailure.bind(this, vendor));
+      this.addCustomEventListener(ep, eventClass.Empty, this._onTtsEmpty.bind(this, vendor));
+    });
+  }
+}
+
+const findSentenceBoundary = (text, limit) => {
+  // Look for punctuation or double newline that signals sentence end.
+  const sentenceEndRegex = /[.!?](?=\s|$)|\n\n/g;
+  let lastSentenceBoundary = -1;
+  let match;
+  while ((match = sentenceEndRegex.exec(text)) && match.index < limit) {
+    const precedingText = text.slice(0, match.index).trim();
+    if (precedingText.length > 0) {
+      if (
+        match[0] === '\n\n' ||
+        (match.index === 0 || !/\d$/.test(text[match.index - 1]))
+      ) {
+        lastSentenceBoundary = match.index + (match[0] === '\n\n' ? 2 : 1);
+      }
+    }
+  }
+  return lastSentenceBoundary;
+};
+
+const findWordBoundary = (text, limit) => {
+  const wordBoundaryRegex = /\s+/g;
+  let lastWordBoundary = -1;
+  let match;
+  while ((match = wordBoundaryRegex.exec(text)) && match.index < limit) {
+    lastWordBoundary = match.index;
+  }
+  return lastWordBoundary;
+};
+
+module.exports = TtsStreamingBuffer;

lib/utils/ws-requestor.js

@@ -1,7 +1,7 @@
 const assert = require('assert');
 const BaseRequestor = require('./base-requestor');
 const short = require('short-uuid');
-const {HookMsgTypes} = require('./constants.json');
+const {HookMsgTypes, WS_CLOSE_CODES} = require('./constants.json');
 const Websocket = require('ws');
 const snakeCaseKeys = require('./snakecase-keys');
 const {
@@ -12,6 +12,20 @@ const {
   JAMBONES_WS_MAX_PAYLOAD,
   HTTP_USER_AGENT_HEADER
 } = require('../config');
+const MTYPE_WANTS_ACK = [
+  'call:status',
+  'verb:status',
+  'jambonz:error',
+  'llm:event',
+  'llm:tool-call',
+  'tts:streaming-event',
+  'tts:tokens-result',
+];
+const MTYPE_NO_DATA = [
+  'llm:tool-output',
+  'tts:flush',
+  'tts:clear'
+];
 
 class WsRequestor extends BaseRequestor {
   constructor(logger, account_sid, hook, secret) {
@@ -41,10 +55,10 @@ class WsRequestor extends BaseRequestor {
    * @param {string} [hook.password] - if basic auth is protecting the endpoint
    * @param {object} [params] - request parameters
    */
-  async request(type, hook, params, httpHeaders = {}) {
+  async request(type, hook, params, httpHeaders = {}, span) {
     assert(HookMsgTypes.includes(type));
     const url = hook.url || hook;
-    const wantsAck = !['call:status', 'verb:status', 'jambonz:error'].includes(type);
+    const wantsAck = !MTYPE_WANTS_ACK.includes(type);
 
     if (this.maliciousClient) {
       this.logger.info({url: this.url}, 'WsRequestor:request - discarding msg to malicious client');
@@ -73,7 +87,7 @@ class WsRequestor extends BaseRequestor {
         this.close();
         this.emit('handover', requestor);
       }
-      return requestor.request(type, hook, params, httpHeaders);
+      return requestor.request(type, hook, params, httpHeaders, span);
     }
 
     /* connect if necessary */
@@ -118,8 +132,8 @@ class WsRequestor extends BaseRequestor {
     assert(this.ws);
 
     /* prepare and send message */
-    let payload = params ? snakeCaseKeys(params, ['customerData', 'sip']) : null;
-    if (type === 'session:new') this._sessionData = payload;
+    let payload = params ? snakeCaseKeys(params, ['customerData', 'sip', 'env_vars', 'args']) : null;
+    if (type === 'session:new' || type === 'session:adulting') this._sessionData = payload;
     if (type === 'session:reconnect') payload = this._sessionData;
     assert.ok(url, 'WsRequestor:request url was not provided');
 
@@ -132,17 +146,23 @@ class WsRequestor extends BaseRequestor {
       type,
       msgid,
       call_sid: this.call_sid,
-      hook: ['verb:hook', 'session:redirect'].includes(type) ? url : undefined,
+      hook: [
+        'verb:hook', 'dial:confirm', 'session:redirect', 'llm:event', 'llm:tool-call'
+      ].includes(type) ? url : undefined,
       data: {...payload},
       ...b3
     };
+    // add msgid to span attributes if it exists
+    if (span) {
+      span.setAttributes({'msgid': msgid});
+    }
 
     const sendQueuedMsgs = () => {
       if (this.queuedMsg.length > 0) {
         for (const {type, hook, params, httpHeaders, promise} of this.queuedMsg) {
           this.logger.debug(`WsRequestor:request - preparing queued ${type} for sending`);
           if (promise) {
-            this.request(type, hook, params, httpHeaders)
+            this.request(type, hook, params, httpHeaders, span)
               .then((res) => promise.resolve(res))
               .catch((err) => promise.reject(err));
           }
@@ -247,13 +267,13 @@ class WsRequestor extends BaseRequestor {
     }
   }
 
-  close() {
+  close(code = WS_CLOSE_CODES.NormalClosure) {
     this.closedGracefully = true;
-    this.logger.debug('WsRequestor:close closing socket');
+    this.logger.debug(`WsRequestor:close closing socket with code ${code}`);
     this._stopPingTimer();
     try {
       if (this.ws) {
-        this.ws.close(1000);
+        this.ws.close(code);
         this.ws.removeAllListeners();
         this.ws = null;
       }
@@ -392,8 +412,9 @@ class WsRequestor extends BaseRequestor {
     /* messages must be JSON format */
     try {
       const obj = JSON.parse(content);
+      this.logger.debug({obj}, 'WsRequestor:_onMessage - received message');
       //const {type, msgid, command, call_sid = this.call_sid, queueCommand = false, data} = obj;
-      const {type, msgid, command, queueCommand = false, data} = obj;
+      const {type, msgid, command, queueCommand = false, tool_call_id, data} = obj;
       const call_sid = obj.callSid || this.call_sid;
 
       //this.logger.debug({obj}, 'WsRequestor:request websocket: received');
@@ -407,8 +428,8 @@ class WsRequestor extends BaseRequestor {
 
         case 'command':
           assert.ok(command, 'command property not supplied');
-          assert.ok(data, 'data property not supplied');
-          this._recvCommand(msgid, command, call_sid, queueCommand, data);
+          assert.ok(data || MTYPE_NO_DATA.includes(command), 'data property not supplied');
+          this._recvCommand(msgid, command, call_sid, queueCommand, tool_call_id, data);
           break;
 
         default:
@@ -416,6 +437,21 @@ class WsRequestor extends BaseRequestor {
       }
     } catch (err) {
       this.logger.info({err, content}, 'WsRequestor:_onMessage - invalid incoming message');
+      const params = {
+        msg: 'InvalidMessage',
+        details: err.message,
+        content: Buffer.from(content).toString('utf-8')
+      };
+      const {writeAlerts, AlertType} = this.Alerter;
+      writeAlerts({
+        account_sid: this.account_sid,
+        alert_type: AlertType.INVALID_APP_PAYLOAD,
+        target_sid: this.call_sid,
+        message: err.message,
+
+      }).catch((err) => this.logger.info({err}, 'Error generating alert for invalid message'));
+      this.request('jambonz:error', '/error', params)
+        .catch((err) => this.logger.debug({err}, 'WsRequestor:_onMessage - Error sending'));
     }
   }
 
@@ -432,10 +468,10 @@ class WsRequestor extends BaseRequestor {
     success && success(data);
   }
 
-  _recvCommand(msgid, command, call_sid, queueCommand, data) {
+  _recvCommand(msgid, command, call_sid, queueCommand, tool_call_id, data) {
     // TODO: validate command
     this.logger.debug({msgid, command, call_sid, queueCommand, data}, 'received command');
-    this.emit('command', {msgid, command, call_sid, queueCommand, data});
+    this.emit('command', {msgid, command, call_sid, queueCommand, tool_call_id, data});
   }
 }
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jambonz-feature-server",
-  "version": "0.9.0",
+  "version": "0.9.4",
   "main": "app.js",
   "engines": {
     "node": ">= 18.x"
@@ -27,14 +27,15 @@
   "dependencies": {
     "@aws-sdk/client-auto-scaling": "^3.549.0",
     "@aws-sdk/client-sns": "^3.549.0",
-    "@jambonz/db-helpers": "^0.9.6",
+    "@jambonz/db-helpers": "^0.9.12",
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/mw-registrar": "^0.2.7",
-    "@jambonz/realtimedb-helpers": "^0.8.8",
-    "@jambonz/speech-utils": "^0.1.11",
+    "@jambonz/realtimedb-helpers": "^0.8.13",
+    "@jambonz/speech-utils": "^0.2.11",
     "@jambonz/stats-collector": "^0.1.10",
-    "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.74",
+    "@jambonz/time-series": "^0.2.13",
+    "@jambonz/verb-specifications": "^0.0.104",
+    "@modelcontextprotocol/sdk": "^1.9.0",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.50.0",
@@ -47,24 +48,22 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.43",
-    "drachtio-srf": "^4.5.35",
+    "drachtio-fsmrf": "^4.0.3",
+    "drachtio-srf": "^5.0.5",
     "express": "^4.19.2",
     "express-validator": "^7.0.1",
-    "ip": "^2.0.1",
     "moment": "^2.30.1",
     "parse-url": "^9.2.0",
     "pino": "^8.20.0",
     "polly-ssml-split": "^0.1.0",
     "proxyquire": "^2.1.3",
-    "sdp-transform": "^2.14.2",
+    "sdp-transform": "^2.15.0",
     "short-uuid": "^5.1.0",
     "sinon": "^17.0.1",
     "to-snake-case": "^1.0.0",
-    "undici": "^6.19.2",
-    "uuid-random": "^1.3.2",
+    "undici": "^7.5.0",
     "verify-aws-sns-signature": "^0.1.0",
-    "ws": "^8.17.1",
+    "ws": "^8.18.0",
     "xml2js": "^0.6.2"
   },
   "devDependencies": {

test/create-call-test.js

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

test/db/jambones-sql.sql

@@ -1,4 +1,5 @@
 /* SQLEditor (MySQL (2))*/
+
 SET FOREIGN_KEY_CHECKS=0;
 
 DROP TABLE IF EXISTS account_static_ips;
@@ -53,6 +54,8 @@ DROP TABLE IF EXISTS signup_history;
 
 DROP TABLE IF EXISTS smpp_addresses;
 
+DROP TABLE IF EXISTS google_custom_voices;
+
 DROP TABLE IF EXISTS speech_credentials;
 
 DROP TABLE IF EXISTS system_information;
@@ -136,6 +139,9 @@ account_sid CHAR(36) NOT NULL,
 is_active BOOLEAN NOT NULL DEFAULT 1,
 username VARCHAR(64),
 password VARCHAR(1024),
+allow_direct_app_calling BOOLEAN NOT NULL DEFAULT 1,
+allow_direct_queue_calling BOOLEAN NOT NULL DEFAULT 1,
+allow_direct_user_calling BOOLEAN NOT NULL DEFAULT 1,
 PRIMARY KEY (client_sid)
 );
 
@@ -338,11 +344,25 @@ label VARCHAR(64),
 PRIMARY KEY (speech_credential_sid)
 );
 
+CREATE TABLE google_custom_voices
+(
+google_custom_voice_sid CHAR(36) NOT NULL UNIQUE ,
+speech_credential_sid CHAR(36) NOT NULL,
+model VARCHAR(512) NOT NULL,
+reported_usage ENUM('REPORTED_USAGE_UNSPECIFIED','REALTIME','OFFLINE') DEFAULT 'REALTIME',
+name VARCHAR(64) NOT NULL,
+voice_cloning_key MEDIUMTEXT,
+use_voice_cloning_key BOOLEAN DEFAULT false,
+PRIMARY KEY (google_custom_voice_sid)
+);
+
 CREATE TABLE system_information
 (
 domain_name VARCHAR(255),
 sip_domain_name VARCHAR(255),
-monitoring_domain_name VARCHAR(255)
+monitoring_domain_name VARCHAR(255),
+private_network_cidr VARCHAR(8192),
+log_level ENUM('info', 'debug') NOT NULL DEFAULT 'info'
 );
 
 CREATE TABLE users
@@ -437,11 +457,14 @@ CREATE TABLE sip_gateways
 sip_gateway_sid CHAR(36),
 ipv4 VARCHAR(128) NOT NULL COMMENT 'ip address or DNS name of the gateway.  For gateways providing inbound calling service, ip address is required.',
 netmask INTEGER NOT NULL DEFAULT 32,
-port INTEGER NOT NULL DEFAULT 5060 COMMENT 'sip signaling port',
+port INTEGER COMMENT 'sip signaling port',
 inbound BOOLEAN NOT NULL COMMENT 'if true, whitelist this IP to allow inbound calls from the gateway',
 outbound BOOLEAN NOT NULL COMMENT 'if true, include in least-cost routing when placing calls to the PSTN',
 voip_carrier_sid CHAR(36) NOT NULL,
 is_active BOOLEAN NOT NULL DEFAULT 1,
+send_options_ping BOOLEAN NOT NULL DEFAULT 0,
+use_sips_scheme BOOLEAN NOT NULL DEFAULT 0,
+pad_crypto BOOLEAN NOT NULL DEFAULT 0,
 protocol ENUM('udp','tcp','tls', 'tls/srtp') DEFAULT 'udp' COMMENT 'Outbound call protocol',
 PRIMARY KEY (sip_gateway_sid)
 ) COMMENT='A whitelisted sip gateway used for origination/termination';
@@ -478,11 +501,19 @@ messaging_hook_sid CHAR(36) COMMENT 'webhook to call for inbound SMS/MMS ',
 app_json TEXT,
 speech_synthesis_vendor VARCHAR(64) NOT NULL DEFAULT 'google',
 speech_synthesis_language VARCHAR(12) NOT NULL DEFAULT 'en-US',
-speech_synthesis_voice VARCHAR(64),
+speech_synthesis_voice VARCHAR(256),
 speech_synthesis_label VARCHAR(64),
 speech_recognizer_vendor VARCHAR(64) NOT NULL DEFAULT 'google',
 speech_recognizer_language VARCHAR(64) NOT NULL DEFAULT 'en-US',
 speech_recognizer_label VARCHAR(64),
+use_for_fallback_speech BOOLEAN DEFAULT false,
+fallback_speech_synthesis_vendor VARCHAR(64),
+fallback_speech_synthesis_language VARCHAR(12),
+fallback_speech_synthesis_voice VARCHAR(256),
+fallback_speech_synthesis_label VARCHAR(64),
+fallback_speech_recognizer_vendor VARCHAR(64),
+fallback_speech_recognizer_language VARCHAR(64),
+fallback_speech_recognizer_label VARCHAR(64),
 created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
 record_all_calls BOOLEAN NOT NULL DEFAULT false,
 PRIMARY KEY (application_sid)
@@ -525,6 +556,7 @@ siprec_hook_sid CHAR(36),
 record_all_calls BOOLEAN NOT NULL DEFAULT false,
 record_format VARCHAR(16) NOT NULL DEFAULT 'mp3',
 bucket_credential VARCHAR(8192) COMMENT 'credential used to authenticate with storage service',
+enable_debug_log BOOLEAN NOT NULL DEFAULT false,
 PRIMARY KEY (account_sid)
 ) COMMENT='An enterprise that uses the platform for comm services';
 
@@ -619,6 +651,10 @@ ALTER TABLE speech_credentials ADD FOREIGN KEY service_provider_sid_idxfk_5 (ser
 CREATE INDEX account_sid_idx ON speech_credentials (account_sid);
 ALTER TABLE speech_credentials ADD FOREIGN KEY account_sid_idxfk_8 (account_sid) REFERENCES accounts (account_sid);
 
+CREATE INDEX google_custom_voice_sid_idx ON google_custom_voices (google_custom_voice_sid);
+CREATE INDEX speech_credential_sid_idx ON google_custom_voices (speech_credential_sid);
+ALTER TABLE google_custom_voices ADD FOREIGN KEY speech_credential_sid_idxfk (speech_credential_sid) REFERENCES speech_credentials (speech_credential_sid) ON DELETE CASCADE;
+
 CREATE INDEX user_sid_idx ON users (user_sid);
 CREATE INDEX email_idx ON users (email);
 CREATE INDEX phone_idx ON users (phone);
@@ -704,4 +740,5 @@ ALTER TABLE accounts ADD FOREIGN KEY queue_event_hook_sid_idxfk (queue_event_hoo
 ALTER TABLE accounts ADD FOREIGN KEY device_calling_application_sid_idxfk (device_calling_application_sid) REFERENCES applications (application_sid);
 
 ALTER TABLE accounts ADD FOREIGN KEY siprec_hook_sid_idxfk (siprec_hook_sid) REFERENCES applications (application_sid);
-SET FOREIGN_KEY_CHECKS=1;
+
+SET FOREIGN_KEY_CHECKS=1;

test/dial-tests.js

@@ -3,9 +3,8 @@ 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 = (ms) => new Promise((r) => setTimeout(r, ms));
+const {provisionCallHook} = require('./utils');
+const { sleepFor } = require('../lib/utils/helpers');
 
 process.on('unhandledRejection', (reason, p) => {
   console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);

test/docker-compose-testbed.yaml

@@ -42,7 +42,7 @@ services:
         ipv4_address: 172.38.0.7
 
   drachtio:
-    image: drachtio/drachtio-server:0.8.25-rc8
+    image: drachtio/drachtio-server:0.8.26
     restart: always
     command: drachtio --contact "sip:*;transport=udp" --mtu 4096 --address 0.0.0.0 --port 9022
     ports:
@@ -57,7 +57,7 @@ services:
         condition: service_healthy
 
   freeswitch:
-    image: drachtio/drachtio-freeswitch-mrf:0.7.3
+    image: drachtio/drachtio-freeswitch-mrf:0.9.2-4
     restart: always
     command: freeswitch --rtp-range-start 20000 --rtp-range-end 20100
     environment:

test/queue-test.js

@@ -3,6 +3,7 @@ const { sippUac } = require('./sipp')('test_fs');
 const clearModule = require('clear-module');
 const {provisionCallHook, provisionActionHook, provisionAnyHook} = require('./utils');
 const bent = require('bent');
+const { sleepFor } = require('../lib/utils/helpers');
 const getJSON = bent('json');
 
 process.on('unhandledRejection', (reason, p) => {
@@ -17,8 +18,6 @@ function connect(connectable) {
   });
 }
 
-const sleepFor = (ms) => new Promise((resolve) => setTimeout(() => resolve(), ms));
-
 test('\'enqueue-dequeue\' tests', async(t) => {
 
   clearModule.all();

test/sip-refer-tests.js

@@ -3,10 +3,9 @@ const { sippUac } = require('./sipp')('test_fs');
 const clearModule = require('clear-module');
 const {provisionCallHook, provisionCustomHook, provisionActionHook} = require('./utils')
 const bent = require('bent');
+const { sleepFor } = require('../lib/utils/helpers');
 const getJSON = bent('json')
 
-const sleepFor = async(ms) => new Promise(resolve => setTimeout(resolve, ms));
-
 process.on('unhandledRejection', (reason, p) => {
   console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
 });

tracer.js

@@ -25,29 +25,38 @@ module.exports = (serviceName) => {
       }),
     });
 
-    let exporter;
+    const exporters = [];
+
     if (OTEL_EXPORTER_JAEGER_AGENT_HOST  || OTEL_EXPORTER_JAEGER_ENDPOINT) {
-      exporter = new JaegerExporter();
-    }
-    else if (OTEL_EXPORTER_ZIPKIN_URL) {
-      exporter = new ZipkinExporter({url:OTEL_EXPORTER_ZIPKIN_URL});
-    }
-    else {
-      exporter = new OTLPTraceExporter({
-        url: OTEL_EXPORTER_COLLECTOR_URL
-      });
+      exporters.push(new JaegerExporter());
     }
 
-    provider.addSpanProcessor(new BatchSpanProcessor(exporter, {
-      // The maximum queue size. After the size is reached spans are dropped.
-      maxQueueSize: 100,
-      // The maximum batch size of every export. It must be smaller or equal to maxQueueSize.
-      maxExportBatchSize: 10,
-      // The interval between two consecutive exports
-      scheduledDelayMillis: 500,
-      // How long the export can run before it is cancelled
-      exportTimeoutMillis: 30000,
-    }));
+    if (OTEL_EXPORTER_ZIPKIN_URL) {
+      exporters.push(new ZipkinExporter({url:OTEL_EXPORTER_ZIPKIN_URL}));
+    }
+
+    if (OTEL_EXPORTER_ZIPKIN_URL) {
+      exporters.push(new ZipkinExporter({url:OTEL_EXPORTER_ZIPKIN_URL}));
+    }
+
+    if (OTEL_EXPORTER_COLLECTOR_URL) {
+      exporters.push(new OTLPTraceExporter({
+        url: OTEL_EXPORTER_COLLECTOR_URL
+      }));
+    }
+
+    exporters.forEach((element) => {
+      provider.addSpanProcessor(new BatchSpanProcessor(element, {
+        // The maximum queue size. After the size is reached spans are dropped.
+        maxQueueSize: 100,
+        // The maximum batch size of every export. It must be smaller or equal to maxQueueSize.
+        maxExportBatchSize: 10,
+        // The interval between two consecutive exports
+        scheduledDelayMillis: 500,
+        // How long the export can run before it is cancelled
+        exportTimeoutMillis: 30000,
+      }));
+    });
 
     // Initialize the OpenTelemetry APIs to use the NodeTracerProvider bindings
     provider.register();