Feature/precache audio (#609)

* wip

* fix for establishing vendor etc

* more fixes

* avoid a pre-caching attempt if synth settings change

.github/workflows/build.yml

@@ -9,7 +9,7 @@ jobs:
       - uses: actions/checkout@v3
       - uses: actions/setup-node@v3
         with:
-          node-version: 16
+          node-version: 18
       - run: npm ci
       - run: npm run jslint
       - run: docker pull drachtio/sipp

README.md

@@ -1,4 +1,4 @@
-# jambones-feature-server ![Build Status](https://github.com/jambonz/jambonz-feature-server/workflows/CI/badge.svg)
+# jambonz-feature-server [![CI](https://github.com/jambonz/jambonz-feature-server/actions/workflows/build.yml/badge.svg)](https://github.com/jambonz/jambonz-feature-server/actions/workflows/build.yml)
 
 This application implements the core feature server of the jambones platform.
 
@@ -37,6 +37,11 @@ Configuration is provided via environment variables:
 |STATS_PORT| listening port for metrics host|no|
 |STATS_PROTOCOL| 'tcp' or 'udp'|no|
 |STATS_TELEGRAF| if 1, metrics will be generated in telegraf format|no|
+|JAMBONZ_RECORD_WS_BASE_URL| recording websocket URL to send the recording audio|no|
+|JAMBONZ_RECORD_WS_USERNAME| recording websocket username|no|
+|JAMBONZ_RECORD_WS_PASSWORD| recording websocket password|no|
+|ANCHOR_MEDIA_ALWAYS| keep media on media server|no|
+|JAMBONZ_DISABLE_DIAL_PAI_HEADER| control P-Asserted-Identity header on B-Leg|no|
 
 ### running under pm2
 Typically, this application runs under [pm2](https://pm2.io) using an [ecosystem.config.js](https://pm2.keymetrics.io/docs/usage/application-declaration/) file similar to this:

app.js

@@ -109,7 +109,7 @@ const disconnect = () => {
     httpServer?.on('close', resolve);
     httpServer?.close();
     srf.disconnect();
-    srf.locals.mediaservers.forEach((ms) => ms.disconnect());
+    srf.locals.mediaservers?.forEach((ms) => ms.disconnect());
   });
 };
 

docs/contributing.md

@@ -45,7 +45,7 @@ The GCP credential is the JSON service key in stringified format.
 
 #### Install Docker
 
-The test suite ralso equires [Docker](https://www.docker.com/) and docker-compose to be installed on your laptop.  Docker is used to set up a network with all of the elements required to test the jambonz-feature-server in a black-box type of fashion.
+The test suite also requires [Docker](https://www.docker.com/) and docker-compose to be installed on your laptop.  Docker is used to set up a network with all of the elements required to test the jambonz-feature-server in a black-box type of fashion.
 
 Once you have docker installed, you can optionally make sure everything Docker-wise is working properly by running this command from the project folder:
 

lib/config.js

@@ -25,13 +25,9 @@ const JAMBONES_MYSQL_USER = process.env.JAMBONES_MYSQL_USER;
 const JAMBONES_MYSQL_PASSWORD = process.env.JAMBONES_MYSQL_PASSWORD;
 const JAMBONES_MYSQL_DATABASE = process.env.JAMBONES_MYSQL_DATABASE;
 const JAMBONES_MYSQL_PORT = parseInt(process.env.JAMBONES_MYSQL_PORT, 10) || 3306;
-const JAMBONES_MYSQL_REFRESH_TTL = process.env.JAMBONES_MYSQL_REFRESH_TTL;
+const JAMBONES_MYSQL_REFRESH_TTL = parseInt(process.env.JAMBONES_MYSQL_REFRESH_TTL, 10) || 0;
 const JAMBONES_MYSQL_CONNECTION_LIMIT = parseInt(process.env.JAMBONES_MYSQL_CONNECTION_LIMIT, 10) || 10;
 
-/* redis */
-const JAMBONES_REDIS_HOST = process.env.JAMBONES_REDIS_HOST;
-const JAMBONES_REDIS_PORT = parseInt(process.env.JAMBONES_REDIS_PORT, 10) || 6379;
-
 /* gather and hints */
 const JAMBONES_GATHER_EARLY_HINTS_MATCH = process.env.JAMBONES_GATHER_EARLY_HINTS_MATCH;
 const JAMBONZ_GATHER_EARLY_HINTS_MATCH = process.env.JAMBONZ_GATHER_EARLY_HINTS_MATCH;
@@ -121,32 +117,20 @@ const HTTP_POOL = process.env.HTTP_POOL && parseInt(process.env.HTTP_POOL);
 const HTTP_POOLSIZE = parseInt(process.env.HTTP_POOLSIZE, 10) || 10;
 const HTTP_PIPELINING = parseInt(process.env.HTTP_PIPELINING, 10) || 1;
 const HTTP_TIMEOUT = 10000;
+const HTTP_PROXY_IP = process.env.JAMBONES_HTTP_PROXY_IP;
+const HTTP_PROXY_PORT = process.env.JAMBONES_HTTP_PROXY_PORT;
+const HTTP_PROXY_PROTOCOL = process.env.JAMBONES_HTTP_PROXY_PROTOCOL || 'http';
+const HTTP_USER_AGENT_HEADER = process.env.JAMBONES_HTTP_USER_AGENT_HEADER || 'jambonz';
 
 const OPTIONS_PING_INTERVAL = parseInt(process.env.OPTIONS_PING_INTERVAL, 10) || 30000;
 
-const JAMBONES_REDIS_SENTINELS = process.env.JAMBONES_REDIS_SENTINELS ? {
+const JAMBONZ_RECORD_WS_BASE_URL = process.env.JAMBONZ_RECORD_WS_BASE_URL || process.env.JAMBONES_RECORD_WS_BASE_URL;
-  sentinels: process.env.JAMBONES_REDIS_SENTINELS.split(',').map((sentinel) => {
+const JAMBONZ_RECORD_WS_USERNAME = process.env.JAMBONZ_RECORD_WS_USERNAME || process.env.JAMBONES_RECORD_WS_USERNAME;
-    let host, port = 26379;
+const JAMBONZ_RECORD_WS_PASSWORD = process.env.JAMBONZ_RECORD_WS_PASSWORD || process.env.JAMBONES_RECORD_WS_PASSWORD;
-    if (sentinel.includes(':')) {
+const JAMBONZ_DISABLE_DIAL_PAI_HEADER = process.env.JAMBONZ_DISABLE_DIAL_PAI_HEADER || false;
-      const arr = sentinel.split(':');
+const JAMBONES_DISABLE_DIRECT_P2P_CALL = process.env.JAMBONES_DISABLE_DIRECT_P2P_CALL || false;
-      host = arr[0];
+
-      port = parseInt(arr[1], 10);
+const JAMBONES_EAGERLY_PRE_CACHE_AUDIO = process.env.JAMBONES_EAGERLY_PRE_CACHE_AUDIO;
-    } else {
-      host = sentinel;
-    }
-    return {host, port};
-  }),
-  name: process.env.JAMBONES_REDIS_SENTINEL_MASTER_NAME,
-  ...(process.env.JAMBONES_REDIS_SENTINEL_PASSWORD && {
-    password: process.env.JAMBONES_REDIS_SENTINEL_PASSWORD
-  }),
-  ...(process.env.JAMBONES_REDIS_SENTINEL_USERNAME && {
-    username: process.env.JAMBONES_REDIS_SENTINEL_USERNAME
-  })
-} : null;
-const JAMBONZ_RECORD_WS_BASE_URL = process.env.JAMBONZ_RECORD_WS_BASE_URL;
-const JAMBONZ_RECORD_WS_USERNAME = process.env.JAMBONZ_RECORD_WS_USERNAME;
-const JAMBONZ_RECORD_WS_PASSWORD = process.env.JAMBONZ_RECORD_WS_PASSWORD;
 
 module.exports = {
   JAMBONES_MYSQL_HOST,
@@ -165,14 +149,12 @@ module.exports = {
   JAMBONZ_GATHER_EARLY_HINTS_MATCH,
   JAMBONES_GATHER_CLEAR_GLOBAL_HINTS_ON_EMPTY_HINTS,
   JAMBONES_FREESWITCH,
-  JAMBONES_REDIS_HOST,
-  JAMBONES_REDIS_PORT,
-  JAMBONES_REDIS_SENTINELS,
   SMPP_URL,
   JAMBONES_NETWORK_CIDR,
   JAMBONES_API_BASE_URL,
   JAMBONES_TIME_SERIES_HOST,
   JAMBONES_INJECT_CONTENT,
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   JAMBONES_ESL_LISTEN_ADDRESS,
   JAMBONES_SBCS,
   JAMBONES_OTEL_ENABLED,
@@ -212,6 +194,10 @@ module.exports = {
   HTTP_POOLSIZE,
   HTTP_PIPELINING,
   HTTP_TIMEOUT,
+  HTTP_PROXY_IP,
+  HTTP_PROXY_PORT,
+  HTTP_PROXY_PROTOCOL,
+  HTTP_USER_AGENT_HEADER,
   OPTIONS_PING_INTERVAL,
   RESPONSE_TIMEOUT_MS,
   JAMBONES_WS_HANDSHAKE_TIMEOUT_MS,
@@ -225,5 +211,7 @@ module.exports = {
   DEEPGRAM_API_KEY,
   JAMBONZ_RECORD_WS_BASE_URL,
   JAMBONZ_RECORD_WS_USERNAME,
-  JAMBONZ_RECORD_WS_PASSWORD
+  JAMBONZ_RECORD_WS_PASSWORD,
+  JAMBONZ_DISABLE_DIAL_PAI_HEADER,
+  JAMBONES_DISABLE_DIRECT_P2P_CALL
 };

lib/dynamic-apps.js

@@ -0,0 +1,54 @@
+const appsMap = {
+  queue: {
+    // Dummy hook to follow later feature server logic.
+    call_hook: {
+      url: 'https://jambonz.org',
+      method: 'GET'
+    },
+    account_sid: '',
+    app_json: [{
+      verb: 'dequeue',
+      name: '',
+      timeout: 5
+    }]
+  },
+  user: {
+    // Dummy hook to follow later feature server logic.
+    call_hook: {
+      url: 'https://jambonz.org',
+      method: 'GET'
+    },
+    account_sid: '',
+    app_json: [{
+      verb: 'dial',
+      callerId: '',
+      answerOnBridge: true,
+      target: [
+        {
+          type: 'user',
+          name: ''
+        }
+      ]
+    }]
+  }
+};
+
+const createJambonzApp = (type, {account_sid, name, caller_id}) => {
+  const app = {...appsMap[type]};
+  app.account_sid = account_sid;
+  switch (type) {
+    case 'queue':
+      app.app_json[0].name = name;
+      break;
+    case 'user':
+      app.app_json[0].callerId = caller_id;
+      app.app_json[0].target[0].name = name;
+      break;
+  }
+  app.app_json = JSON.stringify(app.app_json);
+  return app;
+};
+
+module.exports = {
+  createJambonzApp
+};

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

@@ -5,275 +5,337 @@ const CallInfo = require('../../session/call-info');
 const {CallDirection, CallStatus} = require('../../utils/constants');
 const uuidv4 = require('uuid-random');
 const SipError = require('drachtio-srf').SipError;
+const { validationResult, body } = require('express-validator');
+const { validate } = require('@jambonz/verb-specifications');
 const sysError = require('./error');
 const HttpRequestor = require('../../utils/http-requestor');
 const WsRequestor = require('../../utils/ws-requestor');
 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');
 
-router.post('/', async(req, res) => {
+const removeNullProperties = (obj) => (Object.keys(obj).forEach((key) => obj[key] === null && delete obj[key]), obj);
-  const {logger} = req.app.locals;
+const removeNulls = (req, res, next) => {
-  const accountSid = req.body.account_sid;
+  req.body = removeNullProperties(req.body);
-  const {srf} = require('../../..');
+  next();
+};
 
-  logger.debug({body: req.body}, 'got createCall request');
+router.post('/',
-  try {
+  removeNulls,
-    let uri, cs, to;
+  createCallSchema,
-    // app_json is creaeted by only api-server.
+  body('tag').custom((value) => {
-    // if it available, take it and delete before creating task
+    if (value) {
-    const app_json = req.body.app_json;
+      customSanitizeFunction(value);
-    delete req.body.app_json;
-    const restDial = makeTask(logger, {'rest:dial': req.body});
-    restDial.appJson = app_json;
-    const {lookupAccountDetails, lookupCarrierByPhoneNumber, lookupCarrier} = dbUtils(logger, srf);
-    const {
-      lookupAppBySid
-    }  = srf.locals.dbHelpers;
-    const {getSBC, getFreeswitch} = srf.locals;
-    const sbcAddress = getSBC();
-    if (!sbcAddress) throw new Error('no available SBCs for outbound call creation');
-    const target = restDial.to;
-    const opts = {
-      callingNumber: restDial.from,
-      ...(restDial.callerName && {callingName: restDial.callerName}),
-      headers: req.body.headers || {}
-    };
-
-
-    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 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';
-
-    opts.headers = {
-      ...opts.headers,
-      'X-Jambonz-Routing': target.type,
-      'X-Jambonz-FS-UUID': srf.locals.fsUUID,
-      'X-Call-Sid': callSid,
-      'X-Account-Sid': accountSid,
-      ...(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})
-    };
-
-    switch (target.type) {
-      case 'phone':
-      case 'teams':
-        uri = `sip:${target.number}@${sbcAddress}`;
-        to = target.number;
-        if ('teams' === target.type) {
-          const obj = await lookupTeamsByAccount(accountSid);
-          if (!obj) throw new Error('dial to ms teams not allowed; account must first be configured with teams info');
-          Object.assign(opts.headers, {
-            'X-MS-Teams-FQDN': obj.ms_teams_fqdn,
-            'X-MS-Teams-Tenant-FQDN': target.tenant || obj.tenant_fqdn
-          });
-          if (target.vmail === true) uri = `${uri};opaque=app:voicemail`;
-        }
-        break;
-      case 'user':
-        uri = `sip:${target.name}`;
-        to = target.name;
-        if (target.overrideTo) {
-          Object.assign(opts.headers, {
-            'X-Override-To': target.overrideTo
-          });
-        }
-        break;
-      case 'sip':
-        uri = target.sipUri;
-        to = uri;
-        break;
     }
+    return true;
+  }),
+  async(req, res) => {
+    const {logger} = req.app.locals;
+    const errors = validationResult(req);
+    if (!errors.isEmpty()) {
+      logger.info({errors: errors.array()}, 'POST /Calls: validation errors');
+      return res.status(400).json({ errors: errors.array() });
+    }
+    const accountSid = req.body.account_sid;
+    const {srf} = require('../../..');
 
-    if (target.type === 'phone' && target.trunk) {
+    const app_json = req.body['app_json'];
-      const voip_carrier_sid = await lookupCarrier(req.body.account_sid, target.trunk);
+    try {
-      logger.info(
+    // app_json is created only by api-server.
-        `createCall: selected ${voip_carrier_sid} for requested carrier: ${target.trunk || 'unspecified'})`);
+      if (app_json) {
-      if (voip_carrier_sid) {
+        // if available, delete from req before creating task
-        opts.headers['X-Requested-Carrier-Sid'] = voip_carrier_sid;
+        delete req.body.app_json;
+        // validate possible app_json via verb-specifications
+        validate(logger, JSON.parse(app_json));
       }
+    } catch (err) {
+      logger.debug({ err }, `invalid app_json: ${err.message}`);
     }
 
-    /**
+    logger.debug({body: req.body}, 'got createCall request');
+    try {
+      let uri, cs, to;
+
+      const restDial = makeTask(logger, { 'rest:dial': req.body });
+      restDial.appJson = app_json;
+
+      const {lookupAccountDetails, lookupCarrierByPhoneNumber, lookupCarrier} = dbUtils(logger, srf);
+      const {
+        lookupAppBySid
+      }  = srf.locals.dbHelpers;
+      const {getSBC, getFreeswitch} = srf.locals;
+      const sbcAddress = getSBC();
+      if (!sbcAddress) throw new Error('no available SBCs for outbound call creation');
+      const target = restDial.to;
+      const opts = {
+        callingNumber: restDial.from,
+        ...(restDial.callerName && {callingName: restDial.callerName}),
+        headers: req.body.headers || {}
+      };
+
+
+      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 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';
+      const rootSpan = new RootSpan('rest-call', {
+        callSid,
+        accountSid,
+        ...(req.body?.application_sid && {'X-Application-Sid': req.body.application_sid})
+      });
+
+      opts.headers = {
+        ...opts.headers,
+        'X-Jambonz-Routing': target.type,
+        'X-Jambonz-FS-UUID': srf.locals.fsUUID,
+        'X-Call-Sid': callSid,
+        'X-Account-Sid': accountSid,
+        '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})
+      };
+
+      switch (target.type) {
+        case 'phone':
+        case 'teams':
+          uri = `sip:${target.number}@${sbcAddress}`;
+          to = target.number;
+          if ('teams' === target.type) {
+            const obj = await lookupTeamsByAccount(accountSid);
+            if (!obj) throw new Error('dial to ms teams not allowed; account must first be configured with teams info');
+            Object.assign(opts.headers, {
+              'X-MS-Teams-FQDN': obj.ms_teams_fqdn,
+              'X-MS-Teams-Tenant-FQDN': target.tenant || obj.tenant_fqdn
+            });
+            if (target.vmail === true) uri = `${uri};opaque=app:voicemail`;
+          }
+          break;
+        case 'user':
+          uri = `sip:${target.name}`;
+          to = target.name;
+          if (target.overrideTo) {
+            Object.assign(opts.headers, {
+              'X-Override-To': target.overrideTo
+            });
+          }
+          break;
+        case 'sip':
+          uri = target.sipUri;
+          to = uri;
+          break;
+      }
+
+      if (target.type === 'phone' && target.trunk) {
+        const voip_carrier_sid = await lookupCarrier(req.body.account_sid, target.trunk);
+        logger.info(
+          `createCall: selected ${voip_carrier_sid} for requested carrier: ${target.trunk || 'unspecified'})`);
+        if (voip_carrier_sid) {
+          opts.headers['X-Requested-Carrier-Sid'] = voip_carrier_sid;
+        }
+      }
+
+      /**
      * trunk isn't specified,
      * check if from-number matches any existing numbers on Jambonz
      *  */
-    if (target.type === 'phone' && !target.trunk) {
+      if (target.type === 'phone' && !target.trunk) {
-      const str = restDial.from || '';
+        const str = restDial.from || '';
-      const callingNumber = str.startsWith('+') ? str.substring(1) : str;
+        const callingNumber = str.startsWith('+') ? str.substring(1) : str;
-      const voip_carrier_sid = await lookupCarrierByPhoneNumber(req.body.account_sid, callingNumber);
+        const voip_carrier_sid = await lookupCarrierByPhoneNumber(req.body.account_sid, callingNumber);
-      logger.info(
+        logger.info(
-        `createCall: selected ${voip_carrier_sid} for requested phone number: ${callingNumber || 'unspecified'})`);
+          `createCall: selected ${voip_carrier_sid} for requested phone number: ${callingNumber || 'unspecified'})`);
-      if (voip_carrier_sid) {
+        if (voip_carrier_sid) {
-        opts.headers['X-Requested-Carrier-Sid'] = voip_carrier_sid;
+          opts.headers['X-Requested-Carrier-Sid'] = voip_carrier_sid;
+        }
       }
-    }
 
-    /* create endpoint for outdial */
+      /* create endpoint for outdial */
-    const ms = getFreeswitch();
+      const ms = getFreeswitch();
-    if (!ms) throw new Error('no available Freeswitch for outbound call creation');
+      if (!ms) throw new Error('no available Freeswitch for outbound call creation');
-    const ep = await ms.createEndpoint();
+      const ep = await ms.createEndpoint();
-    logger.debug(`createCall: successfully allocated endpoint, sending INVITE to ${sbcAddress}`);
+      logger.debug(`createCall: successfully allocated endpoint, sending INVITE to ${sbcAddress}`);
 
-    /* launch outdial */
+      /* launch outdial */
-    let sdp, sipLogger;
+      let sdp, sipLogger;
-    const connectStream = async(remoteSdp) => {
+      let dualEp;
-      if (remoteSdp !== sdp) {
+      let localSdp = ep.local.sdp;
-        ep.modify(sdp = remoteSdp);
+
-        return true;
+      if (req.body.dual_streams) {
+        dualEp = await ms.createEndpoint();
+        localSdp = mergeSdpMedia(localSdp, dualEp.local.sdp);
       }
-      return false;
+
-    };
+      const connectStream = async(remoteSdp) => {
-    Object.assign(opts, {
+        if (remoteSdp !== sdp) {
-      proxy: `sip:${sbcAddress}`,
+          sdp = remoteSdp;
-      localSdp: ep.local.sdp
+          if (req.body.dual_streams) {
-    });
+            const [sdpLegA, sdpLebB] = extractSdpMedia(remoteSdp);
-    if (target.auth) opts.auth = target.auth;
+
+            await ep.modify(sdpLegA);
+            await dualEp.modify(sdpLebB);
+            await ep.bridge(dualEp);
+          } else {
+            ep.modify(sdp);
+          }
+          return true;
+        }
+        return false;
+      };
+      Object.assign(opts, {
+        proxy: `sip:${sbcAddress}`,
+        localSdp
+      });
+      if (target.auth) opts.auth = target.auth;
 
 
-    /**
+      /**
      * create our application object -
      * not from the database as per an inbound call,
      * but from the provided params in the request
      */
-    const app = req.body;
+      const app = req.body;
 
-    /**
+      /**
      * attach our requestor and notifier objects
      * these will be used for all http requests we make during this call
      */
-    if ('WS' === app.call_hook?.method || /^wss?:/.test(app.call_hook.url)) {
+      if ('WS' === app.call_hook?.method || /^wss?:/.test(app.call_hook.url)) {
-      logger.debug({call_hook: app.call_hook}, 'creating websocket for call hook');
+        logger.debug({call_hook: app.call_hook}, 'creating websocket for call hook');
-      app.requestor = new WsRequestor(logger, account.account_sid, app.call_hook, account.webhook_secret) ;
+        app.requestor = new WsRequestor(logger, account.account_sid, app.call_hook, account.webhook_secret) ;
-      if (app.call_hook.url === app.call_status_hook.url  || !app.call_status_hook?.url) {
+        if (app.call_hook.url === app.call_status_hook.url  || !app.call_status_hook?.url) {
-        logger.debug('reusing websocket for call status hook');
+          logger.debug('reusing websocket for call status hook');
-        app.notifier = app.requestor;
+          app.notifier = app.requestor;
+        }
+      }
+      else {
+        logger.debug({call_hook: app.call_hook}, 'creating http client for call hook');
+        app.requestor = new HttpRequestor(logger, account.account_sid, app.call_hook, account.webhook_secret);
+      }
+      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');
+      }
+      else if (!app.notifier) {
+        logger.debug('creating null call status hook');
+        app.notifier = {request: () => {}, close: () => {}};
       }
-    }
-    else {
-      logger.debug({call_hook: app.call_hook}, 'creating http client for call hook');
-      app.requestor = new HttpRequestor(logger, account.account_sid, app.call_hook, account.webhook_secret);
-    }
-    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');
-    }
-    else if (!app.notifier) {
-      logger.debug('creating null call status hook');
-      app.notifier = {request: () => {}, close: () => {}};
-    }
 
-    /* now launch the outdial */
+      /* now launch the outdial */
-    try {
+      try {
-      const dlg = await srf.createUAC(uri, {...opts,  followRedirects: true, keepUriOnRedirect: true}, {
+        const dlg = await srf.createUAC(uri, {...opts,  followRedirects: true, keepUriOnRedirect: true}, {
-        cbRequest: (err, inviteReq) => {
+          cbRequest: (err, inviteReq) => {
           /* in case of 302 redirect, this gets called twice, ignore the second
               except to update the req so that it can later be canceled if need be
           */
-          if (res.headersSent) {
+            if (res.headersSent) {
-            logger.info(`create-call: got redirect, updating request to new call-id ${req.get('Call-ID')}`);
+              logger.info(`create-call: got redirect, updating request to new call-id ${req.get('Call-ID')}`);
-            if (cs) cs.req = inviteReq;
+              if (cs) cs.req = inviteReq;
-            return;
+              return;
+            }
+
+            if (err) {
+              logger.error(err, 'createCall Error creating call');
+              res.status(500).send('Call Failure');
+              return;
+            }
+            inviteReq.srf = srf;
+            inviteReq.locals = {
+              ...(inviteReq || {}),
+              callSid,
+              application_sid: app.application_sid
+            };
+            /* ok our outbound INVITE is in flight */
+
+            const tasks = [restDial];
+            sipLogger = logger.child({
+              callSid,
+              callId: inviteReq.get('Call-ID'),
+              accountSid,
+              traceId: rootSpan.traceId
+            });
+            app.requestor.logger = app.notifier.logger = sipLogger;
+            const callInfo = new CallInfo({
+              direction: CallDirection.Outbound,
+              req: inviteReq,
+              to,
+              tag: app.tag,
+              callSid,
+              accountSid: req.body.account_sid,
+              applicationSid: app.application_sid,
+              traceId: rootSpan.traceId
+            });
+            cs = new RestCallSession({
+              logger: sipLogger,
+              application: app,
+              srf,
+              req: inviteReq,
+              ep,
+              ep2: dualEp,
+              tasks,
+              callInfo,
+              accountInfo,
+              rootSpan
+            });
+            cs.exec(req);
+
+            res.status(201).json({sid: cs.callSid, callId: inviteReq.get('Call-ID')});
+
+            sipLogger.info({sid: cs.callSid, callId: inviteReq.get('Call-ID')},
+              `outbound REST call attempt to ${JSON.stringify(target)} has been sent`);
+          },
+          cbProvisional: (prov) => {
+            const callStatus = prov.body ? CallStatus.EarlyMedia : CallStatus.Ringing;
+            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});
           }
-
+        });
-          if (err) {
+        connectStream(dlg.remote.sdp);
-            logger.error(err, 'createCall Error creating call');
+        cs.emit('callStatusChange', {
-            res.status(500).send('Call Failure');
+          callStatus: CallStatus.InProgress,
-            return;
+          sipStatus: 200,
-          }
+          sipReason: 'OK'
-          inviteReq.srf = srf;
+        });
-          inviteReq.locals = {
+        restDial.emit('callStatus', 200);
-            ...(inviteReq || {}),
+        restDial.emit('connect', dlg);
-            callSid,
+      }
-            application_sid: app.application_sid
+      catch (err) {
-          };
+        let callStatus = CallStatus.Failed;
-          /* ok our outbound INVITE is in flight */
+        if (err instanceof SipError) {
-
+          if ([486, 603].includes(err.status)) callStatus = CallStatus.Busy;
-          const tasks = [restDial];
+          else if (487 === err.status) callStatus = CallStatus.NoAnswer;
-          const rootSpan = new RootSpan('rest-call', inviteReq);
+          if (sipLogger) sipLogger.info(`REST outdial failed with ${err.status}`);
-          sipLogger = logger.child({
+          else console.log(`REST outdial failed with ${err.status}`);
-            callSid,
+          if (cs) cs.emit('callStatusChange', {
-            callId: inviteReq.get('Call-ID'),
+            callStatus,
-            accountSid,
+            sipStatus: err.status,
-            traceId: rootSpan.traceId
+            sipReason: err.reason
           });
-          app.requestor.logger = app.notifier.logger = sipLogger;
+          cs.callGone = true;
-          const callInfo = new CallInfo({
-            direction: CallDirection.Outbound,
-            req: inviteReq,
-            to,
-            tag: app.tag,
-            callSid,
-            accountSid: req.body.account_sid,
-            applicationSid: app.application_sid,
-            traceId: rootSpan.traceId
-          });
-          cs = new RestCallSession({
-            logger: sipLogger,
-            application: app,
-            srf,
-            req: inviteReq,
-            ep,
-            tasks,
-            callInfo,
-            accountInfo,
-            rootSpan
-          });
-          cs.exec(req);
-
-          res.status(201).json({sid: cs.callSid, callId: inviteReq.get('Call-ID')});
-
-          sipLogger.info({sid: cs.callSid, callId: inviteReq.get('Call-ID')},
-            `outbound REST call attempt to ${JSON.stringify(target)} has been sent`);
-        },
-        cbProvisional: (prov) => {
-          const callStatus = prov.body ? CallStatus.EarlyMedia : CallStatus.Ringing;
-          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});
         }
-      });
+        else {
-      connectStream(dlg.remote.sdp);
+          if (cs) cs.emit('callStatusChange', {
-      cs.emit('callStatusChange', {
+            callStatus,
-        callStatus: CallStatus.InProgress,
+            sipStatus: 500,
-        sipStatus: 200,
+            sipReason: 'Internal Server Error'
-        sipReason: 'OK'
+          });
-      });
+          if (sipLogger) sipLogger.error({err}, 'REST outdial failed');
-      restDial.emit('callStatus', 200);
+          else console.error(err);
-      restDial.emit('connect', dlg);
+        }
-    }
+        ep.destroy();
-    catch (err) {
+        if (dualEp) {
-      let callStatus = CallStatus.Failed;
+          dualEp.destroy();
-      if (err instanceof SipError) {
+        }
-        if ([486, 603].includes(err.status)) callStatus = CallStatus.Busy;
+        setTimeout(restDial.kill.bind(restDial, cs), 5000);
-        else if (487 === err.status) callStatus = CallStatus.NoAnswer;
-        if (sipLogger) sipLogger.info(`REST outdial failed with ${err.status}`);
-        else console.log(`REST outdial failed with ${err.status}`);
-        if (cs) cs.emit('callStatusChange', {
-          callStatus,
-          sipStatus: err.status,
-          sipReason: err.reason
-        });
       }
-      else {
+    } catch (err) {
-        if (cs) cs.emit('callStatusChange', {
+      sysError(logger, res, err);
-          callStatus,
-          sipStatus: 500,
-          sipReason: 'Internal Server Error'
-        });
-        if (sipLogger) sipLogger.error({err}, 'REST outdial failed');
-        else console.error(err);
-      }
-      ep.destroy();
-      setTimeout(restDial.kill.bind(restDial), 5000);
     }
-  } catch (err) {
+  });
-    sysError(logger, res, err);
-  }
-});
 
 module.exports = router;

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

@@ -9,25 +9,29 @@ const {CallStatus, CallDirection} = require('../../utils/constants');
  */
 function retrieveCallSession(callSid, opts) {
   if (opts.call_status_hook && !opts.call_hook) {
-    throw new DbErrorBadRequest('call_status_hook can be updated only when call_hook is also being updated');
+    throw new DbErrorBadRequest(
+      `call_status_hook can be updated only when call_hook is also being updated for call_sid ${callSid}`);
   }
   const cs = sessionTracker.get(callSid);
   if (!cs) {
-    throw new DbErrorUnprocessableRequest('call session is gone');
+    throw new DbErrorUnprocessableRequest(`call session is gone for call_sid ${callSid}`);
   }
 
   if (opts.call_status === CallStatus.Completed && !cs.hasStableDialog) {
-    throw new DbErrorUnprocessableRequest('current call state is incompatible with requested action');
+    throw new DbErrorUnprocessableRequest(
+      `current call state is incompatible with requested action for call_sid ${callSid}`);
   }
   else if (opts.call_status === CallStatus.NoAnswer) {
     if (cs.direction === CallDirection.Outbound) {
       if (!cs.isOutboundCallRinging) {
-        throw new DbErrorUnprocessableRequest('current call state is incompatible with requested action');
+        throw new DbErrorUnprocessableRequest(
+          `current call state is incompatible with requested action for call_sid ${callSid}`);
       }
     }
     else {
       if (cs.isInboundCallAnswered) {
-        throw new DbErrorUnprocessableRequest('current call state is incompatible with requested action');
+        throw new DbErrorUnprocessableRequest(
+          `current call state is incompatible with requested action for call_sid ${callSid}`);
       }
     }
   }
@@ -41,7 +45,7 @@ function retrieveCallSession(callSid, opts) {
 router.post('/:callSid', async(req, res) => {
   const logger = req.app.locals.logger;
   const callSid = req.params.callSid;
-  logger.debug({body: req.body}, 'got upateCall request');
+  logger.debug({body: req.body}, 'got updateCall request');
   try {
     const cs = retrieveCallSession(callSid, req.body);
     if (!cs) {

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

@@ -0,0 +1,134 @@
+const { checkSchema } = require('express-validator');
+
+/**
+ * @path api-server {{base_url}}/v1/Accounts/:account_sid/Calls
+ * @see https://api.jambonz.org/#243a2edd-7999-41db-bd0d-08082bbab401
+ */
+const createCallSchema = checkSchema({
+  application_sid: {
+    isString: true,
+    optional: true,
+    isLength: { options: { min: 36, max: 36 } },
+    errorMessage: 'Invalid application_sid',
+  },
+  answerOnBridge: {
+    isBoolean: true,
+    optional: true,
+    errorMessage: 'Invalid answerOnBridge',
+  },
+  from: {
+    errorMessage: 'Invalid from',
+    isString: true,
+    isLength: {
+      options: { min: 1, max: 256 },
+    },
+  },
+  fromHost: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid fromHost',
+  },
+  to: {
+    errorMessage: 'Invalid to',
+    isObject: true,
+  },
+  callerName: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid callerName',
+  },
+  amd: {
+    isObject: true,
+    optional: true,
+  },
+  tag: {
+    isObject: true,
+    optional: true,
+    errorMessage: 'Invalid tag',
+  },
+  app_json: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid app_json',
+  },
+  account_sid: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid account_sid',
+    isLength: { options: { min: 36, max: 36 } },
+  },
+  timeout: {
+    isInt: true,
+    optional: true,
+    errorMessage: 'Invalid timeout',
+  },
+  timeLimit: {
+    isInt: true,
+    optional: true,
+    errorMessage: 'Invalid timeLimit',
+  },
+  call_hook: {
+    isObject: true,
+    optional: true,
+    errorMessage: 'Invalid call_hook',
+  },
+  call_status_hook: {
+    isObject: true,
+    optional: true,
+    errorMessage: 'Invalid call_status_hook',
+  },
+  speech_synthesis_vendor: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid speech_synthesis_vendor',
+  },
+  speech_synthesis_language: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid speech_synthesis_language',
+  },
+  speech_synthesis_voice: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid speech_synthesis_voice',
+  },
+  speech_recognizer_vendor: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid speech_recognizer_vendor',
+  },
+  speech_recognizer_language: {
+    isString: true,
+    optional: true,
+    errorMessage: 'Invalid speech_recognizer_language',
+  }
+}, ['body']);
+
+const customSanitizeFunction = (value) => {
+  try {
+    if (Array.isArray(value)) {
+      value = value.map((item) => customSanitizeFunction(item));
+    } else if (typeof value === 'object') {
+      Object.keys(value).forEach((key) => {
+        value[key] = customSanitizeFunction(value[key]);
+      });
+    } else if (typeof value === 'string') {
+      /* trims characters at the beginning and at the end of a string */
+      value = value.trim();
+
+      /* Verify strings including 'http' via new URL */
+      if (value.includes('http')) {
+        value = new URL(value).toString();
+      }
+    }
+  } catch (error) {
+    value = `Error: ${error.message}`;
+  }
+
+  return value;
+};
+
+module.exports = {
+  createCallSchema,
+  customSanitizeFunction
+};

lib/middleware.js

@@ -11,8 +11,10 @@ const dbUtils = require('./utils/db-utils');
 const RootSpan = require('./utils/call-tracer');
 const listTaskNames = require('./utils/summarize-tasks');
 const {
-  JAMBONES_MYSQL_REFRESH_TTL
+  JAMBONES_MYSQL_REFRESH_TTL,
+  JAMBONES_DISABLE_DIRECT_P2P_CALL
 } = require('./config');
+const { createJambonzApp } = require('./dynamic-apps');
 
 module.exports = function(srf, logger) {
   const {
@@ -20,17 +22,21 @@ module.exports = function(srf, logger) {
     lookupAppByRegex,
     lookupAppBySid,
     lookupAppByRealm,
-    lookupAppByTeamsTenant
+    lookupAppByTeamsTenant,
+    registrar,
+    lookupClientByAccountAndUsername
   }  = srf.locals.dbHelpers;
   const {
     writeAlerts,
     AlertType
   } = srf.locals;
-  const {lookupAccountDetails} = dbUtils(logger, srf);
+  const {lookupAccountDetails, lookupGoogleCustomVoice} = dbUtils(logger, srf);
 
-  function initLocals(req, res, next) {
+  async function initLocals(req, res, next) {
     const callId = req.get('Call-ID');
+    const uri = parseUri(req.uri);
     logger.info({
+      uri,
       callId,
       callingNumber: req.callingNumber,
       calledNumber: req.calledNumber
@@ -42,12 +48,52 @@ module.exports = function(srf, logger) {
     const callSid = req.has('X-Retain-Call-Sid') ? req.get('X-Retain-Call-Sid') : uuidv4();
     const account_sid = req.get('X-Account-Sid');
     req.locals = {callSid, account_sid, callId};
-    if (req.has('X-Application-Sid')) {
+
+    let clientDb = null;
+    if (req.has('X-Authenticated-User')) {
+      req.locals.originatingUser = req.get('X-Authenticated-User');
+      let clientSettings;
+      const arr = /^(.*)@(.*)/.exec(req.locals.originatingUser);
+      if (arr) {
+        [clientSettings] = await lookupClientByAccountAndUsername(account_sid, arr[1]);
+      }
+      clientDb = await registrar.query(req.locals.originatingUser);
+      clientDb = {
+        ...clientDb,
+        ...clientSettings,
+      };
+    }
+
+    // check for call to application
+    if (uri.user?.startsWith('app-') && req.locals.originatingUser && clientDb?.allow_direct_app_calling) {
+      const application_sid = uri.user.match(/app-(.*)/)[1];
+      logger.debug(`got application from Request URI header: ${application_sid}`);
+      req.locals.application_sid = application_sid;
+    } else if (req.has('X-Application-Sid')) {
       const application_sid = req.get('X-Application-Sid');
       logger.debug(`got application from X-Application-Sid header: ${application_sid}`);
       req.locals.application_sid = application_sid;
     }
-    if (req.has('X-Authenticated-User')) req.locals.originatingUser = req.get('X-Authenticated-User');
+    // check for call to queue
+    if (uri.user?.startsWith('queue-') && req.locals.originatingUser && clientDb?.allow_direct_queue_calling) {
+      const queue_name = uri.user.match(/queue-(.*)/)[1];
+      logger.debug(`got Queue from Request URI header: ${queue_name}`);
+      req.locals.queue_name = queue_name;
+    }
+    // check for call to registered user
+    if (!JAMBONES_DISABLE_DIRECT_P2P_CALL && req.locals.originatingUser && clientDb?.allow_direct_user_calling) {
+      const arr = /^(.*)@(.*)/.exec(req.locals.originatingUser);
+      if (arr) {
+        const sipRealm = arr[2];
+        const called_user = `${req.calledNumber}@${sipRealm}`;
+        const reg = await registrar.query(called_user);
+        if (reg) {
+          logger.debug(`got called Number is a registered user: ${called_user}`);
+          req.locals.called_user = called_user;
+        }
+      }
+    }
+
     if (req.has('X-MS-Teams-Tenant-FQDN')) req.locals.msTeamsTenant = req.get('X-MS-Teams-Tenant-FQDN');
     if (req.has('X-Cisco-Recording-Participant')) {
       const ciscoParticipants = req.get('X-Cisco-Recording-Participant');
@@ -120,7 +166,7 @@ module.exports = function(srf, logger) {
         };
         logger.info({callId, metadata, sdp}, 'successfully parsed SIPREC payload');
       } catch (err) {
-        logger.info({callId}, 'Error parsing multipart payload');
+        logger.info({err, callId}, 'Error parsing multipart payload');
         return res.send(503);
       }
     }
@@ -184,15 +230,24 @@ module.exports = function(srf, logger) {
     const {span} = rootSpan.startChildSpan('lookupApplication');
     try {
       let app;
-      if (req.locals.application_sid) app = await lookupAppBySid(req.locals.application_sid);
+      if (req.locals.queue_name) {
-      else if (req.locals.originatingUser) {
+        logger.debug(`calling to queue ${req.locals.queue_name}, generating queue app`);
+        app = createJambonzApp('queue', {account_sid, name: req.locals.queue_name});
+      } else if (req.locals.called_user) {
+        logger.debug(`calling to registered user ${req.locals.called_user}, generating dial app`);
+        app = createJambonzApp('user',
+          {account_sid, name: req.locals.called_user, caller_id: req.locals.callingNumber});
+      } else if (req.locals.application_sid) {
+        app = await lookupAppBySid(req.locals.application_sid);
+      } else if (req.locals.originatingUser) {
         const arr = /^(.*)@(.*)/.exec(req.locals.originatingUser);
         if (arr) {
           const sipRealm = arr[2];
           logger.debug(`looking for device calling app for realm ${sipRealm}`);
           app = await lookupAppByRealm(sipRealm);
-          if (app) logger.debug({app}, `retrieved device calling app for realm ${sipRealm}`);
+          if (app) {
-
+            logger.debug({app}, `retrieved device calling app for realm ${sipRealm}`);
+          }
         }
       }
       else if (req.locals.msTeamsTenant) {
@@ -257,7 +312,22 @@ module.exports = function(srf, logger) {
         app2.requestor = new HttpRequestor(logger, account_sid, app.call_hook, accountInfo.account.webhook_secret);
         if (app.call_status_hook) app2.notifier = new HttpRequestor(logger, account_sid, app.call_status_hook,
           accountInfo.account.webhook_secret);
-        else app2.notifier = {request: () => {}};
+        else app2.notifier = {request: () => {}, close: () => {}};
+      }
+
+      // Resolve application.speech_synthesis_voice if it's custom voice
+      if (app2.speech_synthesis_vendor === 'google' && app2.speech_synthesis_voice.startsWith('custom_')) {
+        const arr = /custom_(.*)/.exec(app2.speech_synthesis_voice);
+        if (arr) {
+          const google_custom_voice_sid = arr[1];
+          const [custom_voice] = await lookupGoogleCustomVoice(google_custom_voice_sid);
+          if (custom_voice) {
+            app2.speech_synthesis_voice = {
+              reportedUsage: custom_voice.reported_usage,
+              model: custom_voice.model
+            };
+          }
+        }
       }
 
       req.locals.application = app2;
@@ -299,22 +369,30 @@ module.exports = function(srf, logger) {
       if (app.app_json) {
         json = JSON.parse(app.app_json);
       } else {
+        const defaults = {
+          synthesizer: {
+            vendor: app.speech_synthesis_vendor,
+            ...(app.speech_synthesis_label && {label: app.speech_synthesis_label}),
+            language: app.speech_synthesis_language,
+            voice: app.speech_synthesis_voice,
+            ...(app.fallback_speech_synthesis_vendor && {fallback_vendor: app.fallback_speech_synthesis_vendor}),
+            ...(app.fallback_speech_synthesis_label && {fallback_label: app.fallback_speech_synthesis_label}),
+            ...(app.fallback_speech_synthesis_language && {fallback_language: app.fallback_speech_synthesis_language}),
+            ...(app.fallback_speech_synthesis_voice && {fallback_voice: app.fallback_speech_synthesis_voice})
+          },
+          recognizer: {
+            vendor: app.speech_recognizer_vendor,
+            ...(app.speech_recognizer_label && {label: app.speech_recognizer_label}),
+            language: app.speech_recognizer_language,
+            ...(app.fallback_speech_recognizer_vendor && {fallback_vendor: app.fallback_speech_recognizer_vendor}),
+            ...(app.fallback_speech_recognizer_label && {fallback_label: app.fallback_speech_recognizer_label}),
+            ...(app.fallback_speech_recognizer_language && {fallback_language: app.fallback_speech_recognizer_language})
+          }
+        };
         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: {
-              synthesizer: {
-                vendor: app.speech_synthesis_vendor,
-                language: app.speech_synthesis_language,
-                voice: app.speech_synthesis_voice
-              },
-              recognizer: {
-                vendor: app.speech_recognizer_vendor,
-                language: app.speech_recognizer_language
-              }
-            }
-          });
         logger.debug({ params }, 'sending initial webhook');
         const obj = rootSpan.startChildSpan('performAppWebhook');
         span = obj.span;

lib/session/adulting-call-session.js

@@ -1,4 +1,6 @@
 const CallSession = require('./call-session');
+const {CallStatus} = require('../utils/constants');
+const moment = require('moment');
 
 /**
  * @classdesc Subclass of CallSession.  Represents a CallSession
@@ -19,12 +21,14 @@ class AdultingCallSession extends CallSession {
       rootSpan
     });
     this.sd = singleDialer;
+    this.req = callInfo.req;
 
     this.sd.dlg.on('destroy', () => {
       this.logger.info('AdultingCallSession: called party hung up');
       this._callReleased();
     });
     this.sd.emit('adulting');
+    this.on('callStatusChange', this._notifyCallStatusChange.bind(this));
   }
 
   get dlg() {
@@ -49,6 +53,18 @@ class AdultingCallSession extends CallSession {
   }
 
   _callerHungup() {
+    if (this.dlg.connectTime) {
+      const duration = moment().diff(this.dlg.connectTime, 'seconds');
+      this.rootSpan.setAttributes({'call.termination': 'hangup by caller'});
+      this.callInfo.callTerminationBy = 'caller';
+      this.emit('callStatusChange', {
+        callStatus: CallStatus.Completed,
+        duration
+      });
+    }
+    this.logger.info('InboundCallSession: caller hung up');
+    this._callReleased();
+    this.req.removeAllListeners('cancel');
   }
 }
 

lib/session/call-session.js

@@ -19,11 +19,10 @@ const HttpRequestor = require('../utils/http-requestor');
 const WsRequestor = require('../utils/ws-requestor');
 const {
   JAMBONES_INJECT_CONTENT,
+  JAMBONES_EAGERLY_PRE_CACHE_AUDIO,
   AWS_REGION,
-  JAMBONZ_RECORD_WS_BASE_URL,
-  JAMBONZ_RECORD_WS_USERNAME,
-  JAMBONZ_RECORD_WS_PASSWORD,
 } = require('../config');
+const BackgroundTaskManager = require('../utils/background-task-manager');
 const BADPRECONDITIONS = 'preconditions not met';
 const CALLER_CANCELLED_ERR_MSG = 'Response not sent due to unknown transaction';
 
@@ -66,6 +65,11 @@ class CallSession extends Emitter {
     this.callGone = false;
     this.notifiedComplete = false;
     this.rootSpan = rootSpan;
+    this.backgroundTaskManager = new BackgroundTaskManager({
+      cs: this,
+      logger,
+      rootSpan
+    });
 
     this._origRecognizerSettings = {
       vendor: this.application?.speech_recognizer_vendor,
@@ -136,8 +140,7 @@ class CallSession extends Emitter {
   }
 
   get isBackGroundListen() {
-    return !(this.backgroundListenTask === null ||
+    return this.backgroundTaskManager.isTaskRunning('listen');
-    this.backgroundListenTask === undefined);
   }
 
   /**
@@ -170,6 +173,18 @@ class CallSession extends Emitter {
     return this.application.notifier;
   }
 
+  /**
+   * syntheizer
+   */
+
+  get synthesizer() {
+    return this._synthesizer;
+  }
+
+  set synthesizer(synth) {
+    this._synthesizer = synth;
+  }
+
   /**
    * default vendor to use for speech synthesis if not provided in the app
    */
@@ -179,6 +194,30 @@ class CallSession extends Emitter {
   set speechSynthesisVendor(vendor) {
     this.application.speech_synthesis_vendor = vendor;
   }
+
+  get fallbackSpeechSynthesisVendor() {
+    return this.application.fallback_speech_synthesis_vendor;
+  }
+  set fallbackSpeechSynthesisVendor(vendor) {
+    this.application.fallback_speech_synthesis_vendor = vendor;
+  }
+
+  /**
+   * default label to use for speech synthesis if not provided in the app
+   */
+  get speechSynthesisLabel() {
+    return this.application.speech_synthesis_label;
+  }
+  set speechSynthesisLabel(label) {
+    this.application.speech_synthesis_label = label;
+  }
+
+  get fallbackSpeechSynthesisLabel() {
+    return this.application.fallback_speech_synthesis_label;
+  }
+  set fallbackSpeechSynthesisLabel(label) {
+    this.application.fallback_speech_synthesis_label = label;
+  }
   /**
    * default voice to use for speech synthesis if not provided in the app
    */
@@ -188,6 +227,13 @@ class CallSession extends Emitter {
   set speechSynthesisVoice(voice) {
     this.application.speech_synthesis_voice = voice;
   }
+
+  get fallbackSpeechSynthesisVoice() {
+    return this.application.fallback_speech_synthesis_voice;
+  }
+  set fallbackSpeechSynthesisVoice(voice) {
+    this.application.fallback_speech_synthesis_voice = voice;
+  }
   /**
    * default language to use for speech synthesis if not provided in the app
    */
@@ -198,6 +244,13 @@ class CallSession extends Emitter {
     this.application.speech_synthesis_language = language;
   }
 
+  get fallbackSpeechSynthesisLanguage() {
+    return this.application.fallback_speech_synthesis_language;
+  }
+  set fallbackSpeechSynthesisLanguage(language) {
+    this.application.fallback_speech_synthesis_language = language;
+  }
+
   /**
    * default vendor to use for speech recognition if not provided in the app
    */
@@ -207,6 +260,39 @@ class CallSession extends Emitter {
   set speechRecognizerVendor(vendor) {
     this.application.speech_recognizer_vendor = vendor;
   }
+
+  get fallbackSpeechRecognizerVendor() {
+    return this.application.fallback_speech_recognizer_vendor;
+  }
+  set fallbackSpeechRecognizerVendor(vendor) {
+    this.application.fallback_speech_recognizer_vendor = vendor;
+  }
+  /**
+   * recognizer
+   */
+  get recognizer() {
+    return this._recognizer;
+  }
+
+  set recognizer(rec) {
+    this._recognizer = rec;
+  }
+  /**
+   * default vendor to use for speech recognition if not provided in the app
+   */
+  get speechRecognizerLabel() {
+    return this.application.speech_recognizer_label;
+  }
+  set speechRecognizerLabel(label) {
+    this.application.speech_recognizer_label = label;
+  }
+
+  get fallbackSpeechRecognizerLabel() {
+    return this.application.fallback_speech_recognizer_label;
+  }
+  set fallbackSpeechRecognizerLabel(label) {
+    this.application.fallback_speech_recognizer_label = label;
+  }
   /**
  * default language to use for speech recognition if not provided in the app
  */
@@ -217,6 +303,13 @@ class CallSession extends Emitter {
     this.application.speech_recognizer_language = language;
   }
 
+  get fallbackSpeechRecognizerLanguage() {
+    return this.application.fallback_speech_recognizer_language;
+  }
+  set fallbackSpeechRecognizerLanguage(language) {
+    this.application.fallback_speech_recognizer_language = language;
+  }
+
   /**
    * indicates whether the call currently in progress
    */
@@ -289,11 +382,11 @@ class CallSession extends Emitter {
   }
 
   get isBotModeEnabled() {
-    return this.backgroundGatherTask;
+    return this.backgroundTaskManager.isTaskRunning('bargeIn');
   }
 
   get isListenEnabled() {
-    return this.backgroundListenTask;
+    return this.backgroundTaskManager.isTaskRunning('listen');
   }
 
   get b3() {
@@ -338,6 +431,22 @@ class CallSession extends Emitter {
     return this._globalSttPunctuation;
   }
 
+  get onHoldMusic() {
+    return this._onHoldMusic;
+  }
+
+  set onHoldMusic(url) {
+    this._onHoldMusic = url;
+  }
+
+  get sipRequestWithinDialogHook() {
+    return this._sipRequestWithinDialogHook;
+  }
+
+  set sipRequestWithinDialogHook(url) {
+    this._sipRequestWithinDialogHook = url;
+  }
+
   hasGlobalSttPunctuation() {
     return this._globalSttPunctuation !== undefined;
   }
@@ -506,117 +615,41 @@ class CallSession extends Emitter {
     }
   }
 
-  async startBackgroundListen(opts, bugname) {
-    if (this.isListenEnabled) {
-      this.logger.info('CallSession:startBackgroundListen - listen is already enabled, ignoring request');
-      return;
-    }
-    try {
-      this.logger.debug({opts}, 'CallSession:startBackgroundListen');
-      const t = normalizeJambones(this.logger, [opts]);
-      this.backgroundListenTask = makeTask(this.logger, t[0]);
-      this.backgroundListenTask.bugname = bugname;
-      // Remove unneeded customer data to be sent to api server.
-      this.backgroundListenTask.ignoreCustomerData = true;
-      const resources = await this._evaluatePreconditions(this.backgroundListenTask);
-      const {span, ctx} = this.rootSpan.startChildSpan(`background-listen:${this.backgroundListenTask.summary}`);
-      this.backgroundListenTask.span = span;
-      this.backgroundListenTask.ctx = ctx;
-      this.backgroundListenTask.exec(this, resources)
-        .then(() => {
-          this.logger.info('CallSession:startBackgroundListen: listen completed');
-          this.backgroundListenTask && this.backgroundListenTask.removeAllListeners();
-          this.backgroundListenTask && this.backgroundListenTask.span.end();
-          this.backgroundListenTask = null;
-          return;
-        })
-        .catch((err) => {
-          this.logger.info({err}, 'CallSession:startBackgroundListen: listen threw error');
-          this.backgroundListenTask && this.backgroundListenTask.removeAllListeners();
-          this.backgroundListenTask && this.backgroundListenTask.span.end();
-          this.backgroundListenTask = null;
-        });
-    } catch (err) {
-      this.logger.info({err, opts}, 'CallSession:startBackgroundListen - Error creating listen task');
-    }
-  }
-
-  async stopBackgroundListen() {
-    this.logger.debug('CallSession:stopBackgroundListen');
-    try {
-      if (this.backgroundListenTask) {
-        this.backgroundListenTask.removeAllListeners();
-        this.backgroundListenTask.kill().catch(() => {});
-      }
-    } catch (err) {
-      this.logger.info({err}, 'CallSession:stopBackgroundListen - Error stopping listen task');
-    }
-  }
-
   async enableBotMode(gather, autoEnable) {
     try {
-      const t = normalizeJambones(this.logger, [gather]);
+      let task;
-      const task = makeTask(this.logger, t[0]);
-
       if (this.isBotModeEnabled) {
-        const currInput = this.backgroundGatherTask.input;
+        task = this.backgroundTaskManager.getTask('bargeIn');
-        const newInput = task.input;
+        const currInput = task.input;
+        const t = normalizeJambones(this.logger, [gather]);
+        const tmpTask = makeTask(this.logger, t[0]);
+        const newInput = tmpTask.input;
         if (JSON.stringify(currInput) === JSON.stringify(newInput)) {
           this.logger.info('CallSession:enableBotMode - bot mode currently enabled, ignoring request to start again');
           return;
-        }
+        } else {
-        else {
           this.logger.info({currInput, newInput},
-            'CallSession:enableBotMode - restarting background gather to apply new input type');
+            'CallSession:enableBotMode - restarting background bargeIn to apply new input type');
-          this.backgroundGatherTask.sticky = false;
+          task.sticky = false;
           await this.disableBotMode();
         }
       }
-      this.backgroundGatherTask = task;
+      task = await this.backgroundTaskManager.newTask('bargeIn', gather);
-      this._bargeInEnabled = true;
+      task.sticky = autoEnable;
-      this.backgroundGatherTask
+      task.once('bargeIn-done', () => {
-        .once('dtmf', this._clearTasks.bind(this, this.backgroundGatherTask))
+        if (this.requestor instanceof WsRequestor) {
-        .once('vad', this._clearTasks.bind(this, this.backgroundGatherTask))
+          try {
-        .once('transcription', this._clearTasks.bind(this, this.backgroundGatherTask))
+            this.kill(true);
-        .once('timeout', this._clearTasks.bind(this, this.backgroundGatherTask));
+          } catch (err) {}
-      this.logger.info({gather}, 'CallSession:enableBotMode - starting background gather');
+        }
-      const resources = await this._evaluatePreconditions(this.backgroundGatherTask);
+      });
-      const {span, ctx} = this.rootSpan.startChildSpan(`background-gather:${this.backgroundGatherTask.summary}`);
+      this.logger.info({gather}, 'CallSession:enableBotMode - starting background bargeIn');
-      this.backgroundGatherTask.span = span;
-      this.backgroundGatherTask.ctx = ctx;
-      this.backgroundGatherTask.sticky = autoEnable;
-      this.backgroundGatherTask.exec(this, resources)
-        .then(() => {
-          this.logger.info('CallSession:enableBotMode: gather completed');
-          this.backgroundGatherTask && this.backgroundGatherTask.removeAllListeners();
-          this.backgroundGatherTask && this.backgroundGatherTask.span.end();
-          const sticky = this.backgroundGatherTask?.sticky;
-          this.backgroundGatherTask = null;
-          if (sticky && !this.callGone && !this._stopping && this._bargeInEnabled) {
-            this.logger.info('CallSession:enableBotMode: restarting background gather');
-            setImmediate(() => this.enableBotMode(gather, true));
-          }
-          return;
-        })
-        .catch((err) => {
-          this.logger.info({err}, 'CallSession:enableBotMode: gather threw error');
-          this.backgroundGatherTask && this.backgroundGatherTask.removeAllListeners();
-          this.backgroundGatherTask && this.backgroundGatherTask.span.end();
-          this.backgroundGatherTask = null;
-        });
     } catch (err) {
-      this.logger.info({err, gather}, 'CallSession:enableBotMode - Error creating gather task');
+      this.logger.info({err, gather}, 'CallSession:enableBotMode - Error creating bargeIn task');
     }
   }
   async disableBotMode() {
-    this._bargeInEnabled = false;
+    this.backgroundTaskManager.stop('bargeIn');
-    if (this.backgroundGatherTask) {
-      try {
-        this.backgroundGatherTask.removeAllListeners();
-        await this.backgroundGatherTask.kill();
-      } catch (err) {}
-      this.backgroundGatherTask = null;
-    }
   }
 
   setConferenceDetails(memberId, confName, confUuid) {
@@ -640,15 +673,28 @@ class CallSession extends Emitter {
    * Check for speech credentials for the specified vendor
    * @param {*} vendor - google or aws
    */
-  getSpeechCredentials(vendor, type) {
+  getSpeechCredentials(vendor, type, label = null) {
     const {writeAlerts, AlertType} = this.srf.locals;
     if (this.accountInfo.speech && this.accountInfo.speech.length > 0) {
-      const credential = this.accountInfo.speech.find((s) => s.vendor === vendor);
+      // firstly check if account level has expected credential
+      let credential = this.accountInfo.speech.find((s) => s.vendor === vendor &&
+        s.label === label && s.account_sid);
+      if (!credential) {
+        // check if SP level has expected credential
+        credential = this.accountInfo.speech.find((s) => s.vendor === vendor &&
+        s.label === label && !s.account_sid);
+      }
       if (credential && (
         (type === 'tts' && credential.use_for_tts) ||
         (type === 'stt' && credential.use_for_stt)
       )) {
+        this.logger.info(
+          `Speech vendor: ${credential.vendor} ${credential.label ? `, label: ${credential.label}` : ''} selected`);
         if ('google' === vendor) {
+          if (type === 'tts' && !credential.tts_tested_ok ||
+            type === 'stt' && !credential.stt_tested_ok) {
+            return;
+          }
           try {
             const cred = JSON.parse(credential.service_key.replace(/\n/g, '\\n'));
             return {
@@ -680,8 +726,10 @@ class CallSession extends Emitter {
             region: credential.region,
             use_custom_stt: credential.use_custom_stt,
             custom_stt_endpoint: credential.custom_stt_endpoint,
+            custom_stt_endpoint_url: credential.custom_stt_endpoint_url,
             use_custom_tts: credential.use_custom_tts,
-            custom_tts_endpoint: credential.custom_tts_endpoint
+            custom_tts_endpoint: credential.custom_tts_endpoint,
+            custom_tts_endpoint_url: credential.custom_tts_endpoint_url
           };
         }
         else if ('wellsaid' === vendor) {
@@ -726,7 +774,28 @@ class CallSession extends Emitter {
             riva_server_uri: credential.riva_server_uri
           };
         }
-        else if (vendor.startsWith('custom:')) {
+        else if ('cobalt' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            cobalt_server_uri: credential.cobalt_server_uri
+          };
+        } else if ('elevenlabs' === vendor) {
+          return {
+            api_key: credential.api_key,
+            model_id: credential.model_id,
+            options: credential.options
+          };
+        } else if ('assemblyai' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            api_key: credential.api_key
+          };
+        } else if ('whisper' === vendor) {
+          return {
+            api_key: credential.api_key,
+            model_id: credential.model_id
+          };
+        } else if (vendor.startsWith('custom:')) {
           return {
             speech_credential_sid: credential.speech_credential_sid,
             auth_token: credential.auth_token,
@@ -764,7 +833,7 @@ class CallSession extends Emitter {
         let skip = false;
         this.currentTask = task;
         if (TaskName.Gather === task.name && this.isBotModeEnabled) {
-          if (this.backgroundGatherTask.updateTaskInProgress(task) !== false) {
+          if (this.backgroundTaskManager.getTask('bargeIn').updateTaskInProgress(task) !== false) {
             this.logger.info(`CallSession:exec skipping #${stackNum}:${taskNum}: ${task.name}`);
             skip = true;
           }
@@ -796,23 +865,16 @@ class CallSession extends Emitter {
         }
       }
 
-      if (0 === this.tasks.length && this.requestor instanceof WsRequestor && !this.callGone) {
+      if (0 === this.tasks.length &&
-        //let span;
+        this.requestor instanceof WsRequestor &&
+        !this.requestor.closedGracefully &&
+        !this.callGone &&
+        !this.isConfirmCallSession
+      ) {
         try {
-          //const {span} = this.rootSpan.startChildSpan('waiting for commands');
-          //const {reason, queue, command} = await this._awaitCommandsOrHangup();
-          /*
-          span.setAttributes({
-            'completion.reason': reason,
-            'async.request.queue': queue,
-            'async.request.command': command
-          });
-          span.end();
-          */
           await this._awaitCommandsOrHangup();
           if (this.callGone) break;
         } catch (err) {
-          //span.end();
           this.logger.info(err, 'CallSession:exec - error waiting for new commands');
           break;
         }
@@ -992,6 +1054,39 @@ class CallSession extends Emitter {
     listenTask.updateListen(opts.listen_status);
   }
 
+  /**
+   * perform live call control -- change Transcribe status
+   * @param {object} opts
+   * @param {string} opts.transcribe_status - 'pause' or 'resume'
+  */
+  async _lccTranscribeStatus(opts) {
+    const task = this.currentTask;
+    if (!task || ![TaskName.Dial, TaskName.Transcribe].includes(task.name)) {
+      return this.logger.info(`CallSession:_lccTranscribeStatus - invalid transcribe_status in task ${task.name}`);
+    }
+    const transcribeTask = task.name === TaskName.Transcribe ? task : task.transcribeTask;
+    if (!transcribeTask) {
+      return this.logger
+        .info('CallSession:_lccTranscribeStatus - invalid transcribe_status: Dial does not have a Transcribe');
+    }
+    transcribeTask.updateTranscribe(opts.transcribe_status);
+  }
+
+  /**
+   * perform live call control -- update customer data
+   * @param {object} opts
+   * @param {object} opts.tag - customer data
+  */
+  _lccTag(opts) {
+    const {tag} = opts;
+    if (typeof tag !== 'object' || Array.isArray(tag) || tag === null) {
+      this.logger.info('CallSession:_lccTag - invalid tag data');
+      return;
+    }
+    this.logger.debug({customerData: tag}, 'CallSession:_lccTag set customer data in callInfo');
+    this.callInfo.customerData = tag;
+  }
+
   async _lccMuteStatus(callSid, mute) {
     // this whole thing requires us to be in a Dial or Conference verb
     const task = this.currentTask;
@@ -1042,6 +1137,38 @@ class CallSession extends Emitter {
     }
   }
 
+  /**
+   * perform live call control - send RFC 2833 DTMF
+   * @param {obj} opts
+   * @param {string} opts.dtmf.digit - DTMF digit
+   * @param {string} opts.dtmf.duration - Optional, Duration
+   */
+  async _lccDtmf(opts, callSid) {
+    const {dtmf} = opts;
+    const {digit, duration = 250} = dtmf;
+    if (!this.hasStableDialog) {
+      this.logger.info('CallSession:_lccDtmf - invalid command as we do not have a stable call');
+      return;
+    }
+    try {
+      const dlg = callSid === this.callSid ? this.dlg : this.currentTask.dlg;
+      const res = await dlg.request({
+        method: 'INFO',
+        headers: {
+          'Content-Type': 'application/dtmf',
+          'X-Reason': 'Dtmf'
+        },
+        body: `Signal=${digit}
+Duration=${duration} `
+      });
+      this.logger.debug({res}, `CallSession:_lccDtmf
+ got response to INFO DTMF digit=${digit} and duration=${duration}`);
+      return res;
+    } catch (err) {
+      this.logger.error({err}, 'CallSession:_lccDtmf - error sending INFO RFC 2833 DTMF');
+    }
+  }
+
   /**
    * perform live call control -- whisper to one party or the other on a call
    * @param {array} opts - array of play or say tasks
@@ -1096,7 +1223,7 @@ class CallSession extends Emitter {
   /**
    * perform live call control
    * @param {object} opts - update instructions
-   * @param {string} callSid - identifies call toupdate
+   * @param {string} callSid - identifies call to update
    */
   async updateCall(opts, callSid) {
     this.logger.debug(opts, 'CallSession:updateCall');
@@ -1110,6 +1237,9 @@ class CallSession extends Emitter {
     if (opts.listen_status) {
       await this._lccListenStatus(opts);
     }
+    if (opts.transcribe_status) {
+      await this._lccTranscribeStatus(opts);
+    }
     else if (opts.mute_status) {
       await this._lccMuteStatus(callSid, opts.mute_status === 'mute');
     }
@@ -1122,13 +1252,18 @@ class CallSession extends Emitter {
     else if (opts.sip_request) {
       const res = await this._lccSipRequest(opts, callSid);
       return {status: res.status, reason: res.reason};
+    } else if (opts.dtmf) {
+      await this._lccDtmf(opts, callSid);
     }
     else if (opts.record) {
       await this.notifyRecordOptions(opts.record);
     }
+    else if (opts.tag) {
+      return this._lccTag(opts);
+    }
 
     // whisper may be the only thing we are asked to do, or it may that
-    // we are doing a whisper after having muted, paused reccording etc..
+    // we are doing a whisper after having muted, paused recording etc..
     if (opts.whisper) {
       return this._lccWhisper(opts, callSid);
     }
@@ -1186,7 +1321,6 @@ class CallSession extends Emitter {
             get the full transcription.
           */
           delete t.bargeIn.enable;
-          this._bargeInEnabled = false;
           this.logger.info('CallSession:kill - found bargein disabled in the stack, clearing to that point');
           break;
         }
@@ -1197,6 +1331,35 @@ class CallSession extends Emitter {
     this.taskIdx = 0;
   }
 
+  _preCacheAudio(newTasks) {
+    for (const task of newTasks) {
+      if (task.name === TaskName.Config && task.hasSynthesizer) {
+        /* if they change synthesizer settings don't try to precache */
+        break;
+      }
+      if (task.name === TaskName.Say) {
+        /* identify vendor language, voice, and label */
+        const vendor = task.synthesizer.vendor && task.synthesizer.vendor !== 'default' ?
+          task.synthesizer.vendor :
+          this.speechSynthesisVendor;
+        const language = task.synthesizer.language && task.synthesizer.language !== 'default' ?
+          task.synthesizer.language :
+          this.speechSynthesisLanguage ;
+        const voice =  task.synthesizer.voice && task.synthesizer.voice !== 'default' ?
+          task.synthesizer.voice :
+          this.speechSynthesisVoice;
+        const label = task.synthesizer.label && task.synthesizer.label !== 'default' ?
+          task.synthesizer.label :
+          this.speechSynthesisLabel;
+
+        this.logger.info({vendor, language, voice, label},
+          'CallSession:_preCacheAudio - precaching audio for future prompt');
+        task._synthesizeWithSpecificVendor(this, this.ep, {vendor, language, voice, label, preCache: true})
+          .catch((err) => this.logger.error(err, 'CallSession:_preCacheAudio - error precaching audio'));
+      }
+    }
+  }
+
   /**
    * Append tasks to the current execution stack UNLESS there is a gather in the stack.
    * in that case, insert the tasks before the gather AND if the tasks include
@@ -1254,10 +1417,12 @@ class CallSession extends Emitter {
             this.replaceApplication(t);
           }
           else if (JAMBONES_INJECT_CONTENT) {
+            if (JAMBONES_EAGERLY_PRE_CACHE_AUDIO) this._preCacheAudio(t);
             this._injectTasks(t);
             this.logger.info({tasks: listTaskNames(this.tasks)}, 'CallSession:_onCommand - updated task list');
           }
           else {
+            if (JAMBONES_EAGERLY_PRE_CACHE_AUDIO) this._preCacheAudio(t);
             this.tasks.push(...t);
             this.logger.info({tasks: listTaskNames(this.tasks)}, 'CallSession:_onCommand - updated task list');
           }
@@ -1286,6 +1451,10 @@ class CallSession extends Emitter {
         this._lccListenStatus(data);
         break;
 
+      case 'transcribe:status':
+        this._lccTranscribeStatus(data);
+        break;
+
       case 'whisper':
         this._lccWhisper(data, call_sid);
         break;
@@ -1297,6 +1466,13 @@ class CallSession extends Emitter {
           });
         break;
 
+      case 'dtmf':
+        this._lccDtmf(data, call_sid)
+          .catch((err) => {
+            this.logger.info({err, data}, `CallSession:_onCommand - error sending RFC 2833 DTMF ${data}`);
+          });
+        break;
+
       default:
         this.logger.info(`CallSession:_onCommand - invalid command ${command}`);
     }
@@ -1378,6 +1554,8 @@ class CallSession extends Emitter {
       this.ep = ep;
       this.logger.debug(`allocated endpoint ${ep.uuid}`);
 
+      this._configMsEndpoint();
+
       this.ep.on('destroy', () => {
         this.logger.debug(`endpoint was destroyed!! ${this.ep.uuid}`);
       });
@@ -1448,6 +1626,7 @@ class CallSession extends Emitter {
       return;
     }
     this.ep = await this.ms.createEndpoint({remoteSdp: this.dlg.remote.sdp});
+    this._configMsEndpoint();
 
     await this.dlg.modify(this.ep.local.sdp);
     this.logger.debug('CallSession:replaceEndpoint completed');
@@ -1478,6 +1657,8 @@ class CallSession extends Emitter {
     this.notifier && this.notifier.close();
 
     this.rootSpan && this.rootSpan.end();
+    // close all background tasks
+    this.backgroundTaskManager.stopAll();
   }
 
   /**
@@ -1529,10 +1710,32 @@ class CallSession extends Emitter {
       }
       this.dlg.on('modify', this._onReinvite.bind(this));
       this.dlg.on('refer', this._onRefer.bind(this));
+      if (this.sipRequestWithinDialogHook) {
+        this.dlg.on('info', this._onRequestWithinDialog.bind(this));
+        this.dlg.on('message', this._onRequestWithinDialog.bind(this));
+      }
       this.logger.debug(`CallSession:propagateAnswer - answered callSid ${this.callSid}`);
     }
   }
 
+  async _onRequestWithinDialog(req, res) {
+    if (!this.sipRequestWithinDialogHook) {
+      return;
+    }
+    const sip_method = req.method;
+    if (sip_method === 'INFO') {
+      res.send(200);
+    } else if (sip_method === 'MESSAGE') {
+      res.send(202);
+    } else {
+      this.logger.info(`CallSession:_onRequestWithinDialog unsported method: ${req.method}`);
+      res.send(501);
+      return;
+    }
+    const params = {sip_method, sip_body: req.body, sip_headers: req.headers};
+    this.currentTask.performHook(this, this.sipRequestWithinDialogHook, params);
+  }
+
   async _onReinvite(req, res) {
     try {
       if (this.ep) {
@@ -1541,9 +1744,14 @@ class CallSession extends Emitter {
           res.send(200, {body: this.ep.local.sdp});
         }
         else {
-          const newSdp = await this.ep.modify(req.body);
+          if (this.currentTask.name === TaskName.Dial && this.currentTask.isOnHold) {
-          res.send(200, {body: newSdp});
+            this.logger.info('onholdMusic reINVITE after media has been released');
-          this.logger.info({offer: req.body, answer: newSdp}, 'handling reINVITE');
+            await this.currentTask.handleReinviteAfterMediaReleased(req, res);
+          } else {
+            const newSdp = await this.ep.modify(req.body);
+            res.send(200, {body: newSdp});
+            this.logger.info({offer: req.body, answer: newSdp}, 'handling reINVITE');
+          }
         }
       }
       else if (this.currentTask && this.currentTask.name === TaskName.Dial) {
@@ -1590,6 +1798,7 @@ class CallSession extends Emitter {
     }
     if (!this.ep) {
       this.ep = await this.ms.createEndpoint({remoteSdp: this.req.body});
+      this._configMsEndpoint();
     }
     return {ms: this.ms, ep: this.ep};
   }
@@ -1706,7 +1915,7 @@ class CallSession extends Emitter {
   wrapDialog(dlg) {
     dlg.connectTime = moment();
     const origDestroy = dlg.destroy.bind(dlg);
-    dlg.destroy = () => {
+    dlg.destroy = (opts) => {
       if (dlg.connected) {
         dlg.connected = false;
         dlg.destroy = origDestroy;
@@ -1715,7 +1924,7 @@ class CallSession extends Emitter {
         this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});
         this.logger.debug('CallSession: call terminated by jambonz');
         this.rootSpan.setAttributes({'call.termination': 'hangup by jambonz'});
-        origDestroy().catch((err) => this.logger.info({err}, 'CallSession - error destroying dialog'));
+        origDestroy(opts).catch((err) => this.logger.info({err}, 'CallSession - error destroying dialog'));
         if (this.wakeupResolver) {
           this.wakeupResolver({reason: 'session ended'});
           this.wakeupResolver = null;
@@ -1744,6 +1953,7 @@ class CallSession extends Emitter {
         'X-Reason': 'anchor-media'
       }
     });
+    this._configMsEndpoint();
   }
 
   async handleReinviteAfterMediaReleased(req, res) {
@@ -1765,11 +1975,14 @@ class CallSession extends Emitter {
   async _notifyCallStatusChange({callStatus, sipStatus, sipReason, duration}) {
     if (this.callMoved) return;
 
+    // manage record all call.
     if (callStatus === CallStatus.InProgress) {
-      // nice, call is in progress, good time to enable record
+      if (this.accountInfo.account.record_all_calls ||
-      await this.enableRecordAllCall();
+        this.application.record_all_calls) {
-    } else if (callStatus == CallStatus.Completed && this.isBackGroundListen) {
+        this.backgroundTaskManager.newTask('record');
-      await this.stopBackgroundListen();
+      }
+    } else if (callStatus == CallStatus.Completed) {
+      this.backgroundTaskManager.stop('record');
     }
 
     /* race condition: we hang up at the same time as the caller */
@@ -1806,20 +2019,9 @@ class CallSession extends Emitter {
     }
   }
 
-  async enableRecordAllCall() {
+  _configMsEndpoint() {
-    if (this.accountInfo.account.record_all_calls || this.application.record_all_calls) {
+    if (this.onHoldMusic) {
-      const listenOpts = {
+      this.ep.set({hold_music: `shout://${this.onHoldMusic.replace(/^https?:\/\//, '')}`});
-        url: `${JAMBONZ_RECORD_WS_BASE_URL}/record/${this.accountInfo.account.bucket_credential.vendor}`,
-        wsAuth: {
-          username: JAMBONZ_RECORD_WS_USERNAME,
-          password: JAMBONZ_RECORD_WS_PASSWORD
-        },
-        mixType : 'stereo',
-        passDtmf: true
-      };
-
-      this.logger.debug({listenOpts}, 'Record all calls: enabling listen');
-      await this.startBackgroundListen({verb: 'listen', ...listenOpts}, 'jambonz-session-record');
     }
   }
 
@@ -1859,6 +2061,18 @@ class CallSession extends Emitter {
       } catch (err) {}
     }
   }
+
+  /**
+   * startBackgroundTask - Start background task
+   */
+
+  async startBackgroundTask(type, opts) {
+    await this.backgroundTaskManager.newTask(type, opts);
+  }
+
+  stopBackgroundTask(type) {
+    this.backgroundTaskManager.stop(type);
+  }
 }
 
 module.exports = CallSession;

lib/session/rest-call-session.js

@@ -8,7 +8,7 @@ const moment = require('moment');
  * @extends CallSession
  */
 class RestCallSession extends CallSession {
-  constructor({logger, application, srf, req, ep, tasks, callInfo, accountInfo, rootSpan}) {
+  constructor({logger, application, srf, req, ep, ep2, tasks, callInfo, accountInfo, rootSpan}) {
     super({
       logger,
       application,
@@ -21,6 +21,7 @@ class RestCallSession extends CallSession {
     });
     this.req = req;
     this.ep = ep;
+    this.ep2 = ep2;
     // keep restDialTask reference for closing AMD
     if (tasks.length) {
       this.restDialTask = tasks[0];
@@ -49,7 +50,6 @@ class RestCallSession extends CallSession {
    */
   _callerHungup() {
     if (this.restDialTask) {
-      this.logger.info('RestCallSession: releasing AMD');
       this.restDialTask.turnOffAmd();
     }
     this.callInfo.callTerminationBy = 'caller';

lib/tasks/conference.js

@@ -48,7 +48,7 @@ class Conference extends Task {
     this.confName = this.data.name;
     [
       'beep', 'startConferenceOnEnter', 'endConferenceOnExit', 'joinMuted',
-      'maxParticipants', 'waitHook', 'statusHook', 'endHook', 'enterHook'
+      'maxParticipants', 'waitHook', 'statusHook', 'endHook', 'enterHook', 'endConferenceDuration'
     ].forEach((attr) => this[attr] = this.data[attr]);
     this.record = this.data.record || {};
     this.statusEvents = [];
@@ -344,9 +344,13 @@ class Conference extends Task {
     }
 
     const opts = {};
-    if (this.endConferenceOnExit) Object.assign(opts, {flags: {endconf: true}});
+    if (this.endConferenceOnExit || this.startConferenceOnEnter || this.joinMuted) {
-    if (this.startConferenceOnEnter) Object.assign(opts, {flags: {moderator: true}});
+      Object.assign(opts, {flags: {
-    if (this.joinMuted) Object.assign(opts, {flags: {mute: true}});
+        ...(this.endConferenceOnExit && {endconf: true}),
+        ...(this.startConferenceOnEnter && {moderator: true}),
+        ...(this.joinMuted && {joinMuted: true}),
+      }});
+    }
 
     try {
       const {memberId, confUuid} = await this.ep.join(this.confName, opts);
@@ -389,6 +393,11 @@ class Conference extends Task {
       this.ep.api('conference', `${this.confName} set max_members ${this.maxParticipants}`)
         .catch((err) => this.logger.error(err, `Error setting max participants to ${this.maxParticipants}`));
     }
+
+    if (typeof this.endConferenceDuration === 'number' && this.endConferenceDuration >= 0) {
+      this.ep.api('conference', `${this.confName} set endconference_grace_time ${this.endConferenceDuration}`)
+        .catch((err) => this.logger.error(err, `Error setting end conference time to ${this.endConferenceDuration}`));
+    }
   }
 
   /**

lib/tasks/config.js

@@ -9,7 +9,8 @@ class TaskConfig extends Task {
       'recognizer',
       'bargeIn',
       'record',
-      'listen'
+      'listen',
+      'transcribe'
     ].forEach((k) => this[k] = this.data[k] || {});
 
     if ('notifyEvents' in this.data) {
@@ -30,6 +31,13 @@ class TaskConfig extends Task {
         if (this.bargeIn[k]) this.gatherOpts[k] = this.bargeIn[k];
       });
     }
+    if (this.transcribe?.enable) {
+      this.transcribeOpts = {
+        verb: 'transcribe',
+        ...this.transcribe
+      };
+      delete this.transcribeOpts.enable;
+    }
 
     if (this.data.reset) {
       if (typeof this.data.reset === 'string') this.data.reset = [this.data.reset];
@@ -37,9 +45,15 @@ class TaskConfig extends Task {
     else this.data.reset = [];
 
     if (this.bargeIn.sticky) this.autoEnable = true;
-    this.preconditions = (this.bargeIn.enable || this.record?.action || this.listen?.url || this.data.amd) ?
+    this.preconditions = (this.bargeIn.enable ||
+      this.record?.action ||
+      this.listen?.url ||
+      this.data.amd ||
+      this.transcribe?.enable) ?
       TaskPreconditions.Endpoint :
       TaskPreconditions.None;
+
+    this.onHoldMusic = this.data.onHoldMusic;
   }
 
   get name() { return TaskName.Config; }
@@ -48,6 +62,7 @@ class TaskConfig extends Task {
   get hasRecognizer() { return Object.keys(this.recognizer).length; }
   get hasRecording() { return Object.keys(this.record).length; }
   get hasListen() { return Object.keys(this.listen).length; }
+  get hasTranscribe() { return Object.keys(this.transcribe).length; }
 
   get summary() {
     const phrase = [];
@@ -70,8 +85,12 @@ class TaskConfig extends Task {
     if (this.hasListen) {
       phrase.push(this.listen.enable ? `listen ${this.listen.url}` : 'stop listen');
     }
+    if (this.hasTranscribe) {
+      phrase.push(this.transcribe.enable ? `transcribe ${this.transcribe.transcriptionHook}` : 'stop transcribe');
+    }
     if (this.data.amd) phrase.push('enable amd');
     if (this.notifyEvents) phrase.push(`event notification ${this.notifyEvents ? 'on' : 'off'}`);
+    if (this.onHoldMusic) phrase.push(`onHoldMusic: ${this.onHoldMusic}`);
     return `${this.name}{${phrase.join(',')}}`;
   }
 
@@ -83,6 +102,10 @@ class TaskConfig extends Task {
       cs.notifyEvents = !!this.data.notifyEvents;
     }
 
+    if (this.onHoldMusic) {
+      cs.onHoldMusic = this.onHoldMusic;
+    }
+
     if (this.data.amd) {
       this.startAmd = cs.startAmd;
       this.stopAmd = cs.stopAmd;
@@ -102,24 +125,58 @@ class TaskConfig extends Task {
     });
 
     if (this.hasSynthesizer) {
+      cs.synthesizer = this.synthesizer;
       cs.speechSynthesisVendor = this.synthesizer.vendor !== 'default'
         ? this.synthesizer.vendor
         : cs.speechSynthesisVendor;
+      cs.speechSynthesisLabel = this.synthesizer.label !== 'default'
+        ? this.synthesizer.label
+        : cs.speechSynthesisLabel;
       cs.speechSynthesisLanguage = this.synthesizer.language !== 'default'
         ?  this.synthesizer.language
         : cs.speechSynthesisLanguage;
       cs.speechSynthesisVoice = this.synthesizer.voice !== 'default'
         ? this.synthesizer.voice
         : cs.speechSynthesisVoice;
+
+      // fallback vendor
+      cs.fallbackSpeechSynthesisVendor = this.synthesizer.fallbackVendor !== 'default'
+        ? this.synthesizer.fallbackVendor
+        : cs.fallbackSpeechSynthesisVendor;
+      cs.fallbackSpeechSynthesisLabel = this.synthesizer.fallbackLabel !== 'default'
+        ? this.synthesizer.fallbackLabel
+        : cs.fallbackSpeechSynthesisLabel;
+      cs.fallbackSpeechSynthesisLanguage = this.synthesizer.fallbackLanguage !== 'default'
+        ?  this.synthesizer.fallbackLanguage
+        : cs.fallbackSpeechSynthesisLanguage;
+      cs.fallbackSpeechSynthesisVoice = this.synthesizer.fallbackVoice !== 'default'
+        ? this.synthesizer.fallbackVoice
+        : cs.fallbackSpeechSynthesisVoice;
       this.logger.info({synthesizer: this.synthesizer}, 'Config: updated synthesizer');
     }
     if (this.hasRecognizer) {
+      cs.recognizer = this.recognizer;
       cs.speechRecognizerVendor = this.recognizer.vendor !== 'default'
         ? this.recognizer.vendor
         : cs.speechRecognizerVendor;
+      cs.speechRecognizerLabel = this.recognizer.label !== 'default'
+        ? this.recognizer.label
+        : cs.speechRecognizerLabel;
       cs.speechRecognizerLanguage = this.recognizer.language !== 'default'
         ? this.recognizer.language
         : cs.speechRecognizerLanguage;
+
+      //fallback
+      cs.fallbackSpeechRecognizerVendor = this.recognizer.fallbackVendor !== 'default'
+        ? this.recognizer.fallbackVendor
+        : cs.fallbackSpeechRecognizerVendor;
+      cs.fallbackSpeechRecognizerLabel = this.recognizer.fallbackLabel !== 'default'
+        ? this.recognizer.fallbackLabel
+        : cs.fallbackSpeechRecognizerLabel;
+      cs.fallbackSpeechRecognizerLanguage = this.recognizer.fallbackLanguage !== 'default'
+        ? this.recognizer.fallbackLanguage
+        : cs.fallbackSpeechRecognizerLanguage;
+
       cs.isContinuousAsr = typeof this.recognizer.asrTimeout === 'number' ? true : false;
       if (cs.isContinuousAsr) {
         cs.asrTimeout = this.recognizer.asrTimeout;
@@ -171,12 +228,30 @@ class TaskConfig extends Task {
       const {enable, ...opts} = this.listen;
       if (enable) {
         this.logger.debug({opts}, 'Config: enabling listen');
-        cs.startBackgroundListen({verb: 'listen', ...opts});
+        cs.startBackgroundTask('listen', {verb: 'listen', ...opts});
       } else {
         this.logger.info('Config: disabling listen');
-        cs.stopBackgroundListen();
+        cs.stopBackgroundTask('listen');
       }
     }
+    if (this.hasTranscribe) {
+      if (this.transcribe.enable) {
+        this.transcribeOpts.recognizer = this.hasRecognizer ?
+          this.recognizer :
+          {
+            vendor: cs.speechRecognizerVendor,
+            language: cs.speechRecognizerLanguage
+          };
+        this.logger.debug(this.transcribeOpts, 'Config: enabling transcribe');
+        cs.startBackgroundTask('transcribe', this.transcribeOpts);
+      } else {
+        this.logger.info('Config: disabling transcribe');
+        cs.stopBackgroundTask('transcribe');
+      }
+    }
+    if (this.data.sipRequestWithinDialogHook) {
+      cs.sipRequestWithinDialogHook = this.data.sipRequestWithinDialogHook;
+    }
   }
 
   async kill(cs) {

lib/tasks/dial.js

@@ -12,10 +12,13 @@ const assert = require('assert');
 const placeCall = require('../utils/place-outdial');
 const sessionTracker = require('../session/session-tracker');
 const DtmfCollector = require('../utils/dtmf-collector');
+const ConfirmCallSession = require('../session/confirm-call-session');
 const dbUtils = require('../utils/db-utils');
 const debug = require('debug')('jambonz:feature-server');
 const {parseUri} = require('drachtio-srf');
-const {ANCHOR_MEDIA_ALWAYS} = require('../config');
+const {ANCHOR_MEDIA_ALWAYS, JAMBONZ_DISABLE_DIAL_PAI_HEADER} = require('../config');
+const { isOnhold, isOpusFirst } = require('../utils/sdp-utils');
+const { normalizeJambones } = require('@jambonz/verb-specifications');
 
 function parseDtmfOptions(logger, dtmfCapture) {
   let parentDtmfCollector, childDtmfCollector;
@@ -135,6 +138,10 @@ class TaskDial extends Task {
 
   get name() { return TaskName.Dial; }
 
+  get isOnHold() {
+    return this.isIncomingLegHold || this.isOutgoingLegHold;
+  }
+
   get canReleaseMedia() {
     const keepAnchor = this.data.anchorMedia ||
     this.cs.isBackGroundListen ||
@@ -197,6 +204,7 @@ class TaskDial extends Task {
       await this.performAction(this.results, this.killReason !== KillReason.Replaced);
       this._removeDtmfDetection(cs.dlg);
       this._removeDtmfDetection(this.dlg);
+      this._removeSipIndialogRequestListener(this.dlg);
     } catch (err) {
       this.logger.error({err}, 'TaskDial:exec terminating with error');
       this.kill(cs);
@@ -225,7 +233,7 @@ class TaskDial extends Task {
     }
     this._removeDtmfDetection(cs.dlg);
     this._removeDtmfDetection(this.dlg);
-    this._killOutdials();
+    await this._killOutdials();
     if (this.sd) {
       this.sd.kill();
       this.sd.removeAllListeners();
@@ -314,18 +322,41 @@ class TaskDial extends Task {
         const to = parseUri(req.getParsedHeader('Refer-To').uri);
         const by = parseUri(req.getParsedHeader('Referred-By').uri);
         this.logger.info({to}, 'refer to parsed');
-        await cs.requestor.request('verb:hook', this.referHook, {
+        const json = await cs.requestor.request('verb:hook', this.referHook, {
-          ...callInfo,
+          ...(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.user,
+            refer_to_user: to.scheme === 'tel' ? to.number : to.user,
-            referred_by_user: by.user,
+            referred_by_user: by.scheme === 'tel' ? by.number : by.user,
             referring_call_sid,
             referred_call_sid
           }
         }, httpHeaders);
+        if (json && Array.isArray(json)) {
+          try {
+            const logger = isChild ? this.logger : this.sd.logger;
+            const tasks = normalizeJambones(logger, json).map((tdata) => makeTask(this.logger, tdata));
+            if (tasks && tasks.length > 0) {
+              const legs = isChild ? ['child', 'parent'] : ['parent', 'child'];
+              logger.info(`Dial:handleRefer received REFER on ${legs[0]} leg, setting new app on ${legs[1]} leg`);
+              if (isChild) this.redirect(cs, tasks);
+              else {
+                logger.info({tasks: json}, 'Dial:handleRefer - new application for for child leg');
+                const adultingSession = await this.sd.doAdulting({
+                  logger,
+                  application: cs.application,
+                  tasks
+                });
+                /* need to update the callSid of the child with its own (new) AdultingCallSession */
+                sessionTracker.add(adultingSession.callSid, adultingSession);
+              }
+            }
+          } catch (err) {
+            this.logger.info(err, 'Dial:handleRefer - error setting new application after receiving REFER');
+          }
+        }
         res.send(202);
         this.logger.info('DialTask:handleRefer - sent 202 Accepted');
       } catch (err) {
@@ -346,11 +377,16 @@ class TaskDial extends Task {
     sd.removeAllListeners('callCreateFail');
   }
 
-  _killOutdials() {
+  async _killOutdials() {
     for (const [callSid, sd] of Array.from(this.dials)) {
       this.logger.debug(`Dial:_killOutdials killing callSid ${callSid}`);
-      sd.kill().catch((err) => this.logger.info(err, `Dial:_killOutdials Error killing ${callSid}`));
+      try {
+        await sd.kill();
+      } catch (err) {
+        this.logger.info(err, `Dial:_killOutdials Error killing ${callSid}`);
+      }
       this._removeHandlers(sd);
+      this.logger.debug(`Dial:_killOutdials killed callSid ${callSid}`);
     }
     this.dials.clear();
   }
@@ -363,8 +399,14 @@ class TaskDial extends Task {
   }
 
   _onInfo(cs, dlg, req, res) {
+    // SIP Indialog will be handled by another handler
+    if (cs.sipRequestWithinDialogHook) {
+      return;
+    }
     res.send(200);
-    if (req.get('Content-Type') !== 'application/dtmf-relay') return;
+    if (req.get('Content-Type') !== 'application/dtmf-relay') {
+      return;
+    }
 
     const dtmfDetector = dlg === cs.dlg ? this.parentDtmfCollector : this.childDtmfCollector;
     if (!dtmfDetector) return;
@@ -393,6 +435,20 @@ class TaskDial 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');
+  }
+
+  async _onRequestWithinDialog(cs, req, res) {
+    cs._onRequestWithinDialog(req, res);
+  }
+
   async _initializeInbound(cs) {
     const {ep} = await cs._evalEndpointPrecondition(this);
     this.epOther = ep;
@@ -421,7 +477,8 @@ 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') && {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
+      ...(req && req.has('P-Asserted-Identity') && !JAMBONZ_DISABLE_DIAL_PAI_HEADER &&
+        {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
       ...(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
@@ -431,7 +488,8 @@ class TaskDial extends Task {
       headers: this.headers,
       proxy: `sip:${sbcAddress}`,
       callingNumber: this.callerId || req.callingNumber,
-      ...(this.callerName && {callingName: this.callerName})
+      ...(this.callerName && {callingName: this.callerName}),
+      opusFirst: isOpusFirst(this.cs.ep.remote.sdp)
     };
 
     const t = this.target.find((t) => t.type === 'teams');
@@ -442,10 +500,14 @@ class TaskDial extends Task {
     }
 
     const ms = await cs.getMS();
-    this.timerRing = setTimeout(() => {
+    this.timerRing = setTimeout(async() => {
       this.logger.info(`Dial:_attemptCall: ring no answer timer ${this.timeout}s exceeded`);
       this.timerRing = null;
-      this._killOutdials();
+      try {
+        await this._killOutdials();
+      } catch (err) {
+        this.logger.info(err, 'Dial:_attemptCall - error killing outdials');
+      }
       this.result = {
         dialCallStatus: CallStatus.NoAnswer,
         dialSipStatus: 487
@@ -507,7 +569,8 @@ class TaskDial extends Task {
           callInfo: cs.callInfo,
           accountInfo: cs.accountInfo,
           rootSpan: cs.rootSpan,
-          startSpan: this.startSpan.bind(this)
+          startSpan: this.startSpan.bind(this),
+          dialTask: this
         });
         this.dials.set(sd.callSid, sd);
 
@@ -523,7 +586,8 @@ class TaskDial extends Task {
             }
           })
           .on('callStatusChange', (obj) => {
-            if (this.results.dialCallStatus !== CallStatus.Completed) {
+            if (this.results.dialCallStatus !== CallStatus.Completed &&
+              this.results.dialCallStatus !== CallStatus.NoAnswer) {
               Object.assign(this.results, {
                 dialCallStatus: obj.callStatus,
                 dialSipStatus: obj.sipStatus,
@@ -576,11 +640,7 @@ class TaskDial extends Task {
             }
           })
           .on('reinvite', (req, res) => {
-            try {
+            this._onReinvite(req, res);
-              cs.handleReinviteAfterMediaReleased(req, res);
-            } catch (err) {
-              this.logger.error(err, 'Error in dial einvite from B leg');
-            }
           })
           .on('refer', (callInfo, req, res) => {
 
@@ -616,6 +676,35 @@ class TaskDial extends Task {
     this._killOutdials();       // NB: order is important
   }
 
+  async _onReinvite(req, res) {
+    try {
+      let isHandled = false;
+      if (this.cs.onHoldMusic) {
+        if (isOnhold(req.body) && !this.epOther && !this.ep) {
+          await this.cs.handleReinviteAfterMediaReleased(req, res);
+          // Onhold but media is already released
+          // reconnect A Leg and Response B leg
+          await this.reAnchorMedia(this.cs, this.sd);
+          this.isOutgoingLegHold = true;
+          isHandled = true;
+          this._onHoldHook();
+        } else if (!isOnhold(req.body) && this.epOther && this.ep && this.isOutgoingLegHold && this.canReleaseMedia) {
+          // 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);
+          isHandled = true;
+          this.isOutgoingLegHold = false;
+        }
+      }
+      if (!isHandled) {
+        this.cs.handleReinviteAfterMediaReleased(req, res);
+      }
+    } catch (err) {
+      this.logger.error(err, 'Error in dial einvite from B leg');
+    }
+  }
+
   _onMaxCallDuration(cs) {
     this.logger.info(`Dial:_onMaxCallDuration tearing down call as it has reached ${this.timeLimit}s`);
     this.ep && this.ep.unbridge();
@@ -668,6 +757,7 @@ 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 (this.transcribeTask) this.transcribeTask.exec(cs, {ep: this.epOther, ep2:this.ep});
     if (this.listenTask) this.listenTask.exec(cs, {ep: this.epOther});
@@ -701,9 +791,11 @@ class TaskDial extends Task {
     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);
       const bLegSdp = sd.dlg.remote.sdp;
-      await Promise.all[sd.releaseMediaToSBC(aLegSdp, cs.ep.local.sdp), cs.releaseMediaToSBC(bLegSdp)];
+      await cs.releaseMediaToSBC(bLegSdp);
       this.epOther = null;
       this.logger.info('Dial:_releaseMedia - successfully released media from freewitch');
     } catch (err) {
@@ -720,9 +812,29 @@ class TaskDial extends Task {
   }
 
   async handleReinviteAfterMediaReleased(req, res) {
-    const sdp = await this.dlg.modify(req.body);
+    let isHandled = false;
-    this.logger.info({sdp}, 'Dial:handleReinviteAfterMediaReleased - sent reinvite to B leg');
+    if (isOnhold(req.body) && !this.epOther && !this.ep) {
-    res.send(200, {body: sdp});
+      const sdp = await this.dlg.modify(req.body);
+      res.send(200, {body: sdp});
+      // Onhold but media is already released
+      await this.reAnchorMedia(this.cs, this.sd);
+      isHandled = true;
+      this.isIncomingLegHold = true;
+      this._onHoldHook();
+    } else if (!isOnhold(req.body) && this.epOther && this.ep && this.isIncomingLegHold && this.canReleaseMedia) {
+      // 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);
+      isHandled = true;
+      this.isIncomingLegHold = false;
+    }
+
+    if (!isHandled) {
+      const sdp = await this.dlg.modify(req.body);
+      this.logger.info({sdp}, 'Dial:handleReinviteAfterMediaReleased - sent reinvite to B leg');
+      res.send(200, {body: sdp});
+    }
   }
 
   _onAmdEvent(cs, evt) {
@@ -733,6 +845,48 @@ class TaskDial extends Task {
         this.logger.error({err}, 'Dial:_onAmdEvent - error calling actionHook');
       });
   }
+
+  async _onHoldHook(allowed = [TaskName.Play, TaskName.Say, TaskName.Pause]) {
+    if (this.data.onHoldHook) {
+      // send silence for keep Voice quality
+      await this.epOther.play('silence_stream://500');
+      let allowedTasks;
+      do {
+        try {
+          const b3 = this.getTracingPropagation();
+          const httpHeaders = b3 && {b3};
+          const json = await this.cs.application.requestor.
+            request('verb:hook', this.data.onHoldHook, this.cs.callInfo.toJSON(), httpHeaders);
+          const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
+          allowedTasks = tasks.filter((t) => allowed.includes(t.name));
+          if (tasks.length !== allowedTasks.length) {
+            this.logger.debug({tasks, allowedTasks}, 'unsupported task');
+            throw new Error(`unsupported verb in enqueue waitHook: only ${JSON.stringify(allowed)}`);
+          }
+          this.logger.debug(`DialTask:_onHoldHook: executing ${tasks.length} tasks`);
+          if (tasks.length) {
+            this._playSession = new ConfirmCallSession({
+              logger: this.logger,
+              application: this.cs.application,
+              dlg: this.isIncomingLegHold ? this.dlg : this.cs.dlg,
+              ep: this.isIncomingLegHold ? this.ep : this.cs.ep,
+              callInfo: this.cs.callInfo,
+              accountInfo: this.cs.accountInfo,
+              tasks,
+              rootSpan: this.cs.rootSpan
+            });
+            await this._playSession.exec();
+            this._playSession = null;
+          }
+        } catch (error) {
+          this.logger.info(error, 'DialTask:_onHoldHook: failed retrieving waitHook');
+          this._playSession = null;
+          break;
+        }
+      } while (allowedTasks && allowedTasks.length > 0 && !this.killed && this.isOnHold);
+      this.logger.info('Finish onHoldHook');
+    }
+  }
 }
 
 module.exports = TaskDial;

lib/tasks/dialogflow/index.js

@@ -58,6 +58,13 @@ class Dialogflow extends Task {
       this.vendor = this.data.tts.vendor || 'default';
       this.language = this.data.tts.language || 'default';
       this.voice = this.data.tts.voice || 'default';
+      this.speechSynthesisLabel = this.data.tts.label;
+
+      // fallback tts
+      this.fallbackVendor = this.data.tts.fallbackVendor || 'default';
+      this.fallbackLanguage = this.data.tts.fallbackLanguage || 'default';
+      this.fallbackVoice = this.data.tts.fallbackLanguage || 'default';
+      this.fallbackLabel = this.data.tts.fallbackLabel;
     }
     this.bargein = this.data.bargein;
   }
@@ -118,8 +125,15 @@ class Dialogflow extends Task {
         this.vendor = cs.speechSynthesisVendor;
         this.language = cs.speechSynthesisLanguage;
         this.voice = cs.speechSynthesisVoice;
+        this.speechSynthesisLabel = cs.speechSynthesisLabel;
       }
-      this.ttsCredentials = cs.getSpeechCredentials(this.vendor, 'tts');
+      if (this.fallbackVendor === 'default') {
+        this.fallbackVendor = cs.fallbackSpeechSynthesisVendor;
+        this.fallbackLanguage = cs.fallbackSpeechSynthesisLanguage;
+        this.fallbackVoice = cs.fallbackSpeechSynthesisVoice;
+        this.fallbackLabel = cs.fallbackSpeechSynthesisLabel;
+      }
+      this.ttsCredentials = cs.getSpeechCredentials(this.vendor, 'tts', this.speechSynthesisLabel);
 
       this.ep.addCustomEventListener('dialogflow::intent', this._onIntent.bind(this, ep, cs));
       this.ep.addCustomEventListener('dialogflow::transcription', this._onTranscription.bind(this, ep, cs));
@@ -221,19 +235,8 @@ class Dialogflow extends Task {
       }
 
       try {
-        const obj = {
+        const {filePath} = await this._fallbackSynthAudio(cs, intent, stats, synthAudio);
-          account_sid: cs.accountSid,
-          text: intent.fulfillmentText,
-          vendor: this.vendor,
-          language: this.language,
-          voice: this.voice,
-          salt: cs.callSid,
-          credentials: this.ttsCredentials
-        };
-        this.logger.debug({obj}, 'Dialogflow:_onIntent - playing message via tts');
-        const {filePath, servedFromCache} = await synthAudio(stats, obj);
         if (filePath) cs.trackTmpFile(filePath);
-        if (!this.ttsCredentials && !servedFromCache) cs.billForTts(intent.fulfillmentText.length);
 
         if (this.playInProgress) {
           await ep.api('uuid_break', ep.uuid).catch((err) => this.logger.info(err, 'Error killing audio'));
@@ -277,6 +280,46 @@ class Dialogflow extends Task {
     }
   }
 
+  async _fallbackSynthAudio(cs, intent, stats, synthAudio) {
+    try {
+      const obj = {
+        account_sid: cs.accountSid,
+        text: intent.fulfillmentText,
+        vendor: this.vendor,
+        language: this.language,
+        voice: this.voice,
+        salt: cs.callSid,
+        credentials: this.ttsCredentials
+      };
+      this.logger.debug({obj}, 'Dialogflow:_onIntent - playing message via tts');
+
+      return await synthAudio(stats, obj);
+    } catch (error) {
+      this.logger.info({error}, 'Failed to synthesize audio from primary vendor');
+
+      try {
+        if (this.fallbackVendor) {
+          const credentials = cs.getSpeechCredentials(this.fallbackVendor, 'tts', this.fallbackLabel);
+          const obj = {
+            account_sid: cs.accountSid,
+            text: intent.fulfillmentText,
+            vendor: this.fallbackVendor,
+            language: this.fallbackLanguage,
+            voice: this.fallbackVoice,
+            salt: cs.callSid,
+            credentials
+          };
+          this.logger.debug({obj}, 'Dialogflow:_onIntent - playing message via fallback tts');
+          return await synthAudio(stats, obj);
+        }
+      } catch (err) {
+        this.logger.info({err}, 'Failed to synthesize audio from falllback vendor');
+        throw err;
+      }
+      throw error;
+    }
+  }
+
   /**
    * A transcription - either interim or final - has been returned.
    * If we are doing barge-in based on hotword detection, check for the hotword or phrase.

lib/tasks/enqueue.js

@@ -311,7 +311,8 @@ class TaskEnqueue extends Task {
     }
   }
 
-  async _playHook(cs, dlg, hook, allowed = [TaskName.Play, TaskName.Say, TaskName.Pause, TaskName.Leave]) {
+  async _playHook(cs, dlg, hook,
+    allowed = [TaskName.Play, TaskName.Say, TaskName.Pause, TaskName.Leave, TaskName.Tag]) {
     const {sortedSetLength, sortedSetPositionByPattern} = cs.srf.locals.dbHelpers;
     const b3 = this.getTracingPropagation();
     const httpHeaders = b3 && {b3};
@@ -331,6 +332,7 @@ class TaskEnqueue extends Task {
         queuePosition: queuePosition.length ? queuePosition[0] : 0,
         callSid: this.cs.callSid,
         callId: this.cs.callId,
+        customerData: this.cs.callInfo.customerData
       });
     } catch (err) {
       this.logger.error({err}, `TaskEnqueue:_playHook error retrieving list info for queue ${this.queueName}`);

lib/tasks/gather.js

@@ -1,17 +1,17 @@
-const Task = require('./task');
 const {
   TaskName,
-  TaskPreconditions,
   GoogleTranscriptionEvents,
   NuanceTranscriptionEvents,
   AwsTranscriptionEvents,
   AzureTranscriptionEvents,
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
+  CobaltTranscriptionEvents,
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
-  JambonzTranscriptionEvents
+  JambonzTranscriptionEvents,
-} = require('../utils/constants');
+  AssemblyAiTranscriptionEvents
+} = require('../utils/constants.json');
 const {
   JAMBONES_GATHER_EARLY_HINTS_MATCH,
   JAMBONZ_GATHER_EARLY_HINTS_MATCH,
@@ -19,34 +19,11 @@ const {
 } = require('../config');
 const makeTask = require('./make_task');
 const assert = require('assert');
+const SttTask = require('./stt-task');
 
-const compileTranscripts = (logger, evt, arr) => {
+class TaskGather extends SttTask {
-  if (!Array.isArray(arr) || arr.length === 0) return;
-  let t = '';
-  for (const a of arr) {
-    t += ` ${a.alternatives[0].transcript}`;
-  }
-  t += ` ${evt.alternatives[0].transcript}`;
-  evt.alternatives[0].transcript = t.trim();
-};
-
-class TaskGather extends Task {
   constructor(logger, opts, parentTask) {
-    super(logger, opts);
+    super(logger, opts, parentTask);
-    this.preconditions = TaskPreconditions.Endpoint;
-
-    const {
-      setChannelVarsForStt,
-      normalizeTranscription,
-      removeSpeechListeners,
-      setSpeechCredentialsAtRuntime,
-      compileSonioxTranscripts
-    } = require('../utils/transcription-utils')(logger);
-    this.setChannelVarsForStt = setChannelVarsForStt;
-    this.normalizeTranscription = normalizeTranscription;
-    this.removeSpeechListeners = removeSpeechListeners;
-    this.compileSonioxTranscripts = compileSonioxTranscripts;
-
     [
       'finishOnKey', 'input', 'numDigits', 'minDigits', 'maxDigits',
       'interDigitTimeout', 'partialResultHook', 'bargein', 'dtmfBargein',
@@ -62,27 +39,23 @@ class TaskGather extends Task {
     this.listenDuringPrompt = this.data.listenDuringPrompt === false ? false : true;
     this.minBargeinWordCount = this.data.minBargeinWordCount || 1;
     if (this.data.recognizer) {
-      const recognizer = this.data.recognizer;
-      this.vendor = recognizer.vendor;
-      this.language = recognizer.language;
-
-      /* let credentials be supplied in the recognizer object at runtime */
-      this.sttCredentials = setSpeechCredentialsAtRuntime(recognizer);
-
       /* continuous ASR (i.e. compile transcripts until a special timeout or dtmf key) */
-      this.asrTimeout = typeof recognizer.asrTimeout === 'number' ? recognizer.asrTimeout * 1000 : 0;
+      this.asrTimeout = typeof this.data.recognizer.asrTimeout === 'number' ?
-      if (this.asrTimeout > 0) this.asrDtmfTerminationDigit = recognizer.asrDtmfTerminationDigit;
+        this.data.recognizer.asrTimeout * 1000 : 0;
-      this.isContinuousAsr = this.asrTimeout > 0;
+      if (this.asrTimeout > 0) {
+        this.isContinuousAsr = true;
+        this.asrDtmfTerminationDigit = this.data.recognizer.asrDtmfTerminationDigit;
+      }
 
       if (Array.isArray(this.data.recognizer.hints) &&
         0 == this.data.recognizer.hints.length && JAMBONES_GATHER_CLEAR_GLOBAL_HINTS_ON_EMPTY_HINTS) {
         logger.debug('Gather: an empty hints array was supplied, so we will mask global hints');
         this.maskGlobalSttHints = true;
       }
-      this.data.recognizer.hints = this.data.recognizer.hints || [];
+      // fast Recognition, fire event after a specified time after the last hypothesis.
-      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages || [];
+      this.fastRecognitionTimeout =  typeof this.data.recognizer.fastRecognitionTimeout === 'number' ?
+        this.data.recognizer.fastRecognitionTimeout * 1000 : 0;
     }
-    else this.data.recognizer = {hints: [], altLanguages: []};
 
     this.digitBuffer = '';
     this._earlyMedia = this.data.earlyMedia === true;
@@ -97,12 +70,8 @@ class TaskGather extends Task {
 
     /* buffer speech for continuous asr */
     this._bufferedTranscripts = [];
-
-    /* buffer for soniox transcripts */
-    this._sonioxTranscripts = [];
-
-    this.parentTask = parentTask;
     this.partialTranscriptsCount = 0;
+    this.bugname_prefix = 'gather_';
   }
 
   get name() { return TaskName.Gather; }
@@ -135,13 +104,12 @@ class TaskGather extends Task {
 
   async exec(cs, {ep}) {
     this.logger.debug({options: this.data}, 'Gather:exec');
-    await super.exec(cs);
+    await super.exec(cs, {ep});
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
-    const {getNuanceAccessToken, getIbmAccessToken} = cs.srf.locals.dbHelpers;
 
     if (cs.hasGlobalSttHints && !this.maskGlobalSttHints) {
       const {hints, hintsBoost} = cs.globalSttHints;
-      const setOfHints = new Set(this.data.recognizer.hints
+      const setOfHints = new Set((this.data.recognizer.hints || [])
         .concat(hints)
         .filter((h) => typeof h === 'string' && h.length > 0));
       this.data.recognizer.hints = [...setOfHints];
@@ -149,14 +117,6 @@ class TaskGather extends Task {
       this.logger.debug({hints: this.data.recognizer.hints, hintsBoost: this.data.recognizer.hintsBoost},
         'Gather:exec - applying global sttHints');
     }
-    if (cs.hasAltLanguages) {
-      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages.concat(cs.altLanguages);
-      this.logger.debug({altLanguages: this.altLanguages},
-        'Gather:exec - applying altLanguages');
-    }
-    if (cs.hasGlobalSttPunctuation && !this.data.recognizer.punctuation) {
-      this.data.recognizer.punctuation = cs.globalSttPunctuation;
-    }
     if (!this.isContinuousAsr && cs.isContinuousAsr) {
       this.isContinuousAsr = true;
       this.asrTimeout = cs.asrTimeout * 1000;
@@ -174,67 +134,26 @@ class TaskGather extends Task {
       this.interim = true;
       this.logger.debug('Gather:exec - early hints match enabled');
     }
-
+    const startListening = async(cs, ep) => {
-    this.ep = ep;
-    if ('default' === this.vendor || !this.vendor) {
-      this.vendor = cs.speechRecognizerVendor;
-      if (this.data.recognizer) this.data.recognizer.vendor = this.vendor;
-    }
-    if ('default' === this.language || !this.language) {
-      this.language = cs.speechRecognizerLanguage;
-      if (this.data.recognizer) this.data.recognizer.language = this.language;
-    }
-    if (!this.data.recognizer.vendor) {
-      this.data.recognizer.vendor = this.vendor;
-    }
-    if (this.needsStt && !this.sttCredentials) this.sttCredentials = cs.getSpeechCredentials(this.vendor, 'stt');
-    if (this.needsStt && !this.sttCredentials) {
-      const {writeAlerts, AlertType} = cs.srf.locals;
-      this.logger.info(`TaskGather:exec - ERROR stt using ${this.vendor} requested but creds not supplied`);
-      writeAlerts({
-        account_sid: cs.accountSid,
-        alert_type: AlertType.STT_NOT_PROVISIONED,
-        vendor: this.vendor
-      }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
-      // Notify application that STT vender is wrong.
-      this.notifyError({
-        msg: 'ASR error',
-        details: `No speech-to-text service credentials for ${this.vendor} have been configured`
-      });
-      this.notifyTaskDone();
-      throw new Error(`No speech-to-text service credentials for ${this.vendor} have been configured`);
-    }
-
-    if (this.vendor === 'nuance' && this.sttCredentials.client_id) {
-      /* get nuance access token */
-      const {client_id, secret} = this.sttCredentials;
-      const {access_token, servedFromCache} = await getNuanceAccessToken(client_id, secret, 'asr tts');
-      this.logger.debug({client_id}, `Gather:exec - got nuance access token ${servedFromCache ? 'from cache' : ''}`);
-      this.sttCredentials = {...this.sttCredentials, access_token};
-    }
-    else if (this.vendor == 'ibm' && this.sttCredentials.stt_api_key) {
-      /* get ibm access token */
-      const {stt_api_key, stt_region} = this.sttCredentials;
-      const {access_token, servedFromCache} = await getIbmAccessToken(stt_api_key);
-      this.logger.debug({stt_api_key}, `Gather:exec - got ibm access token ${servedFromCache ? 'from cache' : ''}`);
-      this.sttCredentials = {...this.sttCredentials, access_token, stt_region};
-    }
-    const startListening = (cs, ep) => {
       this._startTimer();
       if (this.isContinuousAsr && 0 === this.timeout) this._startAsrTimer();
       if (this.input.includes('speech') && !this.listenDuringPrompt) {
-        this._initSpeech(cs, ep)
+        try {
-          .then(() => {
+          await this._setSpeechHandlers(cs, ep);
-            if (this.killed) {
+          if (this.killed) {
-              this.logger.info('Gather:exec - task was quickly killed so do not transcribe');
+            this.logger.info('Gather:exec - task was quickly killed so do not transcribe');
-              return;
+            return;
-            }
+          }
-            this._startTranscribing(ep);
+          this._startTranscribing(ep);
-            return updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
+          return updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
-          })
+        } catch (e) {
-          .catch((err) => {
+          if (this.fallbackVendor && this.isHandledByPrimaryProvider) {
-            this.logger.error({err}, 'error in initSpeech');
+            await this._fallback();
-          });
+            startListening(cs, ep);
+          } else {
+            this.logger.error({error: e}, 'error in initSpeech');
+          }
+        }
       }
     };
 
@@ -288,7 +207,7 @@ class TaskGather extends Task {
       }
 
       if (this.input.includes('speech') && this.listenDuringPrompt) {
-        await this._initSpeech(cs, ep);
+        await this._setSpeechHandlers(cs, ep);
         this._startTranscribing(ep);
         updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid)
           .catch(() => {/*already logged error */});
@@ -302,7 +221,7 @@ class TaskGather extends Task {
     } catch (err) {
       this.logger.error(err, 'TaskGather:exec error');
     }
-    this.removeSpeechListeners(ep);
+    this.removeCustomEventListeners();
   }
 
   kill(cs) {
@@ -340,6 +259,16 @@ class TaskGather extends Task {
       this._resolve('dtmf-terminator-key');
     }
     else if (this.input.includes('digits')) {
+      if (this.digitBuffer.length === 0 && this.needsStt) {
+        // DTMF is higher priority than STT.
+        this.removeCustomEventListeners();
+        ep.stopTranscription({
+          vendor: this.vendor,
+          bugname: this.bugname,
+        })
+          .catch((err) => this.logger.error({err},
+            ` Received DTMF, Error stopping transcription for vendor ${this.vendor}`));
+      }
       this.digitBuffer += evt.dtmf;
       const len = this.digitBuffer.length;
       if (len === this.numDigits || len === this.maxDigits) {
@@ -362,38 +291,51 @@ class TaskGather extends Task {
     }
   }
 
-  async _initSpeech(cs, ep) {
+  async _setSpeechHandlers(cs, ep) {
+    if (this._speechHandlersSet) return;
+    this._speechHandlersSet = true;
+
+    /* some special deepgram logic */
+    if (this.vendor === 'deepgram') {
+      if (this.isContinuousAsr) this._doContinuousAsrWithDeepgram(this.asrTimeout);
+      if (this.data.recognizer?.deepgramOptions?.shortUtterance) this.shortUtterance = true;
+    }
+
     const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
-        this.bugname = 'google_transcribe';
+        this.bugname = `${this.bugname_prefix}google_transcribe`;
-        ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
-        ep.addCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance, this._onEndOfUtterance.bind(this, cs, ep));
+          ep, GoogleTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(GoogleTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, GoogleTranscriptionEvents.EndOfUtterance, this._onEndOfUtterance.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, GoogleTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
         break;
 
       case 'aws':
       case 'polly':
-        this.bugname = 'aws_transcribe';
+        this.bugname = `${this.bugname_prefix}aws_transcribe`;
-        ep.addCustomEventListener(AwsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(AwsTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
         break;
       case 'microsoft':
-        this.bugname = 'azure_transcribe';
+        this.bugname = `${this.bugname_prefix}azure_transcribe`;
-        ep.addCustomEventListener(AzureTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
-        ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected,
+          ep, AzureTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AzureTranscriptionEvents.NoSpeechDetected,
           this._onNoSpeechDetected.bind(this, cs, ep));
-        ep.addCustomEventListener(AzureTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AzureTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
         break;
       case 'nuance':
-        this.bugname = 'nuance_transcribe';
+        this.bugname = `${this.bugname_prefix}nuance_transcribe`;
-        ep.addCustomEventListener(NuanceTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.StartOfSpeech,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.StartOfSpeech,
           this._onStartOfSpeech.bind(this, cs, ep));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.TranscriptionComplete,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.TranscriptionComplete,
           this._onTranscriptionComplete.bind(this, cs, ep));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.VadDetected,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.VadDetected,
           this._onVadDetected.bind(this, cs, ep));
 
         /* stall timers until prompt finishes playing */
@@ -403,35 +345,71 @@ class TaskGather extends Task {
         break;
 
       case 'deepgram':
-        this.bugname = 'deepgram_transcribe';
+        this.bugname = `${this.bugname_prefix}deepgram_transcribe`;
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Connect, this._onDeepgramConnect.bind(this, cs, ep));
+          ep, DeepgramTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
-          this._onDeepGramConnectFailure.bind(this, cs, ep));
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+
+        /* if app sets deepgramOptions.utteranceEndMs they essentially want continuous asr */
+        if (opts.DEEPGRAM_SPEECH_UTTERANCE_END_MS) this.isContinuousAsr = true;
         break;
 
       case 'soniox':
-        this.bugname = 'soniox_transcribe';
+        this.bugname = `${this.bugname_prefix}soniox_transcribe`;
-        ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, SonioxTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        break;
+
+      case 'cobalt':
+        this.bugname = `${this.bugname_prefix}cobalt_transcribe`;
+        this.addCustomEventListener(
+          ep, CobaltTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+
+        /* cobalt doesnt have language, it has model, which is required */
+        if (!this.data.recognizer.model) {
+          throw new Error('Cobalt requires a model to be specified');
+        }
+        this.language = this.data.recognizer.model;
+
+        /* special case: if using hints with cobalt we need to compile them */
+        this.hostport = opts.COBALT_SERVER_URI;
+        if (this.vendor === 'cobalt' && opts.COBALT_SPEECH_HINTS) {
+          try {
+            const context = await this.compileHintsForCobalt(
+              ep,
+              this.hostport,
+              this.data.recognizer.model,
+              opts.COBALT_CONTEXT_TOKEN,
+              opts.COBALT_SPEECH_HINTS
+            );
+            if (context) opts.COBALT_COMPILED_CONTEXT_DATA = context;
+            delete opts.COBALT_SPEECH_HINTS;
+          } catch (err) {
+            this.logger.error({err}, 'Error compiling hints for cobalt');
+          }
+        }
+        delete opts.COBALT_SERVER_URI;
         break;
 
       case 'ibm':
-        this.bugname = 'ibm_transcribe';
+        this.bugname = `${this.bugname_prefix}ibm_transcribe`;
-        ep.addCustomEventListener(IbmTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(IbmTranscriptionEvents.Connect, this._onIbmConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
-        ep.addCustomEventListener(IbmTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.ConnectFailure,
-          this._onIbmConnectFailure.bind(this, cs, ep));
+          this._onVendorConnectFailure.bind(this, cs, ep));
         break;
 
       case 'nvidia':
-        this.bugname = 'nvidia_transcribe';
+        this.bugname = `${this.bugname_prefix}nvidia_transcribe`;
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.StartOfSpeech,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.StartOfSpeech,
           this._onStartOfSpeech.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.TranscriptionComplete,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.TranscriptionComplete,
           this._onTranscriptionComplete.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.VadDetected,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.VadDetected,
           this._onVadDetected.bind(this, cs, ep));
 
         /* I think nvidia has this (??) - stall timers until prompt finishes playing */
@@ -440,13 +418,24 @@ class TaskGather extends Task {
         }
         break;
 
+      case 'assemblyai':
+        this.bugname = `${this.bugname_prefix}assemblyai_transcribe`;
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, AssemblyAiTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        break;
       default:
         if (this.vendor.startsWith('custom:')) {
-          this.bugname = `${this.vendor}_transcribe`;
+          this.bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
-          ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+          this.addCustomEventListener(
-          ep.addCustomEventListener(JambonzTranscriptionEvents.Connect, this._onJambonzConnect.bind(this, cs, ep));
+            ep, JambonzTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
-          ep.addCustomEventListener(JambonzTranscriptionEvents.ConnectFailure,
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
-            this._onJambonzConnectFailure.bind(this, cs, ep));
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.ConnectFailure,
+            this._onVendorConnectFailure.bind(this, cs, ep));
           break;
         }
         else {
@@ -457,7 +446,7 @@ class TaskGather extends Task {
     }
 
     /* common handler for all stt engine errors */
-    ep.addCustomEventListener(JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
+    this.addCustomEventListener(ep, JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
     await ep.set(opts)
       .catch((err) => this.logger.info(err, 'Error setting channel variables'));
   }
@@ -469,11 +458,18 @@ class TaskGather extends Task {
       interim: this.interim,
       bugname: this.bugname
     }, 'Gather:_startTranscribing');
+
+    /**
+     * 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
+     * the utterance is finalized (speech_final=true)
+     */
     ep.startTranscription({
       vendor: this.vendor,
       locale: this.language,
       interim: this.interim,
       bugname: this.bugname,
+      hostport: this.hostport,
     }).catch((err) => {
       const {writeAlerts, AlertType} = this.cs.srf.locals;
       this.logger.error(err, 'TaskGather:_startTranscribing error');
@@ -491,7 +487,11 @@ class TaskGather extends Task {
     this._clearTimer();
     this._timeoutTimer = setTimeout(() => {
       if (this.isContinuousAsr) this._startAsrTimer();
-      else this._resolve(this.digitBuffer.length >= this.minDigits ? 'dtmf-num-digits' : 'timeout');
+      else if (this.interDigitTimeout <= 0 ||
+        this.digitBuffer.length < this.minDigits ||
+        this.needsStt && this.digitBuffer.length === 0) {
+        this._resolve(this.digitBuffer.length >= this.minDigits ? 'dtmf-num-digits' : 'timeout');
+      }
     }, this.timeout);
   }
 
@@ -503,11 +503,13 @@ class TaskGather extends Task {
   }
 
   _startAsrTimer() {
+    if (this.vendor === 'deepgram') return; // no need
     assert(this.isContinuousAsr);
     this._clearAsrTimer();
     this._asrTimer = setTimeout(() => {
       this.logger.debug('_startAsrTimer - asr timer went off');
-      this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout');
+      const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+      this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
     }, this.asrTimeout);
     this.logger.debug(`_startAsrTimer: set for ${this.asrTimeout}ms`);
   }
@@ -517,11 +519,28 @@ class TaskGather extends Task {
     this._asrTimer = null;
   }
 
+  _startFastRecognitionTimer(evt) {
+    assert(this.fastRecognitionTimeout > 0);
+    this._clearFastRecognitionTimer();
+    this._fastRecognitionTimer = setTimeout(() => {
+      evt.is_final = true;
+      this._resolve('speech', evt);
+    }, this.fastRecognitionTimeout);
+  }
+
+  _clearFastRecognitionTimer() {
+    if (this._fastRecognitionTimer) {
+      clearTimeout(this._fastRecognitionTimer);
+    }
+    this._fastRecognitionTimer = null;
+  }
+
   _startFinalAsrTimer() {
     this._clearFinalAsrTimer();
     this._finalAsrTimer = setTimeout(() => {
       this.logger.debug('_startFinalAsrTimer - final asr timer went off');
-      this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout');
+      const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+      this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
     }, 1000);
     this.logger.debug('_startFinalAsrTimer: set for 1 second');
   }
@@ -560,11 +579,24 @@ class TaskGather extends Task {
     this.logger.debug({evt, bugname, finished}, `Gather:_onTranscription for vendor ${this.vendor}`);
     if (bugname && this.bugname !== bugname) return;
 
-    if (this.vendor === 'ibm') {
+    if (this.vendor === 'ibm' && evt?.state === 'listening') return;
-      if (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');
+      }
+      else {
+        this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram, return buffered transcript');
+        evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+        this._bufferedTranscripts = [];
+        this._resolve('speech', evt);
+      }
+      return;
     }
 
-    evt = this.normalizeTranscription(evt, this.vendor, 1, this.language);
+    evt = this.normalizeTranscription(evt, this.vendor, 1, this.language,
+      this.shortUtterance, this.data.recognizer.punctuation);
     if (evt.alternatives.length === 0) {
       this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, continue listening');
       return;
@@ -586,15 +618,27 @@ class TaskGather extends Task {
     const bufferedWords = this._sonioxTranscripts.length +
       this._bufferedTranscripts.reduce((count, e) => count + e.alternatives[0]?.transcript.split(' ').length, 0);
 
+    let emptyTranscript = false;
     if (evt.is_final) {
       if (evt.alternatives[0].transcript === '' && !this.callSession.callGone && !this.killed) {
+        emptyTranscript = true;
         if (finished === 'true' && ['microsoft', 'deepgram'].includes(this.vendor)) {
           this.logger.debug({evt}, 'TaskGather:_onTranscription - got empty transcript from old gather, disregarding');
+          return;
         }
-        else {
+        else if (this.vendor !== 'deepgram') {
           this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, continue listening');
+          return;
+        }
+        else if (this.isContinuousAsr) {
+          this.logger.info({evt},
+            'TaskGather:_onTranscription - got empty deepgram transcript during continous asr, continue listening');
+          return;
+        }
+        else if (this.vendor === 'deepgram' && this._bufferedTranscripts.length > 0) {
+          this.logger.info({evt},
+            'TaskGather:_onTranscription - got empty transcript from deepgram, return the buffered transcripts');
         }
-        return;
       }
 
       if (this.isContinuousAsr) {
@@ -606,14 +650,14 @@ class TaskGather extends Task {
             this.logger.debug('TaskGather:_onTranscription - removing trailing punctuation');
             evt.alternatives[0].transcript = t.slice(0, -1);
           }
-          else this.logger.debug({t}, 'TaskGather:_onTranscription - no trailing punctuation');
         }
         this.logger.info({evt}, 'TaskGather:_onTranscription - got transcript during continous asr');
         this._bufferedTranscripts.push(evt);
         this._clearTimer();
         if (this._finalAsrTimer) {
           this._clearFinalAsrTimer();
-          return this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout');
+          const evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+          return this._resolve(this._bufferedTranscripts.length > 0 ? 'speech' : 'timeout', evt);
         }
         this._startAsrTimer();
 
@@ -635,16 +679,25 @@ class TaskGather extends Task {
             evt = this.compileSonioxTranscripts(this._sonioxTranscripts, 1, this.language);
             this._sonioxTranscripts = [];
           }
+          else if (this.vendor === 'deepgram') {
+            /* compile transcripts into one */
+            if (!emptyTranscript) this._bufferedTranscripts.push(evt);
+            if (this.data.recognizer?.deepgramOptions?.utteranceEndMs) {
+              this.logger.debug('TaskGather:_onTranscription - got speech_final waiting for UtteranceEnd event');
+              return;
+            }
+            this.logger.debug({evt}, 'TaskGather:_onTranscription - compiling deepgram transcripts');
+            evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+            this._bufferedTranscripts = [];
+            this.logger.debug({evt}, 'TaskGather:_onTranscription - compiled deepgram transcripts');
+          }
+
+          /* here is where we return a final transcript */
           this._resolve('speech', evt);
         }
       }
     }
     else {
-      /* google has a measure of stability:
-        https://cloud.google.com/speech-to-text/docs/basics#streaming_responses
-        others do not.
-      */
-      //const isStableEnough = typeof evt.stability === 'undefined' || evt.stability > GATHER_STABILITY_THRESHOLD;
       this._clearTimer();
       this._startTimer();
       if (this.bargein && (words + bufferedWords) >= this.minBargeinWordCount) {
@@ -654,6 +707,9 @@ class TaskGather extends Task {
         }
         this._killAudio(cs);
       }
+      if (this.fastRecognitionTimeout) {
+        this._startFastRecognitionTimer(evt);
+      }
       if (this.partialResultHook) {
         const b3 = this.getTracingPropagation();
         const httpHeaders = b3 && {b3};
@@ -667,6 +723,14 @@ class TaskGather extends Task {
           this._sonioxTranscripts.push(evt.vendor.finalWords);
         }
       }
+      /* deepgram can send a non-final transcript but with words that are final, so we need to buffer */
+      if (this.vendor === 'deepgram') {
+        const originalEvent = evt.vendor.evt;
+        if (originalEvent.is_final && evt.alternatives[0].transcript !== '') {
+          this.logger.debug({evt}, 'Gather:_onTranscription - buffering a completed (partial) deepgram transcript');
+          this._bufferedTranscripts.push(evt);
+        }
+      }
     }
   }
   _onEndOfUtterance(cs, ep) {
@@ -681,7 +745,7 @@ class TaskGather extends Task {
      * getting a transcription.  This can happen if someone coughs or mumbles.
      * For that reason don't ask for a single utterance and we'll terminate the transcribe operation
      * once we get a final transcript.
-     * However, if the usr has specified a singleUtterance, then we need to restart here
+     * However, if the user has specified a singleUtterance, then we need to restart here
      * since we dont have a final transcript yet.
      */
     if (!this.resolved && !this.killed && !this._bufferedTranscripts.length && this.wantsSingleUtterance) {
@@ -698,14 +762,25 @@ class TaskGather extends Task {
   _onTranscriptionComplete(cs, ep) {
     this.logger.debug('TaskGather:_onTranscriptionComplete');
   }
-  _onDeepgramConnect(_cs, _ep) {
+  async _onJambonzError(cs, ep, evt) {
-    this.logger.debug('TaskGather:_onDeepgramConnect');
-  }
-  _onJambonzConnect(_cs, _ep) {
-    this.logger.debug('TaskGather:_onJambonzConnect');
-  }
-  _onJambonzError(cs, _ep, evt) {
     this.logger.info({evt}, 'TaskGather:_onJambonzError');
+    if (this.isHandledByPrimaryProvider && this.fallbackVendor) {
+      ep.stopTranscription({
+        vendor: this.vendor,
+        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 {
+        await this._fallback();
+        await this._initSpeech(cs, ep);
+        this._startTranscribing(ep);
+        updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
+        return;
+      } catch (error) {
+        this.logger.info({error}, `There is error while falling back to ${this.fallbackVendor}`);
+      }
+    }
     const {writeAlerts, AlertType} = cs.srf.locals;
 
     if (this.vendor === 'nuance') {
@@ -723,54 +798,16 @@ class TaskGather extends Task {
     this.notifyError({msg: 'ASR error', details:`Custom speech vendor ${this.vendor} error: ${evt.error}`});
   }
 
-  _onDeepGramConnectFailure(cs, _ep, evt) {
+  _onVendorConnectFailure(cs, _ep, evt) {
-    const {reason} = evt;
+    super._onVendorConnectFailure(cs, _ep, evt);
-    const {writeAlerts, AlertType} = cs.srf.locals;
-    this.logger.info({evt}, 'TaskGather:_onDeepgramConnectFailure');
-    writeAlerts({
-      account_sid: cs.accountSid,
-      alert_type: AlertType.STT_FAILURE,
-      message: `Failed connecting to Deepgram speech recognizer: ${reason}`,
-      vendor: 'deepgram',
-    }).catch((err) => this.logger.info({err}, 'Error generating alert for deepgram connection failure'));
-    this.notifyError({msg: 'ASR error', details:`Failed connecting to speech vendor deepgram: ${reason}`});
-    this.notifyTaskDone();
-  }
-  _onJambonzConnectFailure(cs, _ep, evt) {
-    const {reason} = evt;
-    const {writeAlerts, AlertType} = cs.srf.locals;
-    this.logger.info({evt}, 'TaskGather:_onJambonzConnectFailure');
-    writeAlerts({
-      account_sid: cs.accountSid,
-      alert_type: AlertType.STT_FAILURE,
-      message: `Failed connecting to ${this.vendor} speech recognizer: ${reason}`,
-      vendor: this.vendor,
-    }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
-    this.notifyError({msg: 'ASR error', details:`Failed connecting to speech vendor ${this.vendor}: ${reason}`});
     this.notifyTaskDone();
   }
 
-  _onIbmConnect(_cs, _ep) {
+  _onVendorError(cs, _ep, evt) {
-    this.logger.debug('TaskGather:_onIbmConnect');
+    super._onVendorError(cs, _ep, evt);
+    this._resolve('stt-error', evt);
   }
 
-  _onIbmConnectFailure(cs, _ep, evt) {
-    const {reason} = evt;
-    const {writeAlerts, AlertType} = cs.srf.locals;
-    this.logger.info({evt}, 'TaskGather:_onIbmConnectFailure');
-    writeAlerts({
-      account_sid: cs.accountSid,
-      alert_type: AlertType.STT_FAILURE,
-      message: `Failed connecting to IBM watson speech recognizer: ${reason}`,
-      vendor: 'ibm',
-    }).catch((err) => this.logger.info({err}, 'Error generating alert for IBM connection failure'));
-    this.notifyError({msg: 'ASR error', details:`Failed connecting to speech vendor IBM: ${reason}`});
-    this.notifyTaskDone();
-  }
-
-  _onIbmError(cs, _ep, evt) {
-    this.logger.info({evt}, 'TaskGather:_onIbmError');  }
-
   _onVadDetected(cs, ep) {
     if (this.bargein && this.minBargeinWordCount === 0) {
       this.logger.debug('TaskGather:_onVadDetected');
@@ -804,18 +841,7 @@ class TaskGather extends Task {
     }
     clearTimeout(this.interDigitTimer);
     this._clearTimer();
-
+    this._clearFastRecognitionTimer();
-    if (this.isContinuousAsr && reason.startsWith('speech')) {
-      evt = {
-        is_final: true,
-        transcripts: this._bufferedTranscripts
-      };
-      this.logger.debug({evt}, 'TaskGather:resolve continuous asr');
-    }
-    else if (!this.isContinuousAsr && reason.startsWith('speech') && this._bufferedTranscripts.length) {
-      compileTranscripts(this.logger, evt, this._bufferedTranscripts);
-      this.logger.debug({evt}, 'TaskGather:resolve buffered results');
-    }
 
     this.span.setAttributes({
       channel: 1,
@@ -823,7 +849,10 @@ class TaskGather extends Task {
       'stt.result': JSON.stringify(evt)
     });
     if (this.needsStt && this.ep && this.ep.connected) {
-      this.ep.stopTranscription({vendor: this.vendor})
+      this.ep.stopTranscription({
+        vendor: this.vendor,
+        bugname: this.bugname
+      })
         .catch((err) => this.logger.error({err}, 'Error stopping transcription'));
     }
 
@@ -855,6 +884,13 @@ class TaskGather extends Task {
           await this.performAction({reason: 'timeout'});
         }
       }
+      else if (reason.startsWith('stt-error')) {
+        if (this.parentTask) this.parentTask.emit('stt-error', evt);
+        else {
+          this.emit('stt-error', evt);
+          await this.performAction({reason: 'error', details: evt.error});
+        }
+      }
     } catch (err) {  /*already logged error*/ }
     this.notifyTaskDone();
   }

lib/tasks/lex.js

@@ -25,6 +25,13 @@ class Lex extends Task {
       this.vendor = this.data.tts.vendor || 'default';
       this.language = this.data.tts.language || 'default';
       this.voice = this.data.tts.voice || 'default';
+      this.speechCredentialLabel = this.data.tts.label || 'default';
+
+      // fallback tts
+      this.fallbackVendor = this.data.tts.fallbackVendor || 'default';
+      this.fallbackLanguage = this.data.tts.fallbackLanguage || 'default';
+      this.fallbackVoice = this.data.tts.fallbackLanguage || 'default';
+      this.fallbackLabel = this.data.tts.fallbackLabel || 'default';
     }
 
     this.botName = `${this.bot}:${this.alias}:${this.region}`;
@@ -102,8 +109,16 @@ class Lex extends Task {
         this.vendor = cs.speechSynthesisVendor;
         this.language = cs.speechSynthesisLanguage;
         this.voice = cs.speechSynthesisVoice;
+        this.speechCredentialLabel = cs.speechSynthesisLabel;
       }
-      this.ttsCredentials = cs.getSpeechCredentials(this.vendor, 'tts');
+      if (this.fallbackVendor === 'default') {
+        this.fallbackVendor = cs.fallbackSpeechSynthesisVendor;
+        this.fallbackLanguage = cs.fallbackSpeechSynthesisLanguage;
+        this.fallbackVoice = cs.fallbackSpeechSynthesisVoice;
+        this.fallbackLabel = cs.fallbackSpeechSynthesisLabel;
+      }
+
+      this.ttsCredentials = cs.getSpeechCredentials(this.vendor, 'tts', this.speechCredentialLabel);
 
       this.ep.addCustomEventListener('lex::intent', this._onIntent.bind(this, ep, cs));
       this.ep.addCustomEventListener('lex::transcription', this._onTranscription.bind(this, ep, cs));
@@ -168,6 +183,41 @@ class Lex extends Task {
     }
   }
 
+  async _fallbackSynthAudio(cs, msg, stats, synthAudio) {
+    try {
+      const {filePath} = await synthAudio(stats, {
+        account_sid: cs.accountSid,
+        text: msg,
+        vendor: this.vendor,
+        language: this.language,
+        voice: this.voice,
+        salt: cs.callSid,
+        credentials: this.ttsCredentials
+      });
+
+      return filePath;
+    } catch (error) {
+      this.logger.info({error}, 'failed to synth audio from primary vendor');
+      if (this.fallbackVendor) {
+        try {
+          const credential = cs.getSpeechCredentials(this.fallbackVendor, 'tts', this.fallbackLabel);
+          const {filePath} = await synthAudio(stats, {
+            account_sid: cs.accountSid,
+            text: msg,
+            vendor: this.fallbackVendor,
+            language: this.fallbackLanguage,
+            voice: this.fallbackVoice,
+            salt: cs.callSid,
+            credentials: credential
+          });
+          return filePath;
+        } catch (err) {
+          this.logger.info({err}, 'failed to synth audio from fallback vendor');
+        }
+      }
+    }
+  }
+
   /**
    * @param {*} evt - event data
    */
@@ -187,16 +237,7 @@ class Lex extends Task {
 
         try {
           this.logger.debug(`tts with ${this.vendor} ${this.voice}`);
-          // eslint-disable-next-line no-unused-vars
+          const filePath = await this._fallbackSynthAudio(cs, msg, stats, synthAudio);
-          const {filePath, servedFromCache} = await synthAudio(stats, {
-            account_sid: cs.accountSid,
-            text: msg,
-            vendor: this.vendor,
-            language: this.language,
-            voice: this.voice,
-            salt: cs.callSid,
-            credentials: this.ttsCredentials
-          });
           if (filePath) cs.trackTmpFile(filePath);
 
           if (this.events.includes('start-play')) {

lib/tasks/listen.js

@@ -8,6 +8,7 @@ const DTMF_SPAN_NAME = 'dtmf';
 class TaskListen extends Task {
   constructor(logger, opts, parentTask) {
     super(logger, opts);
+    this.disableBidirectionalAudio = opts.disableBidirectionalAudio;
     this.preconditions = TaskPreconditions.Endpoint;
 
     [
@@ -154,7 +155,7 @@ class TaskListen extends Task {
     }
 
     /* support bi-directional audio */
-    if (!this.disableBiDirectionalAudio) {
+    if (!this.disableBidirectionalAudio) {
       ep.addCustomEventListener(ListenEvents.PlayAudio, this._onPlayAudio.bind(this, ep));
     }
     ep.addCustomEventListener(ListenEvents.KillAudio, this._onKillAudio.bind(this, ep));

lib/tasks/rest_dial.js

@@ -16,6 +16,7 @@ class TaskRestDial extends Task {
     this.to = this.data.to;
     this.call_hook = this.data.call_hook;
     this.timeout = this.data.timeout || 60;
+    this.sipRequestWithinDialogHook = this.data.sipRequestWithinDialogHook;
 
     this.on('connect', this._onConnect.bind(this));
     this.on('callStatus', this._onCallStatus.bind(this));
@@ -63,7 +64,8 @@ class TaskRestDial extends Task {
     this.canCancel = false;
     const cs = this.callSession;
     cs.setDialog(dlg);
-
+    this.logger.debug('TaskRestDial:_onConnect - call connected');
+    if (this.sipRequestWithinDialogHook) this._initSipRequestWithinDialogHandler(cs, dlg);
     try {
       const b3 = this.getTracingPropagation();
       const httpHeaders = b3 && {b3};
@@ -90,8 +92,10 @@ class TaskRestDial extends Task {
       }
       let tasks;
       if (this.app_json) {
+        this.logger.debug('TaskRestDial: using app_json from task data');
         tasks = JSON.parse(this.app_json);
       } else {
+        this.logger.debug({call_hook: this.call_hook}, 'TaskRestDial: retrieving application');
         tasks = await cs.requestor.request('session:new', this.call_hook, params, httpHeaders);
       }
       if (tasks && Array.isArray(tasks)) {
@@ -125,7 +129,10 @@ class TaskRestDial extends Task {
   _onCallTimeout() {
     this.logger.debug('TaskRestDial: timeout expired without answer, killing task');
     this.timer = null;
-    this.kill(this.cs);
+    if (this.canCancel) {
+      this.canCancel = false;
+      this.cs?.req?.cancel();
+    }
   }
 
   _onAmdEvent(cs, evt) {
@@ -136,6 +143,16 @@ class TaskRestDial extends Task {
         this.logger.error({err}, 'Rest:dial:_onAmdEvent - error calling actionHook');
       });
   }
+
+  _initSipRequestWithinDialogHandler(cs, dlg) {
+    cs.sipRequestWithinDialogHook = this.sipRequestWithinDialogHook;
+    dlg.on('info', this._onRequestWithinDialog.bind(this, cs));
+    dlg.on('message', this._onRequestWithinDialog.bind(this, cs));
+  }
+
+  async _onRequestWithinDialog(cs, req, res) {
+    cs._onRequestWithinDialog(req, res);
+  }
 }
 
 module.exports = TaskRestDial;

lib/tasks/say.js

@@ -37,6 +37,7 @@ class TaskSay extends Task {
     this.synthesizer = this.data.synthesizer || {};
     this.disableTtsCache = this.data.disableTtsCache;
     this.options = this.synthesizer.options || {};
+    this.isHandledByPrimaryProvider = true;
   }
 
   get name() { return TaskName.Say; }
@@ -49,26 +50,24 @@ class TaskSay extends Task {
     return `${this.name}{${this.text[0]}}`;
   }
 
-  async exec(cs, {ep}) {
+  _validateURL(urlString) {
-    await super.exec(cs);
+    try {
+      new URL(urlString);
+      return true;
+    } catch (e) {
+      return false;
+    }
+  }
 
+  async _synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label, preCache = false}) {
     const {srf} = cs;
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, srf);
     const {writeAlerts, AlertType, stats} = srf.locals;
     const {synthAudio} = srf.locals.dbHelpers;
-    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 ;
-    let voice =  this.synthesizer.voice && this.synthesizer.voice !== 'default' ?
-      this.synthesizer.voice :
-      cs.speechSynthesisVoice;
     const engine = this.synthesizer.engine || 'standard';
     const salt = cs.callSid;
-    let credentials = cs.getSpeechCredentials(vendor, 'tts');
 
+    let credentials = cs.getSpeechCredentials(vendor, 'tts', label);
     /* parse Nuance voices into name and model */
     let model;
     if (vendor === 'nuance' && voice) {
@@ -77,6 +76,8 @@ class TaskSay extends Task {
         voice = arr[1];
         model = arr[2];
       }
+    } else if (vendor === 'deepgram') {
+      model = voice;
     }
 
     /* allow for microsoft custom region voice and api_key to be specified as an override */
@@ -87,10 +88,16 @@ class TaskSay extends Task {
       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;
     }
 
-    this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
+    if (!preCache) this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
-    this.ep = ep;
     try {
       if (!credentials) {
         writeAlerts({
@@ -113,11 +120,15 @@ class TaskSay extends Task {
         if (text.startsWith('silence_stream://')) return text;
 
         /* otel: trace time for tts */
-        const {span} = this.startChildSpan('tts-generation', {
+        let otelSpan;
-          'tts.vendor': vendor,
+        if (!preCache)  {
-          'tts.language': language,
+          const {span} = this.startChildSpan('tts-generation', {
-          'tts.voice': voice
+            'tts.vendor': vendor,
-        });
+            'tts.language': language,
+            'tts.voice': voice
+          });
+          otelSpan = span;
+        }
         try {
           const {filePath, servedFromCache, rtt} = await synthAudio(stats, {
             account_sid: cs.accountSid,
@@ -129,6 +140,7 @@ class TaskSay extends Task {
             model,
             salt,
             credentials,
+            options: this.options,
             disableTtsCache : this.disableTtsCache
           });
           this.logger.debug(`file ${filePath}, served from cache ${servedFromCache}`);
@@ -138,9 +150,9 @@ class TaskSay extends Task {
             updateSpeechCredentialLastUsed(credentials.speech_credential_sid)
               .catch(() => {/*already logged error */});
           }
-          span.setAttributes({'tts.cached': servedFromCache});
+          if (otelSpan) otelSpan.setAttributes({'tts.cached': servedFromCache});
-          span.end();
+          if (otelSpan) otelSpan.end();
-          if (!servedFromCache && rtt) {
+          if (!servedFromCache && rtt && !preCache) {
             this.notifyStatus({
               event: 'synthesized-audio',
               vendor,
@@ -152,7 +164,7 @@ class TaskSay extends Task {
           return filePath;
         } catch (err) {
           this.logger.info({err}, 'Error synthesizing tts');
-          span.end();
+          if (otelSpan) otelSpan.end();
           writeAlerts({
             account_sid: cs.accountSid,
             alert_type: AlertType.TTS_FAILURE,
@@ -160,31 +172,83 @@ class TaskSay extends Task {
             detail: err.message
           }).catch((err) => this.logger.info({err}, 'Error generating alert for tts failure'));
           this.notifyError({msg: 'TTS error', details: err.message || err});
-          return;
+          throw err;
         }
       };
 
-      const arr = this.text.map((t) => generateAudio(t));
+      const arr = this.text.map((t) => (this._validateURL(t) ? t : generateAudio(t)));
-      const filepath = (await Promise.all(arr)).filter((fp) => fp && fp.length);
+      return (await Promise.all(arr)).filter((fp) => fp && fp.length);
-      this.notifyStatus({event: 'start-playback'});
-
-      while (!this.killed && (this.loop === 'forever' || this.loop--) && this.ep?.connected) {
-        let segment = 0;
-        while (!this.killed && segment < filepath.length) {
-          if (cs.isInConference) {
-            const {memberId, confName, confUuid} = cs;
-            await this.playToConfMember(this.ep, memberId, confName, confUuid, filepath[segment]);
-          }
-          else {
-            this.logger.debug(`Say:exec sending command to play file ${filepath[segment]}`);
-            await ep.play(filepath[segment]);
-            this.logger.debug(`Say:exec completed play file ${filepath[segment]}`);
-          }
-          segment++;
-        }
-      }
     } catch (err) {
       this.logger.info(err, 'TaskSay:exec error');
+      throw err;
+    }
+
+  }
+
+  async exec(cs, {ep}) {
+    await super.exec(cs);
+    this.ep = ep;
+
+    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.synthesizer.label && this.synthesizer.label !== 'default' ?
+      this.synthesizer.label :
+      cs.speechSynthesisLabel;
+
+    const fallbackVendor = this.synthesizer.fallbackVendor && this.synthesizer.fallbackVendor !== 'default' ?
+      this.synthesizer.fallbackVendor :
+      cs.fallbackSpeechSynthesisVendor;
+    const fallbackLanguage = this.synthesizer.fallbackLanguage && this.synthesizer.fallbackLanguage !== 'default' ?
+      this.synthesizer.fallbackLanguage :
+      cs.fallbackSpeechSynthesisLanguage ;
+    const fallbackVoice =  this.synthesizer.fallbackVoice && this.synthesizer.fallbackVoice !== 'default' ?
+      this.synthesizer.fallbackVoice :
+      cs.fallbackSpeechSynthesisVoice;
+    const fallbackLabel = this.synthesizer.fallbackLabel && this.synthesizer.fallbackLabel !== 'default' ?
+      this.synthesizer.fallbackLabel :
+      cs.fallbackSpeechSynthesisLabel;
+
+    let filepath;
+    try {
+      filepath = await this._synthesizeWithSpecificVendor(cs, ep, {vendor, language, voice, label});
+    } catch (error) {
+      if (fallbackVendor && this.isHandledByPrimaryProvider) {
+        this.isHandledByPrimaryProvider = false;
+        this.logger.info(`Synthesize error, fallback to ${fallbackVendor}`);
+        filepath = await this._synthesizeWithSpecificVendor(cs, ep,
+          {
+            vendor: fallbackVendor,
+            language: fallbackLanguage,
+            voice: fallbackVoice,
+            label: fallbackLabel
+          });
+      } else {
+        throw error;
+      }
+    }
+    this.notifyStatus({event: 'start-playback'});
+
+    while (!this.killed && (this.loop === 'forever' || this.loop--) && this.ep?.connected) {
+      let segment = 0;
+      while (!this.killed && segment < filepath.length) {
+        if (cs.isInConference) {
+          const {memberId, confName, confUuid} = cs;
+          await this.playToConfMember(this.ep, memberId, confName, confUuid, filepath[segment]);
+        }
+        else {
+          this.logger.debug(`Say:exec sending command to play file ${filepath[segment]}`);
+          await ep.play(filepath[segment]);
+          this.logger.debug(`Say:exec completed play file ${filepath[segment]}`);
+        }
+        segment++;
+      }
     }
     this.emit('playDone');
   }

lib/tasks/stt-task.js

@@ -0,0 +1,278 @@
+const Task = require('./task');
+const assert = require('assert');
+const crypto = require('crypto');
+const { TaskPreconditions, CobaltTranscriptionEvents } = require('../utils/constants');
+
+class SttTask extends Task {
+
+  constructor(logger, data, parentTask) {
+    super(logger, data);
+    this.parentTask = parentTask;
+
+    this.preconditions = TaskPreconditions.Endpoint;
+
+    const {
+      setChannelVarsForStt,
+      normalizeTranscription,
+      setSpeechCredentialsAtRuntime,
+      compileSonioxTranscripts,
+      consolidateTranscripts
+    } = require('../utils/transcription-utils')(logger);
+    this.setChannelVarsForStt = setChannelVarsForStt;
+    this.normalizeTranscription = normalizeTranscription;
+    this.compileSonioxTranscripts = compileSonioxTranscripts;
+    this.consolidateTranscripts = consolidateTranscripts;
+    this.eventHandlers = [];
+    this.isHandledByPrimaryProvider = true;
+    if (this.data.recognizer) {
+      const recognizer = this.data.recognizer;
+      this.vendor = recognizer.vendor;
+      this.language = recognizer.language;
+      this.label = recognizer.label;
+
+      //fallback
+      this.fallbackVendor = recognizer.fallbackVendor || 'default';
+      this.fallbackLanguage = recognizer.fallbackLanguage || 'default';
+      this.fallbackLabel = recognizer.fallbackLabel || 'default';
+
+      /* let credentials be supplied in the recognizer object at runtime */
+      this.sttCredentials = setSpeechCredentialsAtRuntime(recognizer);
+
+      if (!Array.isArray(this.data.recognizer.altLanguages)) {
+        this.data.recognizer.altLanguages = [];
+      }
+    } else {
+      this.data.recognizer = {hints: [], altLanguages: []};
+    }
+
+    /* buffer for soniox transcripts */
+    this._sonioxTranscripts = [];
+    /*bug name prefix */
+    this.bugname_prefix = '';
+
+  }
+
+  async exec(cs, {ep, ep2}) {
+    super.exec(cs);
+    this.ep = ep;
+    this.ep2 = ep2;
+    // copy all value from config verb to this object.
+    if (cs.recognizer) {
+      for (const k in cs.recognizer) {
+        if (Array.isArray(this.data.recognizer[k]) ||
+        Array.isArray(cs.recognizer[k])) {
+          this.data.recognizer[k] = [
+            ...this.data.recognizer[k],
+            ...cs.recognizer[k]
+          ];
+        } else if (typeof this.data.recognizer[k] === 'object' ||
+        typeof cs.recognizer[k] === 'object'
+        ) {
+          this.data.recognizer[k] = {
+            ...this.data.recognizer[k],
+            ...cs.recognizer[k]
+          };
+        } else {
+          this.data.recognizer[k] = cs.recognizer[k] || this.data.recognizer[k];
+        }
+      }
+    }
+    if ('default' === this.vendor || !this.vendor) {
+      this.vendor = cs.speechRecognizerVendor;
+      if (this.data.recognizer) this.data.recognizer.vendor = this.vendor;
+    }
+    if ('default' === this.language || !this.language) {
+      this.language = cs.speechRecognizerLanguage;
+      if (this.data.recognizer) this.data.recognizer.language = this.language;
+    }
+    if ('default' === this.label || !this.label) {
+      this.label = cs.speechRecognizerLabel;
+      if (this.data.recognizer) this.data.recognizer.label = this.label;
+    }
+    // Fallback options
+    if ('default' === this.fallbackVendor || !this.fallbackVendor) {
+      this.fallbackVendor = cs.fallbackSpeechRecognizerVendor;
+      if (this.data.recognizer) this.data.recognizer.fallbackVendor = this.fallbackVendor;
+    }
+    if ('default' === this.fallbackLanguage || !this.fallbackLanguage) {
+      this.fallbackLanguage = cs.fallbackSpeechRecognizerLanguage;
+      if (this.data.recognizer) this.data.recognizer.fallbackLanguage = this.fallbackLanguage;
+    }
+    if ('default' === this.fallbackLabel || !this.fallbackLabel) {
+      this.fallbackLabel = cs.fallbackSpeechRecognizerLabel;
+      if (this.data.recognizer) this.data.recognizer.fallbackLabel = this.fallbackLabel;
+    }
+    if (!this.data.recognizer.vendor) {
+      this.data.recognizer.vendor = this.vendor;
+    }
+    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
+      // By default, application saves cobalt model in language
+      this.data.recognizer.model = cs.speechRecognizerLanguage;
+    }
+
+    if (!this.sttCredentials) {
+      try {
+        this.sttCredentials = await this._initSpeechCredentials(this.cs, this.vendor, this.label);
+      } catch (error) {
+        if (this.fallbackVendor && this.isHandledByPrimaryProvider) {
+          await this._fallback();
+        } else {
+          throw error;
+        }
+      }
+    }
+
+    /* when using cobalt model is required */
+    if (this.vendor === 'cobalt' && !this.data.recognizer.model) {
+      this.notifyError({ msg: 'ASR error', details:'Cobalt requires a model to be specified'});
+      throw new Error('Cobalt requires a model to be specified');
+    }
+
+    if (cs.hasAltLanguages) {
+      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages.concat(cs.altLanguages);
+      this.logger.debug({altLanguages: this.altLanguages},
+        'STT:exec - applying altLanguages');
+    }
+    if (cs.hasGlobalSttPunctuation && !this.data.recognizer.punctuation) {
+      this.data.recognizer.punctuation = cs.globalSttPunctuation;
+    }
+  }
+
+  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 _initSpeechCredentials(cs, vendor, label) {
+    const {getNuanceAccessToken, getIbmAccessToken} = cs.srf.locals.dbHelpers;
+    let credentials = cs.getSpeechCredentials(vendor, 'stt', label);
+
+    if (!credentials) {
+      const {writeAlerts, AlertType} = cs.srf.locals;
+      this.logger.info(`ERROR stt using ${vendor} requested but creds not supplied`);
+      writeAlerts({
+        account_sid: cs.accountSid,
+        alert_type: AlertType.STT_NOT_PROVISIONED,
+        vendor
+      }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
+      // Notify application that STT vender is wrong.
+      this.notifyError({
+        msg: 'ASR error',
+        details: `No speech-to-text service credentials for ${vendor} have been configured`
+      });
+      this.notifyTaskDone();
+      throw new Error(`No speech-to-text service credentials for ${vendor} have been configured`);
+    }
+
+    if (vendor === 'nuance' && credentials.client_id) {
+      /* get nuance access token */
+      const {client_id, secret} = credentials;
+      const {access_token, servedFromCache} = await getNuanceAccessToken(client_id, secret, 'asr tts');
+      this.logger.debug({client_id}, `got nuance access token ${servedFromCache ? 'from cache' : ''}`);
+      credentials = {...credentials, access_token};
+    }
+    else if (vendor == 'ibm' && credentials.stt_api_key) {
+      /* get ibm access token */
+      const {stt_api_key, stt_region} = credentials;
+      const {access_token, servedFromCache} = await getIbmAccessToken(stt_api_key);
+      this.logger.debug({stt_api_key}, `got ibm access token ${servedFromCache ? 'from cache' : ''}`);
+      credentials = {...credentials, access_token, stt_region};
+    }
+    return credentials;
+  }
+
+  async _fallback() {
+    assert(this.fallbackVendor, 'fallback failed without fallbackVendor configuration');
+    this.isHandledByPrimaryProvider = false;
+    this.logger.info(`Failed to use primary STT provider, fallback to ${this.fallbackVendor}`);
+    this.vendor = this.fallbackVendor;
+    this.language = this.fallbackLanguage;
+    this.label = this.fallbackLabel;
+    this.data.recognizer.vendor = this.vendor;
+    this.data.recognizer.language = this.language;
+    this.data.recognizer.label = this.label;
+    this.sttCredentials = await this._initSpeechCredentials(this.cs, this.vendor, this.label);
+  }
+
+  async compileHintsForCobalt(ep, hostport, model, token, hints) {
+    const {retrieveKey} = this.cs.srf.locals.dbHelpers;
+    const hash = crypto.createHash('sha1');
+    hash.update(`${model}:${hints}`);
+    const key = `cobalt:${hash.digest('hex')}`;
+    this.context = await retrieveKey(key);
+    if (this.context) {
+      this.logger.debug({model, hints}, 'found cached cobalt context for supplied hints');
+      return this.context;
+    }
+
+    this.logger.debug({model, hints}, 'compiling cobalt context for supplied hints');
+
+    return new Promise((resolve, reject) => {
+      this.cobaltCompileResolver = resolve;
+      ep.addCustomEventListener(CobaltTranscriptionEvents.CompileContext, this._onCompileContext.bind(this, ep, key));
+      ep.api('uuid_cobalt_compile_context', [ep.uuid, hostport, model, token, hints], (err, evt) => {
+        if (err || 0 !== evt.getBody().indexOf('+OK')) {
+          ep.removeCustomEventListener(CobaltTranscriptionEvents.CompileContext);
+          return reject(err);
+        }
+      });
+    });
+  }
+
+  _onCompileContext(ep, key, evt) {
+    const {addKey} = this.cs.srf.locals.dbHelpers;
+    this.logger.debug({evt}, `received cobalt compile context event, will cache under ${key}`);
+
+    this.cobaltCompileResolver(evt.compiled_context);
+    ep.removeCustomEventListener(CobaltTranscriptionEvents.CompileContext);
+    this.cobaltCompileResolver = null;
+
+    //cache the compiled context
+    addKey(key, evt.compiled_context, 3600 * 12)
+      .catch((err) => this.logger.info({err}, `Error caching cobalt context for ${key}`));
+  }
+
+  _doContinuousAsrWithDeepgram(asrTimeout) {
+    /* deepgram has an utterance_end_ms property that simplifies things */
+    assert(this.vendor === 'deepgram');
+    this.logger.debug(`_doContinuousAsrWithDeepgram - setting utterance_end_ms to ${asrTimeout}`);
+    const dgOptions = this.data.recognizer.deepgramOptions = this.data.recognizer.deepgramOptions || {};
+    dgOptions.utteranceEndMs = dgOptions.utteranceEndMs || asrTimeout;
+  }
+
+  _onVendorConnect(_cs, _ep) {
+    this.logger.debug(`TaskGather:_on${this.vendor}Connect`);
+  }
+
+  _onVendorError(cs, _ep, evt) {
+    this.logger.info({evt}, `${this.name}:_on${this.vendor}Error`);
+    const {writeAlerts, AlertType} = cs.srf.locals;
+    writeAlerts({
+      account_sid: cs.accountSid,
+      alert_type: AlertType.STT_FAILURE,
+      message: 'STT failure reported by vendor',
+      detail: evt.error,
+      vendor: this.vendor,
+    }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
+    this.notifyError({msg: 'ASR error', details:`Failed connecting to speech vendor ${this.vendor}: ${evt.error}`});
+  }
+
+  _onVendorConnectFailure(cs, _ep, evt) {
+    const {reason} = evt;
+    const {writeAlerts, AlertType} = cs.srf.locals;
+    this.logger.info({evt}, `${this.name}:_on${this.vendor}ConnectFailure`);
+    writeAlerts({
+      account_sid: cs.accountSid,
+      alert_type: AlertType.STT_FAILURE,
+      message: `Failed connecting to ${this.vendor} speech recognizer: ${reason}`,
+      vendor: this.vendor,
+    }).catch((err) => this.logger.info({err}, `Error generating alert for ${this.vendor} connection failure`));
+    this.notifyError({msg: 'ASR error', details:`Failed connecting to speech vendor ${this.vendor}: ${reason}`});
+  }
+}
+
+module.exports = SttTask;

lib/tasks/tag.js

@@ -12,7 +12,7 @@ class TaskTag extends Task {
   async exec(cs) {
     super.exec(cs);
     cs.callInfo.customerData = this.data;
-    //this.logger.debug({callInfo: cs.callInfo.toJSON()}, 'TaskTag:exec set customer data in callInfo');
+    this.logger.debug({customerData: cs.callInfo.customerData}, 'TaskTag:exec set customer data in callInfo');
   }
 }
 

lib/tasks/transcribe.js

@@ -1,69 +1,62 @@
-const Task = require('./task');
+const assert = require('assert');
 const {
   TaskName,
-  TaskPreconditions,
   GoogleTranscriptionEvents,
   NuanceTranscriptionEvents,
   AwsTranscriptionEvents,
   AzureTranscriptionEvents,
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
+  CobaltTranscriptionEvents,
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
-  JambonzTranscriptionEvents
+  JambonzTranscriptionEvents,
-} = require('../utils/constants');
+  TranscribeStatus,
+  AssemblyAiTranscriptionEvents
+} = require('../utils/constants.json');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
+const SttTask = require('./stt-task');
 
 const STT_LISTEN_SPAN_NAME = 'stt-listen';
 
-class TaskTranscribe extends Task {
+class TaskTranscribe extends SttTask {
   constructor(logger, opts, parentTask) {
-    super(logger, opts);
+    super(logger, opts, parentTask);
-    this.preconditions = TaskPreconditions.Endpoint;
-    this.parentTask = parentTask;
-
-    const {
-      setChannelVarsForStt,
-      normalizeTranscription,
-      removeSpeechListeners,
-      setSpeechCredentialsAtRuntime,
-      compileSonioxTranscripts
-    } = require('../utils/transcription-utils')(logger);
-    this.setChannelVarsForStt = setChannelVarsForStt;
-    this.normalizeTranscription = normalizeTranscription;
-    this.removeSpeechListeners = removeSpeechListeners;
-    this.compileSonioxTranscripts = compileSonioxTranscripts;
 
     this.transcriptionHook = this.data.transcriptionHook;
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
 
     if (this.data.recognizer) {
-      const recognizer = this.data.recognizer;
+      this.interim = !!this.data.recognizer.interim;
-      this.vendor = recognizer.vendor;
+      this.separateRecognitionPerChannel = this.data.recognizer.separateRecognitionPerChannel;
-      this.language = recognizer.language;
-      /* let credentials be supplied in the recognizer object at runtime */
-      this.sttCredentials = setSpeechCredentialsAtRuntime(recognizer);
-
-      this.interim = !!recognizer.interim;
-      this.separateRecognitionPerChannel = recognizer.separateRecognitionPerChannel;
-
-      this.data.recognizer.hints = this.data.recognizer.hints || [];
-      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages || [];
     }
-    else this.data.recognizer = {hints: [], altLanguages: []};
-
-    /* buffer for soniox transcripts */
-    this._sonioxTranscripts = [];
 
     this.childSpan = [null, null];
+
+    // Continuous asr timeout
+    this.asrTimeout = typeof this.data.recognizer.asrTimeout === 'number' ? this.data.recognizer.asrTimeout * 1000 : 0;
+    if (this.asrTimeout > 0) {
+      this.isContinuousAsr = true;
+    }
+    /* buffer speech for continuous asr */
+    this._bufferedTranscripts = [];
+    this.bugname_prefix = 'transcribe_';
   }
 
   get name() { return TaskName.Transcribe; }
 
   async exec(cs, {ep, ep2}) {
-    super.exec(cs);
+    await super.exec(cs, {ep, ep2});
+
+    if (this.data.recognizer.vendor === 'nuance') {
+      this.data.recognizer.nuanceOptions = {
+        // by default, nuance STT will recognize only 1st utterance.
+        // enable multiple allow nuance detact all utterances
+        utteranceDetectionMode: 'multiple',
+        ...this.data.recognizer.nuanceOptions
+      };
+    }
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
-    const {getNuanceAccessToken, getIbmAccessToken} = cs.srf.locals.dbHelpers;
 
     if (cs.hasGlobalSttHints) {
       const {hints, hintsBoost} = cs.globalSttHints;
@@ -72,57 +65,8 @@ class TaskTranscribe extends Task {
       this.logger.debug({hints: this.data.recognizer.hints, hintsBoost: this.data.recognizer.hintsBoost},
         'Transcribe:exec - applying global sttHints');
     }
-    if (cs.hasAltLanguages) {
-      this.data.recognizer.altLanguages = this.data.recognizer.altLanguages.concat(cs.altLanguages);
-      this.logger.debug({altLanguages: this.altLanguages},
-        'Transcribe:exec - applying altLanguages');
-    }
-    if (cs.hasGlobalSttPunctuation && !this.data.recognizer.punctuation) {
-      this.data.recognizer.punctuation = cs.globalSttPunctuation;
-    }
-
-    this.ep = ep;
-    this.ep2 = ep2;
-    if ('default' === this.vendor || !this.vendor) {
-      this.vendor = cs.speechRecognizerVendor;
-      if (this.data.recognizer) this.data.recognizer.vendor = this.vendor;
-    }
-    if ('default' === this.language || !this.language) {
-      this.language = cs.speechRecognizerLanguage;
-      if (this.data.recognizer) this.data.recognizer.language = this.language;
-    }
-    if (!this.data.recognizer.vendor) {
-      this.data.recognizer.vendor = this.vendor;
-    }
-    if (!this.sttCredentials) this.sttCredentials = cs.getSpeechCredentials(this.vendor, 'stt');
 
     try {
-      if (!this.sttCredentials) {
-        const {writeAlerts, AlertType} = cs.srf.locals;
-        this.logger.info(`TaskTranscribe:exec - ERROR stt using ${this.vendor} requested but creds not supplied`);
-        writeAlerts({
-          account_sid: cs.accountSid,
-          alert_type: AlertType.STT_NOT_PROVISIONED,
-          vendor: this.vendor
-        }).catch((err) => this.logger.info({err}, 'Error generating alert for no stt'));
-        throw new Error('no provisioned speech credentials for TTS');
-      }
-
-      if (this.vendor === 'nuance' && this.sttCredentials.client_id) {
-        /* get nuance access token */
-        const {client_id, secret} = this.sttCredentials;
-        const {access_token, servedFromCache} = await getNuanceAccessToken(client_id, secret, 'asr tts');
-        this.logger.debug({client_id},
-          `Transcribe:exec - got nuance access token ${servedFromCache ? 'from cache' : ''}`);
-        this.sttCredentials = {...this.sttCredentials, access_token};
-      }
-      else if (this.vendor == 'ibm' && this.sttCredentials.stt_api_key) {
-        /* get ibm access token */
-        const {stt_api_key, stt_region} = this.sttCredentials;
-        const {access_token, servedFromCache} = await getIbmAccessToken(stt_api_key);
-        this.logger.debug({stt_api_key}, `Gather:exec - got ibm access token ${servedFromCache ? 'from cache' : ''}`);
-        this.sttCredentials = {...this.sttCredentials, access_token, stt_region};
-      }
       await this._startTranscribing(cs, ep, 1);
       if (this.separateRecognitionPerChannel && ep2) {
         await this._startTranscribing(cs, ep2, 2);
@@ -136,15 +80,17 @@ class TaskTranscribe extends Task {
       this.logger.info(err, 'TaskTranscribe:exec - error');
       this.parentTask && this.parentTask.emit('error', err);
     }
-    this.removeSpeechListeners(ep);
+    this.removeCustomEventListeners();
   }
 
-  async kill(cs) {
+  async _stopTranscription() {
-    super.kill(cs);
     let stopTranscription = false;
     if (this.ep?.connected) {
       stopTranscription = true;
-      this.ep.stopTranscription({vendor: this.vendor})
+      this.ep.stopTranscription({
+        vendor: this.vendor,
+        bugname: this.bugname
+      })
         .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
     }
     if (this.separateRecognitionPerChannel && this.ep2 && this.ep2.connected) {
@@ -152,6 +98,13 @@ class TaskTranscribe extends Task {
       this.ep2.stopTranscription({vendor: this.vendor})
         .catch((err) => this.logger.info(err, 'Error TaskTranscribe:kill'));
     }
+
+    return stopTranscription;
+  }
+
+  async kill(cs) {
+    super.kill(cs);
+    const stopTranscription = this._stopTranscription();
     // hangup after 1 sec if we don't get a final transcription
     if (stopTranscription) this._timer = setTimeout(() => this.notifyTaskDone(), 1500);
     else this.notifyTaskDone();
@@ -159,89 +112,166 @@ class TaskTranscribe extends Task {
     await this.awaitTaskDone();
   }
 
-  async _startTranscribing(cs, ep, channel) {
+  async updateTranscribe(status) {
+    if (!this.killed && this.ep && this.ep.connected) {
+      this.logger.info(`TaskTranscribe:updateTranscribe status ${status}`);
+      switch (status) {
+        case TranscribeStatus.Pause:
+          await this._stopTranscription();
+          break;
+        case TranscribeStatus.Resume:
+          await this._startTranscribing(this.cs, this.ep, 1);
+          if (this.separateRecognitionPerChannel && this.ep2) {
+            await this._startTranscribing(this.cs, this.ep2, 2);
+          }
+          break;
+      }
+    }
+  }
+
+  async _setSpeechHandlers(cs, ep, channel) {
+    if (this[`_speechHandlersSet_${channel}`]) return;
+    this[`_speechHandlersSet_${channel}`] = true;
+
+    /* some special deepgram logic */
+    if (this.vendor === 'deepgram') {
+      if (this.isContinuousAsr) this._doContinuousAsrWithDeepgram(this.asrTimeout);
+    }
+
     const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
-        this.bugname = 'google_transcribe';
+        this.bugname = `${this.bugname_prefix}google_transcribe`;
-        ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, GoogleTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(GoogleTranscriptionEvents.NoAudioDetected,
+        this.addCustomEventListener(ep, GoogleTranscriptionEvents.NoAudioDetected,
           this._onNoAudio.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(GoogleTranscriptionEvents.MaxDurationExceeded,
+        this.addCustomEventListener(ep, GoogleTranscriptionEvents.MaxDurationExceeded,
           this._onMaxDurationExceeded.bind(this, cs, ep, channel));
         break;
 
       case 'aws':
       case 'polly':
-        this.bugname = 'aws_transcribe';
+        this.bugname = `${this.bugname_prefix}aws_transcribe`;
-        ep.addCustomEventListener(AwsTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(AwsTranscriptionEvents.NoAudioDetected,
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.NoAudioDetected,
           this._onNoAudio.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(AwsTranscriptionEvents.MaxDurationExceeded,
+        this.addCustomEventListener(ep, AwsTranscriptionEvents.MaxDurationExceeded,
           this._onMaxDurationExceeded.bind(this, cs, ep, channel));
         break;
       case 'microsoft':
-        this.bugname = 'azure_transcribe';
+        this.bugname = `${this.bugname_prefix}azure_transcribe`;
-        ep.addCustomEventListener(AzureTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, AzureTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected,
+        this.addCustomEventListener(ep, AzureTranscriptionEvents.NoSpeechDetected,
           this._onNoAudio.bind(this, cs, ep, channel));
         break;
       case 'nuance':
-        this.bugname = 'nuance_transcribe';
+        this.bugname = `${this.bugname_prefix}nuance_transcribe`;
-        ep.addCustomEventListener(NuanceTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NuanceTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.StartOfSpeech,
-          this._onStartOfSpeech.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(NuanceTranscriptionEvents.TranscriptionComplete,
-          this._onTranscriptionComplete.bind(this, cs, ep, channel));
         break;
       case 'deepgram':
-        this.bugname = 'deepgram_transcribe';
+        this.bugname = `${this.bugname_prefix}deepgram_transcribe`;
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.Connect,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.Connect,
-          this._onDeepgramConnect.bind(this, cs, ep, channel));
+          this._onVendorConnect.bind(this, cs, ep));
-        ep.addCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure,
+        this.addCustomEventListener(ep, DeepgramTranscriptionEvents.ConnectFailure,
-          this._onDeepGramConnectFailure.bind(this, cs, ep, channel));
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
+
+        /* if app sets deepgramOptions.utteranceEndMs they essentially want continuous asr */
+        if (opts.DEEPGRAM_SPEECH_UTTERANCE_END_MS) this.isContinuousAsr = true;
+
         break;
       case 'soniox':
-        this.bugname = 'soniox_transcribe';
+        this.bugname = `${this.bugname_prefix}soniox_transcribe`;
-        ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, SonioxTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         break;
-      case 'ibm':
+      case 'cobalt':
-        this.bugname = 'ibm_transcribe';
+        this.bugname = `${this.bugname_prefix}cobalt_transcribe`;
-        ep.addCustomEventListener(IbmTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, CobaltTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(IbmTranscriptionEvents.Connect,
+
-          this._onIbmConnect.bind(this, cs, ep, channel));
+        /* cobalt doesnt have language, it has model, which is required */
-        ep.addCustomEventListener(IbmTranscriptionEvents.ConnectFailure,
+        if (!this.data.recognizer.model) {
-          this._onIbmConnectFailure.bind(this, cs, ep, channel));
+          throw new Error('Cobalt requires a model to be specified');
+        }
+        this.language = this.data.recognizer.model;
+
+        /* special case: if using hints with cobalt we need to compile them */
+        this.hostport = opts.COBALT_SERVER_URI;
+        if (this.vendor === 'cobalt' && opts.COBALT_SPEECH_HINTS) {
+          try {
+            const context = await this.compileHintsForCobalt(
+              ep,
+              opts.COBALT_SERVER_URI,
+              this.data.recognizer.model,
+              opts.COBALT_CONTEXT_TOKEN,
+              opts.COBALT_SPEECH_HINTS
+            );
+            if (context) opts.COBALT_COMPILED_CONTEXT_DATA = context;
+            delete opts.COBALT_SPEECH_HINTS;
+          } catch (err) {
+            this.logger.error({err}, 'Error compiling hints for cobalt');
+          }
+        }
+        break;
+
+      case 'ibm':
+        this.bugname = `${this.bugname_prefix}ibm_transcribe`;
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.Connect,
+          this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, IbmTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
         break;
 
       case 'nvidia':
-        this.bugname = 'nvidia_transcribe';
+        this.bugname = `${this.bugname_prefix}nvidia_transcribe`;
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.Transcription,
+        this.addCustomEventListener(ep, NvidiaTranscriptionEvents.Transcription,
-          this._onTranscription.bind(this, cs, ep));
+          this._onTranscription.bind(this, cs, ep, channel));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.StartOfSpeech,
-          this._onStartOfSpeech.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.TranscriptionComplete,
-          this._onTranscriptionComplete.bind(this, cs, ep));
-        ep.addCustomEventListener(NvidiaTranscriptionEvents.VadDetected,
-          this._onVadDetected.bind(this, cs, ep));
         break;
+
+      case 'assemblyai':
+        this.bugname = `${this.bugname_prefix}assemblyai_transcribe`;
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep,
+          AssemblyAiTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, AssemblyAiTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
+        break;
+
       default:
-        throw new Error(`Invalid vendor ${this.vendor}`);
+        if (this.vendor.startsWith('custom:')) {
+          this.bugname = `${this.bugname_prefix}${this.vendor}_transcribe`;
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.Transcription,
+            this._onTranscription.bind(this, cs, ep, channel));
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+          this.addCustomEventListener(ep, JambonzTranscriptionEvents.ConnectFailure,
+            this._onVendorConnectFailure.bind(this, cs, ep));
+          break;
+        }
+        else {
+          this.notifyError({ msg: 'ASR error', details:`Invalid vendor ${this.vendor}`});
+          this.notifyTaskDone();
+          throw new Error(`Invalid vendor ${this.vendor}`);
+        }
     }
 
     /* common handler for all stt engine errors */
-    ep.addCustomEventListener(JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
+    this.addCustomEventListener(ep, JambonzTranscriptionEvents.Error, this._onJambonzError.bind(this, cs, ep));
     await ep.set(opts)
       .catch((err) => this.logger.info(err, 'Error setting channel variables'));
+  }
 
+  async _startTranscribing(cs, ep, channel) {
+    await this._setSpeechHandlers(cs, ep, channel);
     await this._transcribe(ep);
 
     /* start child span for this channel */
@@ -250,12 +280,15 @@ class TaskTranscribe extends Task {
   }
 
   async _transcribe(ep) {
+    this.logger.debug(
+      `TaskTranscribe:_transcribe - starting transcription vendor ${this.vendor} bugname ${this.bugname}`);
     await ep.startTranscription({
       vendor: this.vendor,
       interim: this.interim ? true : false,
       locale: this.language,
       channels: /*this.separateRecognitionPerChannel ? 2 : */ 1,
-      bugname: this.bugname
+      bugname: this.bugname,
+      hostport: this.hostport
     });
   }
 
@@ -264,12 +297,25 @@ class TaskTranscribe extends Task {
     const bugname = fsEvent.getHeader('media-bugname');
     if (bugname && this.bugname !== bugname) return;
 
-    if (this.vendor === 'ibm') {
+    if (this.vendor === 'ibm' && evt?.state === 'listening') return;
-      if (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');
+      }
+      else {
+        this.logger.debug('Gather:_onTranscription - got UtteranceEnd event from deepgram, return buffered transcript');
+        evt = this.consolidateTranscripts(this._bufferedTranscripts, 1, this.language);
+        this._bufferedTranscripts = [];
+        this._resolve('speech', evt);
+      }
+      return;
     }
     this.logger.debug({evt}, 'TaskTranscribe:_onTranscription - before normalization');
 
-    evt = this.normalizeTranscription(evt, this.vendor, channel, this.language);
+    evt = this.normalizeTranscription(evt, this.vendor, channel, this.language, undefined,
+      this.data.recognizer.punctuation);
     this.logger.debug({evt}, 'TaskTranscribe:_onTranscription');
     if (evt.alternatives.length === 0) {
       this.logger.info({evt}, 'TaskTranscribe:_onTranscription - got empty transcript, continue listening');
@@ -296,6 +342,15 @@ class TaskTranscribe extends Task {
       }
     }
 
+    if (this.isContinuousAsr && evt.is_final) {
+      this._bufferedTranscripts.push(evt);
+      this._startAsrTimer(channel);
+    } else {
+      await this._resolve(channel, evt);
+    }
+  }
+
+  async _resolve(channel, evt) {
     /* we've got a 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({
@@ -382,22 +437,49 @@ class TaskTranscribe extends Task {
       this._timer = null;
     }
   }
-  _onDeepgramConnect(_cs, _ep) {
+
-    this.logger.debug('TaskTranscribe:_onDeepgramConnect');
+  async _onJambonzError(cs, _ep, evt) {
+    this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
+    if (this.isHandledByPrimaryProvider && this.fallbackVendor) {
+      _ep.stopTranscription({
+        vendor: this.vendor,
+        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 {
+        await this._fallback();
+        let channel = 1;
+        if (this.ep !== _ep) {
+          channel = 2;
+        }
+        this._startTranscribing(cs, _ep, channel);
+        updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
+        return;
+      } catch (error) {
+        this.logger.info({error}, `There is error while falling back to ${this.fallbackVendor}`);
+      }
+    } else {
+      const {writeAlerts, AlertType} = cs.srf.locals;
+
+      if (this.vendor === 'nuance') {
+        const {code, error} = evt;
+        if (code === 404 && error === 'No speech') return this._resolve('timeout');
+        if (code === 413 && error === 'Too much speech') return this._resolve('timeout');
+      }
+      this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
+      writeAlerts({
+        account_sid: cs.accountSid,
+        alert_type: AlertType.STT_FAILURE,
+        message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
+        vendor: this.vendor,
+      }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
+      this.notifyError({msg: 'ASR error', details:`Custom speech vendor ${this.vendor} error: ${evt.error}`});
+    }
   }
 
-  _onDeepGramConnectFailure(cs, _ep, channel, evt) {
+  _onVendorConnectFailure(cs, _ep, channel, evt) {
-    const {reason} = evt;
+    super._onVendorConnectFailure(cs, _ep, evt);
-    const {writeAlerts, AlertType} = cs.srf.locals;
-    this.logger.info({evt}, 'TaskTranscribe:_onDeepgramConnectFailure');
-    writeAlerts({
-      account_sid: cs.accountSid,
-      alert_type: AlertType.STT_FAILURE,
-      message: `Failed connecting to Deepgram speech recognizer: ${reason}`,
-      vendor: 'deepgram',
-    }).catch((err) => this.logger.info({err}, 'Error generating alert for deepgram connection failure'));
-    this.notifyError(`Failed connecting to speech vendor deepgram: ${reason}`);
-
     if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
       this.childSpan[channel - 1].span.setAttributes({
         channel,
@@ -408,54 +490,23 @@ class TaskTranscribe extends Task {
     this.notifyTaskDone();
   }
 
-  _onIbmConnect(_cs, _ep) {
+  _startAsrTimer(channel) {
-    this.logger.debug('TaskTranscribe:_onIbmConnect');
+    if (this.vendor === 'deepgram') return; // no need
+    assert(this.isContinuousAsr);
+    this._clearAsrTimer(channel);
+    this._asrTimer = setTimeout(() => {
+      this.logger.debug(`TaskTranscribe:_startAsrTimer - asr timer went off for channel: ${channel}`);
+      const evt = this.consolidateTranscripts(this._bufferedTranscripts, channel, this.language);
+      this._bufferedTranscripts = [];
+      this._resolve(channel, evt);
+    }, this.asrTimeout);
+    this.logger.debug(`TaskTranscribe:_startAsrTimer: set for ${this.asrTimeout}ms for channel ${channel}`);
   }
 
-  _onIbmConnectFailure(cs, _ep, channel, evt) {
+  _clearAsrTimer(channel) {
-    const {reason} = evt;
+    if (this._asrTimer) clearTimeout(this._asrTimer);
-    const {writeAlerts, AlertType} = cs.srf.locals;
+    this._asrTimer = null;
-    this.logger.info({evt}, 'TaskTranscribe:_onIbmConnectFailure');
-    writeAlerts({
-      account_sid: cs.accountSid,
-      alert_type: AlertType.STT_FAILURE,
-      message: `Failed connecting to IBM watson speech recognizer: ${reason}`,
-      vendor: 'ibm',
-    }).catch((err) => this.logger.info({err}, 'Error generating alert for IBM connection failure'));
-    this.notifyError(`Failed connecting to speech vendor IBM: ${reason}`);
-
-    if (this.childSpan[channel - 1] && this.childSpan[channel - 1].span) {
-      this.childSpan[channel - 1].span.setAttributes({
-        channel,
-        'stt.resolve': 'connection failure'
-      });
-      this.childSpan[channel - 1].span.end();
-    }
-    this.notifyTaskDone();
   }
-  _onIbmError(cs, _ep, _channel, evt) {
-    this.logger.info({evt}, 'TaskTranscribe:_onIbmError');
-  }
-  _onJambonzError(cs, _ep, evt) {
-    this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
-    const {writeAlerts, AlertType} = cs.srf.locals;
-
-    if (this.vendor === 'nuance') {
-      const {code, error} = evt;
-      if (code === 404 && error === 'No speech') return this._resolve('timeout');
-      if (code === 413 && error === 'Too much speech') return this._resolve('timeout');
-    }
-    this.logger.info({evt}, 'TaskTranscribe:_onJambonzError');
-    writeAlerts({
-      account_sid: cs.accountSid,
-      alert_type: AlertType.STT_FAILURE,
-      message: `Custom speech vendor ${this.vendor} error: ${evt.error}`,
-      vendor: this.vendor,
-    }).catch((err) => this.logger.info({err}, 'Error generating alert for jambonz custom connection failure'));
-    this.notifyError({msg: 'ASR error', details:`Custom speech vendor ${this.vendor} error: ${evt.error}`});
-  }
-
-
 }
 
 module.exports = TaskTranscribe;

lib/utils/amd-utils.js

@@ -9,6 +9,7 @@ const {
   NvidiaTranscriptionEvents,
   IbmTranscriptionEvents,
   SonioxTranscriptionEvents,
+  CobaltTranscriptionEvents,
   DeepgramTranscriptionEvents,
   JambonzTranscriptionEvents,
   AmdEvents,
@@ -54,7 +55,8 @@ class Amd extends Emitter {
     this.language = opts.recognizer?.language || cs.speechRecognizerLanguage;
     if ('default' === this.language) this.language = cs.speechRecognizerLanguage;
 
-    this.sttCredentials = cs.getSpeechCredentials(this.vendor, 'stt');
+    this.sttCredentials = cs.getSpeechCredentials(this.vendor, 'stt',
+      opts.recognizer?.label || cs.speechRecognizerLabel);
 
     if (!this.sttCredentials) throw new Error(`No speech credentials found for vendor ${this.vendor}`);
 
@@ -313,6 +315,10 @@ module.exports = (logger) => {
         ep.addCustomEventListener(NvidiaTranscriptionEvents.Transcription, amd.transcriptionHandler);
         break;
 
+      case 'cobalt':
+        ep.addCustomEventListener(CobaltTranscriptionEvents.Transcription, amd.transcriptionHandler);
+        break;
+
       default:
         if (vendor.startsWith('custom:')) {
           ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription, amd.transcriptionHandler);

lib/utils/background-task-manager.js

@@ -0,0 +1,197 @@
+const { normalizeJambones } = require('@jambonz/verb-specifications');
+const makeTask = require('../tasks/make_task');
+const { JAMBONZ_RECORD_WS_BASE_URL, JAMBONZ_RECORD_WS_USERNAME, JAMBONZ_RECORD_WS_PASSWORD } = require('../config');
+const Emitter = require('events');
+
+class BackgroundTaskManager extends Emitter {
+  constructor({cs, logger, rootSpan}) {
+    super();
+    this.tasks = new Map();
+    this.cs = cs;
+    this.logger = logger;
+    this.rootSpan = rootSpan;
+  }
+
+  isTaskRunning(type) {
+    return this.tasks.has(type);
+  }
+
+  getTask(type) {
+    if (this.tasks.has(type)) {
+      return this.tasks.get(type);
+    }
+  }
+
+  count() {
+    return this.tasks.size;
+  }
+
+  async newTask(type, taskOpts) {
+    this.logger.info({taskOpts}, `initiating Background task ${type}`);
+    if (this.tasks.has(type)) {
+      this.logger.info(`Background task ${type} is running, skiped`);
+      return;
+    }
+    let task;
+    switch (type) {
+      case 'listen':
+        task = await this._initListen(taskOpts);
+        break;
+      case 'bargeIn':
+        task = await this._initBargeIn(taskOpts);
+        break;
+      case 'record':
+        task = await this._initRecord();
+        break;
+      case 'transcribe':
+        task = await this._initTranscribe(taskOpts);
+        break;
+      default:
+        break;
+    }
+    if (task) {
+      this.tasks.set(type, task);
+    }
+    return task;
+  }
+
+  stop(type) {
+    const task = this.getTask(type);
+    if (task) {
+      this.logger.info(`stopping background task: ${type}`);
+      task.removeAllListeners();
+      task.span.end();
+      task.kill();
+      // Remove task from managed List
+      this.tasks.delete(type);
+    } else {
+      this.logger.info(`stopping background task, ${type} is not running, skipped`);
+    }
+  }
+
+  stopAll() {
+    this.logger.info('BackgroundTaskManager:stopAll');
+    for (const key of this.tasks.keys()) {
+      this.stop(key);
+    }
+  }
+
+  // Initiate Task
+  // Initiate Listen
+  async _initListen(opts, bugname = 'jambonz-background-listen', ignoreCustomerData = false, type = 'listen') {
+    let task;
+    try {
+      const t = normalizeJambones(this.logger, [opts]);
+      task = makeTask(this.logger, t[0]);
+      task.bugname = bugname;
+      task.ignoreCustomerData = ignoreCustomerData;
+      const resources = await this.cs._evaluatePreconditions(task);
+      const {span, ctx} = this.rootSpan.startChildSpan(`background-${type}:${task.summary}`);
+      task.span = span;
+      task.ctx = ctx;
+      task.exec(this.cs, resources)
+        .then(this._taskCompleted.bind(this, type, task))
+        .catch(this._taskError.bind(this, type, task));
+    } catch (err) {
+      this.logger.info({err, opts}, `BackgroundTaskManager:_initListen - Error creating ${bugname} task`);
+    }
+    return task;
+  }
+
+  // Initiate Gather
+  async _initBargeIn(opts) {
+    let task;
+    try {
+      const t = normalizeJambones(this.logger, [opts]);
+      task = makeTask(this.logger, t[0]);
+      task
+        .once('dtmf', this._bargeInTaskCompleted.bind(this))
+        .once('vad', this._bargeInTaskCompleted.bind(this))
+        .once('transcription', this._bargeInTaskCompleted.bind(this))
+        .once('timeout', this._bargeInTaskCompleted.bind(this));
+      const resources = await this.cs._evaluatePreconditions(task);
+      const {span, ctx} = this.rootSpan.startChildSpan(`background-bargeIn:${task.summary}`);
+      task.span = span;
+      task.ctx = ctx;
+      task.bugname_prefix = 'background_bargeIn_';
+      task.exec(this.cs, resources)
+        .then(() => {
+          this._taskCompleted('bargeIn', task);
+          if (task.sticky && !this.cs.callGone && !this.cs._stopping) {
+            this.logger.info('BackgroundTaskManager:_initBargeIn: restarting background bargeIn');
+            this.newTask('bargeIn', opts);
+          }
+          return;
+        })
+        .catch(this._taskError.bind(this, 'bargeIn', task));
+    } catch (err) {
+      this.logger.info(err, 'BackgroundTaskManager:_initGather - Error creating bargeIn task');
+    }
+    return task;
+  }
+
+  // Initiate Record
+  async _initRecord() {
+    if (this.cs.accountInfo.account.record_all_calls || this.cs.application.record_all_calls) {
+      if (!JAMBONZ_RECORD_WS_BASE_URL || !this.cs.accountInfo.account.bucket_credential) {
+        this.logger.error(`_initRecord: invalid configuration,
+ missing JAMBONZ_RECORD_WS_BASE_URL or bucket configuration`);
+        return undefined;
+      }
+      const listenOpts = {
+        url: `${JAMBONZ_RECORD_WS_BASE_URL}/record/${this.cs.accountInfo.account.bucket_credential.vendor}`,
+        disableBidirectionalAudio: true,
+        mixType : 'stereo',
+        passDtmf: true
+      };
+      if (JAMBONZ_RECORD_WS_USERNAME && JAMBONZ_RECORD_WS_PASSWORD) {
+        listenOpts.wsAuth = {
+          username: JAMBONZ_RECORD_WS_USERNAME,
+          password: JAMBONZ_RECORD_WS_PASSWORD
+        };
+      }
+      this.logger.debug({listenOpts}, '_initRecord: enabling listen');
+      return await this._initListen({verb: 'listen', ...listenOpts}, 'jambonz-session-record', true, 'record');
+    }
+  }
+
+  // Initiate Transcribe
+  async _initTranscribe(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-transcribe:${task.summary}`);
+      task.span = span;
+      task.ctx = ctx;
+      task.bugname_prefix = 'background_transcribe_';
+      task.exec(this.cs, resources)
+        .then(this._taskCompleted.bind(this, 'transcribe', task))
+        .catch(this._taskError.bind(this, 'transcribe', task));
+    } catch (err) {
+      this.logger.info(err, 'BackgroundTaskManager:_initTranscribe - Error creating transcribe task');
+    }
+    return task;
+  }
+
+  _taskCompleted(type, task) {
+    this.logger.info({type, task}, 'BackgroundTaskManager:_taskCompleted: task completed');
+    task.removeAllListeners();
+    task.span.end();
+    this.tasks.delete(type);
+  }
+  _taskError(type, task, error) {
+    this.logger.info({type, task, error}, 'BackgroundTaskManager:_taskError: task Error');
+    task.removeAllListeners();
+    task.span.end();
+    this.tasks.delete(type);
+  }
+
+  _bargeInTaskCompleted(evt) {
+    this.logger.info({evt}, 'BackgroundTaskManager:_bargeInTaskCompleted on event from background bargeIn');
+    this.emit('bargeIn-done', evt);
+  }
+}
+
+module.exports = BackgroundTaskManager;

lib/utils/call-tracer.js

@@ -2,17 +2,24 @@ const {context, trace} = require('@opentelemetry/api');
 const {Dialog} = require('drachtio-srf');
 class RootSpan {
   constructor(callType, req) {
-    let tracer, callSid, linkedSpanId;
+    const {srf} = require('../../');
+    const tracer = srf.locals.otel.tracer;
+    let callSid, accountSid, applicationSid, linkedSpanId;
 
     if (req instanceof Dialog) {
       const dlg = req;
-      tracer = dlg.srf.locals.otel.tracer;
       callSid = dlg.callSid;
       linkedSpanId = dlg.linkedSpanId;
     }
-    else {
+    else if (req.srf) {
-      tracer = req.srf.locals.otel.tracer;
       callSid = req.locals.callSid;
+      accountSid = req.get('X-Account-Sid'),
+      applicationSid = req.locals.application_sid;
+    }
+    else {
+      callSid = req.callSid;
+      accountSid = req.accountSid;
+      applicationSid = req.applicationSid;
     }
     this._span = tracer.startSpan(callType || 'incoming-call');
     if (req instanceof Dialog) {
@@ -22,13 +29,20 @@ class RootSpan {
         callId: dlg.sip.callId
       });
     }
+    else if (req.srf) {
+      this._span.setAttributes({
+        callSid,
+        accountSid,
+        applicationSid,
+        callId: req.get('Call-ID'),
+        externalCallId: req.get('X-CID')
+      });
+    }
     else {
       this._span.setAttributes({
         callSid,
-        accountSid: req.get('X-Account-Sid'),
+        accountSid,
-        applicationSid: req.locals.application_sid,
+        applicationSid
-        callId: req.get('Call-ID'),
-        externalCallId: req.get('X-CID')
       });
     }
 

lib/utils/constants.json

@@ -29,7 +29,7 @@
     "Tag": "tag",
     "Transcribe": "transcribe"
   },
-  "AllowedSipRecVerbs": ["config", "gather", "transcribe", "listen"],
+  "AllowedSipRecVerbs": ["config", "gather", "transcribe", "listen", "tag"],
   "CallStatus": {
     "Trying": "trying",
     "Ringing": "ringing",
@@ -51,6 +51,11 @@
     "Silence": "silence",
     "Resume": "resume"
   },
+  "TranscribeStatus": {
+    "Pause": "pause",
+    "Silence": "silence",
+    "Resume": "resume"
+  },
   "TaskPreconditions": {
     "None": "none",
     "Endpoint": "endpoint",
@@ -90,6 +95,11 @@
     "Transcription": "soniox_transcribe::transcription",
     "Error": "soniox_transcribe::error"
   },
+  "CobaltTranscriptionEvents": {
+    "Transcription": "cobalt_speech::transcription",
+    "CompileContext": "cobalt_speech::compile_context_response",
+    "Error": "cobalt_speech::error"
+  },
   "IbmTranscriptionEvents": {
     "Transcription": "ibm_transcribe::transcription",
     "ConnectFailure": "ibm_transcribe::connect_failed",
@@ -116,6 +126,12 @@
     "Connect": "jambonz_transcribe::connect",
     "Error": "jambonz_transcribe::error"
   },
+  "AssemblyAiTranscriptionEvents": {
+    "Transcription": "assemblyai_transcribe::transcription",
+    "Error": "assemblyai_transcribe::error",
+    "ConnectFailure": "assemblyai_transcribe::connect_failed",
+    "Connect": "assemblyai_transcribe::connect"
+  },
   "ListenEvents": {
     "Connect": "mod_audio_fork::connect",
     "ConnectFailure": "mod_audio_fork::connect_failed",

lib/utils/db-utils.js

@@ -3,13 +3,10 @@ const {decrypt} = require('./encrypt-decrypt');
 const sqlAccountDetails = `SELECT * 
 FROM accounts account 
 WHERE account.account_sid = ?`;
-const sqlSpeechCredentials = `SELECT * 
+const sqlSpeechCredentialsForAccount = `SELECT * 
 FROM speech_credentials 
-WHERE account_sid = ? `;
+WHERE account_sid = ? OR (account_sid is NULL AND service_provider_sid =  
-const sqlSpeechCredentialsForSP = `SELECT * 
+(SELECT service_provider_sid from accounts where account_sid = ?))`;
-FROM speech_credentials 
-WHERE service_provider_sid =  
-(SELECT service_provider_sid from accounts where account_sid = ?)`;
 const sqlQueryAccountCarrierByName = `SELECT voip_carrier_sid  
 FROM voip_carriers vc 
 WHERE vc.account_sid = ? 
@@ -30,6 +27,9 @@ WHERE pn.account_sid IS NULL
 AND pn.service_provider_sid = 
 (SELECT service_provider_sid from accounts where account_sid = ?) 
 AND pn.number = ?`;
+const sqlQueryGoogleCustomVoices = `SELECT *
+FROM google_custom_voices
+WHERE google_custom_voice_sid = ?`;
 
 const speechMapper = (cred) => {
   const {credential, ...obj} = cred;
@@ -49,8 +49,10 @@ const speechMapper = (cred) => {
       obj.region = o.region;
       obj.use_custom_stt = o.use_custom_stt;
       obj.custom_stt_endpoint = o.custom_stt_endpoint;
+      obj.custom_stt_endpoint_url = o.custom_stt_endpoint_url;
       obj.use_custom_tts = o.use_custom_tts;
       obj.custom_tts_endpoint = o.custom_tts_endpoint;
+      obj.custom_tts_endpoint_url = o.custom_tts_endpoint_url;
     }
     else if ('wellsaid' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
@@ -82,7 +84,22 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.riva_server_uri = o.riva_server_uri;
     }
-    else if (obj.vendor.startsWith('custom:')) {
+    else if ('cobalt' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.cobalt_server_uri = o.cobalt_server_uri;
+    } 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 ('assemblyai' === 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 (obj.vendor.startsWith('custom:')) {
       const o = JSON.parse(decrypt(credential));
       obj.auth_token = o.auth_token;
       obj.custom_stt_url = o.custom_stt_url;
@@ -108,16 +125,9 @@ module.exports = (logger, srf) => {
 
     const [r] = await pp.query({sql: sqlAccountDetails, nestTables: true}, [account_sid]);
     if (0 === r.length) throw new Error(`invalid accountSid: ${account_sid}`);
-    const [r2] = await pp.query(sqlSpeechCredentials, [account_sid]);
+    const [r2] = await pp.query(sqlSpeechCredentialsForAccount, [account_sid, account_sid]);
     const speech = r2.map(speechMapper);
 
-    /* add service provider creds unless we have that vendor at the account level */
-    const [r3] = await pp.query(sqlSpeechCredentialsForSP, [account_sid]);
-    r3.forEach((s) => {
-      if (!speech.find((s2) => s2.vendor === s.vendor)) {
-        speech.push(speechMapper(s));
-      }
-    });
     const account = r[0];
     bucketCredentialDecrypt(account);
 
@@ -162,10 +172,22 @@ module.exports = (logger, srf) => {
     }
   };
 
+  const lookupGoogleCustomVoice = async(google_custom_voice_sid) => {
+    const pp = pool.promise();
+    try {
+      const [r] = await pp.query(sqlQueryGoogleCustomVoices, [google_custom_voice_sid]);
+      return r;
+
+    } catch (err) {
+      logger.error({err}, `lookupGoogleCustomVoices: Error ${google_custom_voice_sid}`);
+    }
+  };
+
   return {
     lookupAccountDetails,
     updateSpeechCredentialLastUsed,
     lookupCarrier,
-    lookupCarrierByPhoneNumber
+    lookupCarrierByPhoneNumber,
+    lookupGoogleCustomVoice
   };
 };

lib/utils/http-listener.js

@@ -6,7 +6,8 @@ const {PORT, HTTP_PORT_MAX} = require('../config');
 const doListen = (logger, app, port, resolve) => {
   const server = app.listen(port, () => {
     const {srf} = app.locals;
-    logger.info(`listening for HTTP requests on port ${PORT}, serviceUrl is ${srf.locals.serviceUrl}`);
+    srf.locals.serviceUrl = `http://${srf.locals.ipv4}:${port}`;
+    logger.info(`listening for HTTP requests on port ${port}, serviceUrl is ${srf.locals.serviceUrl}`);
     resolve({server, app});
   });
   return server;

lib/utils/http-requestor.js

@@ -1,4 +1,4 @@
-const {Client, Pool} = require('undici');
+const {request, getGlobalDispatcher, setGlobalDispatcher, Dispatcher, ProxyAgent, Client, Pool} = require('undici');
 const parseUrl = require('parse-url');
 const assert = require('assert');
 const BaseRequestor = require('./base-requestor');
@@ -10,6 +10,11 @@ const {
   HTTP_POOLSIZE,
   HTTP_PIPELINING,
   HTTP_TIMEOUT,
+  HTTP_PROXY_IP,
+  HTTP_PROXY_PORT,
+  HTTP_PROXY_PROTOCOL,
+  NODE_ENV,
+  HTTP_USER_AGENT_HEADER,
 } = require('../config');
 
 const toBase64 = (str) => Buffer.from(str || '', 'utf8').toString('base64');
@@ -21,6 +26,15 @@ function basicAuth(username, password) {
   return {Authorization: header};
 }
 
+const defaultDispatcher = HTTP_PROXY_IP ?
+  new ProxyAgent(`${HTTP_PROXY_PROTOCOL}://${HTTP_PROXY_IP}${HTTP_PROXY_PORT ? `:${HTTP_PROXY_PORT}` : ''}`) :
+  getGlobalDispatcher();
+
+setGlobalDispatcher(new class extends Dispatcher {
+  dispatch(options, handler) {
+    return defaultDispatcher.dispatch(options, handler);
+  }
+}());
 
 class HttpRequestor extends BaseRequestor {
   constructor(logger, account_sid, hook, secret) {
@@ -60,6 +74,18 @@ class HttpRequestor extends BaseRequestor {
       if (u.port) this.client = new Client(`${u.protocol}://${u.resource}:${u.port}`);
       else this.client = new Client(`${u.protocol}://${u.resource}`);
     }
+
+    if (NODE_ENV == 'test' && process.env.JAMBONES_HTTP_PROXY_IP) {
+      const defDispatcher =
+        new ProxyAgent(`${process.env.JAMBONES_HTTP_PROXY_PROTOCOL}://${process.env.JAMBONES_HTTP_PROXY_IP}${
+          process.env.JAMBONES_HTTP_PROXY_PORT ? `:${process.env.JAMBONES_HTTP_PROXY_PORT}` : ''}`);
+
+      setGlobalDispatcher(new class extends Dispatcher {
+        dispatch(options, handler) {
+          return defDispatcher.dispatch(options, handler);
+        }
+      }());
+    }
   }
 
   get baseUrl() {
@@ -91,6 +117,10 @@ class HttpRequestor extends BaseRequestor {
     const url = hook.url || hook;
     const method = hook.method || 'POST';
     let buf = '';
+    httpHeaders = {
+      ...httpHeaders,
+      ...(HTTP_USER_AGENT_HEADER && {'user-agent' : HTTP_USER_AGENT_HEADER})
+    };
 
     assert.ok(url, 'HttpRequestor:request url was not provided');
     assert.ok, (['GET', 'POST'].includes(method), `HttpRequestor:request method must be 'GET' or 'POST' not ${method}`);
@@ -139,7 +169,18 @@ class HttpRequestor extends BaseRequestor {
       };
       const absUrl = this._isRelativeUrl(url) ? `${this.baseUrl}${url}` : url;
       this.logger.debug({url, absUrl, hdrs}, 'send webhook');
-      const {statusCode, headers, body} =  await client.request({
+      const {statusCode, headers, body} =  HTTP_PROXY_IP ? await request(
+        this.baseUrl,
+        {
+          path,
+          query,
+          method,
+          headers: hdrs,
+          ...('POST' === method && {body: JSON.stringify(payload)}),
+          timeout: HTTP_TIMEOUT,
+          followRedirects: false
+        }
+      ) : await client.request({
         path,
         query,
         method,

lib/utils/install-srf-locals.js

@@ -8,15 +8,13 @@ const {
   JAMBONES_MYSQL_CONNECTION_LIMIT,
   JAMBONES_MYSQL_PORT,
   JAMBONES_FREESWITCH,
-  JAMBONES_REDIS_HOST,
-  JAMBONES_REDIS_PORT,
-  JAMBONES_REDIS_SENTINELS,
   SMPP_URL,
   JAMBONES_TIME_SERIES_HOST,
   JAMBONES_ESL_LISTEN_ADDRESS,
   PORT,
   NODE_ENV,
 } = require('../config');
+const Registrar = require('@jambonz/mw-registrar');
 const assert = require('assert');
 
 function initMS(logger, wrapper, ms) {
@@ -139,7 +137,8 @@ function installSrfLocals(srf, logger) {
     lookupTeamsByAccount,
     lookupAccountBySid,
     lookupAccountCapacitiesBySid,
-    lookupSmppGateways
+    lookupSmppGateways,
+    lookupClientByAccountAndUsername
   } = require('@jambonz/db-helpers')({
     host: JAMBONES_MYSQL_HOST,
     user: JAMBONES_MYSQL_USER,
@@ -173,18 +172,13 @@ function installSrfLocals(srf, logger) {
     retrieveByPatternSortedSet,
     sortedSetLength,
     sortedSetPositionByPattern
-  } = require('@jambonz/realtimedb-helpers')(JAMBONES_REDIS_SENTINELS || {
+  } = require('@jambonz/realtimedb-helpers')({}, logger, tracer);
-    host: JAMBONES_REDIS_HOST,
+  const registrar = new Registrar(logger, client);
-    port: JAMBONES_REDIS_PORT || 6379
-  }, logger, tracer);
   const {
     synthAudio,
     getNuanceAccessToken,
     getIbmAccessToken,
-  } = require('@jambonz/speech-utils')(JAMBONES_REDIS_SENTINELS || {
+  } = require('@jambonz/speech-utils')({}, logger);
-    host: JAMBONES_REDIS_HOST,
-    port: JAMBONES_REDIS_PORT || 6379
-  }, logger, tracer);
   const {
     writeAlerts,
     AlertType
@@ -204,6 +198,7 @@ function installSrfLocals(srf, logger) {
   srf.locals = {...srf.locals,
     dbHelpers: {
       client,
+      registrar,
       pool,
       lookupAppByPhoneNumber,
       lookupAppByRegex,
@@ -214,6 +209,7 @@ function installSrfLocals(srf, logger) {
       lookupAccountBySid,
       lookupAccountCapacitiesBySid,
       lookupSmppGateways,
+      lookupClientByAccountAndUsername,
       updateCallStatus,
       retrieveCall,
       listCalls,

lib/utils/place-outdial.js

@@ -13,9 +13,12 @@ const moment = require('moment');
 const stripCodecs = require('./strip-ancillary-codecs');
 const RootSpan = require('./call-tracer');
 const uuidv4 = require('uuid-random');
+const HttpRequestor = require('./http-requestor');
+const WsRequestor = require('./ws-requestor');
+const {makeOpusFirst} = require('./sdp-utils');
 
 class SingleDialer extends Emitter {
-  constructor({logger, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan}) {
+  constructor({logger, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask}) {
     super();
     assert(target.type);
 
@@ -37,6 +40,7 @@ class SingleDialer extends Emitter {
     this.callGone = false;
 
     this.callSid = uuidv4();
+    this.dialTask = dialTask;
 
     this.on('callStatusChange', this._notifyCallStatusChange.bind(this));
   }
@@ -75,7 +79,8 @@ class SingleDialer extends Emitter {
       ...(this.from.host && {'X-Preferred-From-Host': this.from.host}),
       'X-Jambonz-Routing': this.target.type,
       'X-Call-Sid': this.callSid,
-      ...(this.applicationSid && {'X-Application-Sid': this.applicationSid})
+      ...(this.applicationSid && {'X-Application-Sid': this.applicationSid}),
+      ...(this.target.proxy && {'X-SIP-Proxy': this.target.proxy})
     };
     if (srf.locals.fsUUID) {
       opts.headers = {
@@ -147,7 +152,7 @@ class SingleDialer extends Emitter {
 
       Object.assign(opts, {
         proxy: `sip:${this.sbcAddress}`,
-        localSdp: this.ep.local.sdp
+        localSdp: opts.opusFirst ? makeOpusFirst(this.ep.local.sdp) : this.ep.local.sdp
       });
       if (this.target.auth) opts.auth = this.target.auth;
       inviteSpan = this.startSpan('invite', {
@@ -175,6 +180,7 @@ class SingleDialer extends Emitter {
            *  (a) create a logger for this call
            */
           req.srf = srf;
+          this.req = req;
           this.callInfo = new CallInfo({
             direction: CallDirection.Outbound,
             parentCallInfo: this.parentCallInfo,
@@ -247,9 +253,14 @@ class SingleDialer extends Emitter {
         .on('modify', async(req, res) => {
           try {
             if (this.ep) {
-              const newSdp = await this.ep.modify(req.body);
+              if (this.dialTask && this.dialTask.isOnHold) {
-              res.send(200, {body: newSdp});
+                this.logger.info('dial is onhold, emit event');
-              this.logger.info({offer: req.body, answer: newSdp}, 'SingleDialer:exec: handling reINVITE');
+                this.emit('reinvite', req, res);
+              } else {
+                const newSdp = await this.ep.modify(req.body);
+                res.send(200, {body: newSdp});
+                this.logger.info({offer: req.body, answer: newSdp}, 'SingleDialer:exec: handling reINVITE');
+              }
             }
             else {
               this.logger.info('SingleDialer:exec: handling reINVITE with released media, emit event');
@@ -320,10 +331,15 @@ class SingleDialer extends Emitter {
     try {
       // retrieve set of tasks
       const json = await this.requestor.request('dial:confirm', confirmHook, this.callInfo.toJSON());
+      if (!json || (Array.isArray(json) && json.length === 0)) {
+        this.logger.info('SingleDialer:_executeApp: no tasks returned from confirm hook');
+        return;
+      }
       const tasks = normalizeJambones(this.logger, json).map((tdata) => makeTask(this.logger, tdata));
       // verify it contains only allowed verbs
       const allowedTasks = tasks.filter((task) => {
         return [
+          TaskPreconditions.None,
           TaskPreconditions.StableCall,
           TaskPreconditions.Endpoint
         ].includes(task.preconditions);
@@ -371,15 +387,35 @@ class SingleDialer extends Emitter {
     this.dlg.linkedSpanId = this.rootSpan.traceId;
     const rootSpan = new RootSpan('outbound-call', this.dlg);
     const newLogger = logger.child({traceId: rootSpan.traceId});
+    //clone application from parent call with new requestor
+    //parrent application will be closed in case the parent hangup
+    const app = {...application};
+    if ('WS' === app.call_hook?.method ||
+      app.call_hook?.url.startsWith('ws://') || app.call_hook?.url.startsWith('wss://')) {
+      const requestor = new WsRequestor(logger, this.accountInfo.account.account_sid,
+        app.call_hook, this.accountInfo.account.webhook_secret);
+      app.requestor = requestor;
+      app.notifier = requestor;
+      app.call_hook.method = 'WS';
+    }
+    else {
+      app.requestor = new HttpRequestor(logger, this.accountInfo.account.account_sid,
+        app.call_hook, this.accountInfo.account.webhook_secret);
+      if (app.call_status_hook) app.notifier = new HttpRequestor(logger,
+        this.accountInfo.account.account_sid, app.call_status_hook,
+        this.accountInfo.account.webhook_secret);
+      else app.notifier = {request: () => {}, close: () => {}};
+    }
     const cs = new AdultingCallSession({
       logger: newLogger,
       singleDialer: this,
-      application,
+      application: app,
       callInfo: this.callInfo,
       accountInfo: this.accountInfo,
       tasks,
       rootSpan
     });
+    cs.req = this.req;
     cs.exec().catch((err) => newLogger.error({err}, 'doAdulting: error executing session'));
     return cs;
   }
@@ -430,11 +466,11 @@ class SingleDialer extends Emitter {
 }
 
 function placeOutdial({
-  logger, srf, ms, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan
+  logger, srf, ms, sbcAddress, target, opts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask
 }) {
   const myOpts = deepcopy(opts);
   const sd = new SingleDialer({
-    logger, sbcAddress, target, myOpts, application, callInfo, accountInfo, rootSpan, startSpan
+    logger, sbcAddress, target, myOpts, application, callInfo, accountInfo, rootSpan, startSpan, dialTask
   });
   sd.exec(srf, ms, myOpts);
   return sd;

lib/utils/sdp-utils.js

@@ -0,0 +1,58 @@
+const sdpTransform = require('sdp-transform');
+
+const isOnhold = (sdp) => {
+  return sdp && (sdp.includes('a=sendonly') || sdp.includes('a=inactive'));
+};
+
+const mergeSdpMedia = (sdp1, sdp2) => {
+  const parsedSdp1 = sdpTransform.parse(sdp1);
+  const parsedSdp2 = sdpTransform.parse(sdp2);
+
+  parsedSdp1.media.push(...parsedSdp2.media);
+  return sdpTransform.write(parsedSdp1);
+};
+
+const getCodecPlacement = (parsedSdp, codec) => parsedSdp?.media[0]?.rtp?.findIndex((e) => e.codec === codec);
+
+const isOpusFirst = (sdp) => {
+  return getCodecPlacement(sdpTransform.parse(sdp), 'opus') === 0;
+};
+
+const makeOpusFirst = (sdp) => {
+  const parsedSdp = sdpTransform.parse(sdp);
+  // Find the index of the OPUS codec
+  const opusIndex = getCodecPlacement(parsedSdp, 'opus');
+
+  // Move OPUS codec to the beginning
+  if (opusIndex > 0) {
+    const opusEntry = parsedSdp.media[0].rtp.splice(opusIndex, 1)[0];
+    parsedSdp.media[0].rtp.unshift(opusEntry);
+
+    // Also move the corresponding payload type in the "m" line
+    const opusPayloadType = parsedSdp.media[0].payloads.split(' ')[opusIndex];
+    const otherPayloadTypes = parsedSdp.media[0].payloads.split(' ').filter((pt) => pt != opusPayloadType);
+    parsedSdp.media[0].payloads = [opusPayloadType, ...otherPayloadTypes].join(' ');
+  }
+  return sdpTransform.write(parsedSdp);
+};
+
+const extractSdpMedia = (sdp) => {
+  const parsedSdp1 = sdpTransform.parse(sdp);
+  if (parsedSdp1.media.length > 1) {
+    parsedSdp1.media = [parsedSdp1.media[0]];
+    const parsedSdp2 = sdpTransform.parse(sdp);
+    parsedSdp2.media = [parsedSdp2.media[1]];
+
+    return [sdpTransform.write(parsedSdp1), sdpTransform.write(parsedSdp2)];
+  } else {
+    return [sdp, sdp];
+  }
+};
+
+module.exports = {
+  isOnhold,
+  mergeSdpMedia,
+  extractSdpMedia,
+  isOpusFirst,
+  makeOpusFirst
+};

lib/utils/siprec-utils.js

@@ -97,8 +97,12 @@ const parseSiprecPayload = (req, logger) => {
             obj[`${prefix}participantstreamassoc`].forEach((ps) => {
               const part = participants[ps.$.participant_id];
               if (part) {
-                part.send = ps[`${prefix}send`][0];
+                if (ps.hasOwnProperty(`${prefix}send`)) {
-                part.recv = ps[`${prefix}recv`][0];
+                  part.send = ps[`${prefix}send`][0];
+                }
+                if (ps.hasOwnProperty(`${prefix}recv`)) {
+                  part.recv = ps[`${prefix}recv`][0];
+                }
               }
             });
           }
@@ -109,9 +113,9 @@ const parseSiprecPayload = (req, logger) => {
           obj[`${prefix}stream`].forEach((s) => {
             const streamId = s.$.stream_id;
             let sender;
-            for (const [k, v] of Object.entries(participants)) {
+            for (const v of Object.values(participants)) {
               if (v.send === streamId) {
-                sender = k;
+                sender = v;
                 break;
               }
             }
@@ -121,9 +125,15 @@ const parseSiprecPayload = (req, logger) => {
 
             sender.label = s[`${prefix}label`][0];
 
-            if (-1 !== ['1', 'a_leg', 'inbound'].indexOf(sender.label)) {
+            if (-1 !== ['1', 'a_leg', 'inbound', '10'].indexOf(sender.label)) {
-              opts.caller.aor = sender.aor ;
+              opts.caller.aor = sender.aor;
               if (sender.name) opts.caller.name = sender.name;
+              // Remap the sdp stream base on sender label
+              if (!opts.sdp1.includes(`a=label:${sender.label}`)) {
+                const tmp = opts.sdp1;
+                opts.sdp1 = opts.sdp2;
+                opts.sdp2 = tmp;
+              }
             }
             else {
               opts.callee.aor = sender.aor ;

lib/utils/transcription-utils.js

@@ -1,14 +1,6 @@
 const {
   TaskName,
-  AzureTranscriptionEvents,
+} = require('./constants.json');
-  GoogleTranscriptionEvents,
-  AwsTranscriptionEvents,
-  NuanceTranscriptionEvents,
-  DeepgramTranscriptionEvents,
-  SonioxTranscriptionEvents,
-  NvidiaTranscriptionEvents,
-  JambonzTranscriptionEvents
-} = require('./constants');
 
 const stickyVars = {
   google: [
@@ -33,12 +25,14 @@ const stickyVars = {
     'AZURE_SERVICE_ENDPOINT',
     'AZURE_INITIAL_SPEECH_TIMEOUT_MS',
     'AZURE_USE_OUTPUT_FORMAT_DETAILED',
+    'AZURE_SPEECH_SEGMENTATION_SILENCE_TIMEOUT_MS'
   ],
   deepgram: [
     'DEEPGRAM_SPEECH_KEYWORDS',
     'DEEPGRAM_API_KEY',
     'DEEPGRAM_SPEECH_TIER',
     'DEEPGRAM_SPEECH_MODEL',
+    'DEEPGRAM_SPEECH_ENABLE_SMART_FORMAT',
     'DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION',
     'DEEPGRAM_SPEECH_PROFANITY_FILTER',
     'DEEPGRAM_SPEECH_REDACT',
@@ -49,6 +43,7 @@ const stickyVars = {
     'DEEPGRAM_SPEECH_SEARCH',
     'DEEPGRAM_SPEECH_REPLACE',
     'DEEPGRAM_SPEECH_ENDPOINTING',
+    'DEEPGRAM_SPEECH_UTTERANCE_END_MS',
     'DEEPGRAM_SPEECH_VAD_TURNOFF',
     'DEEPGRAM_SPEECH_TAG'
   ],
@@ -92,12 +87,68 @@ const stickyVars = {
   nvidia: [
     'NVIDIA_HINTS'
   ],
+  cobalt: [
+    'COBALT_SPEECH_HINTS',
+    'COBALT_COMPILED_CONTEXT_DATA',
+    'COBALT_METADATA'
+  ],
   soniox: [
     'SONIOX_PROFANITY_FILTER',
     'SONIOX_MODEL'
+  ],
+  assemblyai: [
+    'ASSEMBLYAI_API_KEY',
+    'ASSEMBLYAI_WORD_BOOST'
   ]
 };
 
+const consolidateTranscripts = (bufferedTranscripts, channel, language) => {
+  if (bufferedTranscripts.length === 1) return bufferedTranscripts[0];
+  let totalConfidence = 0;
+  const finalTranscript = bufferedTranscripts.reduce((acc, evt) => {
+    totalConfidence += evt.alternatives[0].confidence;
+
+    let newTranscript = evt.alternatives[0].transcript;
+
+    // If new transcript consists only of digits, spaces, and a trailing comma or period
+    if (newTranscript.match(/^[\d\s]+[,.]?$/)) {
+      newTranscript = newTranscript.replace(/\s/g, '');  // Remove all spaces
+      if (newTranscript.endsWith(',')) {
+        newTranscript = newTranscript.slice(0, -1);  // Remove the trailing comma
+      } else if (newTranscript.endsWith('.')) {
+        newTranscript = newTranscript.slice(0, -1);  // Remove the trailing period
+      }
+    }
+
+    const lastChar = acc.alternatives[0].transcript.slice(-1);
+    const firstChar = newTranscript.charAt(0);
+
+    if (lastChar.match(/\d/) && firstChar.match(/\d/)) {
+      acc.alternatives[0].transcript += newTranscript;
+    } else {
+      acc.alternatives[0].transcript += ` ${newTranscript}`;
+    }
+
+    return acc;
+  }, {
+    language_code: language,
+    channel_tag: channel,
+    is_final: true,
+    alternatives: [{
+      transcript: ''
+    }]
+  });
+  finalTranscript.alternatives[0].confidence = bufferedTranscripts.length === 1 ?
+    bufferedTranscripts[0].alternatives[0].confidence :
+    totalConfidence / bufferedTranscripts.length;
+  finalTranscript.alternatives[0].transcript = finalTranscript.alternatives[0].transcript.trim();
+  finalTranscript.vendor = {
+    name: 'deepgram',
+    evt: bufferedTranscripts
+  };
+  return finalTranscript;
+};
+
 const compileSonioxTranscripts = (finalWordChunks, channel, language) => {
   const words = finalWordChunks.flat();
   const transcript = words.reduce((acc, word) => {
@@ -155,7 +206,7 @@ const normalizeSoniox = (evt, channel, language) => {
   };
 };
 
-const normalizeDeepgram = (evt, channel, language) => {
+const normalizeDeepgram = (evt, channel, language, shortUtterance) => {
   const copy = JSON.parse(JSON.stringify(evt));
   const alternatives = (evt.channel?.alternatives || [])
     .map((alt) => ({
@@ -163,10 +214,14 @@ const normalizeDeepgram = (evt, channel, language) => {
       transcript: alt.transcript,
     }));
 
+  /**
+   * note difference between is_final and speech_final in Deepgram:
+   * https://developers.deepgram.com/docs/understand-endpointing-interim-results
+   */
   return {
     language_code: language,
     channel_tag: channel,
-    is_final: evt.is_final,
+    is_final: shortUtterance ? evt.is_final : evt.speech_final,
     alternatives: [alternatives[0]],
     vendor: {
       name: 'deepgram',
@@ -225,12 +280,37 @@ const normalizeGoogle = (evt, channel, language) => {
   };
 };
 
-const normalizeCustom = (evt, channel, language) => {
+const normalizeCobalt = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  const alternatives = (evt.alternatives || [])
+    .map((alt) => ({
+      confidence: alt.confidence,
+      transcript: alt.transcript_formatted,
+    }));
+
   return {
     language_code: language,
     channel_tag: channel,
     is_final: evt.is_final,
-    alternatives: [evt.alternatives[0]]
+    alternatives,
+    vendor: {
+      name: 'cobalt',
+      evt: copy
+    }
+  };
+};
+
+const normalizeCustom = (evt, channel, language, vendor) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.is_final,
+    alternatives: [evt.alternatives[0]],
+    vendor: {
+      name: vendor,
+      evt: copy
+    }
   };
 };
 
@@ -248,19 +328,20 @@ const normalizeNuance = (evt, channel, language) => {
   };
 };
 
-const normalizeMicrosoft = (evt, channel, language) => {
+const normalizeMicrosoft = (evt, channel, language, punctuation = true) => {
   const copy = JSON.parse(JSON.stringify(evt));
   const nbest = evt.NBest;
   const language_code = evt.PrimaryLanguage?.Language || language;
   const alternatives = nbest ? nbest.map((n) => {
     return {
       confidence: n.Confidence,
-      transcript: n.Display
+      // remove all puntuation if needed
+      transcript: punctuation ? n.Display : n.Display.replace(/\p{P}/gu, '')
     };
   }) :
     [
       {
-        transcript: evt.DisplayText || evt.Text
+        transcript: punctuation ? evt.DisplayText || evt.Text : (evt.DisplayText || evt.Text).replace(/\p{P}/gu, '')
       }
     ];
 
@@ -290,16 +371,34 @@ const normalizeAws = (evt, channel, language) => {
   };
 };
 
+const normalizeAssemblyAi = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.message_type === 'FinalTranscript',
+    alternatives: [
+      {
+        confidence: evt.confidence,
+        transcript: evt.text,
+      }
+    ],
+    vendor: {
+      name: 'ASSEMBLYAI',
+      evt: copy
+    }
+  };
+};
 
 module.exports = (logger) => {
-  const normalizeTranscription = (evt, vendor, channel, language) => {
+  const normalizeTranscription = (evt, vendor, channel, language, shortUtterance, punctuation) => {
 
     //logger.debug({ evt, vendor, channel, language }, 'normalizeTranscription');
     switch (vendor) {
       case 'deepgram':
-        return normalizeDeepgram(evt, channel, language);
+        return normalizeDeepgram(evt, channel, language, shortUtterance);
       case 'microsoft':
-        return normalizeMicrosoft(evt, channel, language);
+        return normalizeMicrosoft(evt, channel, language, punctuation);
       case 'google':
         return normalizeGoogle(evt, channel, language);
       case 'aws':
@@ -312,9 +411,13 @@ module.exports = (logger) => {
         return normalizeNvidia(evt, channel, language);
       case 'soniox':
         return normalizeSoniox(evt, channel, language);
+      case 'cobalt':
+        return normalizeCobalt(evt, channel, language);
+      case 'assemblyai':
+        return normalizeAssemblyAi(evt, channel, language, shortUtterance);
       default:
         if (vendor.startsWith('custom:')) {
-          return normalizeCustom(evt, channel, language);
+          return normalizeCustom(evt, channel, language, vendor);
         }
         logger.error(`Unknown vendor ${vendor}`);
         return evt;
@@ -356,12 +459,13 @@ module.exports = (logger) => {
         ...(rOpts.punctuation === false && {GOOGLE_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION: 0}),
         ...(rOpts.words  == false && {GOOGLE_SPEECH_ENABLE_WORD_TIME_OFFSETS: 0}),
         ...(rOpts.diarization === false && {GOOGLE_SPEECH_SPEAKER_DIARIZATION: 0}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'string' &&
           {GOOGLE_SPEECH_HINTS: rOpts.hints.join(',')}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'object' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
           {GOOGLE_SPEECH_HINTS: JSON.stringify(rOpts.hints)}),
         ...(typeof rOpts.hintsBoost === 'number' && {GOOGLE_SPEECH_HINTS_BOOST: rOpts.hintsBoost}),
-        ...(rOpts.altLanguages.length > 0 &&
+        // When altLanguages is emptylist, we have to send value to freeswitch to clear the previous settings
+        ...(rOpts.altLanguages &&
           {GOOGLE_SPEECH_ALTERNATIVE_LANGUAGE_CODES: [...new Set(rOpts.altLanguages)].join(',')}),
         ...(rOpts.interactionType &&
           {GOOGLE_SPEECH_METADATA_INTERACTION_TYPE: rOpts.interactionType}),
@@ -384,28 +488,34 @@ module.exports = (logger) => {
       };
     }
     else if ('microsoft' === vendor) {
+      const {azureOptions = {}} = rOpts;
       opts = {
         ...opts,
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'string' &&
           {AZURE_SPEECH_HINTS: rOpts.hints.map((h) => h.trim()).join(',')}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'object' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
           {AZURE_SPEECH_HINTS: rOpts.hints.map((h) => h.phrase).join(',')}),
-        ...(rOpts.altLanguages && rOpts.altLanguages.length > 0 &&
+        // When altLanguages is emptylist, we have to send value to freeswitch to clear the previous settings
+        ...(rOpts.altLanguages &&
           {AZURE_SPEECH_ALTERNATIVE_LANGUAGE_CODES: [...new Set(rOpts.altLanguages)].join(',')}),
         ...(rOpts.requestSnr && {AZURE_REQUEST_SNR: 1}),
         ...(rOpts.profanityOption && {AZURE_PROFANITY_OPTION: rOpts.profanityOption}),
+        ...(sttCredentials.use_custom_stt && sttCredentials.custom_stt_endpoint_url &&
+          {AZURE_SERVICE_ENDPOINT: sttCredentials.custom_stt_endpoint_url}),
         ...(rOpts.azureServiceEndpoint && {AZURE_SERVICE_ENDPOINT: rOpts.azureServiceEndpoint}),
         ...(rOpts.initialSpeechTimeoutMs > 0 &&
           {AZURE_INITIAL_SPEECH_TIMEOUT_MS: rOpts.initialSpeechTimeoutMs}),
         ...(rOpts.requestSnr && {AZURE_REQUEST_SNR: 1}),
         ...(rOpts.audioLogging && {AZURE_AUDIO_LOGGING: 1}),
         ...{AZURE_USE_OUTPUT_FORMAT_DETAILED: 1},
+        ...(azureOptions.speechSegmentationSilenceTimeoutMs &&
+          {AZURE_SPEECH_SEGMENTATION_SILENCE_TIMEOUT_MS: azureOptions.speechSegmentationSilenceTimeoutMs}),
         ...(sttCredentials && {
-          AZURE_SUBSCRIPTION_KEY: sttCredentials.api_key,
+          ...(sttCredentials.api_key && {AZURE_SUBSCRIPTION_KEY: sttCredentials.api_key}),
-          AZURE_REGION: sttCredentials.region,
+          ...(sttCredentials.region && {AZURE_REGION: sttCredentials.region}),
         }),
         ...(sttCredentials.use_custom_stt && sttCredentials.custom_stt_endpoint &&
-          {AZURE_SERVICE_ENDPOINT_ID: sttCredentials.custom_stt_endpoint})
+          {AZURE_SERVICE_ENDPOINT_ID: sttCredentials.custom_stt_endpoint}),
       };
     }
     else if ('nuance' === vendor) {
@@ -468,10 +578,12 @@ module.exports = (logger) => {
           {DEEPGRAM_SPEECH_MODEL: deepgramOptions.model},
         ...(deepgramOptions.punctuate) &&
           {DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION: 1},
+        ...(deepgramOptions.smartFormatting) &&
+          {DEEPGRAM_SPEECH_ENABLE_SMART_FORMAT: 1},
         ...(deepgramOptions.profanityFilter) &&
           {DEEPGRAM_SPEECH_PROFANITY_FILTER: 1},
         ...(deepgramOptions.redact) &&
-          {DEEPGRAM_SPEECH_REDACT: 1},
+          {DEEPGRAM_SPEECH_REDACT: deepgramOptions.redact},
         ...(deepgramOptions.diarize) &&
           {DEEPGRAM_SPEECH_DIARIZE: 1},
         ...(deepgramOptions.diarizeVersion) &&
@@ -486,14 +598,16 @@ module.exports = (logger) => {
           {DEEPGRAM_SPEECH_SEARCH: deepgramOptions.search.join(',')},
         ...(deepgramOptions.replace) &&
           {DEEPGRAM_SPEECH_REPLACE: deepgramOptions.replace.join(',')},
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
+        ...(rOpts.hints && rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
           {DEEPGRAM_SPEECH_KEYWORDS: rOpts.hints.map((h) => h.trim()).join(',')}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'object' &&
+        ...(rOpts.hints && rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'object' &&
           {DEEPGRAM_SPEECH_KEYWORDS: rOpts.hints.map((h) => h.phrase).join(',')}),
         ...(deepgramOptions.keywords) &&
           {DEEPGRAM_SPEECH_KEYWORDS: deepgramOptions.keywords.join(',')},
         ...('endpointing' in deepgramOptions) &&
-          {DEEPGRAM_SPEECH_ENDPOINTING: deepgramOptions.endpointing},
+          {DEEPGRAM_SPEECH_ENDPOINTING: deepgramOptions.endpointing === false ? 'false' : deepgramOptions.endpointing},
+        ...(deepgramOptions.utteranceEndMs) &&
+          {DEEPGRAM_SPEECH_UTTERANCE_END_MS: deepgramOptions.utteranceEndMs},
         ...(deepgramOptions.vadTurnoff) &&
           {DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.vadTurnoff},
         ...(deepgramOptions.tag) &&
@@ -507,9 +621,9 @@ module.exports = (logger) => {
         ...opts,
         ...(sttCredentials.api_key) &&
           {SONIOX_API_KEY: sttCredentials.api_key},
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'string' &&
           {SONIOX_HINTS: rOpts.hints.join(',')}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'object' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
           {SONIOX_HINTS: JSON.stringify(rOpts.hints)}),
         ...(typeof rOpts.hintsBoost === 'number' &&
           {SONIOX_HINTS_BOOST: rOpts.hintsBoost}),
@@ -567,9 +681,9 @@ module.exports = (logger) => {
         ...(rOpts.diarization && rOpts.diarizationMaxSpeakers > 0 &&
           {NVIDIA_DIARIZATION_SPEAKER_COUNT: rOpts.diarizationMaxSpeakers}),
         ...(rOpts.separateRecognitionPerChannel && {NVIDIA_SEPARATE_RECOGNITION_PER_CHANNEL: 1}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'string' &&
           {NVIDIA_HINTS: rOpts.hints.join(',')}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'object' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
           {NVIDIA_HINTS: JSON.stringify(rOpts.hints)}),
         ...(typeof rOpts.hintsBoost === 'number' &&
           {NVIDIA_HINTS_BOOST: rOpts.hintsBoost}),
@@ -577,21 +691,49 @@ module.exports = (logger) => {
           {NVIDIA_CUSTOM_CONFIGURATION: JSON.stringify(nvidiaOptions.customConfiguration)}),
       };
     }
+    else if ('cobalt' === vendor) {
+      const {cobaltOptions = {}} = rOpts;
+      const cobaltUri = cobaltOptions.serverUri || sttCredentials.cobalt_server_uri;
+      opts = {
+        ...opts,
+        ...(rOpts.words && {COBALT_WORD_TIME_OFFSETS: 1}),
+        ...(!rOpts.words && {COBALT_WORD_TIME_OFFSETS: 0}),
+        ...(rOpts.model && {COBALT_MODEL: rOpts.model}),
+        ...(cobaltUri && {COBALT_SERVER_URI: cobaltUri}),
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'string' &&
+          {COBALT_SPEECH_HINTS: rOpts.hints.join(',')}),
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
+          {COBALT_SPEECH_HINTS: JSON.stringify(rOpts.hints)}),
+        ...(rOpts.hints?.length > 0 &&
+          {COBALT_CONTEXT_TOKEN: cobaltOptions.contextToken || 'unk:default'}),
+        ...(cobaltOptions.metadata && {COBALT_METADATA: cobaltOptions.metadata}),
+        ...(cobaltOptions.enableConfusionNetwork && {COBALT_ENABLE_CONFUSION_NETWORK: 1}),
+        ...(cobaltOptions.compiledContextData && {COBALT_COMPILED_CONTEXT_DATA: cobaltOptions.compiledContextData}),
+      };
+    } else if ('assemblyai' === vendor) {
+      opts = {
+        ...opts,
+        ...(sttCredentials.api_key) &&
+          {ASSEMBLYAI_API_KEY: sttCredentials.api_key},
+        ...(rOpts.hints?.length > 0 &&
+          {ASSEMBLYAI_WORD_BOOST: JSON.stringify(rOpts.hints)})
+      };
+    }
     else if (vendor.startsWith('custom:')) {
       let {options = {}} = rOpts;
       const {auth_token, custom_stt_url} = sttCredentials;
       options = {
         ...options,
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'string' &&
         {hints: rOpts.hints}),
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'object' &&
+        ...(rOpts.hints?.length > 0 && typeof rOpts.hints[0] === 'object' &&
         {hints: JSON.stringify(rOpts.hints)}),
         ...(typeof rOpts.hintsBoost === 'number' && {hintsBoost: rOpts.hintsBoost})
       };
 
       opts = {
         ...opts,
-        JAMBONZ_STT_API_KEY: auth_token,
+        ...(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)}),
       };
@@ -603,41 +745,6 @@ module.exports = (logger) => {
     return opts;
   };
 
-  const removeSpeechListeners = (ep) => {
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance);
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(AwsTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(AwsTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(AzureTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected);
-    ep.removeCustomEventListener(AzureTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.TranscriptionComplete);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.StartOfSpeech);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(DeepgramTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(DeepgramTranscriptionEvents.Connect);
-    ep.removeCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure);
-
-    ep.removeCustomEventListener(SonioxTranscriptionEvents.Transcription);
-
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.TranscriptionComplete);
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.StartOfSpeech);
-    ep.removeCustomEventListener(NvidiaTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.Connect);
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.ConnectFailure);
-
-    ep.removeCustomEventListener(JambonzTranscriptionEvents.Error);
-  };
-
   const setSpeechCredentialsAtRuntime = (recognizer) => {
     if (!recognizer) return;
     if (recognizer.vendor === 'nuance') {
@@ -657,6 +764,10 @@ module.exports = (logger) => {
       const {apiKey} = recognizer.sonioxOptions || {};
       if (apiKey) return {api_key: apiKey};
     }
+    else if (recognizer.vendor === 'cobalt') {
+      const {serverUri} = recognizer.cobaltOptions || {};
+      if (serverUri) return {cobalt_server_uri: serverUri};
+    }
     else if (recognizer.vendor === 'ibm') {
       const {ttsApiKey, ttsRegion, sttApiKey, sttRegion, instanceId} = recognizer.ibmOptions || {};
       if (ttsApiKey || sttApiKey) return {
@@ -672,8 +783,8 @@ module.exports = (logger) => {
   return {
     normalizeTranscription,
     setChannelVarsForStt,
-    removeSpeechListeners,
     setSpeechCredentialsAtRuntime,
-    compileSonioxTranscripts
+    compileSonioxTranscripts,
+    consolidateTranscripts
   };
 };

lib/utils/ws-requestor.js

@@ -9,7 +9,8 @@ const {
   JAMBONES_WS_PING_INTERVAL_MS,
   MAX_RECONNECTS,
   JAMBONES_WS_HANDSHAKE_TIMEOUT_MS,
-  JAMBONES_WS_MAX_PAYLOAD
+  JAMBONES_WS_MAX_PAYLOAD,
+  HTTP_USER_AGENT_HEADER
 } = require('../config');
 
 class WsRequestor extends BaseRequestor {
@@ -43,6 +44,7 @@ class WsRequestor extends BaseRequestor {
   async request(type, hook, params, httpHeaders = {}) {
     assert(HookMsgTypes.includes(type));
     const url = hook.url || hook;
+    const wantsAck = !['call:status', 'verb:status', 'jambonz:error'].includes(type);
 
     if (this.maliciousClient) {
       this.logger.info({url: this.url}, 'WsRequestor:request - discarding msg to malicious client');
@@ -73,11 +75,19 @@ class WsRequestor extends BaseRequestor {
       if (this.connectInProgress) {
         this.logger.debug(
           `WsRequestor:request(${this.id}) - queueing ${type} message since we are connecting`);
-        this.queuedMsg.push({type, hook, params, httpHeaders});
+        if (wantsAck) {
+          const p = new Promise((resolve, reject) => {
+            this.queuedMsg.push({type, hook, params, httpHeaders, promise: {resolve, reject}});
+          });
+          return p;
+        }
+        else {
+          this.queuedMsg.push({type, hook, params, httpHeaders});
+        }
         return;
       }
       this.connectInProgress = true;
-      this.logger.debug(`WsRequestor:request(${this.id}) - connecting since we do not have a connection`);
+      this.logger.debug(`WsRequestor:request(${this.id}) - connecting since we do not have a connection for ${type}`);
       if (this.connections >= MAX_RECONNECTS) {
         return Promise.reject(`max attempts connecting to ${this.url}`);
       }
@@ -116,9 +126,14 @@ class WsRequestor extends BaseRequestor {
 
     const sendQueuedMsgs = () => {
       if (this.queuedMsg.length > 0) {
-        for (const {type, hook, params, httpHeaders} of this.queuedMsg) {
+        for (const {type, hook, params, httpHeaders, promise} of this.queuedMsg) {
           this.logger.debug(`WsRequestor:request - preparing queued ${type} for sending`);
-          setImmediate(this.request.bind(this, type, hook, params, httpHeaders));
+          if (promise) {
+            this.request(type, hook, params, httpHeaders)
+              .then((res) => promise.resolve(res))
+              .catch((err) => promise.reject(err));
+          }
+          else setImmediate(this.request.bind(this, type, hook, params, httpHeaders));
         }
         this.queuedMsg.length = 0;
       }
@@ -137,7 +152,7 @@ class WsRequestor extends BaseRequestor {
     }
 
     /* simple notifications */
-    if (['call:status', 'verb:status', 'jambonz:error'].includes(type) || reconnectingWithoutAck) {
+    if (!wantsAck || reconnectingWithoutAck) {
       this.ws?.send(JSON.stringify(obj), () => {
         this.logger.debug({obj}, `WsRequestor:request websocket: sent (${url})`);
         sendQueuedMsgs();
@@ -214,6 +229,9 @@ class WsRequestor extends BaseRequestor {
         maxRedirects: 2,
         handshakeTimeout,
         maxPayload: JAMBONES_WS_MAX_PAYLOAD ? parseInt(JAMBONES_WS_MAX_PAYLOAD) : 24 * 1024,
+        headers: {
+          ...(HTTP_USER_AGENT_HEADER && {'user-agent' : HTTP_USER_AGENT_HEADER})
+        }
       };
       if (this.username && this.password) opts = {...opts, auth: `${this.username}:${this.password}`};
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jambonz-feature-server",
-  "version": "0.8.4",
+  "version": "0.8.5",
   "main": "app.js",
   "engines": {
     "node": ">= 10.16.0"
@@ -25,13 +25,16 @@
     "jslint:fix": "eslint app.js tracer.js lib --fix"
   },
   "dependencies": {
+    "@aws-sdk/client-auto-scaling": "^3.360.0",
+    "@aws-sdk/client-sns": "^3.360.0",
     "@jambonz/db-helpers": "^0.9.1",
     "@jambonz/http-health-check": "^0.0.1",
-    "@jambonz/realtimedb-helpers": "^0.8.6",
+    "@jambonz/mw-registrar": "^0.2.4",
-    "@jambonz/speech-utils": "^0.0.19",
+    "@jambonz/realtimedb-helpers": "^0.8.7",
-    "@jambonz/stats-collector": "^0.1.8",
+    "@jambonz/speech-utils": "^0.0.33",
+    "@jambonz/stats-collector": "^0.1.9",
     "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.26",
+    "@jambonz/verb-specifications": "^0.0.50",
     "@opentelemetry/api": "^1.4.0",
     "@opentelemetry/exporter-jaeger": "^1.9.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
@@ -41,14 +44,13 @@
     "@opentelemetry/sdk-trace-base": "^1.9.0",
     "@opentelemetry/sdk-trace-node": "^1.9.0",
     "@opentelemetry/semantic-conventions": "^1.9.0",
-    "@aws-sdk/client-sns": "^3.360.0",
-    "@aws-sdk/client-auto-scaling": "^3.360.0",
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.23",
+    "drachtio-fsmrf": "^3.0.33",
-    "drachtio-srf": "^4.5.26",
+    "drachtio-srf": "^4.5.31",
     "express": "^4.18.2",
+    "express-validator": "^7.0.1",
     "ip": "^1.1.8",
     "moment": "^2.29.4",
     "parse-url": "^8.1.0",
@@ -59,11 +61,11 @@
     "short-uuid": "^4.2.2",
     "sinon": "^15.0.1",
     "to-snake-case": "^1.0.0",
-    "undici": "^5.19.1",
+    "undici": "^5.26.2",
     "uuid-random": "^1.3.2",
     "verify-aws-sns-signature": "^0.1.0",
     "ws": "^8.9.0",
-    "xml2js": "^0.5.0"
+    "xml2js": "^0.6.2"
   },
   "devDependencies": {
     "clear-module": "^4.1.2",

test/configuration/squid.conf

@@ -0,0 +1,86 @@
+#
+# Recommended minimum configuration:
+#
+
+# Example rule allowing access from your local networks.
+# Adapt to list your (internal) IP networks from where browsing
+# should be allowed
+acl localnet src 0.0.0.1-0.255.255.255	# RFC 1122 "this" network (LAN)
+acl localnet src 10.0.0.0/8		# RFC 1918 local private network (LAN)
+acl localnet src 100.64.0.0/10		# RFC 6598 shared address space (CGN)
+acl localnet src 169.254.0.0/16 	# RFC 3927 link-local (directly plugged) machines
+acl localnet src 172.16.0.0/12		# RFC 1918 local private network (LAN)
+acl localnet src 172.38.0.0/12		# RFC 1918 local private network (LAN)
+acl localnet src 192.168.0.0/16		# RFC 1918 local private network (LAN)
+acl localnet src fc00::/7       	# RFC 4193 local private network range
+acl localnet src fe80::/10      	# RFC 4291 link-local (directly plugged) machines
+
+acl SSL_ports port 443
+acl Safe_ports port 80		# http
+acl Safe_ports port 21		# ftp
+acl Safe_ports port 443		# https
+acl Safe_ports port 70		# gopher
+acl Safe_ports port 210		# wais
+acl Safe_ports port 1025-65535	# unregistered ports
+acl Safe_ports port 280		# http-mgmt
+acl Safe_ports port 488		# gss-http
+acl Safe_ports port 591		# filemaker
+acl Safe_ports port 777		# multiling http
+
+#
+# Recommended minimum Access Permission configuration:
+#
+# Deny requests to certain unsafe ports
+http_access allow !Safe_ports
+
+# Deny CONNECT to other than secure SSL ports
+http_access allow CONNECT !SSL_ports
+
+# Only allow cachemgr access from localhost
+http_access allow localhost manager
+http_access allow manager
+
+# This default configuration only allows localhost requests because a more
+# permissive Squid installation could introduce new attack vectors into the
+# network by proxying external TCP connections to unprotected services.
+http_access allow localhost
+
+# The two deny rules below are unnecessary in this default configuration
+# because they are followed by a "deny all" rule. However, they may become
+# critically important when you start allowing external requests below them.
+
+# Protect web applications running on the same server as Squid. They often
+# assume that only local users can access them at "localhost" ports.
+http_access allow to_localhost
+
+# Protect cloud servers that provide local users with sensitive info about
+# their server via certain well-known link-local (a.k.a. APIPA) addresses.
+# http_access deny to_linklocal
+
+#
+# INSERT YOUR OWN RULE(S) HERE TO ALLOW ACCESS FROM YOUR CLIENTS
+#
+
+# For example, to allow access from your local networks, you may uncomment the
+# following rule (and/or add rules that match your definition of "local"):
+# http_access allow localnet
+
+# And finally deny all other access to this proxy
+http_access allow all
+
+# Squid normally listens to port 3128
+http_port 3128
+
+# Uncomment and adjust the following to add a disk cache directory.
+#cache_dir ufs /usr/local/var/cache/squid 100 16 256
+
+# Leave coredumps in the first cache dir
+coredump_dir /usr/local/var/cache/squid
+
+#
+# Add any of your own refresh_pattern entries above these.
+#
+refresh_pattern ^ftp:		1440	20%	10080
+refresh_pattern ^gopher:	1440	0%	1440
+refresh_pattern -i (/cgi-bin/|\?) 0	0%	0
+refresh_pattern .		0	20%	4320

test/create-call-test.js

@@ -37,7 +37,7 @@ test('test create-call timeout', async(t) => {
       'account_sid':account_sid,
       'timeout': 1,
       "call_hook": {
-        "url": "https://public-apps.jambonz.us/hello-world",
+        "url": "https://public-apps.jambonz.cloud/hello-world",
         "method": "POST"
       },
       "from": "15083718299",
@@ -88,8 +88,8 @@ test('test create-call call-hook basic authentication', async(t) => {
 
     let verbs = [
       {
-        "verb": "say",
+        "verb": "pause",
-        "text": "hello"
+        "length": 1
       }
     ];
     await provisionCallHook(from, verbs);
@@ -99,6 +99,8 @@ test('test create-call call-hook basic authentication', async(t) => {
     let obj = await getJSON(`http:127.0.0.1:3100/lastRequest/${from}`)
     t.ok(obj.headers.Authorization = 'Basic dXNlcm5hbWU6cGFzc3dvcmQ=',
       'create-call: call-hook contains basic authentication header');
+    t.ok(obj.headers['user-agent'] = 'jambonz',
+    'create-call: call-hook contains user-agent header');
     disconnect();
   } catch (err) {
     console.log(`error received: ${err}`);

test/db/create-and-populate-schema.sql

@@ -330,7 +330,7 @@ CREATE TABLE `applications` (
 
 LOCK TABLES `applications` WRITE;
 /*!40000 ALTER TABLE `applications` DISABLE KEYS */;
-INSERT INTO `applications` VALUES ('0dddaabf-0a30-43e3-84e8-426873b1a78b','decline call',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','c71e79db-24f2-4866-a3ee-febb0f97b341','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('0dddaabf-0a30-43e3-84e8-426873b1a78c','app json',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','c71e79db-24f2-4866-a3ee-febb0f97b341','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,'[{\"verb\": \"play\",\"url\": \"silence_stream://5000\"}]','google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('17461c69-56b5-4dab-ad83-1c43a0f93a3d','gather',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','10692465-a511-4277-9807-b7157e4f81e1','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('195d9507-6a42-46a8-825f-f009e729d023','sip info',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','c9113e7a-741f-48b9-96c1-f2f78176eeb3','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('24d0f6af-e976-44dd-a2e8-41c7b55abe33','say account 2',NULL,'622f62e4-303a-49f2-bbe0-eb1e1714e37a','54ab0976-a6c0-45d8-89a4-d90d45bf9d96','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('308b4f41-1a18-4052-b89a-c054e75ce242','say',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','54ab0976-a6c0-45d8-89a4-d90d45bf9d96','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('ae026ab5-3029-47b4-9d7c-236e3a4b4ebe','transcribe account 2',NULL,'622f62e4-303a-49f2-bbe0-eb1e1714e37a','ecb67a8f-f7ce-4919-abf0-bbc69c1001e5','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('baf9213b-5556-4c20-870c-586392ed246f','transcribe',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','ecb67a8f-f7ce-4919-abf0-bbc69c1001e5','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48');
+INSERT INTO `applications` VALUES ('0dddaabf-0a30-43e3-84e8-426873b1a78b','decline call',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','c71e79db-24f2-4866-a3ee-febb0f97b341','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('0dddaabf-0a30-43e3-84e8-426873b1a78c','app json',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','c71e79db-24f2-4866-a3ee-febb0f97b341','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,'[{\"verb\": \"play\",\"url\": \"silence_stream://5000\"}]','google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('17461c69-56b5-4dab-ad83-1c43a0f93a3d','gather',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','10692465-a511-4277-9807-b7157e4f81e1','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('195d9507-6a42-46a8-825f-f009e729d023','sip info',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','c9113e7a-741f-48b9-96c1-f2f78176eeb3','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('24d0f6af-e976-44dd-a2e8-41c7b55abe33','say account 2',NULL,'622f62e4-303a-49f2-bbe0-eb1e1714e37a','54ab0976-a6c0-45d8-89a4-d90d45bf9d96','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('308b4f41-1a18-4052-b89a-c054e75ce242','say',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','54ab0976-a6c0-45d8-89a4-d90d45bf9d96','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('ae026ab5-3029-47b4-9d7c-236e3a4b4ebe','transcribe account 2',NULL,'622f62e4-303a-49f2-bbe0-eb1e1714e37a','ecb67a8f-f7ce-4919-abf0-bbc69c1001e5','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('baf9213b-5556-4c20-870c-586392ed246f','transcribe',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','ecb67a8f-f7ce-4919-abf0-bbc69c1001e5','293904c1-351b-4bca-8d58-1a29b853c7db',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48'),('0dddaabf-0a30-43e3-84e8-426873b1a7ac','http proxy app',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','c71e79db-24f2-4866-a3ee-febb0f97b3ac','293904c1-351b-4bca-8d58-1a29b853c7ac',NULL,NULL,'google','en-US','en-US-Standard-C','google','en-US','2023-05-31 03:52:48');
 /*!40000 ALTER TABLE `applications` ENABLE KEYS */;
 UNLOCK TABLES;
 
@@ -646,7 +646,7 @@ CREATE TABLE `phone_numbers` (
 
 LOCK TABLES `phone_numbers` WRITE;
 /*!40000 ALTER TABLE `phone_numbers` DISABLE KEYS */;
-INSERT INTO `phone_numbers` VALUES ('05eeed62-b29b-4679-bf38-d7a4e318be44','16174000003','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','17461c69-56b5-4dab-ad83-1c43a0f93a3d',NULL),('4b439355-debc-40c7-9cfa-5be58c2bed6b','16174000000','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','0dddaabf-0a30-43e3-84e8-426873b1a78b',NULL),('964d0581-9627-44cb-be20-8118050406b2','16174000006','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','195d9507-6a42-46a8-825f-f009e729d023',NULL),('964d0581-9627-44cb-be20-8118050406b3','16174000007','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','0dddaabf-0a30-43e3-84e8-426873b1a78c',NULL),('9cc9e7fc-b7b0-4101-8f3c-9fe13ce5df0a','16174000001','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','308b4f41-1a18-4052-b89a-c054e75ce242',NULL),('e686a320-0725-418f-be65-532159bdc3ed','16174000002','5145b436-2f38-4029-8d4c-fd8c67831c7a','622f62e4-303a-49f2-bbe0-eb1e1714e37a','24d0f6af-e976-44dd-a2e8-41c7b55abe33',NULL),('f3c53863-b629-4cf6-9dcb-c7fb7072314b','16174000004','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','baf9213b-5556-4c20-870c-586392ed246f',NULL),('f6416c17-829a-4f11-9c32-f0d00e4a9ae9','16174000005','5145b436-2f38-4029-8d4c-fd8c67831c7a','622f62e4-303a-49f2-bbe0-eb1e1714e37a','ae026ab5-3029-47b4-9d7c-236e3a4b4ebe',NULL);
+INSERT INTO `phone_numbers` VALUES ('05eeed62-b29b-4679-bf38-d7a4e318be44','16174000003','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','17461c69-56b5-4dab-ad83-1c43a0f93a3d',NULL),('4b439355-debc-40c7-9cfa-5be58c2bed6b','16174000000','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','0dddaabf-0a30-43e3-84e8-426873b1a78b',NULL),('964d0581-9627-44cb-be20-8118050406b2','16174000006','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','195d9507-6a42-46a8-825f-f009e729d023',NULL),('964d0581-9627-44cb-be20-8118050406b3','16174000007','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','0dddaabf-0a30-43e3-84e8-426873b1a78c',NULL),('9cc9e7fc-b7b0-4101-8f3c-9fe13ce5df0a','16174000001','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','308b4f41-1a18-4052-b89a-c054e75ce242',NULL),('e686a320-0725-418f-be65-532159bdc3ed','16174000002','5145b436-2f38-4029-8d4c-fd8c67831c7a','622f62e4-303a-49f2-bbe0-eb1e1714e37a','24d0f6af-e976-44dd-a2e8-41c7b55abe33',NULL),('f3c53863-b629-4cf6-9dcb-c7fb7072314b','16174000004','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','baf9213b-5556-4c20-870c-586392ed246f',NULL),('f6416c17-829a-4f11-9c32-f0d00e4a9ae9','16174000005','5145b436-2f38-4029-8d4c-fd8c67831c7a','622f62e4-303a-49f2-bbe0-eb1e1714e37a','ae026ab5-3029-47b4-9d7c-236e3a4b4ebe',NULL),('4b439355-debc-40c7-9cfa-5be58c2bedac','16174000015','5145b436-2f38-4029-8d4c-fd8c67831c7a','bb845d4b-83a9-4cde-a6e9-50f3743bab3f','0dddaabf-0a30-43e3-84e8-426873b1a7ac',NULL);
 /*!40000 ALTER TABLE `phone_numbers` ENABLE KEYS */;
 UNLOCK TABLES;
 
@@ -896,7 +896,7 @@ CREATE TABLE `service_providers` (
 
 LOCK TABLES `service_providers` WRITE;
 /*!40000 ALTER TABLE `service_providers` DISABLE KEYS */;
-INSERT INTO `service_providers` VALUES ('2708b1b3-2736-40ea-b502-c53d8396247f','jambonz.us','jambonz.us service provider','yakeeda.com',NULL,NULL);
+INSERT INTO `service_providers` VALUES ('2708b1b3-2736-40ea-b502-c53d8396247f','jambonz.cloud','jambonz.cloud service provider','yakeeda.com',NULL,NULL);
 /*!40000 ALTER TABLE `service_providers` ENABLE KEYS */;
 UNLOCK TABLES;
 
@@ -1045,6 +1045,7 @@ CREATE TABLE `speech_credentials` (
   `tts_tested_ok` tinyint(1) DEFAULT NULL,
   `stt_tested_ok` tinyint(1) DEFAULT NULL,
   `created_at` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP,
+  `label` VARCHAR(64),
   PRIMARY KEY (`speech_credential_sid`),
   UNIQUE KEY `speech_credential_sid` (`speech_credential_sid`),
   UNIQUE KEY `speech_credentials_idx_1` (`vendor`,`account_sid`),
@@ -1063,7 +1064,7 @@ CREATE TABLE `speech_credentials` (
 
 LOCK TABLES `speech_credentials` WRITE;
 /*!40000 ALTER TABLE `speech_credentials` DISABLE KEYS */;
-INSERT INTO `speech_credentials` VALUES ('2add163c-34f2-45c6-a016-f955d218ffb6',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','google','credential-goes-here',1,1,NULL,'2021-04-03 15:42:10',1,1,'2023-05-31 03:44:21'),('2add347f-34f2-45c6-a016-f955d218ffb6',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','microsoft','credential-goes-here',1,1,NULL,'2021-04-03 15:42:10',1,1,'2023-05-31 03:44:21'),('84154212-5c99-4c94-8993-bc2a46288daa',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','aws','credential-goes-here',1,1,NULL,NULL,NULL,NULL,'2023-05-31 03:44:21');
+INSERT INTO `speech_credentials` VALUES ('2add163c-34f2-45c6-a016-f955d218ffb6',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','google','credential-goes-here',1,1,NULL,'2021-04-03 15:42:10',1,1,'2023-05-31 03:44:21', NULL),('2add347f-34f2-45c6-a016-f955d218ffb6',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','microsoft','credential-goes-here',1,1,NULL,'2021-04-03 15:42:10',1,1,'2023-05-31 03:44:21', NULL),('84154212-5c99-4c94-8993-bc2a46288daa',NULL,'bb845d4b-83a9-4cde-a6e9-50f3743bab3f','aws','credential-goes-here',1,1,NULL,NULL,1,1,'2023-05-31 03:44:21', NULL);
 /*!40000 ALTER TABLE `speech_credentials` ENABLE KEYS */;
 UNLOCK TABLES;
 
@@ -1253,7 +1254,7 @@ CREATE TABLE `webhooks` (
 
 LOCK TABLES `webhooks` WRITE;
 /*!40000 ALTER TABLE `webhooks` DISABLE KEYS */;
-INSERT INTO `webhooks` VALUES ('10692465-a511-4277-9807-b7157e4f81e1','http://127.0.0.1:3102/','POST',NULL,NULL),('293904c1-351b-4bca-8d58-1a29b853c7db','http://127.0.0.1:3100/callStatus','POST',NULL,NULL),('54ab0976-a6c0-45d8-89a4-d90d45bf9d96','http://127.0.0.1:3101/','POST',NULL,NULL),('6ac36aeb-6bd0-428a-80a1-aed95640a296','https://flows.jambonz.us/callStatus','POST',NULL,NULL),('c71e79db-24f2-4866-a3ee-febb0f97b341','http://127.0.0.1:3100/','POST',NULL,NULL),('c9113e7a-741f-48b9-96c1-f2f78176eeb3','http://127.0.0.1:3104/','POST',NULL,NULL),('d9c205c6-a129-443e-a9c0-d1bb437d4bb7','https://flows.jambonz.us/testCall','POST',NULL,NULL),('ecb67a8f-f7ce-4919-abf0-bbc69c1001e5','http://127.0.0.1:3103/','POST',NULL,NULL);
+INSERT INTO `webhooks` VALUES ('10692465-a511-4277-9807-b7157e4f81e1','http://127.0.0.1:3102/','POST',NULL,NULL),('293904c1-351b-4bca-8d58-1a29b853c7db','http://127.0.0.1:3100/callStatus','POST',NULL,NULL),('54ab0976-a6c0-45d8-89a4-d90d45bf9d96','http://127.0.0.1:3101/','POST',NULL,NULL),('6ac36aeb-6bd0-428a-80a1-aed95640a296','https://flows.jambonz.cloud/callStatus','POST',NULL,NULL),('c71e79db-24f2-4866-a3ee-febb0f97b341','http://127.0.0.1:3100/','POST',NULL,NULL),('c9113e7a-741f-48b9-96c1-f2f78176eeb3','http://127.0.0.1:3104/','POST',NULL,NULL),('d9c205c6-a129-443e-a9c0-d1bb437d4bb7','https://flows.jambonz.cloud/testCall','POST',NULL,NULL),('ecb67a8f-f7ce-4919-abf0-bbc69c1001e5','http://127.0.0.1:3103/','POST',NULL,NULL),('293904c1-351b-4bca-8d58-1a29b853c7ac','http://172.38.0.60:3000/callStatus','POST',NULL,NULL),('c71e79db-24f2-4866-a3ee-febb0f97b3ac','http://172.38.0.60:3000/','POST',NULL,NULL);
 /*!40000 ALTER TABLE `webhooks` ENABLE KEYS */;
 UNLOCK TABLES;
 /*!40103 SET TIME_ZONE=@OLD_TIME_ZONE */;

test/db/jambones-sql.sql

@@ -13,6 +13,8 @@ DROP TABLE IF EXISTS beta_invite_codes;
 
 DROP TABLE IF EXISTS call_routes;
 
+DROP TABLE IF EXISTS clients;
+
 DROP TABLE IF EXISTS dns_records;
 
 DROP TABLE IF EXISTS lcr;
@@ -127,6 +129,16 @@ application_sid CHAR(36) NOT NULL,
 PRIMARY KEY (call_route_sid)
 ) COMMENT='a regex-based pattern match for call routing';
 
+CREATE TABLE clients
+(
+client_sid CHAR(36) NOT NULL UNIQUE ,
+account_sid CHAR(36) NOT NULL,
+is_active BOOLEAN NOT NULL DEFAULT 1,
+username VARCHAR(64),
+password VARCHAR(1024),
+PRIMARY KEY (client_sid)
+);
+
 CREATE TABLE dns_records
 (
 dns_record_sid CHAR(36) NOT NULL UNIQUE ,
@@ -322,6 +334,7 @@ last_tested DATETIME,
 tts_tested_ok BOOLEAN,
 stt_tested_ok BOOLEAN,
 created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+label VARCHAR(64),
 PRIMARY KEY (speech_credential_sid)
 );
 
@@ -411,7 +424,7 @@ PRIMARY KEY (smpp_gateway_sid)
 CREATE TABLE phone_numbers
 (
 phone_number_sid CHAR(36) UNIQUE ,
-number VARCHAR(132) NOT NULL UNIQUE ,
+number VARCHAR(132) NOT NULL,
 voip_carrier_sid CHAR(36),
 account_sid CHAR(36),
 application_sid CHAR(36),
@@ -466,9 +479,12 @@ 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_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),
 created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+record_all_calls BOOLEAN NOT NULL DEFAULT false,
 PRIMARY KEY (application_sid)
 ) COMMENT='A defined set of behaviors to be applied to phone calls ';
 
@@ -506,6 +522,9 @@ subspace_client_secret VARCHAR(255),
 subspace_sip_teleport_id VARCHAR(255),
 subspace_sip_teleport_destinations VARCHAR(255),
 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',
 PRIMARY KEY (account_sid)
 ) COMMENT='An enterprise that uses the platform for comm services';
 
@@ -526,6 +545,9 @@ ALTER TABLE call_routes ADD FOREIGN KEY account_sid_idxfk_3 (account_sid) REFERE
 
 ALTER TABLE call_routes ADD FOREIGN KEY application_sid_idxfk (application_sid) REFERENCES applications (application_sid);
 
+CREATE INDEX client_sid_idx ON clients (client_sid);
+ALTER TABLE clients ADD CONSTRAINT account_sid_idxfk_13 FOREIGN KEY account_sid_idxfk_13 (account_sid) REFERENCES accounts (account_sid);
+
 CREATE INDEX dns_record_sid_idx ON dns_records (dns_record_sid);
 ALTER TABLE dns_records ADD FOREIGN KEY account_sid_idxfk_4 (account_sid) REFERENCES accounts (account_sid);
 
@@ -590,8 +612,6 @@ CREATE INDEX smpp_address_sid_idx ON smpp_addresses (smpp_address_sid);
 CREATE INDEX service_provider_sid_idx ON smpp_addresses (service_provider_sid);
 ALTER TABLE smpp_addresses ADD FOREIGN KEY service_provider_sid_idxfk_4 (service_provider_sid) REFERENCES service_providers (service_provider_sid);
 
-CREATE UNIQUE INDEX speech_credentials_idx_1 ON speech_credentials (vendor,account_sid);
-
 CREATE INDEX speech_credential_sid_idx ON speech_credentials (speech_credential_sid);
 CREATE INDEX service_provider_sid_idx ON speech_credentials (service_provider_sid);
 ALTER TABLE speech_credentials ADD FOREIGN KEY service_provider_sid_idxfk_5 (service_provider_sid) REFERENCES service_providers (service_provider_sid);
@@ -628,6 +648,8 @@ CREATE INDEX smpp_gateway_sid_idx ON smpp_gateways (smpp_gateway_sid);
 CREATE INDEX voip_carrier_sid_idx ON smpp_gateways (voip_carrier_sid);
 ALTER TABLE smpp_gateways ADD FOREIGN KEY voip_carrier_sid_idxfk (voip_carrier_sid) REFERENCES voip_carriers (voip_carrier_sid);
 
+CREATE UNIQUE INDEX phone_numbers_unique_idx_voip_carrier_number ON phone_numbers (number,voip_carrier_sid);
+
 CREATE INDEX phone_number_sid_idx ON phone_numbers (phone_number_sid);
 CREATE INDEX number_idx ON phone_numbers (number);
 CREATE INDEX voip_carrier_sid_idx ON phone_numbers (voip_carrier_sid);
@@ -682,5 +704,4 @@ 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/docker-compose-testbed.yaml

@@ -42,7 +42,7 @@ services:
         ipv4_address: 172.38.0.7
 
   drachtio:
-    image: drachtio/drachtio-server:0.8.22
+    image: drachtio/drachtio-server:0.8.24
     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:0.4.33
+    image: drachtio/drachtio-freeswitch-mrf:0.6.1
     restart: always
     command: freeswitch --rtp-range-start 20000 --rtp-range-end 20100
     environment:
@@ -92,3 +92,13 @@ services:
     networks:
       fs:
         ipv4_address: 172.38.0.90
+
+  squid:
+    image: ubuntu/squid:edge
+    ports:
+      - "3128:3128"
+    volumes:
+      - ./configuration/squid.conf:/etc/squid/squid.conf
+    networks:
+      fs:
+        ipv4_address: 172.38.0.91

test/hangup-test.js

@@ -0,0 +1,53 @@
+const test = require('tape');
+const { sippUac } = require('./sipp')('test_fs');
+const clearModule = require('clear-module');
+const {provisionCallHook, provisionCustomHook} = require('./utils')
+const bent = require('bent');
+const getJSON = bent('json')
+
+process.on('unhandledRejection', (reason, p) => {
+  console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
+});
+
+function connect(connectable) {
+  return new Promise((resolve, reject) => {
+    connectable.on('connect', () => {
+      return resolve();
+    });
+  });
+}
+
+test('\'hangup\' custom headers', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const verbs = [
+      {
+        verb: 'play',
+        url: 'https://example.com/example.mp3'
+      },
+      {
+        "verb": "hangup",
+        "headers": {
+          "X-Reason" : "maximum call duration exceeded"
+        }
+      }
+    ];
+
+    const from = 'hangup_custom_headers';
+    await provisionCallHook(from, verbs)
+
+    // THEN
+    await sippUac('uac-success-received-bye.xml', '172.38.0.10', from);
+    t.pass('play: succeeds when using single link');
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});

test/http-proxy-test.js

@@ -0,0 +1,72 @@
+const test = require('tape');
+const { sippUac } = require('./sipp')('test_fs');
+const clearModule = require('clear-module');
+const {provisionCallHook, provisionCustomHook} = require('./utils')
+const bent = require('bent');
+const getJSON = bent('json')
+
+process.on('unhandledRejection', (reason, p) => {
+  console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
+});
+
+function connect(connectable) {
+  return new Promise((resolve, reject) => {
+    connectable.on('connect', () => {
+      return resolve();
+    });
+  });
+}
+
+
+test('\'HTTP proxy\' test Info', async(t) => {
+  clearModule.all();
+  process.env.JAMBONES_HTTP_PROXY_IP = "127.0.0.1";
+  process.env.JAMBONES_HTTP_PROXY_PROTOCOL = "http";
+  process.env.JAMBONES_HTTP_PROXY_PORT = 3128;
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const verbs = [
+      {
+        verb: 'config',
+        sipRequestWithinDialogHook: '/customHook'
+      },
+      {
+        verb: 'play',
+        url: 'silence_stream://5000',
+      }
+    ];
+
+    const waitHookVerbs = [
+      {
+        verb: 'hangup'
+      }
+    ];
+
+    const from = 'http_proxy_info';
+    await provisionCustomHook(from, waitHookVerbs)
+    await provisionCallHook(from, verbs);
+    
+
+    // THEN
+    await sippUac('uac-success-info-received-bye.xml', '172.38.0.10', from, "16174000015");
+    t.pass('sip Info: success send Info');
+
+    // Make sure that sipRequestWithinDialogHook is called and success
+    const json = await getJSON(`http:127.0.0.1:3100/lastRequest/${from}_customHook`)
+    t.pass(json.body.sip_method === 'INFO', 'sipRequestWithinDialogHook contains sip_method')
+    t.pass(json.body.sip_body === 'hello jambonz\r\n', 'sipRequestWithinDialogHook contains sip_method')
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  } finally {
+    process.env.JAMBONES_HTTP_PROXY_IP = null;
+    process.env.JAMBONES_HTTP_PROXY_PROTOCOL = null;
+    process.env.JAMBONES_HTTP_PROXY_PORT = null;
+  }
+});

test/in-dialog-test.js

@@ -0,0 +1,65 @@
+const test = require('tape');
+const { sippUac } = require('./sipp')('test_fs');
+const clearModule = require('clear-module');
+const {provisionCallHook, provisionCustomHook} = require('./utils')
+const bent = require('bent');
+const getJSON = bent('json')
+
+process.on('unhandledRejection', (reason, p) => {
+  console.log('Unhandled Rejection at: Promise', p, 'reason:', reason);
+});
+
+function connect(connectable) {
+  return new Promise((resolve, reject) => {
+    connectable.on('connect', () => {
+      return resolve();
+    });
+  });
+}
+
+
+test('\'sip Indialog\' test Info', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const verbs = [
+      {
+        verb: 'config',
+        sipRequestWithinDialogHook: '/customHook'
+      },
+      {
+        verb: 'play',
+        url: 'silence_stream://5000',
+      }
+    ];
+
+    const waitHookVerbs = [
+      {
+        verb: 'hangup'
+      }
+    ];
+
+    const from = 'sip_indialog_info';
+    await provisionCustomHook(from, waitHookVerbs)
+    await provisionCallHook(from, verbs);
+    
+
+    // THEN
+    await sippUac('uac-success-info-received-bye.xml', '172.38.0.10', from);
+    t.pass('sip Info: success send Info');
+
+    // Make sure that sipRequestWithinDialogHook is called and success
+    const json = await getJSON(`http:127.0.0.1:3100/lastRequest/${from}_customHook`)
+    t.pass(json.body.sip_method === 'INFO', 'sipRequestWithinDialogHook contains sip_method')
+    t.pass(json.body.sip_body === 'hello jambonz\r\n', 'sipRequestWithinDialogHook contains sip_method')
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});

test/index.js

@@ -15,5 +15,9 @@ require('./sip-refer-tests');
 require('./listen-tests');
 require('./config-test');
 require('./queue-test');
+require('./in-dialog-test');
+require('./hangup-test');
+require('./sdp-utils-test');
+require('./http-proxy-test');
 require('./remove-test-db');
 require('./docker_stop');

test/play-tests.js

@@ -188,7 +188,7 @@ test('\'play\' tests with seekOffset and actionHook', async(t) => {
     const seconds = parseInt(obj.body.playback_seconds);
     const milliseconds = parseInt(obj.body.playback_milliseconds);
     const lastOffsetPos = parseInt(obj.body.playback_last_offset_pos);
-    //console.log({obj}, 'lastRequest');
+    console.log({obj}, 'lastRequest');
     t.ok(obj.body.reason === "playCompleted", "play: actionHook success received");
     t.ok(seconds === 2, "playback_seconds: actionHook success received");
     t.ok(milliseconds === 2048, "playback_milliseconds: actionHook success received");

test/say-tests.js

@@ -84,6 +84,46 @@ test('\'config\' reset synthesizer tests', async(t) => {
   }
 });
 
+test('Say verb array test', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const verbs = [
+      {
+        "verb": "config",
+        "synthesizer": {
+          "vendor": "microsft",
+          "voice": "foobar"
+        },
+      },
+      {
+        "verb": "config",
+        "reset": 'synthesizer',
+      },
+      {
+        verb: 'say',
+        text: ['hello', 'https://samplelib.com/lib/preview/mp3/sample-3s.mp3']
+      }
+    ];
+
+    const from = 'say_test_success';
+    await provisionCallHook(from, verbs)
+
+    // THEN
+    await sippUac('uac-success-received-bye.xml', '172.38.0.10', from);
+    t.pass('say: succeeds when using using account credentials');
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
 const {MICROSOFT_CUSTOM_API_KEY, MICROSOFT_DEPLOYMENT_ID, MICROSOFT_CUSTOM_REGION, MICROSOFT_CUSTOM_VOICE} = process.env;
 if (MICROSOFT_CUSTOM_API_KEY && MICROSOFT_DEPLOYMENT_ID && MICROSOFT_CUSTOM_REGION && MICROSOFT_CUSTOM_VOICE) {
   test('\'say\' tests - microsoft custom voice', async(t) => {

test/scenarios/uac-success-info-received-bye.xml

@@ -0,0 +1,114 @@
+<?xml version="1.0" encoding="ISO-8859-1" ?>
+<!DOCTYPE scenario SYSTEM "sipp.dtd">
+
+
+<scenario name="Basic Sipstone UAC">
+  <!-- In client mode (sipp placing calls), the Call-ID MUST be         -->
+  <!-- generated by sipp. To do so, use [call_id] keyword.                -->
+  <send retrans="500">
+    <![CDATA[
+
+      INVITE sip:[to]@[remote_ip]:[remote_port] SIP/2.0
+      Via: SIP/2.0/[transport] [local_ip]:[local_port];branch=[branch]
+      From: [from] <sip:[from]@[local_ip]:[local_port]>;tag=[pid]SIPpTag00[call_number]
+      To: <sip:[to]@[remote_ip]:[remote_port]>
+      Call-ID: [call_id]
+      CSeq: 1 INVITE
+      Contact: sip:[from]@[local_ip]:[local_port]
+      Max-Forwards: 70
+      X-Account-Sid: bb845d4b-83a9-4cde-a6e9-50f3743bab3f
+      Subject: uac-say
+      Content-Type: application/sdp
+      Content-Length: [len]
+
+      v=0
+      o=user1 53655765 2353687637 IN IP[local_ip_type] [local_ip]
+      s=-
+      c=IN IP[media_ip_type] [media_ip]
+      t=0 0
+      m=audio [media_port] RTP/AVP 0
+      a=rtpmap:0 PCMU/8000
+
+    ]]>
+  </send>
+
+  <recv response="100"
+        optional="true">
+  </recv>
+
+  <recv response="180" optional="true">
+  </recv>
+
+  <recv response="183" optional="true">
+  </recv>
+
+  <!-- By adding rrs="true" (Record Route Sets), the route sets         -->
+  <!-- are saved and used for following messages sent. Useful to test   -->
+  <!-- against stateful SIP proxies/B2BUAs.                             -->
+  <recv response="200" rtd="true">
+  </recv>
+
+  <!-- Packet lost can be simulated in any send/recv message by         -->
+  <!-- by adding the 'lost = "10"'. Value can be [1-100] percent.       -->
+  <send>
+    <![CDATA[
+
+      ACK sip:[to]@[remote_ip]:[remote_port] SIP/2.0
+      Via: SIP/2.0/[transport] [local_ip]:[local_port];branch=[branch]
+      From: [from] <sip:[from]@[local_ip]:[local_port]>;tag=[pid]SIPpTag00[call_number]
+      To: [to] <sip:[to]@[remote_ip]:[remote_port]>[peer_tag_param]
+      Call-ID: [call_id]
+      CSeq: 1 ACK
+      Contact: sip:[from]@[local_ip]:[local_port]
+      Max-Forwards: 70
+      Subject: uac-say
+      Content-Length: 0
+
+    ]]>
+  </send>
+
+   <pause milliseconds="2000"/>
+
+  <!-- Send an INFO message -->
+  <send>
+    <![CDATA[
+      INFO sip:[service]@[remote_ip]:[remote_port] SIP/2.0
+      Via: SIP/2.0/[transport] [local_ip]:[local_port];branch=[branch]
+      From: [from] <sip:[from]@[local_ip]:[local_port]>;tag=[pid]SIPpTag00[call_number]
+      To: [to] <sip:[to]@[remote_ip]:[remote_port]>[peer_tag_param]
+      Call-ID: [call_id]
+      CSeq: 2 INFO
+      Contact: sip:[from]@[local_ip]:[local_port]
+      Max-Forwards: 70
+      Subject: Performance Test
+      Content-Type: text/plain
+      Content-Length: [len]
+
+      hello jambonz
+    ]]>
+  </send>
+
+  <!-- Receive 200 OK -->
+  <recv response="200">
+  </recv>
+
+  <recv request="BYE">
+  </recv>
+
+  <send>
+    <![CDATA[
+
+      SIP/2.0 200 OK
+      [last_Via:]
+      [last_From:]
+      [last_To:]
+      [last_Call-ID:]
+      [last_CSeq:]
+      Contact: <sip:[local_ip]:[local_port];transport=[transport]>
+      Content-Length: 0
+
+    ]]>
+  </send>
+
+</scenario>
+

test/sdp-utils-test.js

@@ -0,0 +1,26 @@
+const test = require('tape');
+const {makeOpusFirst, isOpusFirst} = require('../lib/utils/sdp-utils');
+const sdpTransform = require('sdp-transform');
+
+test('test opus first', (t) => {
+    const sdp = 'v=0\r\no=- 3348584794228993675 2 IN IP4 127.0.0.1\r\ns=-\r\nt=0 0\r\na=group:BUNDLE 0\r\na=extmap-allow-mixed\r\na=msid-semantic: WMS caca8b77-5ae5-4e73-a4d5-de1fce930335\r\nm=audio 57088 UDP/TLS/RTP/SAVPF 111 63 9 0 8 13 110 126\r\nc=IN IP4 14.238.89.50\r\na=rtcp:9 IN IP4 0.0.0.0\r\na=candidate:1401281302 1 udp 2122260223 10.231.36.146 57088 typ host generation 0 network-id 1 network-cost 10\r\na=candidate:2173263513 1 udp 1686052607 14.238.89.50 57088 typ srflx raddr 10.231.36.146 rport 57088 generation 0 network-id 1 network-cost 10\r\na=ice-ufrag:k5nc\r\na=ice-pwd:J0qwMs6HrIcFNZbDG5m8Kqpk\r\na=ice-options:trickle\r\na=fingerprint:sha-256 66:DE:9A:76:CE:11:2D:65:C4:08:C7:87:B4:90:7E:F1:8D:07:B9:F4:FF:E3:81:D7:E7:7D:C6:56:47:01:6E:55\r\na=setup:actpass\r\na=mid:0\r\na=extmap:1 urn:ietf:params:rtp-hdrext:ssrc-audio-level\r\na=extmap:2 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time\r\na=extmap:3 http://www.ietf.org/id/draft-holmer-rmcat-transport-wide-cc-extensions-01\r\na=extmap:4 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=sendrecv\r\na=msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\na=rtcp-mux\r\na=rtcp-fb:111 transport-cc\r\na=fmtp:111 minptime=10;useinbandfec=1\r\na=rtpmap:63 red/48000/2\r\na=fmtp:63 111/111\r\na=rtpmap:9 G722/8000\r\na=rtpmap:0 PCMU/8000\r\na=rtpmap:8 PCMA/8000\r\na=rtpmap:13 CN/8000\r\na=rtpmap:111 opus/48000/2\r\na=rtpmap:110 telephone-event/48000\r\na=rtpmap:126 telephone-event/8000\r\na=ssrc:3207459321 cname:4nyPJ6KXvseBUIhu\r\na=ssrc:3207459321 msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\n';
+    const opusSdp = makeOpusFirst(sdp);
+    const parsedSdp = sdpTransform.parse(opusSdp);
+    const opusIndex = parsedSdp.media[0].rtp.findIndex((entry) => entry.codec === 'opus');
+    t.ok(opusIndex === 0, 'succesffuly move opus to be first offer')
+    t.end();
+});
+
+
+test('test is opus first', (t) => {
+
+    const sdp = 'v=0\r\no=- 3348584794228993675 2 IN IP4 127.0.0.1\r\ns=-\r\nt=0 0\r\na=group:BUNDLE 0\r\na=extmap-allow-mixed\r\na=msid-semantic: WMS caca8b77-5ae5-4e73-a4d5-de1fce930335\r\nm=audio 57088 UDP/TLS/RTP/SAVPF 111 63 9 0 8 13 110 126\r\nc=IN IP4 14.238.89.50\r\na=rtcp:9 IN IP4 0.0.0.0\r\na=candidate:1401281302 1 udp 2122260223 10.231.36.146 57088 typ host generation 0 network-id 1 network-cost 10\r\na=candidate:2173263513 1 udp 1686052607 14.238.89.50 57088 typ srflx raddr 10.231.36.146 rport 57088 generation 0 network-id 1 network-cost 10\r\na=ice-ufrag:k5nc\r\na=ice-pwd:J0qwMs6HrIcFNZbDG5m8Kqpk\r\na=ice-options:trickle\r\na=fingerprint:sha-256 66:DE:9A:76:CE:11:2D:65:C4:08:C7:87:B4:90:7E:F1:8D:07:B9:F4:FF:E3:81:D7:E7:7D:C6:56:47:01:6E:55\r\na=setup:actpass\r\na=mid:0\r\na=extmap:1 urn:ietf:params:rtp-hdrext:ssrc-audio-level\r\na=extmap:2 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time\r\na=extmap:3 http://www.ietf.org/id/draft-holmer-rmcat-transport-wide-cc-extensions-01\r\na=extmap:4 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=sendrecv\r\na=msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\na=rtcp-mux\r\na=rtpmap:111 opus/48000/2\r\na=rtcp-fb:111 transport-cc\r\na=fmtp:111 minptime=10;useinbandfec=1\r\na=rtpmap:63 red/48000/2\r\na=fmtp:63 111/111\r\na=rtpmap:9 G722/8000\r\na=rtpmap:0 PCMU/8000\r\na=rtpmap:8 PCMA/8000\r\na=rtpmap:13 CN/8000\r\na=rtpmap:110 telephone-event/48000\r\na=rtpmap:126 telephone-event/8000\r\na=ssrc:3207459321 cname:4nyPJ6KXvseBUIhu\r\na=ssrc:3207459321 msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\n';
+    t.ok(isOpusFirst(sdp), "opus is first offer");
+
+    const sdp2 = 'v=0\r\no=- 3348584794228993675 2 IN IP4 127.0.0.1\r\ns=-\r\nt=0 0\r\na=group:BUNDLE 0\r\na=extmap-allow-mixed\r\na=msid-semantic: WMS caca8b77-5ae5-4e73-a4d5-de1fce930335\r\nm=audio 57088 UDP/TLS/RTP/SAVPF 111 63 9 0 8 13 110 126\r\nc=IN IP4 14.238.89.50\r\na=rtcp:9 IN IP4 0.0.0.0\r\na=candidate:1401281302 1 udp 2122260223 10.231.36.146 57088 typ host generation 0 network-id 1 network-cost 10\r\na=candidate:2173263513 1 udp 1686052607 14.238.89.50 57088 typ srflx raddr 10.231.36.146 rport 57088 generation 0 network-id 1 network-cost 10\r\na=ice-ufrag:k5nc\r\na=ice-pwd:J0qwMs6HrIcFNZbDG5m8Kqpk\r\na=ice-options:trickle\r\na=fingerprint:sha-256 66:DE:9A:76:CE:11:2D:65:C4:08:C7:87:B4:90:7E:F1:8D:07:B9:F4:FF:E3:81:D7:E7:7D:C6:56:47:01:6E:55\r\na=setup:actpass\r\na=mid:0\r\na=extmap:1 urn:ietf:params:rtp-hdrext:ssrc-audio-level\r\na=extmap:2 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time\r\na=extmap:3 http://www.ietf.org/id/draft-holmer-rmcat-transport-wide-cc-extensions-01\r\na=extmap:4 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=sendrecv\r\na=msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\na=rtcp-mux\r\na=rtcp-fb:111 transport-cc\r\na=fmtp:111 minptime=10;useinbandfec=1\r\na=rtpmap:63 red/48000/2\r\na=fmtp:63 111/111\r\na=rtpmap:9 G722/8000\r\na=rtpmap:0 PCMU/8000\r\na=rtpmap:8 PCMA/8000\r\na=rtpmap:13 CN/8000\r\na=rtpmap:111 opus/48000/2\r\na=rtpmap:110 telephone-event/48000\r\na=rtpmap:126 telephone-event/8000\r\na=ssrc:3207459321 cname:4nyPJ6KXvseBUIhu\r\na=ssrc:3207459321 msid:caca8b77-5ae5-4e73-a4d5-de1fce930335 52ad01f1-b1df-4b8e-a208-9201e98b6f7b\r\n';
+    t.ok(!isOpusFirst(sdp2), "opus is not first offer")
+
+    const sdp3 = 'v=0\r\no=xhoaluu2 1314 1504 IN IP4 192.168.1.4\r\ns=Talk\r\nc=IN IP4 192.168.1.4\r\nt=0 0\r\na=ice-pwd:397d063ea23fdc05164e3ee4\r\na=ice-ufrag:16c449a3\r\na=rtcp-xr:rcvr-rtt=all:10000 stat-summary=loss,dup,jitt,TTL voip-metrics\r\na=group:BUNDLE as\r\na=record:off\r\nm=audio 56542 RTP/AVPF 0 8\r\nc=IN IP4 14.226.233.151\r\na=rtcp-mux\r\na=mid:as\r\na=extmap:1 urn:ietf:params:rtp-hdrext:sdes:mid\r\na=rtcp:63076 IN IP4 192.168.1.4\r\na=candidate:1 1 UDP 2130706303 192.168.1.4 56542 typ host\r\na=candidate:1 2 UDP 2130706302 192.168.1.4 63076 typ host\r\na=candidate:2 1 UDP 2130706431 2001:ee0:d744:dcf0:c1d3:d73f:7a93:dc9f 56542 typ host\r\na=candidate:2 2 UDP 2130706430 2001:ee0:d744:dcf0:c1d3:d73f:7a93:dc9f 63076 typ host\r\na=candidate:3 1 UDP 2130706431 2001:ee0:d744:dcf0:15:6be3:8e6b:b736 56542 typ host\r\na=candidate:3 2 UDP 2130706430 2001:ee0:d744:dcf0:15:6be3:8e6b:b736 63076 typ host\r\na=candidate:4 1 UDP 1694498687 14.226.233.151 56542 typ srflx raddr 192.168.1.4 rport 56542\r\na=rtcp-fb:* trr-int 1000\r\na=rtcp-fb:* ccm tmmbr';
+    t.ok(!isOpusFirst(sdp2), "opus is not first offer")
+    t.end();
+});

test/transcribe-tests.js

@@ -52,6 +52,7 @@ test('\'transcribe\' test - google', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using google credentials');
 
@@ -89,6 +90,7 @@ test('\'transcribe\' test - microsoft', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using  microsoft credentials');
 
@@ -126,6 +128,7 @@ test('\'transcribe\' test - aws', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using aws credentials');
 
@@ -137,6 +140,71 @@ test('\'transcribe\' test - aws', async(t) => {
   }
 });
 
+
+test('\'transcribe\' test - deepgram config options', async(t) => {
+  if (!DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "config",
+        "recognizer": {
+          "vendor": "deepgram",
+          "language": "en-US",
+          "altLanguages": [
+            "en-US"
+          ],
+          "deepgramOptions": {
+            "model": "2-ea",
+            "tier": "nova",
+            "numerals": true,
+            "ner": true,
+            "vadTurnoff": 10,
+            "keywords": [
+              "CPT"
+            ]
+          }
+        }
+      },
+      {
+        "verb": "transcribe",
+        "transcriptionHook": "/transcriptionHook",
+        "recognizer": {
+          "vendor": "deepgram",
+          "altLanguages": [
+            "en-AU"
+          ],
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "deepgramOptions": {
+            "apiKey": DEEPGRAM_API_KEY,
+          }
+        }
+      }
+    ];
+    let from = "gather_success";
+    await provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
+      'transcribe: succeeds when using deepgram credentials');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
 test('\'transcribe\' test - deepgram', async(t) => {
   if (!DEEPGRAM_API_KEY ) {
     t.pass('skipping deepgram tests');
@@ -166,6 +234,7 @@ test('\'transcribe\' test - deepgram', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
       'transcribe: succeeds when using deepgram credentials');
 
@@ -210,6 +279,168 @@ test('\'transcribe\' test - soniox', async(t) => {
     t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using soniox credentials');
 
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+test('\'transcribe\' test - google with asrTimeout', async(t) => {
+  if (!GCP_JSON_KEY) {
+    t.pass('skipping google tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "transcribe",
+        "recognizer": {
+          "vendor": "google",
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "asrTimeout": 4
+        },
+        "transcriptionHook": "/transcriptionHook"
+      }
+    ];
+    let from = "gather_success";
+    await provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
+      'transcribe: succeeds when using google credentials');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+test('\'transcribe\' test - deepgram config options altLanguages', async(t) => {
+  if (!DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "config",
+        "recognizer": {
+          "vendor": "deepgram",
+          "language": "en-US",
+          "altLanguages": [
+            "en-US"
+          ],
+          "deepgramOptions": {
+            "model": "2-ea",
+            "tier": "nova",
+            "numerals": true,
+            "ner": true,
+            "vadTurnoff": 10,
+            "keywords": [
+              "CPT"
+            ]
+          }
+        }
+      },
+      {
+        "verb": "transcribe",
+        "transcriptionHook": "/transcriptionHook",
+        "recognizer": {
+          "vendor": "deepgram",
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "deepgramOptions": {
+            "apiKey": DEEPGRAM_API_KEY,
+          }
+        }
+      }
+    ];
+    let from = "gather_success_no_altLanguages";
+    await provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
+      'transcribe: succeeds when using deepgram credentials');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+test('\'transcribe\' test - deepgram config options altLanguages', async(t) => {
+  if (!DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "config",
+        "recognizer": {
+          "vendor": "deepgram",
+          "language": "en-US",
+          "altLanguages": [
+            "en-US"
+          ],
+          "deepgramOptions": {
+            "model": "2-ea",
+            "tier": "nova",
+            "numerals": true,
+            "ner": true,
+            "vadTurnoff": 10,
+            "keywords": [
+              "CPT"
+            ]
+          }
+        }
+      },
+      {
+        "verb": "transcribe",
+        "transcriptionHook": "/transcriptionHook",
+        "recognizer": {
+          "vendor": "deepgram",
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "altLanguages": [],
+          "deepgramOptions": {
+            "apiKey": DEEPGRAM_API_KEY,
+          }
+        }
+      }
+    ];
+    let from = "gather_success_has_altLanguages";
+    await provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
+      'transcribe: succeeds when using deepgram credentials');
+
     disconnect();
   } catch (err) {
     console.log(`error received: ${err}`);