fix: transcribe 2 channels (#479)

* fix: transcribe 2 channels

* fix

* fix

.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

@@ -37,6 +37,9 @@ 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|
 
 ### 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:

lib/config.js

@@ -144,9 +144,10 @@ const JAMBONES_REDIS_SENTINELS = process.env.JAMBONES_REDIS_SENTINELS ? {
     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;
+const JAMBONZ_RECORD_WS_BASE_URL = process.env.JAMBONZ_RECORD_WS_BASE_URL || process.env.JAMBONES_RECORD_WS_BASE_URL;
+const JAMBONZ_RECORD_WS_USERNAME = process.env.JAMBONZ_RECORD_WS_USERNAME || process.env.JAMBONES_RECORD_WS_USERNAME;
+const JAMBONZ_RECORD_WS_PASSWORD = process.env.JAMBONZ_RECORD_WS_PASSWORD || process.env.JAMBONES_RECORD_WS_PASSWORD;
+const JAMBONZ_DISABLE_DIAL_PAI_HEADER = process.env.JAMBONZ_DISABLE_DIAL_PAI_HEADER || false;
 
 module.exports = {
   JAMBONES_MYSQL_HOST,
@@ -225,5 +226,6 @@ 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
 };

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

@@ -10,6 +10,7 @@ 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');
 
 router.post('/', async(req, res) => {
   const {logger} = req.app.locals;
@@ -127,16 +128,33 @@ router.post('/', async(req, res) => {
 
     /* launch outdial */
     let sdp, sipLogger;
+    let dualEp;
+    let localSdp = ep.local.sdp;
+
+    if (req.body.dual_streams) {
+      dualEp = await ms.createEndpoint();
+      localSdp = mergeSdpMedia(localSdp, dualEp.local.sdp);
+    }
+
     const connectStream = async(remoteSdp) => {
       if (remoteSdp !== sdp) {
-        ep.modify(sdp = remoteSdp);
+        sdp = remoteSdp;
+        if (req.body.dual_streams) {
+          const [sdpLegA, sdpLebB] = extractSdpMedia(remoteSdp);
+
+          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: ep.local.sdp
+      localSdp
     });
     if (target.auth) opts.auth = target.auth;
 
@@ -223,6 +241,7 @@ router.post('/', async(req, res) => {
             srf,
             req: inviteReq,
             ep,
+            ep2: dualEp,
             tasks,
             callInfo,
             accountInfo,
@@ -275,7 +294,10 @@ router.post('/', async(req, res) => {
         else console.error(err);
       }
       ep.destroy();
-      setTimeout(restDial.kill.bind(restDial), 5000);
+      if (dualEp) {
+        dualEp.destroy();
+      }
+      setTimeout(restDial.kill.bind(restDial, cs), 5000);
     }
   } catch (err) {
     sysError(logger, res, err);

lib/middleware.js

@@ -299,22 +299,30 @@ 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_synthesis_label && {label: app.speech_synthesis_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: {
-              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
-              }
-            }
-          });
+          { defaults });
         logger.debug({ params }, 'sending initial webhook');
         const obj = rootSpan.startChildSpan('performAppWebhook');
         span = obj.span;

lib/session/call-session.js

@@ -414,6 +414,14 @@ class CallSession extends Emitter {
     this._onHoldMusic = url;
   }
 
+  get sipRequestWithinDialogHook() {
+    return this._sipRequestWithinDialogHook;
+  }
+
+  set sipRequestWithinDialogHook(url) {
+    this._sipRequestWithinDialogHook = url;
+  }
+
   hasGlobalSttPunctuation() {
     return this._globalSttPunctuation !== undefined;
   }
@@ -731,8 +739,8 @@ class CallSession extends Emitter {
         (type === 'tts' && credential.use_for_tts) ||
         (type === 'stt' && credential.use_for_stt)
       )) {
-        this.logger.info(`Speech credential vendor: ${credential.vendor} 
-${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
+        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) {
@@ -817,6 +825,12 @@ ${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
             riva_server_uri: credential.riva_server_uri
           };
         }
+        else if ('cobalt' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            cobalt_server_uri: credential.cobalt_server_uri
+          };
+        }
         else if (vendor.startsWith('custom:')) {
           return {
             speech_credential_sid: credential.speech_credential_sid,
@@ -1640,10 +1654,32 @@ ${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
       }
       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};
+    this.currentTask.performHook(this, this.sipRequestWithinDialogHook, params);
+  }
+
   async _onReinvite(req, res) {
     try {
       if (this.ep) {
@@ -1927,17 +1963,22 @@ ${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
 
   async enableRecordAllCall() {
     if (this.accountInfo.account.record_all_calls || this.application.record_all_calls) {
+      if (!JAMBONZ_RECORD_WS_BASE_URL || !this.accountInfo.account.bucket_credential) {
+        this.logger.error('Record all calls: invalid configuration');
+        return;
+      }
       const listenOpts = {
         url: `${JAMBONZ_RECORD_WS_BASE_URL}/record/${this.accountInfo.account.bucket_credential.vendor}`,
-        wsAuth: {
-          username: JAMBONZ_RECORD_WS_USERNAME,
-          password: JAMBONZ_RECORD_WS_PASSWORD
-        },
         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}, 'Record all calls: enabling listen');
       await this.startBackgroundListen({verb: 'listen', ...listenOpts}, 'jambonz-session-record');
     }

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];

lib/tasks/config.js

@@ -216,6 +216,9 @@ class TaskConfig extends Task {
         cs.stopBackgroundListen();
       }
     }
+    if (this.data.sipRequestWithinDialogHook) {
+      cs.sipRequestWithinDialogHook = this.data.sipRequestWithinDialogHook;
+    }
   }
 
   async kill(cs) {

lib/tasks/dial.js

@@ -16,7 +16,7 @@ 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 } = require('../utils/sdp-utils');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 
@@ -204,6 +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);
@@ -232,7 +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();
@@ -327,8 +328,8 @@ class TaskDial extends Task {
             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,
-            referred_by_user: by.user,
+            refer_to_user: to.scheme === 'tel' ? to.number : to.user,
+            referred_by_user: by.scheme === 'tel' ? by.number : by.user,
             referring_call_sid,
             referred_call_sid
           }
@@ -353,11 +354,16 @@ 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();
   }
@@ -370,8 +376,14 @@ 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;
@@ -400,6 +412,20 @@ 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;
@@ -428,7 +454,8 @@ 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
@@ -449,10 +476,14 @@ 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
@@ -531,7 +562,8 @@ 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,
@@ -701,6 +733,7 @@ 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});

lib/tasks/dialogflow/index.js

@@ -235,9 +235,8 @@ class Dialogflow extends Task {
       }
 
       try {
-        const {filePath, servedFromCache} = await this._fallbackSynthAudio(cs, intent, stats, synthAudio);
+        const {filePath} = await this._fallbackSynthAudio(cs, intent, stats, synthAudio);
         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'));

lib/tasks/gather.js

@@ -6,6 +6,7 @@ const {
   AzureTranscriptionEvents,
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
+  CobaltTranscriptionEvents,
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents
@@ -58,6 +59,9 @@ class TaskGather extends SttTask {
         logger.debug('Gather: an empty hints array was supplied, so we will mask global hints');
         this.maskGlobalSttHints = true;
       }
+      // fast Recognition, fire event after a specified time after the last hypothesis.
+      this.fastRecognitionTimeout =  typeof this.data.recognizer.fastRecognitionTimeout === 'number' ?
+        this.data.recognizer.fastRecognitionTimeout * 1000 : 0;
     }
 
     this.digitBuffer = '';
@@ -111,7 +115,7 @@ class TaskGather extends SttTask {
 
     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];
@@ -174,6 +178,10 @@ class TaskGather extends SttTask {
     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.needsStt && !this.sttCredentials) {
       try {
@@ -187,12 +195,18 @@ class TaskGather extends SttTask {
       }
     }
 
+    /* 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');
+    }
+
     const startListening = async(cs, ep) => {
       this._startTimer();
       if (this.isContinuousAsr && 0 === this.timeout) this._startAsrTimer();
       if (this.input.includes('speech') && !this.listenDuringPrompt) {
         try {
-          await this._initSpeech(cs, ep);
+          await this._setSpeechHandlers(cs, ep);
           if (this.killed) {
             this.logger.info('Gather:exec - task was quickly killed so do not transcribe');
             return;
@@ -260,7 +274,7 @@ class TaskGather extends SttTask {
       }
 
       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 */});
@@ -334,7 +348,9 @@ class TaskGather extends SttTask {
     }
   }
 
-  async _initSpeech(cs, ep) {
+  async _setSpeechHandlers(cs, ep) {
+    if (this._speechHandlersSet) return;
+    this._speechHandlersSet = true;
     const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
@@ -387,6 +403,36 @@ class TaskGather extends SttTask {
         ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
         break;
 
+      case 'cobalt':
+        this.bugname = 'cobalt_transcribe';
+        ep.addCustomEventListener(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';
         ep.addCustomEventListener(IbmTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
@@ -446,6 +492,7 @@ class TaskGather extends SttTask {
       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');
@@ -489,6 +536,22 @@ class TaskGather extends SttTask {
     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(() => {
@@ -626,6 +689,9 @@ class TaskGather extends SttTask {
         }
         this._killAudio(cs);
       }
+      if (this.fastRecognitionTimeout) {
+        this._startFastRecognitionTimer(evt);
+      }
       if (this.partialResultHook) {
         const b3 = this.getTracingPropagation();
         const httpHeaders = b3 && {b3};
@@ -676,16 +742,16 @@ class TaskGather extends SttTask {
   _onJambonzConnect(_cs, _ep) {
     this.logger.debug('TaskGather:_onJambonzConnect');
   }
-  async _onJambonzError(cs, _ep, evt) {
+  async _onJambonzError(cs, ep, evt) {
     this.logger.info({evt}, 'TaskGather:_onJambonzError');
     if (this.isHandledByPrimaryProvider && this.fallbackVendor) {
-      _ep.stopTranscription({vendor: this.vendor})
+      ep.stopTranscription({vendor: this.vendor})
         .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);
+        await this._initSpeech(cs, ep);
+        this._startTranscribing(ep);
         updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid);
         return;
       } catch (error) {
@@ -790,6 +856,7 @@ class TaskGather extends SttTask {
     }
     clearTimeout(this.interDigitTimer);
     this._clearTimer();
+    this._clearFastRecognitionTimer();
 
     if (this.isContinuousAsr && reason.startsWith('speech')) {
       evt = {

lib/tasks/stt-task.js

@@ -1,6 +1,7 @@
 const Task = require('./task');
 const assert = require('assert');
-const { TaskPreconditions } = require('../utils/constants');
+const crypto = require('crypto');
+const { TaskPreconditions, CobaltTranscriptionEvents } = require('../utils/constants');
 
 class SttTask extends Task {
 
@@ -95,6 +96,44 @@ class SttTask extends Task {
     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}`));
+  }
 }
 
 module.exports = SttTask;

lib/tasks/transcribe.js

@@ -7,6 +7,7 @@ const {
   AzureTranscriptionEvents,
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
+  CobaltTranscriptionEvents,
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
@@ -90,6 +91,10 @@ class TaskTranscribe extends SttTask {
     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 {
@@ -103,6 +108,12 @@ class TaskTranscribe extends SttTask {
       }
     }
 
+    /* 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');
+    }
+
     try {
       await this._startTranscribing(cs, ep, 1);
       if (this.separateRecognitionPerChannel && ep2) {
@@ -163,7 +174,9 @@ class TaskTranscribe extends SttTask {
     }
   }
 
-  async _startTranscribing(cs, ep, channel) {
+  async _setSpeechHandlers(cs, ep, channel) {
+    if (ep._speechHandlersSet) return;
+    ep._speechHandlersSet = true;
     const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
@@ -216,6 +229,36 @@ class TaskTranscribe extends SttTask {
         ep.addCustomEventListener(SonioxTranscriptionEvents.Transcription,
           this._onTranscription.bind(this, cs, ep, channel));
         break;
+      case 'cobalt':
+        this.bugname = 'cobalt_transcribe';
+        ep.addCustomEventListener(CobaltTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+
+        /* 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,
+              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 = 'ibm_transcribe';
         ep.addCustomEventListener(IbmTranscriptionEvents.Transcription,
@@ -237,10 +280,12 @@ class TaskTranscribe extends SttTask {
         ep.addCustomEventListener(NvidiaTranscriptionEvents.VadDetected,
           this._onVadDetected.bind(this, cs, ep));
         break;
+
       default:
         if (this.vendor.startsWith('custom:')) {
           this.bugname = `${this.vendor}_transcribe`;
-          ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+          ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription,
+            this._onTranscription.bind(this, cs, ep, channel));
           ep.addCustomEventListener(JambonzTranscriptionEvents.Connect, this._onJambonzConnect.bind(this, cs, ep));
           ep.addCustomEventListener(JambonzTranscriptionEvents.ConnectFailure,
             this._onJambonzConnectFailure.bind(this, cs, ep));
@@ -257,7 +302,10 @@ class TaskTranscribe extends SttTask {
     ep.addCustomEventListener(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 */
@@ -271,7 +319,8 @@ class TaskTranscribe extends SttTask {
       interim: this.interim ? true : false,
       locale: this.language,
       channels: /*this.separateRecognitionPerChannel ? 2 : */ 1,
-      bugname: this.bugname
+      bugname: this.bugname,
+      hostport: this.hostport
     });
   }
 

lib/utils/amd-utils.js

@@ -9,6 +9,7 @@ const {
   NvidiaTranscriptionEvents,
   IbmTranscriptionEvents,
   SonioxTranscriptionEvents,
+  CobaltTranscriptionEvents,
   DeepgramTranscriptionEvents,
   JambonzTranscriptionEvents,
   AmdEvents,
@@ -314,6 +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/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",
@@ -95,6 +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",

lib/utils/db-utils.js

@@ -81,6 +81,10 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.riva_server_uri = o.riva_server_uri;
     }
+    else if ('cobalt' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.cobalt_server_uri = o.cobalt_server_uri;
+    }
     else if (obj.vendor.startsWith('custom:')) {
       const o = JSON.parse(decrypt(credential));
       obj.auth_token = o.auth_token;

lib/utils/sdp-utils.js

@@ -1,7 +1,32 @@
+const sdpTransform = require('sdp-transform');
+
 const isOnhold = (sdp) => {
   return sdp && (sdp.includes('a=sendonly') || sdp.includes('a=inactive'));
 };
 
-module.exports = {
-  isOnhold
+const mergeSdpMedia = (sdp1, sdp2) => {
+  const parsedSdp1 = sdpTransform.parse(sdp1);
+  const parsedSdp2 = sdpTransform.parse(sdp2);
+
+  parsedSdp1.media.push(...parsedSdp2.media);
+  return sdpTransform.write(parsedSdp1);
+};
+
+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
 };

lib/utils/transcription-utils.js

@@ -7,6 +7,7 @@ const {
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
   NvidiaTranscriptionEvents,
+  CobaltTranscriptionEvents,
   JambonzTranscriptionEvents
 } = require('./constants');
 
@@ -92,6 +93,11 @@ const stickyVars = {
   nvidia: [
     'NVIDIA_HINTS'
   ],
+  cobalt: [
+    'COBALT_SPEECH_HINTS',
+    'COBALT_COMPILED_CONTEXT_DATA',
+    'COBALT_METADATA'
+  ],
   soniox: [
     'SONIOX_PROFANITY_FILTER',
     'SONIOX_MODEL'
@@ -225,6 +231,26 @@ const normalizeGoogle = (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,
+    vendor: {
+      name: 'cobalt',
+      evt: copy
+    }
+  };
+};
+
 const normalizeCustom = (evt, channel, language, vendor) => {
   const copy = JSON.parse(JSON.stringify(evt));
   return {
@@ -317,6 +343,8 @@ module.exports = (logger) => {
         return normalizeNvidia(evt, channel, language);
       case 'soniox':
         return normalizeSoniox(evt, channel, language);
+      case 'cobalt':
+        return normalizeCobalt(evt, channel, language);
       default:
         if (vendor.startsWith('custom:')) {
           return normalizeCustom(evt, channel, language, vendor);
@@ -391,9 +419,9 @@ module.exports = (logger) => {
     else if ('microsoft' === vendor) {
       opts = {
         ...opts,
-        ...(rOpts.hints.length > 0 && typeof rOpts.hints[0] === 'string' &&
+        ...(rOpts.hints && 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 && 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 &&
           {AZURE_SPEECH_ALTERNATIVE_LANGUAGE_CODES: [...new Set(rOpts.altLanguages)].join(',')}),
@@ -478,7 +506,7 @@ module.exports = (logger) => {
         ...(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) &&
@@ -584,6 +612,25 @@ 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 (vendor.startsWith('custom:')) {
       let {options = {}} = rOpts;
       const {auth_token, custom_stt_url} = sttCredentials;
@@ -633,6 +680,9 @@ module.exports = (logger) => {
 
     ep.removeCustomEventListener(SonioxTranscriptionEvents.Transcription);
 
+    ep.removeCustomEventListener(CobaltTranscriptionEvents.Transcription);
+    ep.removeCustomEventListener(CobaltTranscriptionEvents.CompileContext);
+
     ep.removeCustomEventListener(NvidiaTranscriptionEvents.Transcription);
     ep.removeCustomEventListener(NvidiaTranscriptionEvents.TranscriptionComplete);
     ep.removeCustomEventListener(NvidiaTranscriptionEvents.StartOfSpeech);
@@ -664,6 +714,10 @@ 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 {

package-lock.json

@@ -14,10 +14,10 @@
         "@jambonz/db-helpers": "^0.9.1",
         "@jambonz/http-health-check": "^0.0.1",
         "@jambonz/realtimedb-helpers": "^0.8.6",
-        "@jambonz/speech-utils": "^0.0.20",
+        "@jambonz/speech-utils": "^0.0.21",
         "@jambonz/stats-collector": "^0.1.9",
         "@jambonz/time-series": "^0.2.8",
-        "@jambonz/verb-specifications": "^0.0.30",
+        "@jambonz/verb-specifications": "^0.0.37",
         "@opentelemetry/api": "^1.4.0",
         "@opentelemetry/exporter-jaeger": "^1.9.0",
         "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
@@ -30,8 +30,8 @@
         "bent": "^7.3.12",
         "debug": "^4.3.4",
         "deepcopy": "^2.1.0",
-        "drachtio-fsmrf": "^3.0.24",
-        "drachtio-srf": "^4.5.26",
+        "drachtio-fsmrf": "^3.0.27",
+        "drachtio-srf": "^4.5.29",
         "express": "^4.18.2",
         "ip": "^1.1.8",
         "moment": "^2.29.4",
@@ -2983,9 +2983,9 @@
       }
     },
     "node_modules/@jambonz/speech-utils": {
-      "version": "0.0.20",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.20.tgz",
-      "integrity": "sha512-vyMgfCAdv5+zOVisyJKhVQ1+lY1i1Ygv5amgw3JgyUy+lDbWmzANctLZXpJpoUGbTw8ZOwippEnk0F4XWFmtlg==",
+      "version": "0.0.21",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.21.tgz",
+      "integrity": "sha512-I9ULe9PHMIISZ5gr58SZ+9I5hL1F24XjFPqPlulfaoe6vJ4EgSKs1l1fWKPCuZQmv6Cf259gK5H+bjGTqjVRQw==",
       "dependencies": {
         "@aws-sdk/client-polly": "^3.359.0",
         "@google-cloud/text-to-speech": "^4.2.1",
@@ -3019,9 +3019,9 @@
       }
     },
     "node_modules/@jambonz/verb-specifications": {
-      "version": "0.0.30",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.30.tgz",
-      "integrity": "sha512-IbusgtBBo2V5Tc1FvDJvkWogHOhR2tNZN6Iyb2PjUomMI48BsWKmHW/wegppKTDpBBeN3ABY+L96pvX4N0+mCw==",
+      "version": "0.0.37",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.37.tgz",
+      "integrity": "sha512-iunu0AoholuUlkrRIPgbNFtKj2irlJbkLnZX/ghEAlNQANZOndKnlKUdAmR9Xgrjo3G7Vole/Guu372m1PRCfw==",
       "dependencies": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"
@@ -5152,9 +5152,9 @@
       }
     },
     "node_modules/drachtio-fsmrf": {
-      "version": "3.0.24",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.24.tgz",
-      "integrity": "sha512-MsTwHDOOg5yrKCcQ2buNTVSEwoMZzTcVeU5/BZ2Km0TPxBI1lErW6IiyZxA5pyoLfSdBtbqSh9ZwXrtxWTp6kA==",
+      "version": "3.0.27",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.27.tgz",
+      "integrity": "sha512-k5jwc6RpiSoasKEMq0ROxXVga+yaMplujNUKjAidKjheso+PijdQtjEwNANl0F+JyTBELaRe81PEgeRHuAdv1w==",
       "dependencies": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",
@@ -5209,9 +5209,9 @@
       }
     },
     "node_modules/drachtio-srf": {
-      "version": "4.5.26",
-      "resolved": "https://registry.npmjs.org/drachtio-srf/-/drachtio-srf-4.5.26.tgz",
-      "integrity": "sha512-Dqq3E0qY5yxjDAkrgtwrkmQYuaGUwKqUu98kDS9xzEdX8oFqKZeV5onUZq0veznw4ctuGQolY8BI6zs1e5ynHg==",
+      "version": "4.5.29",
+      "resolved": "https://registry.npmjs.org/drachtio-srf/-/drachtio-srf-4.5.29.tgz",
+      "integrity": "sha512-Hj2OW+SyQxAyLpyHngJwW26FX9V4fDYZGX0mlU4YOF4ml7I7b7XITcRPxKhroYDOrLgKqhNeh5BcPoKqPhEK7A==",
       "dependencies": {
         "debug": "^3.2.7",
         "delegates": "^0.1.0",
@@ -12949,9 +12949,9 @@
       }
     },
     "@jambonz/speech-utils": {
-      "version": "0.0.20",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.20.tgz",
-      "integrity": "sha512-vyMgfCAdv5+zOVisyJKhVQ1+lY1i1Ygv5amgw3JgyUy+lDbWmzANctLZXpJpoUGbTw8ZOwippEnk0F4XWFmtlg==",
+      "version": "0.0.21",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.21.tgz",
+      "integrity": "sha512-I9ULe9PHMIISZ5gr58SZ+9I5hL1F24XjFPqPlulfaoe6vJ4EgSKs1l1fWKPCuZQmv6Cf259gK5H+bjGTqjVRQw==",
       "requires": {
         "@aws-sdk/client-polly": "^3.359.0",
         "@google-cloud/text-to-speech": "^4.2.1",
@@ -12985,9 +12985,9 @@
       }
     },
     "@jambonz/verb-specifications": {
-      "version": "0.0.30",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.30.tgz",
-      "integrity": "sha512-IbusgtBBo2V5Tc1FvDJvkWogHOhR2tNZN6Iyb2PjUomMI48BsWKmHW/wegppKTDpBBeN3ABY+L96pvX4N0+mCw==",
+      "version": "0.0.37",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.37.tgz",
+      "integrity": "sha512-iunu0AoholuUlkrRIPgbNFtKj2irlJbkLnZX/ghEAlNQANZOndKnlKUdAmR9Xgrjo3G7Vole/Guu372m1PRCfw==",
       "requires": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"
@@ -14610,9 +14610,9 @@
       }
     },
     "drachtio-fsmrf": {
-      "version": "3.0.24",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.24.tgz",
-      "integrity": "sha512-MsTwHDOOg5yrKCcQ2buNTVSEwoMZzTcVeU5/BZ2Km0TPxBI1lErW6IiyZxA5pyoLfSdBtbqSh9ZwXrtxWTp6kA==",
+      "version": "3.0.27",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.27.tgz",
+      "integrity": "sha512-k5jwc6RpiSoasKEMq0ROxXVga+yaMplujNUKjAidKjheso+PijdQtjEwNANl0F+JyTBELaRe81PEgeRHuAdv1w==",
       "requires": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",
@@ -14662,9 +14662,9 @@
       }
     },
     "drachtio-srf": {
-      "version": "4.5.26",
-      "resolved": "https://registry.npmjs.org/drachtio-srf/-/drachtio-srf-4.5.26.tgz",
-      "integrity": "sha512-Dqq3E0qY5yxjDAkrgtwrkmQYuaGUwKqUu98kDS9xzEdX8oFqKZeV5onUZq0veznw4ctuGQolY8BI6zs1e5ynHg==",
+      "version": "4.5.29",
+      "resolved": "https://registry.npmjs.org/drachtio-srf/-/drachtio-srf-4.5.29.tgz",
+      "integrity": "sha512-Hj2OW+SyQxAyLpyHngJwW26FX9V4fDYZGX0mlU4YOF4ml7I7b7XITcRPxKhroYDOrLgKqhNeh5BcPoKqPhEK7A==",
       "requires": {
         "debug": "^3.2.7",
         "delegates": "^0.1.0",

package.json

@@ -30,10 +30,10 @@
     "@jambonz/db-helpers": "^0.9.1",
     "@jambonz/http-health-check": "^0.0.1",
     "@jambonz/realtimedb-helpers": "^0.8.6",
-    "@jambonz/speech-utils": "^0.0.20",
+    "@jambonz/speech-utils": "^0.0.21",
     "@jambonz/stats-collector": "^0.1.9",
     "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.30",
+    "@jambonz/verb-specifications": "^0.0.37",
     "@opentelemetry/api": "^1.4.0",
     "@opentelemetry/exporter-jaeger": "^1.9.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
@@ -46,8 +46,8 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.24",
-    "drachtio-srf": "^4.5.26",
+    "drachtio-fsmrf": "^3.0.27",
+    "drachtio-srf": "^4.5.29",
     "express": "^4.18.2",
     "ip": "^1.1.8",
     "moment": "^2.29.4",

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 +15,6 @@ require('./sip-refer-tests');
 require('./listen-tests');
 require('./config-test');
 require('./queue-test');
+require('./in-dialog-test');
 require('./remove-test-db');
 require('./docker_stop');

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>
+