fix #445

.github/workflows/build.yml

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

README.md

@@ -37,9 +37,6 @@ 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

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

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

@@ -275,7 +275,7 @@ router.post('/', async(req, res) => {
         else console.error(err);
       }
       ep.destroy();
-      setTimeout(restDial.kill.bind(restDial, cs), 5000);
+      setTimeout(restDial.kill.bind(restDial), 5000);
     }
   } catch (err) {
     sysError(logger, res, err);

lib/middleware.js

@@ -299,30 +299,22 @@ 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 });
+          {
+            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/call-session.js

@@ -731,8 +731,8 @@ class CallSession extends Emitter {
         (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`);
+        this.logger.info(`Speech credential vendor: ${credential.vendor} 
+${credential.label ? `, label: ${credential.label}` : ''} is chosen`);
         if ('google' === vendor) {
           if (type === 'tts' && !credential.tts_tested_ok ||
             type === 'stt' && !credential.stt_tested_ok) {
@@ -817,12 +817,6 @@ class CallSession extends Emitter {
             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,
@@ -1933,22 +1927,17 @@ class CallSession extends Emitter {
 
   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/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, JAMBONZ_DISABLE_DIAL_PAI_HEADER} = require('../config');
+const {ANCHOR_MEDIA_ALWAYS} = require('../config');
 const { isOnhold } = require('../utils/sdp-utils');
 const { normalizeJambones } = require('@jambonz/verb-specifications');
 
@@ -433,8 +433,7 @@ class TaskDial extends Task {
     this.headers = {
       'X-Account-Sid': cs.accountSid,
       ...(req && req.has('X-CID') && {'X-CID': req.get('X-CID')}),
-      ...(req && req.has('P-Asserted-Identity') && !JAMBONZ_DISABLE_DIAL_PAI_HEADER &&
-        {'P-Asserted-Identity': req.get('P-Asserted-Identity')}),
+      ...(req && req.has('P-Asserted-Identity') && {'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

lib/tasks/gather.js

@@ -6,7 +6,6 @@ const {
   AzureTranscriptionEvents,
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
-  CobaltTranscriptionEvents,
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents
@@ -59,9 +58,6 @@ 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 = '';
@@ -115,7 +111,7 @@ 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];
@@ -191,18 +187,12 @@ 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._setSpeechHandlers(cs, ep);
+          await this._initSpeech(cs, ep);
           if (this.killed) {
             this.logger.info('Gather:exec - task was quickly killed so do not transcribe');
             return;
@@ -270,7 +260,7 @@ class TaskGather extends SttTask {
       }
 
       if (this.input.includes('speech') && this.listenDuringPrompt) {
-        await this._setSpeechHandlers(cs, ep);
+        await this._initSpeech(cs, ep);
         this._startTranscribing(ep);
         updateSpeechCredentialLastUsed(this.sttCredentials.speech_credential_sid)
           .catch(() => {/*already logged error */});
@@ -344,9 +334,7 @@ class TaskGather extends SttTask {
     }
   }
 
-  async _setSpeechHandlers(cs, ep) {
-    if (this._speechHandlersSet) return;
-    this._speechHandlersSet = true;
+  async _initSpeech(cs, ep) {
     const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
@@ -399,36 +387,6 @@ 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));
@@ -488,7 +446,6 @@ 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');
@@ -532,22 +489,6 @@ 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(() => {
@@ -685,9 +626,6 @@ class TaskGather extends SttTask {
         }
         this._killAudio(cs);
       }
-      if (this.fastRecognitionTimeout) {
-        this._startFastRecognitionTimer(evt);
-      }
       if (this.partialResultHook) {
         const b3 = this.getTracingPropagation();
         const httpHeaders = b3 && {b3};
@@ -738,16 +676,16 @@ 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) {
@@ -852,7 +790,6 @@ class TaskGather extends SttTask {
     }
     clearTimeout(this.interDigitTimer);
     this._clearTimer();
-    this._clearFastRecognitionTimer();
 
     if (this.isContinuousAsr && reason.startsWith('speech')) {
       evt = {

lib/tasks/stt-task.js

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

@@ -7,7 +7,6 @@ const {
   AzureTranscriptionEvents,
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
-  CobaltTranscriptionEvents,
   IbmTranscriptionEvents,
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
@@ -104,12 +103,6 @@ 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) {
@@ -170,9 +163,7 @@ class TaskTranscribe extends SttTask {
     }
   }
 
-  async _setSpeechHandlers(cs, ep, channel) {
-    if (this._speechHandlersSet) return;
-    this._speechHandlersSet = true;
+  async _startTranscribing(cs, ep, channel) {
     const opts = this.setChannelVarsForStt(this, this.sttCredentials, this.data.recognizer);
     switch (this.vendor) {
       case 'google':
@@ -225,36 +216,6 @@ 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,
@@ -276,12 +237,10 @@ 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, channel));
+          ep.addCustomEventListener(JambonzTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
           ep.addCustomEventListener(JambonzTranscriptionEvents.Connect, this._onJambonzConnect.bind(this, cs, ep));
           ep.addCustomEventListener(JambonzTranscriptionEvents.ConnectFailure,
             this._onJambonzConnectFailure.bind(this, cs, ep));
@@ -298,10 +257,7 @@ 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 */
@@ -315,8 +271,7 @@ class TaskTranscribe extends SttTask {
       interim: this.interim ? true : false,
       locale: this.language,
       channels: /*this.separateRecognitionPerChannel ? 2 : */ 1,
-      bugname: this.bugname,
-      hostport: this.hostport
+      bugname: this.bugname
     });
   }
 

lib/utils/amd-utils.js

@@ -9,7 +9,6 @@ const {
   NvidiaTranscriptionEvents,
   IbmTranscriptionEvents,
   SonioxTranscriptionEvents,
-  CobaltTranscriptionEvents,
   DeepgramTranscriptionEvents,
   JambonzTranscriptionEvents,
   AmdEvents,
@@ -315,10 +314,6 @@ 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

@@ -95,11 +95,6 @@
     "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

@@ -81,10 +81,6 @@ 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/transcription-utils.js

@@ -7,7 +7,6 @@ const {
   DeepgramTranscriptionEvents,
   SonioxTranscriptionEvents,
   NvidiaTranscriptionEvents,
-  CobaltTranscriptionEvents,
   JambonzTranscriptionEvents
 } = require('./constants');
 
@@ -93,11 +92,6 @@ const stickyVars = {
   nvidia: [
     'NVIDIA_HINTS'
   ],
-  cobalt: [
-    'COBALT_SPEECH_HINTS',
-    'COBALT_COMPILED_CONTEXT_DATA',
-    'COBALT_METADATA'
-  ],
   soniox: [
     'SONIOX_PROFANITY_FILTER',
     'SONIOX_MODEL'
@@ -231,26 +225,6 @@ 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 {
@@ -343,8 +317,6 @@ 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);
@@ -419,9 +391,9 @@ module.exports = (logger) => {
     else if ('microsoft' === vendor) {
       opts = {
         ...opts,
-        ...(rOpts.hints && 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 && 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 &&
           {AZURE_SPEECH_ALTERNATIVE_LANGUAGE_CODES: [...new Set(rOpts.altLanguages)].join(',')}),
@@ -612,25 +584,6 @@ 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;
@@ -680,9 +633,6 @@ 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);
@@ -714,10 +664,6 @@ 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": "^0.0.21",
+        "@jambonz/speech-utils": "^0.0.20",
         "@jambonz/stats-collector": "^0.1.9",
         "@jambonz/time-series": "^0.2.8",
-        "@jambonz/verb-specifications": "^0.0.34",
+        "@jambonz/verb-specifications": "^0.0.30",
         "@opentelemetry/api": "^1.4.0",
         "@opentelemetry/exporter-jaeger": "^1.9.0",
         "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
@@ -30,7 +30,7 @@
         "bent": "^7.3.12",
         "debug": "^4.3.4",
         "deepcopy": "^2.1.0",
-        "drachtio-fsmrf": "^3.0.27",
+        "drachtio-fsmrf": "^3.0.24",
         "drachtio-srf": "^4.5.26",
         "express": "^4.18.2",
         "ip": "^1.1.8",
@@ -2983,9 +2983,9 @@
       }
     },
     "node_modules/@jambonz/speech-utils": {
-      "version": "0.0.21",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.21.tgz",
-      "integrity": "sha512-I9ULe9PHMIISZ5gr58SZ+9I5hL1F24XjFPqPlulfaoe6vJ4EgSKs1l1fWKPCuZQmv6Cf259gK5H+bjGTqjVRQw==",
+      "version": "0.0.20",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.20.tgz",
+      "integrity": "sha512-vyMgfCAdv5+zOVisyJKhVQ1+lY1i1Ygv5amgw3JgyUy+lDbWmzANctLZXpJpoUGbTw8ZOwippEnk0F4XWFmtlg==",
       "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": "0.0.34",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.34.tgz",
-      "integrity": "sha512-6/oplhOMpxz6qbIdz0K3Y2+KXtnb/DLziYmnGUK8NALWD7rhDvp2RdTsuEjOOkOhwLyRJUhuIAM17GcpAq/lCw==",
+      "version": "0.0.30",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.30.tgz",
+      "integrity": "sha512-IbusgtBBo2V5Tc1FvDJvkWogHOhR2tNZN6Iyb2PjUomMI48BsWKmHW/wegppKTDpBBeN3ABY+L96pvX4N0+mCw==",
       "dependencies": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"
@@ -5152,9 +5152,9 @@
       }
     },
     "node_modules/drachtio-fsmrf": {
-      "version": "3.0.27",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.27.tgz",
-      "integrity": "sha512-k5jwc6RpiSoasKEMq0ROxXVga+yaMplujNUKjAidKjheso+PijdQtjEwNANl0F+JyTBELaRe81PEgeRHuAdv1w==",
+      "version": "3.0.24",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.24.tgz",
+      "integrity": "sha512-MsTwHDOOg5yrKCcQ2buNTVSEwoMZzTcVeU5/BZ2Km0TPxBI1lErW6IiyZxA5pyoLfSdBtbqSh9ZwXrtxWTp6kA==",
       "dependencies": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",
@@ -12949,9 +12949,9 @@
       }
     },
     "@jambonz/speech-utils": {
-      "version": "0.0.21",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.21.tgz",
-      "integrity": "sha512-I9ULe9PHMIISZ5gr58SZ+9I5hL1F24XjFPqPlulfaoe6vJ4EgSKs1l1fWKPCuZQmv6Cf259gK5H+bjGTqjVRQw==",
+      "version": "0.0.20",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.0.20.tgz",
+      "integrity": "sha512-vyMgfCAdv5+zOVisyJKhVQ1+lY1i1Ygv5amgw3JgyUy+lDbWmzANctLZXpJpoUGbTw8ZOwippEnk0F4XWFmtlg==",
       "requires": {
         "@aws-sdk/client-polly": "^3.359.0",
         "@google-cloud/text-to-speech": "^4.2.1",
@@ -12985,9 +12985,9 @@
       }
     },
     "@jambonz/verb-specifications": {
-      "version": "0.0.34",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.34.tgz",
-      "integrity": "sha512-6/oplhOMpxz6qbIdz0K3Y2+KXtnb/DLziYmnGUK8NALWD7rhDvp2RdTsuEjOOkOhwLyRJUhuIAM17GcpAq/lCw==",
+      "version": "0.0.30",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.30.tgz",
+      "integrity": "sha512-IbusgtBBo2V5Tc1FvDJvkWogHOhR2tNZN6Iyb2PjUomMI48BsWKmHW/wegppKTDpBBeN3ABY+L96pvX4N0+mCw==",
       "requires": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"
@@ -14610,9 +14610,9 @@
       }
     },
     "drachtio-fsmrf": {
-      "version": "3.0.27",
-      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.27.tgz",
-      "integrity": "sha512-k5jwc6RpiSoasKEMq0ROxXVga+yaMplujNUKjAidKjheso+PijdQtjEwNANl0F+JyTBELaRe81PEgeRHuAdv1w==",
+      "version": "3.0.24",
+      "resolved": "https://registry.npmjs.org/drachtio-fsmrf/-/drachtio-fsmrf-3.0.24.tgz",
+      "integrity": "sha512-MsTwHDOOg5yrKCcQ2buNTVSEwoMZzTcVeU5/BZ2Km0TPxBI1lErW6IiyZxA5pyoLfSdBtbqSh9ZwXrtxWTp6kA==",
       "requires": {
         "camel-case": "^4.1.2",
         "debug": "^2.6.9",

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": "^0.0.21",
+    "@jambonz/speech-utils": "^0.0.20",
     "@jambonz/stats-collector": "^0.1.9",
     "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.34",
+    "@jambonz/verb-specifications": "^0.0.30",
     "@opentelemetry/api": "^1.4.0",
     "@opentelemetry/exporter-jaeger": "^1.9.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
@@ -46,7 +46,7 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.27",
+    "drachtio-fsmrf": "^3.0.24",
     "drachtio-srf": "^4.5.26",
     "express": "^4.18.2",
     "ip": "^1.1.8",