reset variables like hints so that previous hints do not automatically carry over

lib/session/call-session.js

@@ -270,6 +270,10 @@ class CallSession extends Emitter {
     return this.backgroundGatherTask;
   }
 
+  get isListenEnabled() {
+    return this.backgroundListenTask;
+  }
+
   get b3() {
     return this.rootSpan?.getTracingPropagation();
   }
@@ -461,6 +465,50 @@ class CallSession extends Emitter {
     }
   }
 
+  async startBackgroundListen(opts) {
+    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]);
+      const resources = await this._evaluatePreconditions(this.backgroundListenTask);
+      const {span, ctx} = this.rootSpan.startChildSpan(`background-gather:${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() {
+    try {
+      if (this.backgroundListenTask) {
+        this.backgroundListenTask.removeAllListeners();
+        this.backgroundListenTask.kill().catch(() => {});
+      }
+    } catch (err) {
+      this.logger.info({err}, 'CallSession:stopBackgroundListen - Error stopping listen task');
+    }
+    this.backgroundListenTask = null;
+  }
+
   async enableBotMode(gather, autoEnable) {
     try {
       if (this.backgroundGatherTask) {

lib/tasks/config.js

@@ -8,7 +8,8 @@ class TaskConfig extends Task {
       'synthesizer',
       'recognizer',
       'bargeIn',
-      'record'
+      'record',
+      'listen'
     ].forEach((k) => this[k] = this.data[k] || {});
 
     if ('notifyEvents' in this.data) {
@@ -30,7 +31,7 @@ class TaskConfig extends Task {
       });
     }
     if (this.bargeIn.sticky) this.autoEnable = true;
-    this.preconditions = (this.bargeIn.enable || this.record?.action || this.data.amd) ?
+    this.preconditions = (this.bargeIn.enable || this.record?.action || this.listen?.url || this.data.amd) ?
       TaskPreconditions.Endpoint :
       TaskPreconditions.None;
   }
@@ -38,8 +39,9 @@ class TaskConfig extends Task {
   get name() { return TaskName.Config; }
 
   get hasSynthesizer() { return Object.keys(this.synthesizer).length; }
-
   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 summary() {
     const phrase = [];
@@ -54,6 +56,10 @@ class TaskConfig extends Task {
       const s = `{${v},${l}}`;
       phrase.push(`set recognizer${s}`);
     }
+    if (this.hasRecording) phrase.push(this.record.action);
+    if (this.hasListen) {
+      phrase.push(this.listen.enable ? `listen ${this.listen.url}` : 'stop listen');
+    }
     if (this.data.amd) phrase.push('enable amd');
     if (this.notifyEvents) phrase.push(`event notification ${this.notifyEvents ? 'on' : 'off'}`);
     return `${this.name}{${phrase.join(',')}`;
@@ -64,8 +70,7 @@ class TaskConfig extends Task {
 
     if (this.notifyEvents) {
       this.logger.debug(`turning event notification ${this.notifyEvents ? 'on' : 'off'}`);
-      cs.notifyEvents = !!this.data.notifEvents;
-
+      cs.notifyEvents = !!this.data.notifyEvents;
     }
 
     if (this.data.amd) {
@@ -147,11 +152,21 @@ class TaskConfig extends Task {
         this.logger.info({err}, 'Config: error starting recording');
       }
     }
+    if (this.hasListen) {
+      const {enable, ...opts} = this.listen;
+      if (enable) {
+        this.logger.debug({opts}, 'Config: enabling listen');
+        cs.startBackgroundListen({verb: 'listen', ...opts});
+      } else {
+        this.logger.info('Config: disabling listen');
+        cs.stopBackgroundListen();
+      }
+    }
   }
 
   async kill(cs) {
     super.kill(cs);
-    if (this.ep && this.stopAmd) this.stopAmd(this.ep, this);
+    //if (this.ep && this.stopAmd) this.stopAmd(this.ep, this);
   }
 
   _onAmdEvent(cs, evt) {

lib/tasks/listen.js

@@ -10,7 +10,7 @@ class TaskListen extends Task {
 
     [
       'action', 'auth', 'method', 'url', 'finishOnKey', 'maxLength', 'metadata', 'mixType', 'passDtmf', 'playBeep',
-      'sampleRate', 'timeout', 'transcribe', 'wsAuth'
+      'sampleRate', 'timeout', 'transcribe', 'wsAuth', 'disableBidirectionalAudio'
     ].forEach((k) => this[k] = this.data[k]);
 
     this.mixType = this.mixType || 'mono';
@@ -136,7 +136,9 @@ class TaskListen extends Task {
     }
 
     /* support bi-directional audio */
-    ep.addCustomEventListener(ListenEvents.PlayAudio, this._onPlayAudio.bind(this, ep));
+    if (!this.disableBiDirectionalAudio) {
+      ep.addCustomEventListener(ListenEvents.PlayAudio, this._onPlayAudio.bind(this, ep));
+    }
     ep.addCustomEventListener(ListenEvents.KillAudio, this._onKillAudio.bind(this, ep));
     ep.addCustomEventListener(ListenEvents.Disconnect, this._onDisconnect.bind(this, ep));
   }

lib/tasks/specs.json

@@ -42,11 +42,24 @@
       "recognizer": "#recognizer",
       "bargeIn": "#bargeIn",
       "record": "#recordOptions",
+      "listen": "#listenOptions",
       "amd": "#amd",
       "notifyEvents": "boolean"
     },
     "required": []
   },
+  "listenOptions": {
+    "properties": {
+      "enable": "boolean",
+      "url": "string",
+      "sampleRate": "number",
+      "wsAuth": "#auth",
+      "metadata": "object"
+    },
+    "required": [
+      "enable"
+    ]
+  },
   "bargeIn": {
     "properties": {
       "enable": "boolean",
@@ -275,6 +288,7 @@
       },
       "passDtmf": "boolean",
       "playBeep": "boolean",
+      "disableBidirectionalAudio": "boolean",
       "sampleRate": "number",
       "timeout": "number",
       "transcribe": "#transcribe",

lib/utils/transcription-utils.js

@@ -7,6 +7,86 @@ const {
   DeepgramTranscriptionEvents,
 } = require('./constants');
 
+const stickyVars = {
+  google: [
+    'GOOGLE_SPEECH_HINTS',
+    'GOOGLE_SPEECH_SEPARATE_RECOGNITION_PER_CHANNEL',
+    'GOOGLE_SPEECH_PROFANITY_FILTER',
+    'GOOGLE_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION',
+    'GOOGLE_SPEECH_ENABLE_WORD_TIME_OFFSETS',
+    'GOOGLE_SPEECH_SINGLE_UTTERANCE',
+    'GOOGLE_SPEECH_SPEAKER_DIARIZATION',
+    'GOOGLE_SPEECH_USE_ENHANCED',
+    'GOOGLE_SPEECH_ALTERNATIVE_LANGUAGE_CODES',
+    'GOOGLE_SPEECH_METADATA_INTERACTION_TYPE',
+    'GOOGLE_SPEECH_METADATA_INDUSTRY_NAICS_CODE'
+  ],
+  microsoft: [
+    'AZURE_SPEECH_HINTS',
+    'AZURE_SERVICE_ENDPOINT_ID',
+    'AZURE_REQUEST_SNR',
+    'AZURE_PROFANITY_OPTION',
+    'AZURE_SERVICE_ENDPOINT',
+    'AZURE_INITIAL_SPEECH_TIMEOUT_MS',
+    'AZURE_USE_OUTPUT_FORMAT_DETAILED',
+  ],
+  deepgram: [
+    'DEEPGRAM_SPEECH_KEYWORDS',
+    'DEEPGRAM_API_KEY',
+    'DEEPGRAM_SPEECH_TIER',
+    'DEEPGRAM_SPEECH_MODEL',
+    'DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION',
+    'DEEPGRAM_SPEECH_PROFANITY_FILTER',
+    'DEEPGRAM_SPEECH_REDACT',
+    'DEEPGRAM_SPEECH_DIARIZE',
+    'DEEPGRAM_SPEECH_NER',
+    'DEEPGRAM_SPEECH_ALTERNATIVES',
+    'DEEPGRAM_SPEECH_NUMERALS',
+    'DEEPGRAM_SPEECH_SEARCH',
+    'DEEPGRAM_SPEECH_REPLACE',
+    'DEEPGRAM_SPEECH_ENDPOINTING',
+    'DEEPGRAM_SPEECH_VAD_TURNOFF',
+    'DEEPGRAM_SPEECH_TAG'
+  ],
+  aws: [
+    'AWS_VOCABULARY_NAME',
+    'AWS_VOCABULARY_FILTER_METHOD',
+    'AWS_VOCABULARY_FILTER_NAME'
+  ],
+  nuance: [
+    'NUANCE_ACCESS_TOKEN',
+    'NUANCE_KRYPTON_ENDPOINT',
+    'NUANCE_TOPIC',
+    'NUANCE_UTTERANCE_DETECTION_MODE',
+    'NUANCE_FILTER_PROFANITY',
+    'NUANCE_INCLUDE_TOKENIZATION',
+    'NUANCE_DISCARD_SPEAKER_ADAPTATION',
+    'NUANCE_SUPPRESS_CALL_RECORDING',
+    'NUANCE_MASK_LOAD_FAILURES',
+    'NUANCE_SUPPRESS_INITIAL_CAPITALIZATION',
+    'NUANCE_ALLOW_ZERO_BASE_LM_WEIGHT',
+    'NUANCE_FILTER_WAKEUP_WORD',
+    'NUANCE_NO_INPUT_TIMEOUT_MS',
+    'NUANCE_RECOGNITION_TIMEOUT_MS',
+    'NUANCE_UTTERANCE_END_SILENCE_MS',
+    'NUANCE_MAX_HYPOTHESES',
+    'NUANCE_SPEECH_DOMAIN',
+    'NUANCE_FORMATTING',
+    'NUANCE_RESOURCES'
+  ],
+  ibm: [
+    'IBM_ACCESS_TOKEN',
+    'IBM_SPEECH_REGION',
+    'IBM_SPEECH_INSTANCE_ID',
+    'IBM_SPEECH_MODEL',
+    'IBM_SPEECH_LANGUAGE_CUSTOMIZATION_ID',
+    'IBM_SPEECH_ACOUSTIC_CUSTOMIZATION_ID',
+    'IBM_SPEECH_BASE_MODEL_VERSION',
+    'IBM_SPEECH_WATSON_METADATA',
+    'IBM_SPEECH_WATSON_LEARNING_OPT_OUT'
+  ]
+};
+
 const normalizeDeepgram = (evt, channel, language) => {
   const copy = JSON.parse(JSON.stringify(evt));
   const alternatives = (evt.channel?.alternatives || [])
@@ -19,7 +99,7 @@ const normalizeDeepgram = (evt, channel, language) => {
     language_code: language,
     channel_tag: channel,
     is_final: evt.is_final,
-    alternatives,
+    alternatives: [alternatives[0]],
     vendor: {
       name: 'deepgram',
       evt: copy
@@ -50,7 +130,7 @@ const normalizeGoogle = (evt, channel, language) => {
     language_code: language,
     channel_tag: channel,
     is_final: evt.is_final,
-    alternatives: evt.alternatives,
+    alternatives: [evt.alternatives[0]],
     vendor: {
       name: 'google',
       evt: copy
@@ -64,7 +144,7 @@ const normalizeNuance = (evt, channel, language) => {
     language_code: language,
     channel_tag: channel,
     is_final: evt.is_final,
-    alternatives: evt.alternatives,
+    alternatives: [evt.alternatives[0]],
     vendor: {
       name: 'nuance',
       evt: copy
@@ -92,7 +172,7 @@ const normalizeMicrosoft = (evt, channel, language) => {
     language_code,
     channel_tag: channel,
     is_final: evt.RecognitionStatus === 'Success',
-    alternatives,
+    alternatives: [alternatives[0]],
     vendor: {
       name: 'microsoft',
       evt: copy
@@ -198,7 +278,7 @@ module.exports = (logger) => {
           {GOOGLE_SPEECH_ALTERNATIVE_LANGUAGE_CODES: rOpts.altLanguages.join(',')}),
         ...(rOpts.interactionType &&
           {GOOGLE_SPEECH_METADATA_INTERACTION_TYPE: rOpts.interactionType}),
-        ...{GOOGLE_SPEECH_MODEL: rOpts.model || (task.name === TaskName.Gather ? 'command_and_search' : 'phone_call')},
+        ...{GOOGLE_SPEECH_MODEL: rOpts.model || (task.name === TaskName.Gather ? 'latest_short' : 'phone_call')},
         ...(rOpts.naicsCode > 0 &&
           {GOOGLE_SPEECH_METADATA_INDUSTRY_NAICS_CODE: rOpts.naicsCode}),
       };
@@ -333,7 +413,7 @@ module.exports = (logger) => {
         ...(deepgramOptions.vadTurnoff) &&
           {DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.vadTurnoff},
         ...(deepgramOptions.tag) &&
-          {DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.tag}
+          {DEEPGRAM_SPEECH_TAG: deepgramOptions.tag}
       };
     }
     else if ('ibm' === rOpts.vendor) {
@@ -360,7 +440,9 @@ module.exports = (logger) => {
           {IBM_SPEECH_WATSON_LEARNING_OPT_OUT: ibmOptions.watsonLearningOptOut}
       };
     }
-
+    stickyVars[rOpts.vendor].forEach((key) => {
+      if (!opts[key]) opts[key] = '';
+    });
     logger.debug({opts}, 'recognizer channel vars');
     return opts;
   };

test/config-test.js

@@ -0,0 +1,102 @@
+const test = require('tape');
+const { sippUac } = require('./sipp')('test_fs');
+const bent = require('bent');
+const getJSON = bent('json')
+const clearModule = require('clear-module');
+const {provisionCallHook} = require('./utils')
+
+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('\'config: listen\'', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const from = "config_listen_success";
+    let verbs = [
+      {
+        "verb": "config",
+        "listen": {
+          "enable": true,
+          "url": `ws://172.38.0.60:3000/${from}`
+        }
+      },
+      {
+        "verb": "pause",
+        "length": 5
+      }
+    ];
+
+    provisionCallHook(from, verbs);
+
+    // THEN
+    await sippUac('uac-gather-account-creds-success-send-bye.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/ws_packet_count/${from}`);
+    t.pass('config: successfully started background listen');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+test('\'config: listen - stop\'', async(t) => {
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+  try {
+    await connect(srf);
+
+    // GIVEN
+    const from = "config_listen_success";
+    let verbs = [
+      {
+        "verb": "config",
+        "listen": {
+          "enable": true,
+          "url": `ws://172.38.0.60:3000/${from}`
+        }
+      },
+      {
+        "verb": "pause",
+        "length": 1
+      },
+      {
+        "verb": "config",
+        "listen": {
+          "enable": false
+        }
+      },
+      {
+        "verb": "pause",
+        "length": 3
+      }
+    ];
+
+    provisionCallHook(from, verbs);
+
+    // THEN
+    await sippUac('uac-gather-account-creds-success-send-bye.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/ws_packet_count/${from}`);
+    t.pass('config: successfully started then stopped background listen');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});

test/docker-compose-testbed.yaml

@@ -44,7 +44,7 @@ services:
   drachtio:
     image: drachtio/drachtio-server:latest
     restart: always
-    command: drachtio --contact "sip:*;transport=udp,tcp" --address 0.0.0.0 --port 9022
+    command: drachtio --contact "sip:*;transport=udp" --mtu 4096 --address 0.0.0.0 --port 9022
     ports:
       - "9060:9022/tcp"
     networks:

test/index.js

@@ -12,5 +12,6 @@ require('./create-call-test');
 require('./play-tests');
 require('./sip-refer-tests');
 require('./listen-tests');
+require('./config-test');
 require('./remove-test-db');
 require('./docker_stop');

test/listen-tests.js

@@ -40,7 +40,7 @@ test('\'listen-success\'', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success-send-bye.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/ws_packet_count/${from}`);
-    t.ok(38000 <= obj.count, 'listen: success incomming call audio');
+    t.ok(38000 <= obj.count, 'listen: success incoming call audio');
 
     obj = await getJSON(`http://127.0.0.1:3100/ws_metadata/${from}`);
     t.ok(obj.metadata.from === from && obj.metadata.sampleRate === 8000, 'listen: success metadata');