Feature/deepgram stt (#190)

* initial changes to support deepgram stt

* fixes for normalizing vendor-specific transcriptions

* update to latest drachtio-fsmrf with support for deepgram stt

* deepgram parsing error

* hints support for deepgram

* handling deepgram errors

* ignore late arriving transcripts for deepgram

* handling of empty transcripts

* transcribe changes

* allow deepgram stt credentials to be provided at run time

* bind channel in transcription handler

* fixes for transcribe when handling empty transcripts

* more empty transcript fixes

* update tests to latest modules

* add test cases for deepgram speech recognition

lib/session/call-session.js

@@ -569,6 +569,12 @@ class CallSession extends Emitter {
             secret: credential.secret
           };
         }
+        else if ('deepgram' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            api_key: credential.api_key
+          };
+        }
       }
       else {
         writeAlerts({

lib/tasks/gather.js

@@ -5,7 +5,8 @@ const {
   GoogleTranscriptionEvents,
   NuanceTranscriptionEvents,
   AwsTranscriptionEvents,
-  AzureTranscriptionEvents
+  AzureTranscriptionEvents,
+  DeepgramTranscriptionEvents
 } = require('../utils/constants');
 
 const makeTask = require('./make_task');
@@ -54,11 +55,14 @@ class TaskGather extends Task {
       this.vendor = recognizer.vendor;
       this.language = recognizer.language;
 
+      /* let credentials be supplied in the recognizer object at runtime */
       if (recognizer.vendor === 'nuance') {
         const {clientId, secret} = recognizer.nuanceOptions;
-        if (clientId && secret) {
+        if (clientId && secret) this.sttCredentials = {client_id: clientId, secret};
-          this.sttCredentials = {client_id: clientId, secret};
+      }
-        }
+      else if (recognizer.vendor === 'deepgram') {
+        const {apiKey} = recognizer.deepgramOptions;
+        if (apiKey) this.sttCredentials = {api_key: apiKey};
       }
 
       /* continuous ASR (i.e. compile transcripts until a special timeout or dtmf key) */
@@ -338,8 +342,16 @@ class TaskGather extends Task {
         if ((this.sayTask || this.playTask) && this.listenDuringPrompt) {
           opts.NUANCE_STALL_TIMERS = 1;
         }
-
         break;
+
+      case 'deepgram':
+        this.bugname = 'deepgram_transcribe';
+        ep.addCustomEventListener(DeepgramTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
+        ep.addCustomEventListener(DeepgramTranscriptionEvents.Connect, this._onDeepgramConnect.bind(this, cs, ep));
+        ep.addCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure,
+          this._onDeepGramConnectFailure.bind(this, cs, ep));
+        break;
+
       default:
         throw new Error(`Invalid vendor ${this.vendor}`);
     }
@@ -441,30 +453,31 @@ class TaskGather extends Task {
 
   _onTranscription(cs, ep, evt, fsEvent) {
     // make sure this is not a transcript from answering machine detection
-    this.logger.debug({evt}, 'Gather:_onTranscription');
     const bugname = fsEvent.getHeader('media-bugname');
     const finished = fsEvent.getHeader('transcription-session-finished');
+    this.logger.debug({evt, bugname, finished}, 'Gather:_onTranscription');
     if (bugname && this.bugname !== bugname) return;
 
     evt = this.normalizeTranscription(evt, this.vendor, 1, this.language);
 
     /* count words for bargein feature */
-    const words = evt.alternatives[0].transcript.split(' ').length;
+    const words = evt.alternatives[0]?.transcript.split(' ').length;
     const bufferedWords = this._bufferedTranscripts.reduce((count, e) => {
-      return count + e.alternatives[0].transcript.split(' ').length;
+      return count + e.alternatives[0]?.transcript.split(' ').length;
     }, 0);
 
     if (evt.is_final) {
       if (evt.alternatives[0].transcript === '' && !this.callSession.callGone && !this.killed) {
-        if ('microsoft' === this.vendor && finished === 'true') {
+        if (finished === 'true' && ['microsoft', 'deepgram'].includes(this.vendor)) {
           this.logger.debug({evt}, 'TaskGather:_onTranscription - got empty transcript from old gather, disregarding');
         }
         else {
-          this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, listen again');
+          this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, continue listening');
-          this._startTranscribing(ep);
+          //this._startTranscribing(ep);
         }
         return;
       }
+
       if (this.isContinuousAsr) {
         /* append the transcript and start listening again for asrTimeout */
         const t = evt.alternatives[0].transcript;
@@ -548,6 +561,23 @@ class TaskGather extends Task {
       return this._resolve('timeout');
     }
   }
+  _onDeepgramConnect(_cs, _ep) {
+    this.logger.debug('TaskGather:_onDeepgramConnect');
+  }
+
+  _onDeepGramConnectFailure(cs, _ep, evt) {
+    const {reason} = evt;
+    const {writeAlerts, AlertType} = cs.srf.locals;
+    this.logger.info({evt}, 'TaskGather:_onDeepgramConnectFailure');
+    writeAlerts({
+      account_sid: cs.accountSid,
+      alert_type: AlertType.STT_FAILURE,
+      message: `Failed connecting to Deepgram speech recognizer: ${reason}`,
+      vendor: 'deepgram',
+    }).catch((err) => this.logger.info({err}, 'Error generating alert for deepgram connection failure'));
+    this.notifyError(`Failed connecting to speech vendor deepgram: ${reason}`);
+    this.notifyTaskDone();
+  }
 
   _onVadDetected(cs, ep) {
     if (this.bargein && this.minBargeinWordCount === 0) {

lib/tasks/specs.json

@@ -445,7 +445,7 @@
     "properties": {
       "vendor": {
         "type": "string",
-        "enum": ["google", "aws", "microsoft", "nuance", "default"]
+        "enum": ["google", "aws", "microsoft", "nuance", "deepgram", "default"]
       },
       "language": "string",
       "vad": "#vad",
@@ -510,12 +510,63 @@
       "azureSttEndpointId": "string",
       "asrDtmfTerminationDigit": "string",
       "asrTimeout": "number",
-      "nuanceOptions": "#nuanceOptions"
+      "nuanceOptions": "#nuanceOptions",
+      "deepgramOptions": "#deepgramOptions"
     },
     "required": [
       "vendor"
     ]
   },
+  "deepgramOptions": {
+    "properties": {
+      "apiKey": "string",
+      "tier": {
+        "type": "string",
+        "enum": [
+          "enhanced",
+          "base"
+        ]
+      },
+      "model": {
+        "type": "string",
+        "enum": [
+          "general",
+          "meeting",
+          "phonecall",
+          "voicemail",
+          "finance",
+          "conversationalai",
+          "video",
+          "custom"
+        ]
+      },
+      "customModel": "string",
+      "version": "string",
+      "punctuate": "boolean",
+      "profanityFilter": "boolean",
+      "redact": {
+        "type": "string",
+        "enum": [
+          "pci",
+          "numbers",
+          "true",
+          "ssn"
+        ]
+      },
+      "diarize": "boolean",
+      "diarizeVersion": "string",
+      "ner": "boolean",
+      "multichannel": "boolean",
+      "alternatives": "number",
+      "numerals": "boolean",
+      "search": "array",
+      "replace": "array",
+      "keywords": "array",
+      "endpointing": "boolean",
+      "vadTurnoff": "number",
+      "tag": "string"
+    }
+  },
   "nuanceOptions": {
     "properties": {
       "clientId": "string",

lib/tasks/transcribe.js

@@ -5,7 +5,8 @@ const {
   GoogleTranscriptionEvents,
   AzureTranscriptionEvents,
   AwsTranscriptionEvents,
-  NuanceTranscriptionEvents
+  NuanceTranscriptionEvents,
+  DeepgramTranscriptionEvents
 } = require('../utils/constants');
 const normalizeJambones = require('../utils/normalize-jambones');
 
@@ -15,9 +16,14 @@ class TaskTranscribe extends Task {
     this.preconditions = TaskPreconditions.Endpoint;
     this.parentTask = parentTask;
 
-    const {setChannelVarsForStt, normalizeTranscription} = require('../utils/transcription-utils')(logger);
+    const {
+      setChannelVarsForStt,
+      normalizeTranscription,
+      removeSpeechListeners
+    } = require('../utils/transcription-utils')(logger);
     this.setChannelVarsForStt = setChannelVarsForStt;
     this.normalizeTranscription = normalizeTranscription;
+    this.removeSpeechListeners = removeSpeechListeners;
 
     this.transcriptionHook = this.data.transcriptionHook;
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
@@ -28,12 +34,17 @@ class TaskTranscribe extends Task {
     this.interim = !!recognizer.interim;
     this.separateRecognitionPerChannel = recognizer.separateRecognitionPerChannel;
 
+    /* let credentials be supplied in the recognizer object at runtime */
     if (recognizer.vendor === 'nuance') {
       const {clientId, secret} = recognizer.nuanceOptions;
       if (clientId && secret) {
         this.sttCredentials = {client_id: clientId, secret};
       }
     }
+    else if (recognizer.vendor === 'deepgram') {
+      const {apiKey} = recognizer.deepgramOptions;
+      if (apiKey) this.sttCredentials = {api_key: apiKey};
+    }
 
     recognizer.hints = recognizer.hints || [];
     recognizer.altLanguages = recognizer.altLanguages || [];
@@ -69,7 +80,7 @@ class TaskTranscribe extends Task {
     if (!this.data.recognizer.vendor) {
       this.data.recognizer.vendor = this.vendor;
     }
-    this.sttCredentials = cs.getSpeechCredentials(this.vendor, 'stt');
+    if (!this.sttCredentials) this.sttCredentials = cs.getSpeechCredentials(this.vendor, 'stt');
 
     try {
       if (!this.sttCredentials) {
@@ -105,22 +116,7 @@ class TaskTranscribe extends Task {
       this.logger.info(err, 'TaskTranscribe:exec - error');
       this.parentTask && this.parentTask.emit('error', err);
     }
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.Transcription);
+    this.removeSpeechListeners(ep);
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.EndOfUtterance);
-    ep.removeCustomEventListener(GoogleTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(AwsTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(AwsTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(AzureTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected);
-    ep.removeCustomEventListener(AzureTranscriptionEvents.VadDetected);
-
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.Transcription);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.TranscriptionComplete);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.StartOfSpeech);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.Error);
-    ep.removeCustomEventListener(NuanceTranscriptionEvents.VadDetected);
   }
 
   async kill(cs) {
@@ -184,6 +180,15 @@ class TaskTranscribe extends Task {
         ep.addCustomEventListener(AzureTranscriptionEvents.Error,
           this._onNuanceError.bind(this, cs, ep, channel));
         break;
+      case 'deepgram':
+        this.bugname = 'deepgram_transcribe';
+        ep.addCustomEventListener(DeepgramTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        ep.addCustomEventListener(DeepgramTranscriptionEvents.Connect,
+          this._onDeepgramConnect.bind(this, cs, ep, channel));
+        ep.addCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure,
+          this._onDeepGramConnectFailure.bind(this, cs, ep, channel));
+        break;
       default:
         throw new Error(`Invalid vendor ${this.vendor}`);
     }
@@ -215,9 +220,15 @@ class TaskTranscribe extends Task {
 
     this.logger.debug({evt}, 'TaskTranscribe:_onTranscription');
 
-    if (evt.alternatives[0].transcript === '' && !cs.callGone && !this.killed) {
+    if (evt.alternatives[0]?.transcript === '' && !cs.callGone && !this.killed) {
-      this.logger.info({evt}, 'TaskGather:_onTranscription - got empty transcript, listen again');
+      if (['microsoft', 'deepgram'].includes(this.vendor)) {
-      return this._transcribe(ep);
+        this.logger.info({evt}, 'TaskTranscribe:_onTranscription - got empty transcript, continue listening');
+      }
+      else {
+        this.logger.info({evt}, 'TaskTranscribe:_onTranscription - got empty transcript, listen again');
+        this._transcribe(ep);
+      }
+      return;
     }
 
     if (this.transcriptionHook) {
@@ -268,6 +279,34 @@ class TaskTranscribe extends Task {
       this._timer = null;
     }
   }
+  _onNuanceError(_cs, _ep, evt) {
+    const {code, error, details} = evt;
+    if (code === 404 && error === 'No speech') {
+      this.logger.debug({code, error, details}, 'TaskTranscribe:_onNuanceError');
+      return this._resolve('timeout');
+    }
+    this.logger.info({code, error, details}, 'TaskTranscribe:_onNuanceError');
+    if (code === 413 && error === 'Too much speech') {
+      return this._resolve('timeout');
+    }
+  }
+  _onDeepgramConnect(_cs, _ep) {
+    this.logger.debug('TaskTranscribe:_onDeepgramConnect');
+  }
+
+  _onDeepGramConnectFailure(cs, _ep, evt) {
+    const {reason} = evt;
+    const {writeAlerts, AlertType} = cs.srf.locals;
+    this.logger.info({evt}, 'TaskTranscribe:_onDeepgramConnectFailure');
+    writeAlerts({
+      account_sid: cs.accountSid,
+      alert_type: AlertType.STT_FAILURE,
+      message: `Failed connecting to Deepgram speech recognizer: ${reason}`,
+      vendor: 'deepgram',
+    }).catch((err) => this.logger.info({err}, 'Error generating alert for deepgram connection failure'));
+    this.notifyError(`Failed connecting to speech vendor deepgram: ${reason}`);
+    this.notifyTaskDone();
+  }
 }
 
 module.exports = TaskTranscribe;

lib/utils/constants.json

@@ -74,6 +74,11 @@
     "Error": "nuance_transcribe::error",
     "VadDetected": "nuance_transcribe::vad_detected"
   },
+  "DeepgramTranscriptionEvents": {
+    "Transcription": "deepgram_transcribe::transcription",
+    "ConnectFailure": "deepgram_transcribe::connect_failed",
+    "Connect": "deepgram_transcribe::connect"
+  },
   "AwsTranscriptionEvents": {
     "Transcription": "aws_transcribe::transcription",
     "EndOfTranscript": "aws_transcribe::end_of_transcript",

lib/utils/db-utils.js

@@ -51,6 +51,10 @@ const speechMapper = (cred) => {
       obj.client_id = o.client_id;
       obj.secret = o.secret;
     }
+    else if ('deepgram' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+    }
   } catch (err) {
     console.log(err);
   }
@@ -74,6 +78,7 @@ module.exports = (logger, srf) => {
     const haveMicrosoft = speech.find((s) => s.vendor === 'microsoft');
     const haveWellsaid = speech.find((s) => s.vendor === 'wellsaid');
     const haveNuance = speech.find((s) => s.vendor === 'nuance');
+    const haveDeepgram = speech.find((s) => s.vendor === 'deepgram');
     if (!haveGoogle || !haveAws || !haveMicrosoft || !haveWellsaid || !haveNuance) {
       const [r3] = await pp.query(sqlSpeechCredentialsForSP, account_sid);
       if (r3.length) {
@@ -97,6 +102,10 @@ module.exports = (logger, srf) => {
           const nuance = r3.find((s) => s.vendor === 'nuance');
           if (nuance) speech.push(speechMapper(nuance));
         }
+        if (!haveDeepgram) {
+          const deepgram = r3.find((s) => s.vendor === 'deepgram');
+          if (deepgram) speech.push(speechMapper(deepgram));
+        }
       }
     }
 

lib/utils/transcription-utils.js

@@ -3,52 +3,120 @@ const {
   AzureTranscriptionEvents,
   GoogleTranscriptionEvents,
   AwsTranscriptionEvents,
-  NuanceTranscriptionEvents
+  NuanceTranscriptionEvents,
+  DeepgramTranscriptionEvents,
 } = require('./constants');
 
+const normalizeDeepgram = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  const alternatives = (evt.channel?.alternatives || [])
+    .map((alt) => ({
+      confidence: alt.confidence,
+      transcript: alt.transcript,
+    }));
+
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.is_final,
+    alternatives,
+    vendor: {
+      name: 'deepgram',
+      evt: copy
+    }
+  };
+};
+
+const normalizeGoogle = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.is_final,
+    alternatives: evt.alternatives,
+    vendor: {
+      name: 'google',
+      evt: copy
+    }
+  };
+};
+
+const normalizeNuance = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.is_final,
+    alternatives: evt.alternatives,
+    vendor: {
+      name: 'nuance',
+      evt: copy
+    }
+  };
+};
+
+const normalizeMicrosoft = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  const nbest = evt.NBest;
+  const language_code = evt.PrimaryLanguage?.Language || language;
+  const alternatives = nbest ? nbest.map((n) => {
+    return {
+      confidence: n.Confidence,
+      transcript: n.Display
+    };
+  }) :
+    [
+      {
+        transcript: evt.DisplayText || evt.Text
+      }
+    ];
+
+  return {
+    language_code,
+    channel_tag: channel,
+    is_final: evt.RecognitionStatus === 'Success',
+    alternatives,
+    vendor: {
+      name: 'microsoft',
+      evt: copy
+    }
+  };
+};
+
+const normalizeAws = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt[0].is_final,
+    alternatives: evt[0].alternatives,
+    vendor: {
+      name: 'aws',
+      evt: copy
+    }
+  };
+};
+
+
 module.exports = (logger) => {
   const normalizeTranscription = (evt, vendor, channel, language) => {
-    let newEvent = JSON.parse(JSON.stringify(evt));
 
-    /* add in channel_tag and provide the full vendor-specific event */
+    logger.debug({ evt, vendor, channel, language }, 'normalizeTranscription');
-    newEvent = {
+    switch (vendor) {
-      ...(vendor === 'aws' ? newEvent[0] : newEvent),
+      case 'deepgram':
-      language_code: language,
+        return normalizeDeepgram(evt, channel, language);
-      channel_tag: channel
+      case 'microsoft':
-    };
+        return normalizeMicrosoft(evt, channel, language);
-
+      case 'google':
-
+        return normalizeGoogle(evt, channel, language);
-    if ('aws' === vendor && Array.isArray(evt) && evt.length > 0) {
+      case 'aws':
-      newEvent = {
+        return normalizeAws(evt, channel, language);
-        ...newEvent,
+      case 'nuance':
-        vendor: {event: evt, name: vendor}
+        return normalizeNuance(evt, channel, language);
-      };
+      default:
+        logger.error(`Unknown vendor ${vendor}`);
+        return evt;
     }
-    else if ('microsoft' === vendor) {
-      const nbest = evt.NBest;
-      const language_code = evt.PrimaryLanguage?.Language || language;
-      const alternatives = nbest ? nbest.map((n) => {
-        return {
-          confidence: n.Confidence,
-          transcript: n.Display
-        };
-      }) :
-        [
-          {
-            transcript: evt.DisplayText || evt.Text
-          }
-        ];
-
-      newEvent = {
-        ...newEvent,
-        is_final: evt.RecognitionStatus === 'Success',
-        channel,
-        language_code,
-        alternatives,
-        vendor: {event: evt, name: vendor}
-      };
-    }
-    return newEvent;
   };
 
   const setChannelVarsForStt = (task, sttCredentials, rOpts = {}) => {
@@ -201,6 +269,48 @@ module.exports = (logger) => {
           {NUANCE_RESOURCES: JSON.stringify(nuanceOptions.resources)},
       };
     }
+    else if ('deepgram' === rOpts.vendor) {
+      const {deepgramOptions = {}} = rOpts;
+      opts = {
+        ...opts,
+        ...(sttCredentials.api_key) &&
+          {DEEPGRAM_API_KEY: sttCredentials.api_key},
+        ...(deepgramOptions.tier) &&
+          {DEEPGRAM_SPEECH_TIER: deepgramOptions.tier},
+        ...(deepgramOptions.model) &&
+          {DEEPGRAM_SPEECH_MODEL: deepgramOptions.model},
+        ...(deepgramOptions.punctuate) &&
+          {DEEPGRAM_SPEECH_ENABLE_AUTOMATIC_PUNCTUATION: 1},
+        ...(deepgramOptions.profanityFilter) &&
+          {DEEPGRAM_SPEECH_PROFANITY_FILTER: 1},
+        ...(deepgramOptions.redact) &&
+          {DEEPGRAM_SPEECH_REDACT: 1},
+        ...(deepgramOptions.diarize) &&
+          {DEEPGRAM_SPEECH_DIARIZE: 1},
+        ...(deepgramOptions.diarizeVersion) &&
+          {DEEPGRAM_SPEECH_DIARIZE_VERSION: deepgramOptions.diarizeVersion},
+        ...(deepgramOptions.ner) &&
+          {DEEPGRAM_SPEECH_NER: 1},
+        ...(deepgramOptions.alternatives) &&
+          {DEEPGRAM_SPEECH_ALTERNATIVES: deepgramOptions.alternatives},
+        ...(deepgramOptions.numerals) &&
+          {DEEPGRAM_SPEECH_NUMERALS: deepgramOptions.numerals},
+        ...(deepgramOptions.search) &&
+          {DEEPGRAM_SPEECH_SEARCH: deepgramOptions.search.join(',')},
+        ...(deepgramOptions.replace) &&
+          {DEEPGRAM_SPEECH_REPLACE: deepgramOptions.replace.join(',')},
+        ...(rOpts.hints.length > 0 &&
+          {DEEPGRAM_SPEECH_KEYWORDS: rOpts.hints.join(',')}),
+        ...(deepgramOptions.keywords) &&
+          {DEEPGRAM_SPEECH_KEYWORDS: deepgramOptions.keywords.join(',')},
+        ...('endpointing' in deepgramOptions) &&
+          {DEEPGRAM_SPEECH_ENDPOINTING: deepgramOptions.endpointing},
+        ...(deepgramOptions.vadTurnoff) &&
+          {DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.vadTurnoff},
+        ...(deepgramOptions.tag) &&
+          {DEEPGRAM_SPEECH_VAD_TURNOFF: deepgramOptions.tag}
+      };
+    }
     logger.debug({opts}, 'recognizer channel vars');
     return opts;
   };
@@ -223,6 +333,11 @@ module.exports = (logger) => {
     ep.removeCustomEventListener(NuanceTranscriptionEvents.Error);
     ep.removeCustomEventListener(NuanceTranscriptionEvents.VadDetected);
 
+    ep.removeCustomEventListener(DeepgramTranscriptionEvents.Transcription);
+    ep.removeCustomEventListener(DeepgramTranscriptionEvents.Connect);
+    ep.removeCustomEventListener(DeepgramTranscriptionEvents.ConnectFailure);
+
+
   };
   return {
     normalizeTranscription,

package.json

@@ -26,7 +26,7 @@
   "dependencies": {
     "@jambonz/db-helpers": "^0.7.3",
     "@jambonz/http-health-check": "^0.0.1",
-    "@jambonz/realtimedb-helpers": "^0.5.7",
+    "@jambonz/realtimedb-helpers": "^0.6.0",
     "@jambonz/stats-collector": "^0.1.6",
     "@jambonz/time-series": "^0.2.5",
     "@opentelemetry/api": "^1.2.0",
@@ -42,7 +42,7 @@
     "bent": "^7.3.12",
     "debug": "^4.3.4",
     "deepcopy": "^2.1.0",
-    "drachtio-fsmrf": "^3.0.5",
+    "drachtio-fsmrf": "^3.0.6",
     "drachtio-srf": "^4.5.18",
     "express": "^4.18.2",
     "helmet": "^5.1.1",

test/docker-compose-testbed.yaml

@@ -57,7 +57,7 @@ services:
         condition: service_healthy
 
   freeswitch:
-    image: drachtio/drachtio-freeswitch-mrf:0.4.15
+    image: drachtio/drachtio-freeswitch-mrf:0.4.18
     restart: always
     command: freeswitch --rtp-range-start 20000 --rtp-range-end 20100
     environment:

test/gather-tests.js

@@ -45,7 +45,8 @@ test('\'gather\' test - google', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript = 'I\'d like to speak to customer support',
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'gather: succeeds when using google credentials');
 
     disconnect();
@@ -80,7 +81,8 @@ test('\'gather\' test - default (google)', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript = 'I\'d like to speak to customer support',
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase() === 'i\'d like to speak to customer support',
       'gather: succeeds when using default (google) credentials');
 
     disconnect();
@@ -119,7 +121,8 @@ test('\'gather\' test - microsoft', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript = 'I\'d like to speak to customer support',
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'gather: succeeds when using  microsoft credentials');
 
     disconnect();
@@ -158,7 +161,8 @@ test('\'gather\' test - aws', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript = 'I\'d like to speak to customer support',
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'gather: succeeds when using aws credentials');
 
     disconnect();
@@ -167,4 +171,47 @@ test('\'gather\' test - aws', async(t) => {
     disconnect();
     t.error(err);
   }
-});
+});
+
+test('\'gather\' test - deepgram', async(t) => {
+  if (!process.env.DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "gather",
+        "input": ["speech"],
+        "recognizer": {
+          "vendor": "deepgram",
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "deepgramOptions": {
+            "apiKey": process.env.DEEPGRAM_API_KEY
+          }
+        },
+        "timeout": 10,
+        "actionHook": "/actionHook"
+      }
+    ];
+    let from = "gather_success";
+    provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    //console.log(JSON.stringify(obj));
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
+      'gather: succeeds when using  deepgram credentials');
+
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});

test/transcribe-tests.js

@@ -43,7 +43,7 @@ test('\'transcribe\' test - google', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript = 'I\'d like to speak to customer support',
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using google credentials');
 
     disconnect();
@@ -80,7 +80,7 @@ test('\'transcribe\' test - microsoft', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript = 'I\'d like to speak to customer support',
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using  microsoft credentials');
 
     disconnect();
@@ -117,9 +117,49 @@ test('\'transcribe\' test - aws', async(t) => {
     // THEN
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
-    t.ok(obj.body.speech.alternatives[0].transcript = 'I\'d like to speak to customer support',
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
       'transcribe: succeeds when using aws credentials');
 
+    disconnect();
+  } catch (err) {
+    console.log(`error received: ${err}`);
+    disconnect();
+    t.error(err);
+  }
+});
+
+test('\'transcribe\' test - deepgram', async(t) => {
+  if (!process.env.DEEPGRAM_API_KEY ) {
+    t.pass('skipping deepgram tests');
+    return t.end();
+  }
+  clearModule.all();
+  const {srf, disconnect} = require('../app');
+
+  try {
+    await connect(srf);
+    // GIVEN
+    let verbs = [
+      {
+        "verb": "transcribe",
+        "recognizer": {
+          "vendor": "aws",
+          "hints": ["customer support", "sales", "human resources", "HR"],
+          "deepgramOptions": {
+            "apiKey": process.env.DEEPGRAM_API_KEY
+          }
+        },
+        "transcriptionHook": "/transcriptionHook"
+      }
+    ];
+    let from = "gather_success";
+    provisionCallHook(from, verbs);
+    // THEN
+    await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
+    let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
+      'transcribe: succeeds when using deepgram credentials');
+
     disconnect();
   } catch (err) {
     console.log(`error received: ${err}`);