From dbdc1cd43d1522483239245967232c86357032ec Mon Sep 17 00:00:00 2001
From: Hoan Luu Huu <110280845+xquanluu@users.noreply.github.com>
Date: Wed, 5 Feb 2025 20:33:35 +0700
Subject: [PATCH] support voxist stt (#1066)

* support voxist stt

* wip
---
 lib/session/call-session.js      |  6 ++++++
 lib/tasks/gather.js              | 12 ++++++++++++
 lib/tasks/transcribe.js          | 12 ++++++++++++
 lib/utils/constants.json         |  6 ++++++
 lib/utils/db-utils.js            |  4 ++++
 lib/utils/transcription-utils.js | 31 +++++++++++++++++++++++++++++++
 6 files changed, 71 insertions(+)

diff --git a/lib/session/call-session.js b/lib/session/call-session.js
index 43a73bfe..885aa672 100644
--- a/lib/session/call-session.js
+++ b/lib/session/call-session.js
@@ -1084,6 +1084,12 @@ class CallSession extends Emitter {
             api_key: credential.api_key
           };
         }
+        else if ('voxist' === vendor) {
+          return {
+            speech_credential_sid: credential.speech_credential_sid,
+            api_key: credential.api_key
+          };
+        }
         else if ('whisper' === vendor) {
           return {
             api_key: credential.api_key,
diff --git a/lib/tasks/gather.js b/lib/tasks/gather.js
index 8f1ecc47..7013805d 100644
--- a/lib/tasks/gather.js
+++ b/lib/tasks/gather.js
@@ -11,6 +11,7 @@ const {
   NvidiaTranscriptionEvents,
   JambonzTranscriptionEvents,
   AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
   VadDetection,
   VerbioTranscriptionEvents,
   SpeechmaticsTranscriptionEvents
@@ -524,6 +525,17 @@ class TaskGather extends SttTask {
           this._onVendorConnectFailure.bind(this, cs, ep));
         break;
 
+      case 'voxist':
+        this.bugname = `${this.bugname_prefix}voxist_transcribe`;
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep));
+        this.addCustomEventListener(
+          ep, VoxistTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep));
+        break;
+
       case 'speechmatics':
         this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
         this.addCustomEventListener(
diff --git a/lib/tasks/transcribe.js b/lib/tasks/transcribe.js
index 4933152a..051c0816 100644
--- a/lib/tasks/transcribe.js
+++ b/lib/tasks/transcribe.js
@@ -13,6 +13,7 @@ const {
   JambonzTranscriptionEvents,
   TranscribeStatus,
   AssemblyAiTranscriptionEvents,
+  VoxistTranscriptionEvents,
   VerbioTranscriptionEvents,
   SpeechmaticsTranscriptionEvents
 } = require('../utils/constants.json');
@@ -300,6 +301,17 @@ class TaskTranscribe extends SttTask {
           this._onVendorConnectFailure.bind(this, cs, ep, channel));
         break;
 
+      case 'voxist':
+        this.bugname = `${this.bugname_prefix}voxist_transcribe`;
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Transcription,
+          this._onTranscription.bind(this, cs, ep, channel));
+        this.addCustomEventListener(ep,
+          VoxistTranscriptionEvents.Connect, this._onVendorConnect.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.Error, this._onVendorError.bind(this, cs, ep));
+        this.addCustomEventListener(ep, VoxistTranscriptionEvents.ConnectFailure,
+          this._onVendorConnectFailure.bind(this, cs, ep, channel));
+        break;
+
       case 'speechmatics':
         this.bugname = `${this.bugname_prefix}speechmatics_transcribe`;
         this.addCustomEventListener(
diff --git a/lib/utils/constants.json b/lib/utils/constants.json
index 2d292f3e..cdaa26db 100644
--- a/lib/utils/constants.json
+++ b/lib/utils/constants.json
@@ -149,6 +149,12 @@
     "ConnectFailure": "assemblyai_transcribe::connect_failed",
     "Connect": "assemblyai_transcribe::connect"
   },
+  "VoxistTranscriptionEvents": {
+    "Transcription": "voxist_transcribe::transcription",
+    "Error": "voxist_transcribe::error",
+    "ConnectFailure": "voxist_transcribe::connect_failed",
+    "Connect": "voxist_transcribe::connect"
+  },
   "VadDetection": {
     "Detection": "vad_detect:detection"
   },
diff --git a/lib/utils/db-utils.js b/lib/utils/db-utils.js
index 9884f53c..2b4b78c9 100644
--- a/lib/utils/db-utils.js
+++ b/lib/utils/db-utils.js
@@ -122,6 +122,10 @@ const speechMapper = (cred) => {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
     }
+    else if ('voxist' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = o.api_key;
+    }
     else if ('whisper' === obj.vendor) {
       const o = JSON.parse(decrypt(credential));
       obj.api_key = o.api_key;
diff --git a/lib/utils/transcription-utils.js b/lib/utils/transcription-utils.js
index f2b083c5..cca1b561 100644
--- a/lib/utils/transcription-utils.js
+++ b/lib/utils/transcription-utils.js
@@ -105,6 +105,9 @@ const stickyVars = {
     'ASSEMBLYAI_API_KEY',
     'ASSEMBLYAI_WORD_BOOST'
   ],
+  voxist: [
+    'VOXIST_API_KEY',
+  ],
   speechmatics: [
     'SPEECHMATICS_API_KEY',
     'SPEECHMATICS_HOST',
@@ -517,6 +520,25 @@ const normalizeAssemblyAi = (evt, channel, language) => {
   };
 };
 
+const normalizeVoxist = (evt, channel, language) => {
+  const copy = JSON.parse(JSON.stringify(evt));
+  return {
+    language_code: language,
+    channel_tag: channel,
+    is_final: evt.type === 'final',
+    alternatives: [
+      {
+        confidence: 1.00,
+        transcript: evt.text,
+      }
+    ],
+    vendor: {
+      name: 'voxist',
+      evt: copy
+    }
+  };
+};
+
 const normalizeSpeechmatics = (evt, channel, language) => {
   const copy = JSON.parse(JSON.stringify(evt));
   const is_final = evt.message === 'AddTranscript';
@@ -567,6 +589,8 @@ module.exports = (logger) => {
         return normalizeCobalt(evt, channel, language);
       case 'assemblyai':
         return normalizeAssemblyAi(evt, channel, language, shortUtterance);
+      case 'voxist':
+        return normalizeVoxist(evt, channel, language);
       case 'verbio':
         return normalizeVerbio(evt, channel, language);
       case 'speechmatics':
@@ -926,6 +950,13 @@ module.exports = (logger) => {
           {ASSEMBLYAI_WORD_BOOST: JSON.stringify(rOpts.hints)})
       };
     }
+    else if ('voxist' === vendor) {
+      opts = {
+        ...opts,
+        ...(sttCredentials.api_key) &&
+          {VOXIST_API_KEY: sttCredentials.api_key},
+      };
+    }
     else if ('verbio' === vendor) {
       const {verbioOptions = {}} = rOpts;
       opts = {