support gladia stt (#503)

* support gladia stt

* wip

* update verb specification

lib/utils/speech-utils.js

@@ -6,6 +6,7 @@ const { SpeechClient } = require('@soniox/soniox-node');
 const fs = require('fs');
 const { AssemblyAI } = require('assemblyai');
 const Houndify = require('houndify');
+const { GladiaClient } = require('@gladiaio/sdk');
 const {decrypt, obscureKey} = require('./encrypt-decrypt');
 const { RealtimeSession } = require('speechmatics');
 
@@ -50,6 +51,7 @@ const SttHoundifyLanguagesVoices = require('./speech-data/stt-houndify');
 const SttVoxistLanguagesVoices = require('./speech-data/stt-voxist');
 const SttVerbioLanguagesVoices = require('./speech-data/stt-verbio');
 const SttOpenaiLanguagesVoices = require('./speech-data/stt-openai');
+const SttGladiaLanguagesVoices = require('./speech-data/stt-gladia');
 
 
 const SttModelOpenai = require('./speech-data/stt-model-openai');
@@ -170,6 +172,65 @@ const testGoogleStt = async(logger, credentials) => {
   }
 };
 
+const testGladiaStt = async(logger, credentials) => {
+  const {api_key} = credentials;
+
+  try {
+    const gladiaClient = new GladiaClient({
+      apiKey: api_key,
+    });
+    const gladiaConfig = {
+      model: 'solaria-1',
+      encoding: 'wav/pcm',
+      sample_rate: 16000,
+      bit_depth: 16,
+      channels: 1,
+      language_config: {
+        languages: ['en'],
+        code_switching: false,
+      },
+    };
+    // Start the live session
+    const liveSession = gladiaClient.liveV2().startSession(gladiaConfig);
+    // Read the test audio file
+    const audioBuffer = fs.readFileSync(`${__dirname}/../../data/test_audio.wav`);
+
+    // Wait for final transcript
+    return new Promise((resolve, reject) => {
+      liveSession.on('message', (message) => {
+        if (message.type === 'transcript' && message.data.is_final) {
+          logger.debug(`${message.data.id}: ${message.data.utterance.text}`);
+          liveSession.stopRecording();
+          resolve(message.data.utterance.text);
+        }
+      });
+
+      liveSession.on('error', (error) => {
+        logger.error({error}, 'Gladia Live STT error');
+        reject(error);
+      });
+
+      // Send audio in chunks
+      const chunkSize = 1024;
+      for (let i = 0; i < audioBuffer.length; i += chunkSize) {
+        const chunk = audioBuffer.slice(i, i + chunkSize);
+        liveSession.sendAudio(chunk);
+      }
+      // Stop recording after sending all audio
+      liveSession.stopRecording();
+
+      // Set a timeout to prevent hanging
+      setTimeout(() => {
+        reject(new Error('Gladia STT test timeout'));
+      }, 30000); // 30 second timeout
+    });
+
+  } catch (error) {
+    logger.error({error}, 'Failed to create Gladia Live STT session');
+    throw error;
+  }
+};
+
 const testDeepgramStt = async(logger, credentials) => {
   const {api_key, deepgram_stt_uri, deepgram_stt_use_tls} = credentials;
   const deepgram = new Deepgram(api_key, deepgram_stt_uri, deepgram_stt_uri && deepgram_stt_use_tls);
@@ -758,6 +819,10 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
     const o = JSON.parse(decrypt(credential));
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
   }
+  else if ('gladia' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+  }
   else if ('ibm' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
     obj.tts_api_key = isObscureKey ? obscureKey(o.tts_api_key) : o.tts_api_key;
@@ -881,6 +946,8 @@ async function getLanguagesAndVoicesForVendor(logger, vendor, credential, getTts
       return await getLanguagesVoicesForNuane(credential, getTtsVoices, logger);
     case 'deepgram':
       return await getLanguagesVoicesForDeepgram(credential, getTtsVoices, logger);
+    case 'gladia':
+      return await getLanguagesVoicesForGladia(credential, getTtsVoices, logger);
     case 'ibm':
       return await getLanguagesVoicesForIbm(credential, getTtsVoices, logger);
     case 'nvidia':
@@ -1052,6 +1119,11 @@ async function getLanguagesVoicesForDeepgram(credential, getTtsVoices, logger) {
     TtsModelDeepgram, sttModelDeepgram.sort((a, b) => a.name.localeCompare(b.name)));
 }
 
+async function getLanguagesVoicesForGladia(credential, getTtsVoices, logger) {
+  return tranform(undefined, SttGladiaLanguagesVoices.sort((a, b) => a.name.localeCompare(b.name)),
+    undefined, undefined);
+}
+
 async function getLanguagesVoicesForIbm(credential, getTtsVoices, logger) {
   if (credential) {
     try {
@@ -1706,6 +1778,7 @@ module.exports = {
   testNuanceTts,
   testNuanceStt,
   testDeepgramStt,
+  testGladiaStt,
   testIbmTts,
   testIbmStt,
   testSonioxStt,