add support for soniox speech (#120)

Co-authored-by: eglehelms <e.helms@cognigy.com>

lib/routes/api/speech-credentials.js

@@ -17,6 +17,7 @@ const {
   testNuanceStt,
   testNuanceTts,
   testDeepgramStt,
+  testSonioxStt,
   testIbmTts,
   testIbmStt
 } = require('../../utils/speech-utils');
@@ -112,6 +113,11 @@ const encryptCredential = (obj) => {
       const nvidiaData = JSON.stringify({ riva_server_uri });
       return encrypt(nvidiaData);
 
+    case 'soniox':
+      assert(api_key, 'invalid soniox speech credential: api_key is required');
+      const sonioxData = JSON.stringify({api_key});
+      return encrypt(sonioxData);
+
     default:
       assert(false, `invalid or missing vendor: ${vendor}`);
   }
@@ -220,6 +226,10 @@ router.get('/', async(req, res) => {
         const o = JSON.parse(decrypt(credential));
         obj.riva_server_uri = o.riva_server_uri;
       }
+      else if ('soniox' === obj.vendor) {
+        const o = JSON.parse(decrypt(credential));
+        obj.api_key = obscureKey(o.api_key);
+      }
       return obj;
     }));
   } catch (err) {
@@ -582,6 +592,18 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.sttTestResult(sid, false);
         }
       }
+    } else if (cred.vendor === 'soniox') {
+      const {api_key} = credential;
+      if (cred.use_for_stt) {
+        try {
+          await testSonioxStt(logger, {api_key});
+          results.stt.status = 'ok';
+          SpeechCredential.sttTestResult(sid, true);
+        } catch (err) {
+          results.stt = {status: 'fail', reason: err.message};
+          SpeechCredential.sttTestResult(sid, false);
+        }
+      }
     }
 
     res.status(200).json(results);

lib/utils/speech-utils.js

@@ -4,9 +4,27 @@ const Polly = require('aws-sdk/clients/polly');
 const AWS = require('aws-sdk');
 const { Deepgram } = require('@deepgram/sdk');
 const sdk = require('microsoft-cognitiveservices-speech-sdk');
+const { SpeechClient } = require('@soniox/soniox-node');
 const bent = require('bent');
 const fs = require('fs');
 
+
+const testSonioxStt = async(logger, credentials) => {
+  const api_key = credentials;
+  const soniox = new SpeechClient(api_key);
+
+  return new Promise(async(resolve, reject) => {
+    try {
+      const result = await soniox.transcribeFileShort('data/test_audio.wav');
+      if (result.words.length > 0) resolve(result);
+      else reject(new Error('no transcript returned'));
+    } catch (error) {
+      logger.info({error}, 'failed to get soniox transcript');
+      reject(error);
+    }
+  });
+};
+
 const testNuanceTts = async(logger, getTtsVoices, credentials) => {
   const voices = await getTtsVoices({vendor: 'nuance', credentials});
   return voices;
@@ -225,5 +243,6 @@ module.exports = {
   testNuanceStt,
   testDeepgramStt,
   testIbmTts,
-  testIbmStt
+  testIbmStt,
+  testSonioxStt
 };