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
 };

package-lock.json

@@ -16,6 +16,7 @@
         "@jambonz/realtimedb-helpers": "^0.6.0",
         "@jambonz/time-series": "^0.2.5",
         "@jambonz/verb-specifications": "^0.0.3",
+        "@soniox/soniox-node": "^1.1.0",
         "argon2-ffi": "^2.0.0",
         "aws-sdk": "^2.1302.0",
         "bent": "^7.3.12",
@@ -841,6 +842,15 @@
       "resolved": "https://registry.npmjs.org/@protobufjs/utf8/-/utf8-1.1.0.tgz",
       "integrity": "sha512-Vvn3zZrhQZkkBE8LSuW3em98c0FwgO4nxzv6OdSxPKJIEKY2bGbHn+mhGIPerzI4twdxaP8/0+06HBpwf345Lw=="
     },
+    "node_modules/@soniox/soniox-node": {
+      "version": "1.1.0",
+      "resolved": "https://registry.npmjs.org/@soniox/soniox-node/-/soniox-node-1.1.0.tgz",
+      "integrity": "sha512-LtE1JZZ3PFAWyqJpp/UwYmMx1IbZ+JR/OzFmyQ1Oj4mrMrvGpH+lTMIpdG8s6/lDhhefoKP5OTziks8nQNkdew==",
+      "dependencies": {
+        "@grpc/grpc-js": "^1.6.10",
+        "@grpc/proto-loader": "^0.7.2"
+      }
+    },
     "node_modules/@tokenizer/token": {
       "version": "0.3.0",
       "resolved": "https://registry.npmjs.org/@tokenizer/token/-/token-0.3.0.tgz",
@@ -8273,6 +8283,15 @@
       "resolved": "https://registry.npmjs.org/@protobufjs/utf8/-/utf8-1.1.0.tgz",
       "integrity": "sha512-Vvn3zZrhQZkkBE8LSuW3em98c0FwgO4nxzv6OdSxPKJIEKY2bGbHn+mhGIPerzI4twdxaP8/0+06HBpwf345Lw=="
     },
+    "@soniox/soniox-node": {
+      "version": "1.1.0",
+      "resolved": "https://registry.npmjs.org/@soniox/soniox-node/-/soniox-node-1.1.0.tgz",
+      "integrity": "sha512-LtE1JZZ3PFAWyqJpp/UwYmMx1IbZ+JR/OzFmyQ1Oj4mrMrvGpH+lTMIpdG8s6/lDhhefoKP5OTziks8nQNkdew==",
+      "requires": {
+        "@grpc/grpc-js": "^1.6.10",
+        "@grpc/proto-loader": "^0.7.2"
+      }
+    },
     "@tokenizer/token": {
       "version": "0.3.0",
       "resolved": "https://registry.npmjs.org/@tokenizer/token/-/token-0.3.0.tgz",

package.json

@@ -25,6 +25,7 @@
     "@jambonz/realtimedb-helpers": "^0.6.0",
     "@jambonz/time-series": "^0.2.5",
     "@jambonz/verb-specifications": "^0.0.3",
+    "@soniox/soniox-node": "^1.1.0",
     "argon2-ffi": "^2.0.0",
     "aws-sdk": "^2.1302.0",
     "bent": "^7.3.12",

test/speech-credentials.js

@@ -300,6 +300,38 @@ test('speech credentials tests', async(t) => {
       t.ok(result.statusCode === 204, 'successfully deleted speech credential');
     }
 
+    /* add a credential for Siniox */
+    if (process.env.SONIOX_API_KEY) {
+      result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          vendor: 'soniox',
+          use_for_stt: true,
+          api_key: process.env.SONIOX_API_KEY
+        }
+      });
+      t.ok(result.statusCode === 201, 'successfully added speech credential for soniox');
+      const ms_sid = result.body.sid;
+
+      /* test the speech credential */
+      result = await request.get(`/Accounts/${account_sid}/SpeechCredentials/${ms_sid}/test`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,   
+      });
+      console.log(JSON.stringify(result));
+      t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for soniox');
+
+      /* delete the credential */
+      result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${ms_sid}`, {
+        auth: authUser,
+        resolveWithFullResponse: true,
+      });
+      t.ok(result.statusCode === 204, 'successfully deleted speech credential');
+    }
+
     /* add a credential for nvidia */
     result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
       resolveWithFullResponse: true,