support whisper tts (#255)

* support whisper tts

* wip

* update speech util version

lib/routes/api/speech-credentials.js

@@ -5,7 +5,7 @@ const SpeechCredential = require('../../models/speech-credential');
 const sysError = require('../error');
 const {decrypt, encrypt} = require('../../utils/encrypt-decrypt');
 const {parseAccountSid, parseServiceProviderSid, parseSpeechCredentialSid} = require('./utils');
-const {decryptCredential} = require('../../utils/speech-utils');
+const {decryptCredential, testWhisper} = require('../../utils/speech-utils');
 const {DbErrorUnprocessableRequest, DbErrorForbidden} = require('../../utils/errors');
 const {
   testGoogleTts,
@@ -218,6 +218,12 @@ const encryptCredential = (obj) => {
       const assemblyaiData = JSON.stringify({api_key});
       return encrypt(assemblyaiData);
 
+    case 'whisper':
+      assert(api_key, 'invalid whisper speech credential: api_key is required');
+      assert(model_id, 'invalid whisper speech credential: model_id is required');
+      const whisperData = JSON.stringify({api_key, model_id});
+      return encrypt(whisperData);
+
     default:
       if (vendor.startsWith('custom:')) {
         const customData = JSON.stringify({auth_token, custom_stt_url, custom_tts_url});
@@ -405,6 +411,7 @@ router.put('/:sid', async(req, res) => {
           custom_stt_url,
           custom_tts_url,
           cobalt_server_uri,
+          model_id
         } = req.body;
 
         const newCred = {
@@ -425,7 +432,8 @@ router.put('/:sid', async(req, res) => {
           nuance_tts_uri,
           custom_stt_url,
           custom_tts_url,
-          cobalt_server_uri
+          cobalt_server_uri,
+          model_id
         };
         logger.info({o, newCred}, 'updating speech credential with this new credential');
         obj.credential = encryptCredential(newCred);
@@ -454,7 +462,7 @@ router.put('/:sid', async(req, res) => {
  * Test a credential
  */
 router.get('/:sid/test', async(req, res) => {
-  const logger = req.app.locals.logger;
+  const {logger, synthAudio} = req.app.locals;
   try {
     const sid = parseSpeechCredentialSid(req);
     const creds = await SpeechCredential.retrieve(sid);
@@ -706,6 +714,17 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.sttTestResult(sid, false);
         }
       }
+    } else if (cred.vendor === 'whisper') {
+      if (cred.use_for_tts) {
+        try {
+          await testWhisper(logger, synthAudio, credential);
+          results.tts.status = 'ok';
+          SpeechCredential.ttsTestResult(sid, true);
+        } catch (err) {
+          results.tts = {status: 'fail', reason: err.message};
+          SpeechCredential.ttsTestResult(sid, false);
+        }
+      }
     }
 
     res.status(200).json(results);