support elevenlabs different endpoint (#502)

* support elevenlabs different endpoint

* wip

* wip

* wip

lib/routes/api/speech-credentials.js

@@ -162,6 +162,7 @@ const encryptCredential = (obj) => {
     voice_engine,
     engine_version,
     service_version,
+    api_uri,
     options
   } = obj;
 
@@ -238,10 +239,10 @@ const encryptCredential = (obj) => {
       });
       return encrypt(resembleData);
 
-    case 'deepgramriver':
-      assert(api_key, 'invalid deepgram river speech credential: api_key is required');
-      const deepgramriverData = JSON.stringify({api_key});
-      return encrypt(deepgramriverData);
+    case 'deepgramflux':
+      assert(api_key, 'invalid deepgram flux speech credential: api_key is required');
+      const deepgramfluxData = JSON.stringify({api_key});
+      return encrypt(deepgramfluxData);
 
     case 'ibm':
       const ibmData = JSON.stringify({tts_api_key, tts_region, stt_api_key, stt_region, instance_id});
@@ -265,7 +266,11 @@ const encryptCredential = (obj) => {
     case 'elevenlabs':
       assert(api_key, 'invalid elevenLabs speech credential: api_key is required');
       assert(model_id, 'invalid elevenLabs speech credential: model_id is required');
-      const elevenlabsData = JSON.stringify({api_key, model_id, options});
+      const elevenlabsData = JSON.stringify({
+        api_key,
+        model_id,
+        ...(api_uri && {api_uri}),
+        options});
       return encrypt(elevenlabsData);
 
     case 'speechmatics':
@@ -537,7 +542,8 @@ router.put('/:sid', async(req, res) => {
           service_version,
           speechmatics_stt_uri,
           resemble_tts_use_tls,
-          resemble_tts_uri
+          resemble_tts_uri,
+          api_uri
         } = req.body;
 
         const newCred = {
@@ -572,7 +578,8 @@ router.put('/:sid', async(req, res) => {
           service_version,
           speechmatics_stt_uri,
           resemble_tts_uri,
-          resemble_tts_use_tls
+          resemble_tts_use_tls,
+          api_uri
         };
         logger.info({o, newCred}, 'updating speech credential with this new credential');
         obj.credential = encryptCredential(newCred);
@@ -806,7 +813,7 @@ router.get('/:sid/test', async(req, res) => {
         }
       }
     }
-    else if (cred.vendor === 'deepgramriver') {
+    else if (cred.vendor === 'deepgramflux') {
       const {api_key} = credential;
       if (cred.use_for_stt && api_key) {
         try {
@@ -865,10 +872,10 @@ router.get('/:sid/test', async(req, res) => {
         }
       }
     } else if (cred.vendor === 'elevenlabs') {
-      const {api_key, model_id} = credential;
+      const {api_key, model_id, api_uri} = credential;
       if (cred.use_for_tts) {
         try {
-          await testElevenlabs(logger, {api_key, model_id});
+          await testElevenlabs(logger, {api_key, model_id, api_uri});
           results.tts.status = 'ok';
           SpeechCredential.ttsTestResult(sid, true);
         } catch (err) {

lib/utils/speech-utils.js

@@ -316,8 +316,8 @@ const testWellSaidTts = async(logger, credentials) => {
 };
 
 const testElevenlabs = async(logger, credentials) => {
-  const {api_key, model_id} = credentials;
-  const response = await fetch('https://api.elevenlabs.io/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM', {
+  const {api_key, model_id, api_uri} = credentials;
+  const response = await fetch(`https://${api_uri || 'api.elevenlabs.io'}/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM`, {
     method: 'POST',
     headers: {
       'xi-api-key': api_key,
@@ -686,7 +686,7 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
     obj.deepgram_tts_uri = o.deepgram_tts_uri;
     obj.model_id = o.model_id;
   }
-  else if ('deepgramriver' === obj.vendor) {
+  else if ('deepgramflux' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
   }
@@ -714,6 +714,7 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
     const o = JSON.parse(decrypt(credential));
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
     obj.model_id = o.model_id;
+    obj.api_uri = o.api_uri;
     obj.options = o.options;
   } else if ('playht' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
@@ -1014,10 +1015,12 @@ async function getLanguagesVoicesForElevenlabs(credential) {
       'xi-api-key': credential.api_key
     };
 
-    const getModelPromise = fetch('https://api.elevenlabs.io/v1/models', {
+    const api_uri = credential.api_uri || 'api.elevenlabs.io';
+
+    const getModelPromise = fetch(`https://${api_uri}/v1/models`, {
       headers
     });
-    const getVoicePromise = fetch('https://api.elevenlabs.io/v1/voices', {
+    const getVoicePromise = fetch(`https://${api_uri}/v1/voices`, {
       headers
     });
     const [langResp, voiceResp] = await Promise.all([getModelPromise, getVoicePromise]);

package-lock.json

@@ -20,9 +20,9 @@
         "@jambonz/lamejs": "^1.2.2",
         "@jambonz/mw-registrar": "^0.2.7",
         "@jambonz/realtimedb-helpers": "^0.8.15",
-        "@jambonz/speech-utils": "^0.2.23",
+        "@jambonz/speech-utils": "^0.2.25",
         "@jambonz/time-series": "^0.2.8",
-        "@jambonz/verb-specifications": "^0.0.111",
+        "@jambonz/verb-specifications": "^0.0.115",
         "@soniox/soniox-node": "^1.2.2",
         "ajv": "^8.17.1",
         "argon2": "^0.40.1",
@@ -4086,9 +4086,9 @@
       }
     },
     "node_modules/@jambonz/speech-utils": {
-      "version": "0.2.23",
-      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.2.23.tgz",
-      "integrity": "sha512-o28IBoKzdnQoUUSC1XljHVkDPWhkTH+rFnI9OWYC6p1/f8px++4Y23/JMIAJVbxqKB1CIf531JhTwy4tCnQP7g==",
+      "version": "0.2.25",
+      "resolved": "https://registry.npmjs.org/@jambonz/speech-utils/-/speech-utils-0.2.25.tgz",
+      "integrity": "sha512-dtMY4SoGhpvKM6slPvTlMz61Fd0Rxnajwm3sJVxlAYZnFtzfIvqg9rrYUN4h05tJbo/4TLs15/dVeicw7dTnew==",
       "license": "MIT",
       "dependencies": {
         "23": "^0.0.0",
@@ -4254,9 +4254,9 @@
       }
     },
     "node_modules/@jambonz/verb-specifications": {
-      "version": "0.0.111",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.111.tgz",
-      "integrity": "sha512-P2lDki5wCHGPPXpPTeMdht4NfGujF6NrLyrI8o/4U6l+2elfc/1mKl/Lz/PWGo9rQM4mTOVIhkk9ECo0NvOVoA==",
+      "version": "0.0.115",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.115.tgz",
+      "integrity": "sha512-T2J5skCZBADaxePkFqpm0JYwr5Iu6e9zgvk6j0/4vD6lwK29GSoSdk2LCKUY6VH2GYdaN6r3dyIw4xpFVXuMfQ==",
       "license": "MIT",
       "dependencies": {
         "debug": "^4.3.4",

package.json

@@ -31,9 +31,9 @@
     "@jambonz/lamejs": "^1.2.2",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.15",
-    "@jambonz/speech-utils": "^0.2.23",
+    "@jambonz/speech-utils": "^0.2.25",
     "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.111",
+    "@jambonz/verb-specifications": "^0.0.115",
     "@soniox/soniox-node": "^1.2.2",
     "ajv": "^8.17.1",
     "argon2": "^0.40.1",

test/speech-credentials.js

@@ -656,7 +656,8 @@ test('speech credentials tests', async(t) => {
         use_for_stt: true,
         use_for_tts: false,
         api_key: 'asdasdasdasddsadasda',
-        model_id: 'eleven_multilingual_v2'
+        model_id: 'eleven_multilingual_v2',
+        api_uri: 'api.elevenlabs.io'
       }
     });
     t.ok(result.statusCode === 201, 'successfully added speech credential for elevenlabs');
@@ -930,21 +931,21 @@ test('speech credentials tests', async(t) => {
       auth: authUser,
       json: true,
       body: {
-        vendor: 'deepgramriver',
+        vendor: 'deepgramflux',
         use_for_tts: false,
         use_for_stt: true,
         api_key: 'api_key',
       }
     });
     t.ok(result.statusCode === 201, 'successfully added speech credential for Verbio');
-    const deepgramriverSid = result.body.sid;
+    const deepgramfluxSid = result.body.sid;
 
     /* delete the credential */
-    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${deepgramriverSid}`, {
+    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${deepgramfluxSid}`, {
       auth: authUser,
       resolveWithFullResponse: true,
     });
-    t.ok(result.statusCode === 204, 'successfully deleted speech credential deepgramriver');
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential deepgramflux');
 
     /* Check google supportedLanguagesAndVoices */
     result = await request.get(`/Accounts/${account_sid}/SpeechCredentials/speech/supportedLanguagesAndVoices?vendor=google`, {