add eleven labs (#240)

* add eleven labs

* wip

* add voices and languages endpoints

lib/models/speech-credential.js

@@ -20,7 +20,7 @@ class SpeechCredential extends Model {
     return rows;
   }
 
-  static async isAvailableVendorAndLabel(service_provider_sid, account_sid, vendor, label) {
+  static async getSpeechCredentialsByVendorAndLabel(service_provider_sid, account_sid, vendor, label) {
     let sql;
     if (account_sid) {
       sql = 'SELECT * FROM speech_credentials WHERE account_sid = ? AND vendor = ? AND label = ?';

lib/routes/api/speech-credentials.js

@@ -19,8 +19,10 @@ const {
   testDeepgramStt,
   testSonioxStt,
   testIbmTts,
-  testIbmStt
+  testIbmStt,
+  testElevenlabs
 } = require('../../utils/speech-utils');
+const bent = require('bent');
 const {promisePool} = require('../../db');
 
 const validateAdd = async(req) => {
@@ -127,7 +129,8 @@ const encryptCredential = (obj) => {
     custom_stt_url,
     custom_tts_url,
     auth_token = '',
-    cobalt_server_uri
+    cobalt_server_uri,
+    model_id
   } = obj;
 
   switch (vendor) {
@@ -202,6 +205,12 @@ const encryptCredential = (obj) => {
       const cobaltData = JSON.stringify({cobalt_server_uri});
       return encrypt(cobaltData);
 
+    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});
+      return encrypt(elevenlabsData);
+
     default:
       if (vendor.startsWith('custom:')) {
         const customData = JSON.stringify({auth_token, custom_stt_url, custom_tts_url});
@@ -236,7 +245,7 @@ router.post('/', async(req, res) => {
 
     // Check if vendor and label is already used for account or SP
     if (label) {
-      const existingSpeech = await SpeechCredential.isAvailableVendorAndLabel(
+      const existingSpeech = await SpeechCredential.getSpeechCredentialsByVendorAndLabel(
         service_provider_sid, account_sid, vendor, label);
       if (existingSpeech.length > 0) {
         throw new DbErrorUnprocessableRequest(`Label ${label} is already in use for another speech credential`);
@@ -317,6 +326,10 @@ function decryptCredential(obj, credential, logger) {
   } else if ('soniox' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
     obj.api_key = obscureKey(o.api_key);
+  } else if ('elevenlabs' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = obscureKey(o.api_key);
+    obj.model_id = o.model_id;
   } else if (obj.vendor.startsWith('custom:')) {
     const o = JSON.parse(decrypt(credential));
     obj.auth_token = obscureKey(o.auth_token);
@@ -454,7 +467,7 @@ router.put('/:sid', async(req, res) => {
           custom_stt_endpoint_url,
           custom_stt_url,
           custom_tts_url,
-          cobalt_server_uri
+          cobalt_server_uri,
         } = req.body;
 
         const newCred = {
@@ -732,6 +745,18 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.sttTestResult(sid, false);
         }
       }
+    } else if (cred.vendor === 'elevenlabs') {
+      const {api_key, model_id} = credential;
+      if (cred.use_for_tts) {
+        try {
+          await testElevenlabs(logger, {api_key, model_id});
+          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);
@@ -741,4 +766,81 @@ router.get('/:sid/test', async(req, res) => {
   }
 });
 
+/**
+ * Fetch speech voices and languages
+ */
+
+router.post('/voices', async(req, res) => {
+  const logger = req.app.locals.logger;
+  const {vendor, label} = req.body;
+  const account_sid = req.user.account_sid || req.body.account_sid;
+  const service_provider_sid = req.user.service_provider_sid ||
+    req.body.service_provider_sid || parseServiceProviderSid(req);
+  try {
+    res.status(200).json(await getTtsVoices(vendor, label, service_provider_sid, account_sid));
+  } catch (err) {
+    sysError(logger, res, err);
+  }
+});
+
+router.post('/languages', async(req, res) => {
+  const logger = req.app.locals.logger;
+  const {vendor, label} = req.body;
+  const account_sid = req.user.account_sid || req.body.account_sid;
+  const service_provider_sid = req.user.service_provider_sid ||
+    req.body.service_provider_sid || parseServiceProviderSid(req);
+  try {
+    res.status(200).json(await getTtsLanguages(vendor, label, service_provider_sid, account_sid));
+  } catch (err) {
+    sysError(logger, res, err);
+  }
+});
+
+const getTtsVoices = async(vendor, label, service_provider_sid, account_sid) => {
+  const credentials = await SpeechCredential.isAvailableVendorAndLabel(
+    service_provider_sid, account_sid, vendor, label);
+  const cred = credentials && credentials.length > 0 ? credentials[0] : null;
+  if (vendor === 'elevenlabs') {
+    const get = bent('https://api.elevenlabs.io', 'GET', 'json', {
+      ...(cred && {
+        'xi-api-key' : cred.api_key
+      })
+    });
+    const resp = await get('/v1/voices');
+    return resp ? resp.voices.map((v) => {
+      return {
+        value: v.voice_id,
+        name: `${v.name} - ${v.labels.accent}, ${v.labels.description},
+ ${v.labels.age}, ${v.labels.gender}, ${v.labels['use case']}`
+      };
+    }) : [];
+  }
+  return [];
+};
+
+const getTtsLanguages = async(vendor, label, service_provider_sid, account_sid) => {
+  const credentials = await SpeechCredential.isAvailableVendorAndLabel(
+    service_provider_sid, account_sid, vendor, label);
+  const cred = credentials && credentials.length > 0 ? credentials[0] : null;
+  if (vendor === 'elevenlabs') {
+    if (!cred) {
+      return [];
+    }
+    const get = bent('https://api.elevenlabs.io', 'GET', 'json', {
+      'xi-api-key' : cred.api_key
+    });
+    const resp = await get('/v1/models');
+    if (!resp || resp.length === 0) {
+      return [];
+    }
+    const model = resp.find((m) => m.model_id === cred.model_id);
+    return model ? model.languages.map((l) => {
+      return {
+        value: l.language_id,
+        name: l.name
+      };
+    }) : [];
+  }
+};
+
 module.exports = router;

lib/utils/speech-utils.js

@@ -203,6 +203,29 @@ const testWellSaidTts = async(logger, credentials) => {
   }
 };
 
+const testElevenlabs = async(logger, credentials) => {
+  const {api_key, model_id} = credentials;
+  try {
+    const post = bent('https://api.elevenlabs.io', 'POST', 'buffer', {
+      'xi-api-key': api_key,
+      'Accept': 'audio/mpeg',
+      'Content-Type': 'application/json'
+    });
+    const mp3 = await post('/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM', {
+      text: 'Hello',
+      model_id,
+      voice_settings: {
+        stability: 0.5,
+        similarity_boost: 0.5
+      }
+    });
+    return mp3;
+  } catch (err) {
+    logger.info({err}, 'synthEvenlabs returned error');
+    throw err;
+  }
+};
+
 const testIbmTts = async(logger, getTtsVoices, credentials) => {
   const {tts_api_key, tts_region} = credentials;
   const voices = await getTtsVoices({vendor: 'ibm', credentials: {tts_api_key, tts_region}});
@@ -251,5 +274,6 @@ module.exports = {
   testDeepgramStt,
   testIbmTts,
   testIbmStt,
-  testSonioxStt
+  testSonioxStt,
+  testElevenlabs
 };