Merge branch 'main' into feature/pagination

lib/routes/api/speech-credentials.js

@@ -752,7 +752,7 @@ router.get('/:sid/test', async(req, res) => {
  */
 
 router.get('/speech/supportedLanguagesAndVoices', async(req, res) => {
-  const logger = req.app.locals.logger;
+  const {logger, getTtsVoices} = req.app.locals;
   try {
     const {vendor, label} = req.query;
     if (!vendor) {
@@ -767,7 +767,7 @@ router.get('/speech/supportedLanguagesAndVoices', async(req, res) => {
     const tmp = credentials && credentials.length > 0 ? credentials[0] : null;
     const cred = tmp ? JSON.parse(decrypt(tmp.credential)) : null;
     try {
-      const data = await getLanguagesAndVoicesForVendor(logger, vendor, cred);
+      const data = await getLanguagesAndVoicesForVendor(logger, vendor, cred, getTtsVoices);
       res.status(200).json(data);
     } catch (err) {
       throw new DbErrorUnprocessableRequest(err.message);

lib/swagger/swagger.yaml

@@ -382,11 +382,35 @@ paths:
             application/json:
               schema:
                 $ref: '#/components/schemas/GeneralError'
+  /login:
+    post:
+      tags:
+        - Authentication
+      summary: login and retrieve a JWT
+      operationId: login
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/Login'
+      responses:
+        200:
+          description: user logged in
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/SuccessfulLogin' 
+        500:
+          description: system error
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/GeneralError'
   /logout:
     post:
       tags:
         - Authentication
-      summary: log out and deactivate jwt
+      summary: log out and deactivate the JWT
       operationId: logoutUser
       responses:
         204:
@@ -584,10 +608,9 @@ paths:
           content:
             application/json:
               schema:
-                type:
-                  array
+                type: array
                 items:
-                  $ref: '#/components/schemas/Users'
+                  $ref: '#/components/schemas/UserList'
         403:
           description: unauthorized
         500:
@@ -610,27 +633,13 @@ paths:
         - Users
       summary: retrieve user information
       operationId: getUser
-      requestBody:
-        content:
-          application/json:
-            schema:
-              type: object
-              properties:
-                name:
-                  type: string
-                email: 
-                  type: string
-                is_active:
-                  type: boolean
-                force_change:
-                  type: boolean
-                scope:
-                  type: string
-                permissions:
-                  type: array
       responses:
-        204:
+        200:
           description: user information
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/UserProfile'
         403:
           description: user information
           content:
@@ -674,6 +683,8 @@ paths:
                   type: string
                 permissions:
                   type: array
+                  items:
+                    type: string
       responses:
         204:
           description: user updated
@@ -712,6 +723,8 @@ paths:
                   type: string
                 permissions:
                   type: array
+                  items:
+                    type: string
                 old_password:
                   type: string
                   description: existing password, which is to be replaced
@@ -998,7 +1011,7 @@ paths:
             application/json:
               schema:
                 $ref: '#/components/schemas/GeneralError'
-  /AccountTest/:ServiceProviderSid:
+  /AccountTest/{ServiceProviderSid}:
     parameters:
     - name: ServiceProviderSid
       in: path
@@ -1971,7 +1984,7 @@ paths:
       tags:
         - Service Providers
       summary: add a VoiPCarrier to a service provider based on PredefinedCarrier template
-      operationId: createVoipCarrierFromTemplate
+      operationId: createVoipCarrierFromTemplateBySP
       responses:
         201:
           description:  voip carrier successfully created
@@ -2079,6 +2092,12 @@ paths:
       summary: get supported languages, voices and models
       operationId: supportedLanguagesAndVoices
       parameters:
+        - name: ServiceProviderSid
+          in: path
+          required: true
+          schema:
+            type: string
+            format: uuid
         - name: vendor
           in: query
           required: true
@@ -2920,7 +2939,7 @@ paths:
       tags:
         - Accounts
       summary: get a specific speech credential
-      operationId: getSpeechCredential
+      operationId: getSpeechCredentialByAccount
       responses:
         200:
           description: retrieve speech credentials for a specified account
@@ -2934,7 +2953,7 @@ paths:
       tags:
         - Accounts
       summary: update a speech credential
-      operationId: updateSpeechCredential
+      operationId: updateSpeechCredentialByAccount
       requestBody:
         content:
           application/json:
@@ -2955,7 +2974,7 @@ paths:
       tags:
         - Accounts
       summary: delete a speech credential
-      operationId: deleteSpeechCredential
+      operationId: deleteSpeechCredentialByAccount
       responses:
         204:
           description: credential successfully deleted
@@ -2966,8 +2985,14 @@ paths:
       tags:
         - Accounts
       summary: get supported languages, voices and models
-      operationId: supportedLanguagesAndVoices
+      operationId: supportedLanguagesAndVoicesByAccount
       parameters:
+        - name: AccountSid
+          in: path
+          required: true
+          schema:
+            type: string
+            format: uuid
         - name: vendor
           in: query
           required: true
@@ -2995,7 +3020,7 @@ paths:
       tags:
         - Accounts
       summary: test a speech credential
-      operationId: testSpeechCredential
+      operationId: testSpeechCredentialByAccount
       parameters:
         - name: AccountSid
           in: path
@@ -3241,7 +3266,7 @@ paths:
       tags:
         - Service Providers
       summary: retrieve pcap for a call
-      operationId: getRecentCallTrace
+      operationId: getRecentCallTraceBySP
       responses:
         200:
           description: retrieve sip trace data
@@ -3327,7 +3352,7 @@ paths:
       tags:
         - Service Providers
       summary: retrieve recent calls for an account
-      operationId: listRecentCalls
+      operationId: listRecentCallsBySP
       responses:
         200:
           description: retrieve recent call records for a specified account
@@ -3428,7 +3453,7 @@ paths:
       tags:
         - Service Providers
       summary: retrieve sip trace detail for a call
-      operationId: getRecentCallTrace
+      operationId: getRecentCallTraceByCallId
       responses:
         200:
           description: retrieve sip trace data
@@ -3455,7 +3480,7 @@ paths:
       tags:
         - Accounts
       summary: retrieve pcap for a call
-      operationId: getRecentCallTrace
+      operationId: getRecentCallTraceByAccount
       responses:
         200:
           description: retrieve sip trace data
@@ -3641,7 +3666,7 @@ paths:
       tags:
        - Accounts
       summary: retrieve alerts for an account
-      operationId: listAlerts
+      operationId: listAlertsByAccount
       responses:
         200:
           description: retrieve alerts for a specified account
@@ -4240,7 +4265,7 @@ paths:
       tags:
         - Accounts
       summary: retrieve online sip users for an account
-      operationId: listQueues
+      operationId: listRegisteredSipUsers
       responses:
         200:
           description: retrieve online sip users for an account
@@ -4254,7 +4279,7 @@ paths:
       tags:
         - Accounts
       summary: retrieve online sip users for an account by list of sip username
-      operationId: listRegisteredSipUsers
+      operationId: listRegisteredSipUsersByUsername
       requestBody:
         content:
           application/json:
@@ -4273,6 +4298,12 @@ paths:
                   $ref: '#/components/schemas/RegisteredClient'
   /Accounts/{AccountSid}/RegisteredSipUsers/{Client}:
     parameters:
+      - name: AccountSid
+        in: path
+        required: true
+        schema:
+          type: string
+          format: uuid
       - name: Client
         in: path
         required: true
@@ -4293,6 +4324,13 @@ paths:
               schema:
                 $ref: '#/components/schemas/RegisteredClient'
   /Accounts/{AccountSid}/TtsCache/Synthesize:
+    parameters:
+      - name: AccountSid
+        in: path
+        required: true
+        schema:
+          type: string
+          format: uuid
     post:
       tags:
         - Accounts
@@ -5031,17 +5069,32 @@ components:
       scheme: bearer
       bearerFormat: token
   schemas:
+    SuccessfulLogin:
+      type: object
+      required:
+        - username
+        - password
+      properties:
+        token: 
+          type: string
+        user_sid:
+          type: string
+        scope:
+          type: string
+        force_change:
+          type: boolean
+    
     Login:
       type: object
       properties:
-        user_sid: 
+        username: 
           type: string
-        api_token:
-          type: string
-        change_password:
-          type: boolean
+        password:
+          type: string        
       required:
-        - user_sid
+        - username
+        - password
+    
     SuccessfulApiKeyAdd:
       type: object
       required:
@@ -6096,8 +6149,23 @@ components:
           type: array
           items:
             $ref: '#/components/schemas/TtsModel'
-        
-
+    UserList:
+      type: object
+      properties:      
+        name:
+          type: string
+        email: 
+          type: string
+        is_active:
+          type: boolean
+        force_change:
+          type: boolean
+        scope:
+          type: string
+        permissions:
+          type: array
+          items:
+            type: string
 
 security:
   - bearerAuth: []

lib/utils/speech-utils.js

@@ -464,49 +464,90 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
       ]
     }
  */
-async function getLanguagesAndVoicesForVendor(logger, vendor, credential) {
+async function getLanguagesAndVoicesForVendor(logger, vendor, credential, getTtsVoices) {
   switch (vendor) {
     case 'google':
-      return await getLanguagesVoicesForGoogle(credential);
+      return await getLanguagesVoicesForGoogle(credential, getTtsVoices, logger);
     case 'aws':
-      return await getLanguagesVoicesForAws(credential);
+      return await getLanguagesVoicesForAws(credential, getTtsVoices, logger);
     case 'microsoft':
-      return await getLanguagesVoicesForMicrosoft(credential);
+      return await getLanguagesVoicesForMicrosoft(credential, getTtsVoices, logger);
     case 'wellsaid':
-      return await getLanguagesVoicesForWellsaid(credential);
+      return await getLanguagesVoicesForWellsaid(credential, getTtsVoices, logger);
     case 'nuance':
-      return await getLanguagesVoicesForNuane(credential);
+      return await getLanguagesVoicesForNuane(credential, getTtsVoices, logger);
     case 'deepgram':
-      return await getLanguagesVoicesForDeepgram(credential);
+      return await getLanguagesVoicesForDeepgram(credential, getTtsVoices, logger);
     case 'ibm':
-      return await getLanguagesVoicesForIbm(credential);
+      return await getLanguagesVoicesForIbm(credential, getTtsVoices, logger);
     case 'nvidia':
-      return await getLanguagesVoicesForNvida(credential);
+      return await getLanguagesVoicesForNvida(credential, getTtsVoices, logger);
     case 'cobalt':
-      return await getLanguagesVoicesForCobalt(credential);
+      return await getLanguagesVoicesForCobalt(credential, getTtsVoices, logger);
     case 'soniox':
-      return await getLanguagesVoicesForSoniox(credential);
+      return await getLanguagesVoicesForSoniox(credential, getTtsVoices, logger);
     case 'elevenlabs':
-      return await getLanguagesVoicesForElevenlabs(credential);
+      return await getLanguagesVoicesForElevenlabs(credential, getTtsVoices, logger);
     case 'assemblyai':
-      return await getLanguagesVoicesForAssemblyAI(credential);
+      return await getLanguagesVoicesForAssemblyAI(credential, getTtsVoices, logger);
     case 'whisper':
-      return await getLanguagesVoicesForWhisper(credential);
+      return await getLanguagesVoicesForWhisper(credential, getTtsVoices, logger);
     default:
       logger.info(`invalid vendor ${vendor}, return empty result`);
       throw new Error(`Invalid vendor ${vendor}`);
   }
 }
 
-async function getLanguagesVoicesForGoogle(credential) {
+async function getLanguagesVoicesForGoogle(credential, getTtsVoices, logger) {
+  if (credential) {
+    try {
+      const [result] = await getTtsVoices({
+        vendor: 'google',
+        credentials: credential
+      });
+      const tts = parseGooglelanguagesVoices(result.voices);
+      return tranform(tts, SttGoogleLanguagesVoices);
+    } catch (err) {
+      logger.info('Error while fetching google languages, voices, return predefined values', err);
+    }
+  }
   return tranform(TtsGoogleLanguagesVoices, SttGoogleLanguagesVoices);
 }
 
-async function getLanguagesVoicesForAws(credential) {
+async function getLanguagesVoicesForAws(credential, getTtsVoices, logger) {
+  if (credential) {
+    try {
+      const result = await getTtsVoices({
+        vendor: 'aws',
+        credentials: {
+          accessKeyId: credential.access_key_id,
+          secretAccessKey: credential.secret_access_key,
+          region: credential.aws_region || process.env.AWS_REGION
+        }
+      });
+      const tts = parseAwsLanguagesVoices(result.Voices);
+      return tranform(tts, SttAwsLanguagesVoices);
+    } catch (err) {
+      logger.info('Error while fetching AWS languages, voices, return predefined values', err);
+    }
+  }
   return tranform(TtsAwsLanguagesVoices, SttAwsLanguagesVoices);
 }
 
-async function getLanguagesVoicesForMicrosoft(credential) {
+async function getLanguagesVoicesForMicrosoft(credential, getTtsVoices, logger) {
+  if (credential) {
+    try {
+      const get = bent('https://westus.tts.speech.microsoft.com', 'GET', 'json', {
+        'Ocp-Apim-Subscription-Key' : credential.api_key
+      });
+
+      const voices = await get('/cognitiveservices/voices/list');
+      const tts = parseMicrosoftLanguagesVoices(voices);
+      return tranform(tts, SttMicrosoftLanguagesVoices);
+    } catch (err) {
+      logger.info('Error while fetching Microsoft languages, voices, return predefined values', err);
+    }
+  }
   return tranform(TtsMicrosoftLanguagesVoices, SttMicrosoftLanguagesVoices);
 }
 
@@ -514,7 +555,19 @@ async function getLanguagesVoicesForWellsaid(credential) {
   return tranform(TtsWellsaidLanguagesVoices);
 }
 
-async function getLanguagesVoicesForNuane(credential) {
+async function getLanguagesVoicesForNuane(credential, getTtsVoices, logger) {
+  if (credential) {
+    try {
+      const result = await getTtsVoices({
+        vendor: 'nuance',
+        credentials: credential
+      });
+      const tts = parseNuanceLanguagesVoices(result.result.voices);
+      return tranform(tts, SttNuanceLanguagesVoices);
+    } catch (err) {
+      logger.info('Error while fetching IBM languages, voices, return predefined values', err);
+    }
+  }
   return tranform(TtsNuanceLanguagesVoices, SttNuanceLanguagesVoices);
 }
 
@@ -522,7 +575,19 @@ async function getLanguagesVoicesForDeepgram(credential) {
   return tranform(undefined, SttDeepgramLanguagesVoices, TtsModelDeepgram);
 }
 
-async function getLanguagesVoicesForIbm(credential) {
+async function getLanguagesVoicesForIbm(credential, getTtsVoices, logger) {
+  if (credential) {
+    try {
+      const result = await getTtsVoices({
+        vendor: 'ibm',
+        credentials: credential
+      });
+      const tts = parseIBMLanguagesVoices(result.result.voices);
+      return tranform(tts, SttIbmLanguagesVoices);
+    } catch (err) {
+      logger.info('Error while fetching IBM languages, voices, return predefined values', err);
+    }
+  }
   return tranform(TtsIbmLanguagesVoices, SttIbmLanguagesVoices);
 }
 
@@ -596,6 +661,125 @@ function tranform(tts, stt, models) {
   };
 }
 
+function parseGooglelanguagesVoices(data) {
+  return data.reduce((acc, voice) => {
+    const languageCode = voice.languageCodes[0];
+    const existingLanguage = acc.find((lang) => lang.value === languageCode);
+
+    if (existingLanguage) {
+      existingLanguage.voices.push({
+        value: voice.name,
+        name: `${voice.name.substring(languageCode.length + 1, voice.name.length)} (${voice.ssmlGender})`
+      });
+    } else {
+      acc.push({
+        value: languageCode,
+        name: SttGoogleLanguagesVoices.find((lang) => lang.value === languageCode)?.name || languageCode,
+        voices: [{
+          value: voice.name,
+          name: `${voice.name.substring(languageCode.length + 1, voice.name.length)} (${voice.ssmlGender})`
+        }]
+      });
+    }
+
+    return acc;
+  }, []);
+}
+
+function parseIBMLanguagesVoices(data) {
+  return data.reduce((acc, voice) => {
+    const languageCode = voice.language;
+    const existingLanguage = acc.find((lang) => lang.value === languageCode);
+    if (existingLanguage) {
+      existingLanguage.voices.push({
+        value: voice.name,
+        name: `(${voice.gender}) ${voice.description}`
+      });
+    } else {
+      acc.push({
+        value: languageCode,
+        name: SttGoogleLanguagesVoices.find((lang) => lang.value === languageCode)?.name || languageCode,
+        voices: [{
+          value: voice.name,
+          name: `(${voice.gender}) ${voice.description}`
+        }]
+      });
+    }
+    return acc;
+  }, []);
+}
+
+function parseAwsLanguagesVoices(data) {
+  return data.reduce((acc, voice) => {
+    const languageCode = voice.LanguageCode;
+    const existingLanguage = acc.find((lang) => lang.value === languageCode);
+    if (existingLanguage) {
+      existingLanguage.voices.push({
+        value: voice.Id,
+        name: `(${voice.Gender}) ${voice.Name}`
+      });
+    } else {
+      acc.push({
+        value: languageCode,
+        name: voice.LanguageName,
+        voices: [{
+          value: voice.Id,
+          name: `(${voice.Gender}) ${voice.Name}`
+        }]
+      });
+    }
+    return acc;
+  }, []);
+}
+
+function parseNuanceLanguagesVoices(data) {
+  return data.reduce((acc, voice) => {
+    const languageCode = voice.language;
+    const existingLanguage = acc.find((lang) => lang.value === languageCode);
+    if (existingLanguage) {
+      existingLanguage.voices.push({
+        value: voice.name,
+        name: voice.name,
+        model: voice.model
+      });
+    } else {
+      acc.push({
+        value: languageCode,
+        name: SttGoogleLanguagesVoices.find((lang) => lang.value === languageCode)?.name || languageCode,
+        voices: [{
+          value: voice.name,
+          name: voice.name,
+          model: voice.model
+        }]
+      });
+    }
+    return acc;
+  }, []);
+}
+
+function parseMicrosoftLanguagesVoices(data) {
+  return data.reduce((acc, voice) => {
+    const languageCode = voice.Locale;
+    const existingLanguage = acc.find((lang) => lang.value === languageCode);
+    if (existingLanguage) {
+      existingLanguage.voices.push({
+        value: voice.ShortName,
+        name: `${voice.DisplayName} (${voice.Gender})`,
+      });
+    } else {
+      acc.push({
+        value: voice.Locale,
+        name: voice.LocaleName,
+        voices: [{
+          value: voice.ShortName,
+          name: `${voice.DisplayName} (${voice.Gender})`,
+        }]
+      });
+    }
+    return acc;
+  }, []);
+}
+
 module.exports = {
   testGoogleTts,
   testGoogleStt,

package.json

@@ -29,7 +29,7 @@
     "@jambonz/lamejs": "^1.2.2",
     "@jambonz/mw-registrar": "^0.2.7",
     "@jambonz/realtimedb-helpers": "^0.8.7",
-    "@jambonz/speech-utils": "^0.0.33",
+    "@jambonz/speech-utils": "^0.0.35",
     "@jambonz/time-series": "^0.2.8",
     "@jambonz/verb-specifications": "^0.0.45",
     "@soniox/soniox-node": "^1.1.1",