support mod_playht_tts (#304)

* support mod_playht_tts

* wip

* wip

* wip

* wip

* wip

* update speech utils version

db/jambones-sql.sql

@@ -162,7 +162,7 @@ regex VARCHAR(32) NOT NULL COMMENT 'regex-based pattern match against dialed num
 description VARCHAR(1024),
 priority INTEGER NOT NULL COMMENT 'lower priority routes are attempted first',
 PRIMARY KEY (lcr_route_sid)
-) COMMENT='An ordered list of  digit patterns in an LCR table.  The pat';
+) COMMENT='An ordered list of  digit patterns in an LCR table.  The patterns are tested in sequence until one matches';
 
 CREATE TABLE lcr
 (
@@ -173,7 +173,7 @@ default_carrier_set_entry_sid CHAR(36) COMMENT 'default carrier/route to use whe
 service_provider_sid CHAR(36),
 account_sid CHAR(36),
 PRIMARY KEY (lcr_sid)
-) COMMENT='An LCR (least cost routing) table that is used by a service ';
+) COMMENT='An LCR (least cost routing) table that is used by a service provider or account to make decisions about routing outbound calls when multiple carriers are available.';
 
 CREATE TABLE password_settings
 (
@@ -458,6 +458,7 @@ inbound BOOLEAN NOT NULL COMMENT 'if true, whitelist this IP to allow inbound ca
 outbound BOOLEAN NOT NULL COMMENT 'if true, include in least-cost routing when placing calls to the PSTN',
 voip_carrier_sid CHAR(36) NOT NULL,
 is_active BOOLEAN NOT NULL DEFAULT 1,
+send_options_ping BOOLEAN NOT NULL DEFAULT 0,
 pad_crypto BOOLEAN NOT NULL DEFAULT 0,
 protocol ENUM('udp','tcp','tls', 'tls/srtp') DEFAULT 'udp' COMMENT 'Outbound call protocol',
 PRIMARY KEY (sip_gateway_sid)
@@ -495,7 +496,7 @@ messaging_hook_sid CHAR(36) COMMENT 'webhook to call for inbound SMS/MMS ',
 app_json TEXT,
 speech_synthesis_vendor VARCHAR(64) NOT NULL DEFAULT 'google',
 speech_synthesis_language VARCHAR(12) NOT NULL DEFAULT 'en-US',
-speech_synthesis_voice VARCHAR(64),
+speech_synthesis_voice VARCHAR(256),
 speech_synthesis_label VARCHAR(64),
 speech_recognizer_vendor VARCHAR(64) NOT NULL DEFAULT 'google',
 speech_recognizer_language VARCHAR(64) NOT NULL DEFAULT 'en-US',
@@ -503,7 +504,7 @@ speech_recognizer_label VARCHAR(64),
 use_for_fallback_speech BOOLEAN DEFAULT false,
 fallback_speech_synthesis_vendor VARCHAR(64),
 fallback_speech_synthesis_language VARCHAR(12),
-fallback_speech_synthesis_voice VARCHAR(64),
+fallback_speech_synthesis_voice VARCHAR(256),
 fallback_speech_synthesis_label VARCHAR(64),
 fallback_speech_recognizer_vendor VARCHAR(64),
 fallback_speech_recognizer_language VARCHAR(64),

db/jambones.sqs

@@ -551,7 +551,7 @@
         </location>
         <size>
             <width>293.00</width>
-            <height>540.00</height>
+            <height>560.00</height>
         </size>
         <zorder>6</zorder>
         <SQLField>
@@ -2332,7 +2332,7 @@
         </location>
         <size>
             <width>281.00</width>
-            <height>240.00</height>
+            <height>260.00</height>
         </size>
         <zorder>7</zorder>
         <SQLField>
@@ -2399,10 +2399,18 @@
             <notNull><![CDATA[1]]></notNull>
             <uid><![CDATA[27D4A5BD-8093-4ADD-B5B5-D546844206F9]]></uid>
         </SQLField>
+        <SQLField>
+            <name><![CDATA[send_options_ping]]></name>
+            <type><![CDATA[BOOLEAN]]></type>
+            <defaultValue><![CDATA[0]]></defaultValue>
+            <notNull><![CDATA[1]]></notNull>
+            <uid><![CDATA[E04C19A2-12BF-443F-AB61-96990224A18D]]></uid>
+        </SQLField>
         <SQLField>
             <name><![CDATA[pad_crypto]]></name>
             <type><![CDATA[BOOLEAN]]></type>
             <defaultValue><![CDATA[0]]></defaultValue>
+            <forcedUnique><![CDATA[0]]></forcedUnique>
             <notNull><![CDATA[1]]></notNull>
             <uid><![CDATA[C5C0043B-100A-4476-BF01-BE0777AE27C0]]></uid>
         </SQLField>
@@ -2560,7 +2568,7 @@
         </SQLField>
         <SQLField>
             <name><![CDATA[speech_synthesis_voice]]></name>
-            <type><![CDATA[VARCHAR(64)]]></type>
+            <type><![CDATA[VARCHAR(256)]]></type>
             <notNull><![CDATA[0]]></notNull>
             <uid><![CDATA[929D66F0-64B9-4D7C-AB4B-24F131E1178F]]></uid>
         </SQLField>
@@ -2610,7 +2618,7 @@
         </SQLField>
         <SQLField>
             <name><![CDATA[fallback_speech_synthesis_voice]]></name>
-            <type><![CDATA[VARCHAR(64)]]></type>
+            <type><![CDATA[VARCHAR(256)]]></type>
             <notNull><![CDATA[0]]></notNull>
             <uid><![CDATA[6A0E92C9-32B9-4179-A893-3DADF5DD7728]]></uid>
         </SQLField>
@@ -3097,17 +3105,17 @@
         <overviewPanelHidden><![CDATA[0]]></overviewPanelHidden>
         <pageBoundariesVisible><![CDATA[0]]></pageBoundariesVisible>
         <PageGridVisible><![CDATA[0]]></PageGridVisible>
-        <RightSidebarWidth><![CDATA[1924.000000]]></RightSidebarWidth>
+        <RightSidebarWidth><![CDATA[1235.000000]]></RightSidebarWidth>
         <sidebarIndex><![CDATA[2]]></sidebarIndex>
         <snapToGrid><![CDATA[0]]></snapToGrid>
-        <SourceSidebarWidth><![CDATA[0.000000]]></SourceSidebarWidth>
+        <SourceSidebarWidth><![CDATA[312.000000]]></SourceSidebarWidth>
         <SQLEditorFileFormatVersion><![CDATA[4]]></SQLEditorFileFormatVersion>
         <uid><![CDATA[58C99A00-06C9-478C-A667-C63842E088F3]]></uid>
-        <windowHeight><![CDATA[985.000000]]></windowHeight>
-        <windowLocationX><![CDATA[-1307.000000]]></windowLocationX>
-        <windowLocationY><![CDATA[1008.000000]]></windowLocationY>
-        <windowScrollOrigin><![CDATA[{1.5, 786}]]></windowScrollOrigin>
-        <windowWidth><![CDATA[2201.000000]]></windowWidth>
+        <windowHeight><![CDATA[870.000000]]></windowHeight>
+        <windowLocationX><![CDATA[-1164.000000]]></windowLocationX>
+        <windowLocationY><![CDATA[1131.000000]]></windowLocationY>
+        <windowScrollOrigin><![CDATA[{0.5, 0}]]></windowScrollOrigin>
+        <windowWidth><![CDATA[1512.000000]]></windowWidth>
     </SQLDocumentInfo>
     <AllowsIndexRenamingOnInsert><![CDATA[1]]></AllowsIndexRenamingOnInsert>
     <defaultLabelExpanded><![CDATA[1]]></defaultLabelExpanded>

db/upgrade-jambonz-db.js

@@ -190,7 +190,12 @@ const sql = {
     'ALTER TABLE google_custom_voices ADD FOREIGN KEY speech_credential_sid_idxfk (speech_credential_sid) REFERENCES speech_credentials (speech_credential_sid) ON DELETE CASCADE',
     'ALTER TABLE clients ADD COLUMN allow_direct_queue_calling BOOLEAN NOT NULL DEFAULT 1',
     'ALTER TABLE clients ADD COLUMN allow_direct_user_calling BOOLEAN NOT NULL DEFAULT 1',
-    'ALTER TABLE clients ADD COLUMN allow_direct_app_calling BOOLEAN NOT NULL DEFAULT 1'
+    'ALTER TABLE clients ADD COLUMN allow_direct_app_calling BOOLEAN NOT NULL DEFAULT 1',
+  ],
+  9000: [
+    'ALTER TABLE sip_gateways ADD COLUMN send_options_ping BOOLEAN NOT NULL DEFAULT 0',
+    'ALTER TABLE applications MODIFY COLUMN speech_synthesis_voice VARCHAR(256)',
+    'ALTER TABLE applications MODIFY COLUMN fallback_speech_synthesis_voice VARCHAR(256)',
   ]
 };
 
@@ -223,6 +228,7 @@ const doIt = async() => {
         if (val < 8003) upgrades.push(...sql['8003']);
         if (val < 8004) upgrades.push(...sql['8004']);
         if (val < 8005) upgrades.push(...sql['8005']);
+        if (val < 9000) upgrades.push(...sql['9000']);
 
         // perform all upgrades
         logger.info({upgrades}, 'applying schema upgrades..');

lib/models/account.js

@@ -199,8 +199,9 @@ class Account extends Model {
     debug(r3, 'Account.activateSubscription - replaced old subscription');
 
     /* update account.plan to paid, if it isnt already */
+    /* update account.is_active to 1, if account is deactivated */
     await promisePool.execute(
-      'UPDATE accounts SET plan_type = \'paid\' WHERE account_sid = ?',
+      'UPDATE accounts SET plan_type = \'paid\', is_active = 1 WHERE account_sid = ?',
       [account_sid]);
     return true;
   }

lib/models/voip-carrier.js

@@ -61,6 +61,10 @@ VoipCarrier.fields = [
     name: 'requires_register',
     type: 'number'
   },
+  {
+    name: 'register_use_tls',
+    type: 'number'
+  },
   {
     name: 'register_username',
     type: 'string'

lib/record/azure-storage.js

@@ -1,28 +1,58 @@
 const { Writable } = require('stream');
 const { BlobServiceClient } = require('@azure/storage-blob');
 const { v4: uuidv4 } = require('uuid');
+const streamBuffers = require('stream-buffers');
 
 class AzureStorageUploadStream extends Writable {
   constructor(logger, opts) {
     super(opts);
     const blobServiceClient = BlobServiceClient.fromConnectionString(opts.connection_string);
     this.blockBlobClient = blobServiceClient.getContainerClient(opts.bucketName).getBlockBlobClient(opts.Key);
+
     this.metadata = opts.metadata;
     this.blocks = [];
+
+    this.bufferSize = 2 * 1024 * 1024; // Buffer size set to 2MB
+    this.buffer = new streamBuffers.WritableStreamBuffer({
+      initialSize: this.bufferSize,
+      incrementAmount: this.bufferSize
+    });
   }
 
   async _write(chunk, encoding, callback) {
-    const blockID = uuidv4().replace(/-/g, '');
-    this.blocks.push(blockID);
-    try {
-      await this.blockBlobClient.stageBlock(blockID, chunk, chunk.length);
+    this.buffer.write(chunk, encoding);
+
+    if (this.buffer.size() >= this.bufferSize) {
+      const blockID = uuidv4().replace(/-/g, '');
+      this.blocks.push(blockID);
+
+      try {
+        const dataToWrite = this.buffer.getContents();
+        await this.blockBlobClient.stageBlock(blockID, dataToWrite, dataToWrite.length);
+        callback();
+      } catch (error) {
+        callback(error);
+      }
+    } else {
       callback();
-    } catch (error) {
-      callback(error);
     }
   }
 
   async _final(callback) {
+    // Write any remaining data in buffer
+    if (this.buffer.size() > 0) {
+      const remainingData = this.buffer.getContents();
+      const blockID = uuidv4().replace(/-/g, '');
+      this.blocks.push(blockID);
+
+      try {
+        await this.blockBlobClient.stageBlock(blockID, remainingData, remainingData.length);
+      } catch (error) {
+        callback(error);
+        return;
+      }
+    }
+
     try {
       await this.blockBlobClient.commitBlockList(this.blocks);
       // remove all null/undefined props

lib/record/google-storage.js

@@ -1,5 +1,6 @@
 const { Storage } = require('@google-cloud/storage');
 const { Writable } = require('stream');
+const streamBuffers = require('stream-buffers');
 
 class GoogleStorageUploadStream extends Writable {
 
@@ -12,18 +13,38 @@ class GoogleStorageUploadStream extends Writable {
     this.gcsFile = storage.bucket(opts.bucketName).file(opts.Key);
     this.writeStream = this.gcsFile.createWriteStream();
 
+    this.bufferSize = 2 * 1024 * 1024; // Buffer size set to 2MB
+    this.buffer = new streamBuffers.WritableStreamBuffer({
+      initialSize: this.bufferSize,
+      incrementAmount: this.bufferSize
+    });
+
     this.writeStream.on('error', (err) => this.logger.error(err));
     this.writeStream.on('finish', () => {
-      this.logger.info('google storage Upload completed.');
+      this.logger.info('Google storage Upload completed.');
       this._addMetadata();
     });
   }
 
   _write(chunk, encoding, callback) {
-    this.writeStream.write(chunk, encoding, callback);
+    this.buffer.write(chunk, encoding);
+
+    // Write to GCS when buffer reaches desired size
+    if (this.buffer.size() >= this.bufferSize) {
+      const dataToWrite = this.buffer.getContents();
+      this.writeStream.write(dataToWrite, callback);
+    } else {
+      callback();
+    }
   }
 
   _final(callback) {
+    // Write any remaining data in the buffer to GCS
+    if (this.buffer.size() > 0) {
+      const remainingData = this.buffer.getContents();
+      this.writeStream.write(remainingData);
+    }
+
     this.writeStream.end();
     this.writeStream.once('finish', callback);
   }
@@ -33,7 +54,7 @@ class GoogleStorageUploadStream extends Writable {
       await this.gcsFile.setMetadata({metadata: this.metadata});
       this.logger.info('Google storage Upload and metadata setting completed.');
     } catch (err) {
-      this.logger.error(err, 'Google storage  An error occurred while setting metadata');
+      this.logger.error(err, 'Google storage An error occurred while setting metadata');
     }
   }
 }

lib/record/s3-multipart-upload-stream.js

@@ -16,7 +16,7 @@ class S3MultipartUploadStream extends Writable {
     this.partNumber = 1;
     this.multipartETags = [];
     this.buffer = Buffer.alloc(0);
-    this.minPartSize = 5 * 1024 * 1024; // 5 MB
+    this.minPartSize = 2 * 1024 * 1024; // 5 MB
     this.s3 = new S3Client(opts.bucketCredential);
     this.metadata = opts.metadata;
   }

lib/routes/api/speech-credentials.js

@@ -6,7 +6,8 @@ const sysError = require('../error');
 const {decrypt, encrypt} = require('../../utils/encrypt-decrypt');
 const {parseAccountSid, parseServiceProviderSid, parseSpeechCredentialSid} = require('./utils');
 const {decryptCredential, testWhisper, testDeepgramTTS,
-  getLanguagesAndVoicesForVendor} = require('../../utils/speech-utils');
+  getLanguagesAndVoicesForVendor,
+  testPlayHT} = require('../../utils/speech-utils');
 const {DbErrorUnprocessableRequest, DbErrorForbidden, DbErrorBadRequest} = require('../../utils/errors');
 const {
   testGoogleTts,
@@ -116,6 +117,8 @@ const encryptCredential = (obj) => {
     secret,
     nuance_tts_uri,
     nuance_stt_uri,
+    deepgram_stt_uri,
+    deepgram_stt_use_tls,
     use_custom_tts,
     custom_tts_endpoint,
     custom_tts_endpoint_url,
@@ -133,6 +136,8 @@ const encryptCredential = (obj) => {
     auth_token = '',
     cobalt_server_uri,
     model_id,
+    user_id,
+    voice_engine,
     options
   } = obj;
 
@@ -185,8 +190,11 @@ const encryptCredential = (obj) => {
       return encrypt(nuanceData);
 
     case 'deepgram':
-      assert(api_key, 'invalid deepgram speech credential: api_key is required');
-      const deepgramData = JSON.stringify({api_key});
+      // API key is optional if onprem
+      if (!deepgram_stt_uri) {
+        assert(api_key, 'invalid deepgram speech credential: api_key is required');
+      }
+      const deepgramData = JSON.stringify({api_key, deepgram_stt_uri, deepgram_stt_use_tls});
       return encrypt(deepgramData);
 
     case 'ibm':
@@ -214,6 +222,13 @@ const encryptCredential = (obj) => {
       const elevenlabsData = JSON.stringify({api_key, model_id, options});
       return encrypt(elevenlabsData);
 
+    case 'playht':
+      assert(api_key, 'invalid playht speech credential: api_key is required');
+      assert(user_id, 'invalid playht speech credential: user_id is required');
+      assert(voice_engine, 'invalid voice_engine speech credential: voice_engine is required');
+      const playhtData = JSON.stringify({api_key, user_id, voice_engine, options});
+      return encrypt(playhtData);
+
     case 'assemblyai':
       assert(api_key, 'invalid assemblyai speech credential: api_key is required');
       const assemblyaiData = JSON.stringify({api_key});
@@ -413,7 +428,10 @@ router.put('/:sid', async(req, res) => {
           custom_tts_url,
           cobalt_server_uri,
           model_id,
-          options
+          voice_engine,
+          options,
+          deepgram_stt_uri,
+          deepgram_stt_use_tls,
         } = req.body;
 
         const newCred = {
@@ -436,7 +454,10 @@ router.put('/:sid', async(req, res) => {
           custom_tts_url,
           cobalt_server_uri,
           model_id,
-          options
+          voice_engine,
+          options,
+          deepgram_stt_uri,
+          deepgram_stt_use_tls,
         };
         logger.info({o, newCred}, 'updating speech credential with this new credential');
         obj.credential = encryptCredential(newCred);
@@ -647,7 +668,7 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.ttsTestResult(sid, false);
         }
       }
-      if (cred.use_for_stt) {
+      if (cred.use_for_stt && api_key) {
         try {
           await testDeepgramStt(logger, {api_key});
           results.stt.status = 'ok';
@@ -715,6 +736,17 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.ttsTestResult(sid, false);
         }
       }
+    } else if (cred.vendor === 'playht') {
+      if (cred.use_for_tts) {
+        try {
+          await testPlayHT(logger, synthAudio, credential);
+          results.tts.status = 'ok';
+          SpeechCredential.ttsTestResult(sid, true);
+        } catch (err) {
+          results.tts = {status: 'fail', reason: err.message};
+          SpeechCredential.ttsTestResult(sid, false);
+        }
+      }
     } else if (cred.vendor === 'assemblyai') {
       const {api_key} = credential;
       if (cred.use_for_stt) {
@@ -752,7 +784,7 @@ 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 +799,7 @@ 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
@@ -1095,6 +1108,9 @@ paths:
                 requires_register:
                   type: boolean
                   description: wehther this provider requires us to send a REGISTER to them in order to receive calls
+                register_use_tls:
+                  type: boolean
+                  description: wehther this provider requires us to send a REGISTER use TLS protocol
                 register_username:
                   type: string
                   description: sip username to authenticate with, if registration is required
@@ -1971,7 +1987,7 @@ 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 +2095,12 @@ 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 +2942,7 @@ 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 +2956,7 @@ paths:
       tags:
         - Accounts
       summary: update a speech credential
-      operationId: updateSpeechCredential
+      operationId: updateSpeechCredentialByAccount
       requestBody:
         content:
           application/json:
@@ -2955,7 +2977,7 @@ paths:
       tags:
         - Accounts
       summary: delete a speech credential
-      operationId: deleteSpeechCredential
+      operationId: deleteSpeechCredentialByAccount
       responses:
         204:
           description: credential successfully deleted
@@ -2966,8 +2988,14 @@ 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 +3023,7 @@ paths:
       tags:
         - Accounts
       summary: test a speech credential
-      operationId: testSpeechCredential
+      operationId: testSpeechCredentialByAccount
       parameters:
         - name: AccountSid
           in: path
@@ -3241,7 +3269,7 @@ paths:
       tags:
         - Service Providers
       summary: retrieve pcap for a call
-      operationId: getRecentCallTrace
+      operationId: getRecentCallTraceBySP
       responses:
         200:
           description: retrieve sip trace data
@@ -3327,7 +3355,7 @@ 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 +3456,7 @@ 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 +3483,7 @@ paths:
       tags:
         - Accounts
       summary: retrieve pcap for a call
-      operationId: getRecentCallTrace
+      operationId: getRecentCallTraceByAccount
       responses:
         200:
           description: retrieve sip trace data
@@ -3641,7 +3669,7 @@ paths:
       tags:
        - Accounts
       summary: retrieve alerts for an account
-      operationId: listAlerts
+      operationId: listAlertsByAccount
       responses:
         200:
           description: retrieve alerts for a specified account
@@ -4240,7 +4268,7 @@ 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 +4282,7 @@ 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 +4301,12 @@ 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 +4327,13 @@ 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 +5072,32 @@ 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 +6152,23 @@ 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-data/tts-model-deepgram.js

@@ -1,9 +1,14 @@
 module.exports = [
-  { name: 'Aurora English (US) Female', value: 'alpha-aurora-en-v2' },
-  { name: 'Asteria English (US) Female', value: 'alpha-asteria-en-v2' },
-  { name: 'Artemis English (UK) Female', value: 'alpha-artemis-en-v3' },
-  { name: 'Andromeda English (US) Female', value: 'alpha-andromeda-en-v3' },
-  { name: 'Stella	English (UK) Female', value: 'alpha-stella-en-v2' },
-  { name: 'Orion	English (US)	Male', value: 'alpha-orion-en-v2' },
-  { name: 'Atlas	English (US)	Male', value: 'alpha-atlas-en-v3' },
+  { name: 'Asteria English (US) Female', value: 'aura-asteria-en' },
+  { name: 'Luna English (US) Female', value: 'aura-luna-en' },
+  { name: 'Stella English (US) Female', value: 'aura-stella-en' },
+  { name: 'Stella English (UK) Female', value: 'aura-athena-en' },
+  { name: 'Hera English (US) Female', value: 'aura-hera-en' },
+  { name: 'Orion English (US) Male', value: 'aura-orion-en' },
+  { name: 'Arcas English (US) Male', value: 'aura-arcas-en' },
+  { name: 'Perseus English (US) Male', value: 'aura-perseus-en' },
+  { name: 'Angus English (Ireland) Male', value: 'aura-angus-en' },
+  { name: 'Orpheus English (US) Male', value: 'aura-orpheus-en' },
+  { name: 'Helios English (UK) Male', value: 'aura-helios-en' },
+  { name: 'Zeus English (US) Male', value: 'aura-zeus-en' },
 ];

lib/utils/speech-data/tts-model-playht.js

@@ -0,0 +1,6 @@
+module.exports = [
+  { name: 'PlayHT2.0-turbo', value: 'PlayHT2.0-turbo' },
+  { name: 'PlayHT2.0', value: 'PlayHT2.0' },
+  { name: 'PlayHT1.0', value: 'PlayHT1.0' },
+];
+

lib/utils/speech-utils.js

@@ -21,6 +21,7 @@ const TtsWhisperLanguagesVoices = require('./speech-data/tts-whisper');
 const TtsModelDeepgram = require('./speech-data/tts-model-deepgram');
 const TtsModelElevenLabs = require('./speech-data/tts-model-elevenlabs');
 const TtsModelWhisper = require('./speech-data/tts-model-whisper');
+const TtsModelPlayHT = require('./speech-data/tts-model-playht');
 
 const SttGoogleLanguagesVoices = require('./speech-data/stt-google');
 const SttAwsLanguagesVoices = require('./speech-data/stt-aws');
@@ -240,6 +241,27 @@ const testElevenlabs = async(logger, credentials) => {
   }
 };
 
+const testPlayHT = async(logger, synthAudio, credentials) => {
+  try {
+    await synthAudio(
+      {
+        increment: () => {},
+        histogram: () => {}
+      },
+      {
+        vendor: 'playht',
+        credentials,
+        language: 'en-US',
+        voice: 's3://voice-cloning-zero-shot/d9ff78ba-d016-47f6-b0ef-dd630f59414e/female-cs/manifest.json',
+        text: 'Hi there and welcome to jambones!'
+      }
+    );
+  } catch (err) {
+    logger.info({err}, 'synth Playht returned error');
+    throw err;
+  }
+};
+
 const testWhisper = async(logger, synthAudio, credentials) => {
   try {
     await synthAudio({increment: () => {}, histogram: () => {}},
@@ -263,9 +285,7 @@ const testDeepgramTTS = async(logger, synthAudio, credentials) => {
       {
         vendor: 'deepgram',
         credentials,
-        language: 'en-US',
-        voice: 'alpha-aurora-en-v2',
-        model: 'alpha-aurora-en-v2',
+        model: 'aura-asteria-en',
         text: 'Hi there and welcome to jambones!'
       }
     );
@@ -406,6 +426,8 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
   else if ('deepgram' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+    obj.deepgram_stt_uri = o.deepgram_stt_uri;
+    obj.deepgram_stt_use_tls = o.deepgram_stt_use_tls;
   }
   else if ('ibm' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
@@ -428,6 +450,12 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
     obj.model_id = o.model_id;
     obj.options = o.options;
+  } else if ('playht' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+    obj.user_id = o.user_id;
+    obj.voice_engine = o.voice_engine;
+    obj.options = o.options;
   } else if (obj.vendor.startsWith('custom:')) {
     const o = JSON.parse(decrypt(credential));
     obj.auth_token = isObscureKey ? obscureKey(o.auth_token) : o.auth_token;
@@ -464,49 +492,92 @@ 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 'playht':
+      return await getLanguagesVoicesForPlayHT(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 +585,19 @@ 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 +605,19 @@ 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);
 }
 
@@ -580,6 +675,49 @@ async function getLanguagesVoicesForElevenlabs(credential) {
   }
 }
 
+const concat = (a) => {
+  return a ? ` ${a},` : '';
+};
+
+async function getLanguagesVoicesForPlayHT(credential) {
+  if (credential) {
+    const get = bent('https://api.play.ht', 'GET', 'json', {
+      'AUTHORIZATION' : credential.api_key,
+      'X-USER-ID': credential.user_id,
+      'Accept': 'application/json'
+    });
+
+    const voices = await get('/api/v2/voices');
+
+    const buildVoice = (d) => {
+      let name = `${d.name} -${concat(d.accent)}${concat(d.age)}${concat(d.gender)}
+${concat(d.loudness)}${concat(d.style)}${concat(d.tempo)}${concat(d.texture)}` ;
+      name = name.endsWith(',') ? name.slice(0, -1) : name;
+      return {
+        value: `${d.id}`,
+        name
+      };
+    };
+
+    const ttsVoices = voices.reduce((acc, voice) => {
+      const languageCode = voice.language_code;
+      const existingLanguage = acc.find((lang) => lang.value === languageCode);
+      if (existingLanguage) {
+        existingLanguage.voices.push(buildVoice(voice));
+      } else {
+        acc.push({
+          value: voice.language_code,
+          name: voice.language,
+          voices: [buildVoice(voice)]
+        });
+      }
+      return acc;
+    }, []);
+    return tranform(ttsVoices, undefined, TtsModelPlayHT);
+  }
+  return tranform(undefined, undefined, TtsModelPlayHT);
+}
+
 async function getLanguagesVoicesForAssemblyAI(credential) {
   return tranform(undefined, SttAssemblyaiLanguagesVoices);
 }
@@ -596,6 +734,125 @@ 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,
@@ -612,6 +869,7 @@ module.exports = {
   testIbmStt,
   testSonioxStt,
   testElevenlabs,
+  testPlayHT,
   testAssemblyStt,
   testDeepgramTTS,
   getSpeechCredential,

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jambonz-api-server",
-  "version": "0.8.5",
+  "version": "0.9.0",
   "description": "",
   "main": "app.js",
   "scripts": {
@@ -19,53 +19,54 @@
     "url": "https://github.com/jambonz/jambonz-api-server.git"
   },
   "dependencies": {
-    "@aws-sdk/client-s3": "^3.363.0",
-    "@aws-sdk/client-transcribe": "^3.363.0",
-    "@azure/storage-blob": "^12.15.0",
+    "@aws-sdk/client-s3": "^3.550.0",
+    "@aws-sdk/client-transcribe": "^3.549.0",
+    "@azure/storage-blob": "^12.17.0",
     "@deepgram/sdk": "^1.21.0",
-    "@google-cloud/speech": "^5.2.0",
-    "@google-cloud/storage": "^6.12.0",
-    "@jambonz/db-helpers": "^0.9.1",
+    "@google-cloud/speech": "^6.5.0",
+    "@google-cloud/storage": "^7.9.0",
+    "@jambonz/db-helpers": "^0.9.3",
     "@jambonz/lamejs": "^1.2.2",
     "@jambonz/mw-registrar": "^0.2.7",
-    "@jambonz/realtimedb-helpers": "^0.8.7",
-    "@jambonz/speech-utils": "^0.0.33",
+    "@jambonz/realtimedb-helpers": "^0.8.8",
+    "@jambonz/speech-utils": "^0.0.50",
     "@jambonz/time-series": "^0.2.8",
-    "@jambonz/verb-specifications": "^0.0.45",
-    "@soniox/soniox-node": "^1.1.1",
-    "argon2": "^0.30.3",
-    "assemblyai": "^3.0.1",
+    "@jambonz/verb-specifications": "^0.0.69",
+    "@soniox/soniox-node": "^1.2.2",
+    "argon2": "^0.40.1",
+    "assemblyai": "^4.3.4",
     "bent": "^7.3.12",
     "cors": "^2.8.5",
     "debug": "^4.3.4",
-    "express": "^4.18.1",
-    "express-rate-limit": "^6.4.0",
-    "form-data": "^2.5.1",
-    "helmet": "^5.1.0",
-    "ibm-watson": "^7.1.2",
-    "jsonwebtoken": "^9.0.0",
-    "mailgun.js": "^9.1.2",
-    "microsoft-cognitiveservices-speech-sdk": "1.31.0",
-    "mysql2": "^2.3.3",
-    "nocache": "3.0.4",
-    "passport": "^0.6.0",
+    "express": "^4.19.2",
+    "express-rate-limit": "^7.2.0",
+    "form-data": "^4.0.0",
+    "helmet": "^7.1.0",
+    "ibm-watson": "^9.0.1",
+    "jsonwebtoken": "^9.0.2",
+    "mailgun.js": "^10.2.1",
+    "microsoft-cognitiveservices-speech-sdk": "1.36.0",
+    "mysql2": "^3.9.3",
+    "nocache": "4.0.0",
+    "passport": "^0.7.0",
     "passport-http-bearer": "^1.0.1",
-    "pino": "^5.17.0",
-    "short-uuid": "^4.1.0",
-    "stripe": "^8.222.0",
-    "swagger-ui-express": "^4.4.0",
-    "uuid": "^8.3.2",
+    "pino": "^8.20.0",
+    "short-uuid": "^4.2.2",
+    "stream-buffers": "^3.0.2",
+    "stripe": "^14.24.0",
+    "swagger-ui-express": "^5.0.0",
+    "uuid": "^9.0.1",
     "wav": "^1.0.2",
-    "ws": "^8.12.1",
+    "ws": "^8.16.0",
     "yamljs": "^0.3.0"
   },
   "devDependencies": {
     "eslint": "^8.39.0",
     "eslint-plugin-promise": "^6.1.1",
-    "husky": "7.0.4",
+    "husky": "9.0.11",
     "nyc": "^15.1.0",
     "request": "^2.88.2",
     "request-promise-native": "^1.0.9",
-    "tape": "^5.5.3"
+    "tape": "^5.7.5"
   }
 }

test/docker-compose-testbed.yaml

@@ -10,7 +10,7 @@ networks:
 
 services:  
   mysql:
-    platform: linux/x86_64
+    # platform: linux/x86_64
     image: mysql:5.7
     ports:
       - "3360:3306"
@@ -36,7 +36,7 @@ services:
         ipv4_address: 172.58.0.3
     
   influxdb:
-    platform: linux/x86_64
+    # platform: linux/x86_64
     image: influxdb:1.8
     ports:
       - "8086:8086"

test/speech-credentials.js

@@ -292,6 +292,60 @@ test('speech credentials tests', async(t) => {
       });
       t.ok(result.statusCode === 204, 'successfully deleted speech credential');
     }
+    // test create deepgram onprem
+    result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,
+      body: {
+        vendor: 'deepgram',
+        use_for_stt: true,
+        deepgram_stt_uri: "127.0.0.1:50002",
+        deepgram_stt_use_tls: true
+      }
+    });
+    t.ok(result.statusCode === 201, 'successfully added speech credential for deepgram');
+    const dg_sid = result.body.sid;
+
+    result = await request.get(`/Accounts/${account_sid}/SpeechCredentials/${dg_sid}`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,   
+    });
+    //console.log(JSON.stringify(result));
+    t.ok(result.statusCode === 200, 'successfully get speech credential for deepgram');
+    t.ok(result.body.deepgram_stt_uri === '127.0.0.1:50002', "deepgram_stt_uri is correct for deepgram");
+    t.ok(result.body.deepgram_stt_use_tls === true, "deepgram_stt_use_tls is correct for deepgram");
+
+    result = await request.put(`/Accounts/${account_sid}/SpeechCredentials/${dg_sid}`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,
+      body: {
+        vendor: 'deepgram',
+        use_for_stt: true,
+        deepgram_stt_uri: "127.0.0.2:50002",
+        deepgram_stt_use_tls: false
+      }
+    });
+    t.ok(result.statusCode === 204, 'successfully updated speech credential for deepgram onprem');
+
+    result = await request.get(`/Accounts/${account_sid}/SpeechCredentials/${dg_sid}`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,   
+    });
+    //console.log(JSON.stringify(result));
+    t.ok(result.statusCode === 200, 'successfully get speech credential for deepgram onprem');
+    t.ok(result.body.deepgram_stt_uri === '127.0.0.2:50002', "deepgram_stt_uri is correct for deepgram onprem");
+    t.ok(result.body.deepgram_stt_use_tls === false, "deepgram_stt_use_tls is correct for deepgram onprem");
+
+    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${dg_sid}`, {
+      auth: authUser,
+      resolveWithFullResponse: true,
+    });
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential for deepgram onprem');
+
     /* add a credential for ibm tts */
     if (process.env.IBM_TTS_API_KEY && process.env.IBM_TTS_REGION) {
       result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
@@ -482,7 +536,7 @@ test('speech credentials tests', async(t) => {
         model_id: 'eleven_multilingual_v2'
       }
     });
-    t.ok(result.statusCode === 201, 'successfully added speech credential for Cobalt');
+    t.ok(result.statusCode === 201, 'successfully added speech credential for elevenlabs');
     const elevenlabs_sid = result.body.sid;
 
     /* delete the credential */
@@ -490,7 +544,31 @@ test('speech credentials tests', async(t) => {
       auth: authUser,
       resolveWithFullResponse: true,
     });
-    t.ok(result.statusCode === 204, 'successfully deleted speech credential for Cobalt');
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential for elevenlabs');
+
+    /* add a credential for playht */
+    result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,
+      body: {
+        vendor: 'playht',
+        use_for_stt: false,
+        use_for_tts: true,
+        api_key: 'asdasdasdasddsadasda',
+        user_id: 'user_id',
+        voice_engine: 'PlayHT2.0-turbo'
+      }
+    });
+    t.ok(result.statusCode === 201, 'successfully added speech credential for playht');
+    const playht_sid = result.body.sid;
+
+    /* delete the credential */
+    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${playht_sid}`, {
+      auth: authUser,
+      resolveWithFullResponse: true,
+    });
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential for playht');
 
 
     /* add a credential for custom voices google */