update to speech utils with azure 1.36.0 (#303 )

update to speech utils with azure 1.36.0
back out column addition of -register_use_tls
2026-01-25 02:08:24 +00:00 · 2024-04-07 17:45:33 -04:00 · 2024-04-07 12:16:55 -04:00 · 2024-04-06 13:48:26 -04:00 · 2024-04-06 13:27:32 -04:00 · 2024-04-04 08:02:29 -04:00
14 changed files with 3425 additions and 13147 deletions
--- a/db/jambones.sqs
+++ b/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>
--- a/db/upgrade-jambonz-db.js
+++ b/db/upgrade-jambonz-db.js
@@ -190,7 +190,10 @@ 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',
  ]
 };

@@ -223,6 +226,7 @@ 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..');
--- a/lib/models/account.js
+++ b/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;
  }
--- a/lib/models/voip-carrier.js
+++ b/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'
--- a/lib/record/azure-storage.js
+++ b/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
--- a/lib/record/google-storage.js
+++ b/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');
    }
  }
 }
--- a/lib/record/s3-multipart-upload-stream.js
+++ b/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;
  }
--- a/lib/routes/api/speech-credentials.js
+++ b/lib/routes/api/speech-credentials.js
@@ -116,6 +116,8 @@ 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,
@@ -185,8 +187,11 @@ 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':
@@ -413,7 +418,9 @@ router.put('/:sid', async(req, res) => {
          custom_tts_url,
          cobalt_server_uri,
          model_id,
-          options
+          options,
+          deepgram_stt_uri,
+          deepgram_stt_use_tls,
        } = req.body;

        const newCred = {
@@ -436,7 +443,9 @@ router.put('/:sid', async(req, res) => {
          custom_tts_url,
          cobalt_server_uri,
          model_id,
-          options
+          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 +656,7 @@ 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';
@@ -752,7 +761,7 @@ 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 +776,7 @@ 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);
--- a/lib/swagger/swagger.yaml
+++ b/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: []
--- a/lib/utils/speech-data/tts-model-deepgram.js
+++ b/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' },
 ];
--- a/lib/utils/speech-utils.js
+++ b/lib/utils/speech-utils.js
@@ -263,9 +263,7 @@ 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 +404,8 @@ 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));
@@ -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,
--- a/package-lock.json
+++ b/package-lock.json
--- a/package.json
+++ b/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",
-    "@deepgram/sdk": "^1.21.0",
-    "@google-cloud/speech": "^5.2.0",
-    "@google-cloud/storage": "^6.12.0",
-    "@jambonz/db-helpers": "^0.9.1",
+    "@aws-sdk/client-s3": "^3.550.0",
+    "@aws-sdk/client-transcribe": "^3.549.0",
+    "@azure/storage-blob": "^12.17.0",
+    "@deepgram/sdk": "^3.2.0",
+    "@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": "^0.0.49",
    "@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"
  }
 }
--- a/test/speech-credentials.js
+++ b/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`, {
Author	SHA1	Message	Date
Dave Horton	3a299bc3ca	update to speech utils with azure 1.36.0 (#303 )	2024-04-07 17:45:33 -04:00
Dave Horton	70c9407742	update to speech utils with azure 1.36.0	2024-04-07 12:16:55 -04:00
Dave Horton	dba66d58fc	back out column addition of -register_use_tls	2024-04-06 13:48:26 -04:00
Dave Horton	0ff3d22faf	Revert "feat send options ping for sip gateway (#273 )" This reverts commit `a4792a521f`.	2024-04-06 13:27:32 -04:00
Hoan Luu Huu	187a428a75	register use tls (#302 )	2024-04-04 08:02:29 -04:00
Hoan Luu Huu	a4792a521f	feat send options ping for sip gateway (#273 ) * feat send options ping for sip gateway * update upgrade db script to have 8006	2024-03-30 09:14:29 -04:00
Dave Horton	3ac9693735	update speech-utils with fixes for deepgram production api and tts streaming	2024-03-24 08:15:00 -04:00
Dave Horton	3ad54a0e72	update to released deepgram tts voices (#299 )	2024-03-13 09:16:12 -04:00
Hoan Luu Huu	bd8fb2f9db	remove use_streaming from speech credential (#294 ) * remove use_streaming from speech credential * wip	2024-02-20 08:01:33 -05:00
Dave Horton	32b317ae68	update to latest speech-utils	2024-02-12 21:11:49 -05:00
Hoan Luu Huu	40e8d08727	support deepgram tts onprem (#292 ) * support deepgram tts onprem * wip * wip * deepgram disable speech test if api_key is missng	2024-02-12 09:27:13 -05:00
Hoan Luu Huu	256ca440a0	add use_streaming flag for elevenlabs and whisper (#290 ) * add use_streaming flag for elevenlabs (not for whisper yet) --------- Co-authored-by: Dave Horton <daveh@beachdognet.com>	2024-02-12 09:18:49 -05:00
Markus Frindt	68d73345ef	Improve Swagger file, add login route, fix swagger linting (#291 ) Co-authored-by: Markus Frindt <m.frindt@cognigy.com>	2024-02-06 12:10:31 -05:00
Hoan Luu Huu	54dd72ff66	fetch list of tts voices from provider (#289 ) * fetch list of tts voices from provider * revert serve-integration * fix * fix for aws * fix for aws * fix for aws * update speech-utils version	2024-01-25 12:03:02 -05:00
Dave Horton	832a4e8032	update db-helpers	2024-01-17 13:24:51 -05:00
Hoan Luu Huu	33c3b99e2e	update paid account to active if it's in deactivated (#287 ) * update paid account to active if it's in deactivated * fix review comment	2024-01-17 09:20:38 -05:00
Hoan Luu Huu	8b2a2e196e	Feat/record upload buffer (#285 ) * uploader with buffer for google and azure * wip * wip * wip	2024-01-15 09:51:10 -05:00