feat TTS synthAduio from API (#252)

* feat TTS synthAduio from API * fix failing testcase * wip * wip * wip * wip * uip * wip * fix * fix * swagger
2026-07-24 21:12:10 +00:00 · 2023-11-06 19:48:52 +07:00
parent 1e9f388f51
commit f725d5f0a1
8 changed files with 280 additions and 104 deletions
@@ -6,6 +6,7 @@ const { SpeechClient } = require('@soniox/soniox-node');
 const bent = require('bent');
 const fs = require('fs');
 const { AssemblyAI } = require('assemblyai');
+const {decrypt, obscureKey} = require('./encrypt-decrypt');


 const testSonioxStt = async(logger, credentials) => {
@@ -287,6 +288,107 @@ const testAssemblyStt = async(logger, credentials) => {
  });
 };

+const getSpeechCredential = (credential, logger) => {
+  const {vendor} = credential;
+  logger.info(
+    `Speech vendor: ${credential.vendor} ${credential.label ? `, label: ${credential.label}` : ''} selected`);
+  if ('google' === vendor) {
+    try {
+      const cred = JSON.parse(credential.service_key.replace(/\n/g, '\\n'));
+      return {
+        ...credential,
+        credentials: cred
+      };
+    } catch (err) {
+      logger.info({err}, `malformed google service_key provisioned for account ${credential.speech_credential_sid}`);
+    }
+  }
+  else if (['aws', 'polly'].includes(vendor)) {
+    return {
+      ...credential,
+      accessKeyId: credential.access_key_id,
+      secretAccessKey: credential.secret_access_key,
+      region: credential.aws_region || 'us-east-1'
+    };
+  }
+  return credential;
+};
+
+function decryptCredential(obj, credential, logger, isObscureKey = true) {
+  if ('google' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    const key_header = '-----BEGIN PRIVATE KEY-----\n';
+    const obscured = {
+      ...o,
+      private_key: `${key_header}${isObscureKey ?
+        obscureKey(o.private_key.slice(key_header.length, o.private_key.length)) :
+        o.private_key.slice(key_header.length, o.private_key.length)}`
+    };
+    obj.service_key = JSON.stringify(obscured);
+  }
+  else if ('aws' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.access_key_id = o.access_key_id;
+    obj.secret_access_key = isObscureKey ? obscureKey(o.secret_access_key) : o.secret_access_key;
+    obj.aws_region = o.aws_region;
+    logger.info({obj, o}, 'retrieving aws speech credential');
+  }
+  else if ('microsoft' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+    obj.region = o.region;
+    obj.use_custom_tts = o.use_custom_tts;
+    obj.custom_tts_endpoint = o.custom_tts_endpoint;
+    obj.custom_tts_endpoint_url = o.custom_tts_endpoint_url;
+    obj.use_custom_stt = o.use_custom_stt;
+    obj.custom_stt_endpoint = o.custom_stt_endpoint;
+    obj.custom_stt_endpoint_url = o.custom_stt_endpoint_url;
+    logger.info({obj, o}, 'retrieving azure speech credential');
+  }
+  else if ('wellsaid' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+  }
+  else if ('nuance' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.client_id = o.client_id;
+    obj.secret = o.secret ? (isObscureKey ? obscureKey(o.secret) : o.secret) : null;
+  }
+  else if ('deepgram' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+  }
+  else if ('ibm' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.tts_api_key = isObscureKey ? obscureKey(o.tts_api_key) : o.tts_api_key;
+    obj.tts_region = o.tts_region;
+    obj.stt_api_key = isObscureKey ? obscureKey(o.stt_api_key) : o.stt_api_key;
+    obj.stt_region = o.stt_region;
+    obj.instance_id = o.instance_id;
+  } else if ('nvidia' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.riva_server_uri = o.riva_server_uri;
+  } else if ('cobalt' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.cobalt_server_uri = o.cobalt_server_uri;
+  } else if ('soniox' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+  } else if ('elevenlabs' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+    obj.model_id = o.model_id;
+  } else if (obj.vendor.startsWith('custom:')) {
+    const o = JSON.parse(decrypt(credential));
+    obj.auth_token = isObscureKey ? obscureKey(o.auth_token) : o.auth_token;
+    obj.custom_stt_url = o.custom_stt_url;
+    obj.custom_tts_url = o.custom_tts_url;
+  } else if ('assemblyai' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+  }
+}
+
 module.exports = {
  testGoogleTts,
  testGoogleStt,
@@ -303,5 +405,7 @@ module.exports = {
  testIbmStt,
  testSonioxStt,
  testElevenlabs,
-  testAssemblyStt
+  testAssemblyStt,
+  getSpeechCredential,
+  decryptCredential
 };