support assembly ai (#248)

* support assembly ai

* wip

* wip

lib/routes/api/speech-credentials.js

@@ -20,7 +20,8 @@ const {
   testSonioxStt,
   testIbmTts,
   testIbmStt,
-  testElevenlabs
+  testElevenlabs,
+  testAssemblyStt
 } = require('../../utils/speech-utils');
 const bent = require('bent');
 const {promisePool} = require('../../db');
@@ -211,6 +212,11 @@ const encryptCredential = (obj) => {
       const elevenlabsData = JSON.stringify({api_key, model_id});
       return encrypt(elevenlabsData);
 
+    case 'assemblyai':
+      assert(api_key, 'invalid assemblyai speech credential: api_key is required');
+      const assemblyaiData = JSON.stringify({api_key});
+      return encrypt(assemblyaiData);
+
     default:
       if (vendor.startsWith('custom:')) {
         const customData = JSON.stringify({auth_token, custom_stt_url, custom_tts_url});
@@ -335,6 +341,9 @@ function decryptCredential(obj, credential, logger) {
     obj.auth_token = obscureKey(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 = obscureKey(o.api_key);
   }
 }
 
@@ -757,6 +766,18 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.ttsTestResult(sid, false);
         }
       }
+    } else if (cred.vendor === 'assemblyai') {
+      const {api_key} = credential;
+      if (cred.use_for_stt) {
+        try {
+          await testAssemblyStt(logger, {api_key});
+          results.stt.status = 'ok';
+          SpeechCredential.sttTestResult(sid, true);
+        } catch (err) {
+          results.stt = {status: 'fail', reason: err.message};
+          SpeechCredential.sttTestResult(sid, false);
+        }
+      }
     }
 
     res.status(200).json(results);

lib/utils/speech-utils.js

@@ -5,6 +5,7 @@ const sdk = require('microsoft-cognitiveservices-speech-sdk');
 const { SpeechClient } = require('@soniox/soniox-node');
 const bent = require('bent');
 const fs = require('fs');
+const { AssemblyAI } = require('assemblyai');
 
 
 const testSonioxStt = async(logger, credentials) => {
@@ -260,6 +261,32 @@ const testWellSaidStt = async(logger, credentials) => {
   return true;
 };
 
+const testAssemblyStt = async(logger, credentials) => {
+  const {api_key} = credentials;
+
+  const assemblyai = new AssemblyAI({
+    apiKey: api_key
+  });
+
+  const audioUrl = `${__dirname}/../../data/test_audio.wav`;
+
+  return new Promise((resolve, reject) => {
+    assemblyai.transcripts
+      .create({ audio_url: audioUrl })
+      .then((transcript) => {
+        logger.debug({transcript}, 'got transcription from AssemblyAi');
+        if (transcript.status === 'error') {
+          return reject({message: transcript.error});
+        }
+        return resolve(transcript.text);
+      })
+      .catch((err) => {
+        logger.info({err}, 'failed to get assemblyAI transcription');
+        reject(err);
+      });
+  });
+};
+
 module.exports = {
   testGoogleTts,
   testGoogleStt,
@@ -275,5 +302,6 @@ module.exports = {
   testIbmTts,
   testIbmStt,
   testSonioxStt,
-  testElevenlabs
+  testElevenlabs,
+  testAssemblyStt
 };

package-lock.json

@@ -24,6 +24,7 @@
         "@jambonz/verb-specifications": "^0.0.29",
         "@soniox/soniox-node": "^1.1.1",
         "argon2": "^0.30.3",
+        "assemblyai": "^3.0.1",
         "bent": "^7.3.12",
         "cors": "^2.8.5",
         "debug": "^4.3.4",
@@ -3314,6 +3315,14 @@
         "safer-buffer": "~2.1.0"
       }
     },
+    "node_modules/assemblyai": {
+      "version": "3.0.1",
+      "resolved": "https://registry.npmjs.org/assemblyai/-/assemblyai-3.0.1.tgz",
+      "integrity": "sha512-7EW3v00o8s3TSweQERJjoNbXMDU8abwYwpE3gYKQhNY4xlcxb2in344rQi6MsMLVxRtLe8HBXGVRUlZkzfxC3w==",
+      "dependencies": {
+        "ws": "^8.13.0"
+      }
+    },
     "node_modules/assert-plus": {
       "version": "1.0.0",
       "resolved": "https://registry.npmjs.org/assert-plus/-/assert-plus-1.0.0.tgz",
@@ -12416,6 +12425,14 @@
         "safer-buffer": "~2.1.0"
       }
     },
+    "assemblyai": {
+      "version": "3.0.1",
+      "resolved": "https://registry.npmjs.org/assemblyai/-/assemblyai-3.0.1.tgz",
+      "integrity": "sha512-7EW3v00o8s3TSweQERJjoNbXMDU8abwYwpE3gYKQhNY4xlcxb2in344rQi6MsMLVxRtLe8HBXGVRUlZkzfxC3w==",
+      "requires": {
+        "ws": "^8.13.0"
+      }
+    },
     "assert-plus": {
       "version": "1.0.0",
       "resolved": "https://registry.npmjs.org/assert-plus/-/assert-plus-1.0.0.tgz",

package.json

@@ -34,6 +34,7 @@
     "@jambonz/verb-specifications": "^0.0.29",
     "@soniox/soniox-node": "^1.1.1",
     "argon2": "^0.30.3",
+    "assemblyai": "^3.0.1",
     "bent": "^7.3.12",
     "cors": "^2.8.5",
     "debug": "^4.3.4",

test/speech-credentials.js

@@ -537,6 +537,27 @@ test('speech credentials tests', async(t) => {
     });
     t.ok(result.statusCode === 204, 'successfully deleted speech credential for custom voice google');
 
+    /* add a credential for assemblyAI */
+    result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+      resolveWithFullResponse: true,
+      auth: authUser,
+      json: true,
+      body: {
+        vendor: 'assemblyai',
+        use_for_stt: true,
+        api_key: "APIKEY"
+      }
+    });
+    t.ok(result.statusCode === 201, 'successfully added speech credential for assemblyai');
+    const assemblyAiSid = result.body.sid;
+
+    /* delete the credential */
+    result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${assemblyAiSid}`, {
+      auth: authUser,
+      resolveWithFullResponse: true,
+    });
+    t.ok(result.statusCode === 204, 'successfully deleted speech credential');
+
     await deleteObjectBySid(request, '/Accounts', account_sid);
     await deleteObjectBySid(request, '/ServiceProviders', service_provider_sid);
     t.end();