support speechmatics region authentication

lib/routes/api/speech-credentials.js

@@ -123,6 +123,7 @@ const encryptCredential = (obj) => {
     secret,
     nuance_tts_uri,
     nuance_stt_uri,
+    speechmatics_stt_uri,
     deepgram_stt_uri,
     deepgram_stt_use_tls,
     deepgram_tts_uri,
@@ -239,7 +240,8 @@ const encryptCredential = (obj) => {
 
     case 'speechmatics':
       assert(api_key, 'invalid speechmatics speech credential: api_key is required');
-      const speechmaticsData = JSON.stringify({api_key, options});
+      assert(speechmatics_stt_uri, 'invalid speechmatics speech credential: speechmatics_stt_uri is required');
+      const speechmaticsData = JSON.stringify({api_key, speechmatics_stt_uri, options});
       return encrypt(speechmaticsData);
 
     case 'playht':

lib/utils/speech-utils.js

@@ -7,7 +7,7 @@ const bent = require('bent');
 const fs = require('fs');
 const { AssemblyAI } = require('assemblyai');
 const {decrypt, obscureKey} = require('./encrypt-decrypt');
-const { Speechmatics } = require('speechmatics');
+const { RealtimeSession } = require('speechmatics');
 
 const TtsGoogleLanguagesVoices = require('./speech-data/tts-google');
 const TtsAwsLanguagesVoices = require('./speech-data/tts-aws');
@@ -56,25 +56,53 @@ const testSonioxStt = async(logger, credentials) => {
 };
 
 const testSpeechmaticsStt = async(logger, credentials) => {
-  const {api_key} = credentials;
-  const sm = new Speechmatics(api_key);
-
+  const {api_key, speechmatics_stt_uri} = credentials;
   return new Promise(async(resolve, reject) => {
     try {
-      const inputFile = new Blob([
-        fs.readFileSync(`${__dirname}/../../data/test_audio.wav`),
-      ]);
-      const result = await sm.batch.transcribe({
-        data: inputFile,
-        fileName: 'test_audio.wav'
-      },
-      {
-        transcription_config: {
-          language: 'en'
-        }},
-      'text');
-      if (result.length > 0) resolve(result);
-      else reject(new Error('no transcript returned'));
+      const session = new RealtimeSession({ apiKey: api_key, realtimeUrl: speechmatics_stt_uri });
+      let transcription = '';
+      session.addListener('Error', (error) => {
+        reject(error);
+      });
+
+      session.addListener('AddTranscript', (message) => {
+        transcription += message.metadata.transcript;
+      });
+
+      session.addListener('EndOfTranscript', () => {
+        resolve(transcription);
+      });
+
+      session
+        .start({
+          transcription_config: {
+            language: 'en',
+            operating_point: 'enhanced',
+            enable_partials: true,
+            max_delay: 2,
+          },
+          audio_format: { type: 'file' },
+        })
+        .then(() => {
+          //prepare file stream
+          const fileStream = fs.createReadStream(`${__dirname}/../../data/test_audio.wav`);
+
+          //send it
+          fileStream.on('data', (sample) => {
+            session.sendAudio(sample);
+          });
+
+          //end the session
+          fileStream.on('end', () => {
+            session.stop();
+          });
+
+          return;
+
+        })
+        .catch((error) => {
+          reject(error);
+        });
     } catch (error) {
       logger.info({error}, 'failed to get speechmatics transcript');
       reject(error);
@@ -560,6 +588,10 @@ function decryptCredential(obj, credential, logger, isObscureKey = true) {
   } else if ('soniox' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+  } else if ('speechmatics' === obj.vendor) {
+    const o = JSON.parse(decrypt(credential));
+    obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;
+    obj.speechmatics_stt_uri = o.speechmatics_stt_uri;
   } else if ('elevenlabs' === obj.vendor) {
     const o = JSON.parse(decrypt(credential));
     obj.api_key = isObscureKey ? obscureKey(o.api_key) : o.api_key;