support stt speechmatics (#353)

* support stt speechmatics

* support speechmatics region authentication

* update testcase for speechmatics_stt_uri

lib/routes/api/speech-credentials.js

@@ -10,7 +10,8 @@ const {decryptCredential, testWhisper, testDeepgramTTS,
   testPlayHT,
   testRimelabs,
   testVerbioTts,
-  testVerbioStt} = require('../../utils/speech-utils');
+  testVerbioStt,
+  testSpeechmaticsStt} = require('../../utils/speech-utils');
 const {DbErrorUnprocessableRequest, DbErrorForbidden, DbErrorBadRequest} = require('../../utils/errors');
 const {
   testGoogleTts,
@@ -122,6 +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,
@@ -236,6 +238,12 @@ const encryptCredential = (obj) => {
       const elevenlabsData = JSON.stringify({api_key, model_id, options});
       return encrypt(elevenlabsData);
 
+    case 'speechmatics':
+      assert(api_key, 'invalid speechmatics speech credential: api_key is required');
+      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':
       assert(api_key, 'invalid playht speech credential: api_key is required');
       assert(user_id, 'invalid playht speech credential: user_id is required');
@@ -768,6 +776,18 @@ router.get('/:sid/test', async(req, res) => {
           SpeechCredential.ttsTestResult(sid, false);
         }
       }
+    } else if (cred.vendor === 'speechmatics') {
+      const {api_key} = credential;
+      if (cred.use_for_stt) {
+        try {
+          await testSpeechmaticsStt(logger, {api_key});
+          results.stt.status = 'ok';
+          SpeechCredential.ttsTestResult(sid, true);
+        } catch (err) {
+          results.stt = {status: 'fail', reason: err.message};
+          SpeechCredential.ttsTestResult(sid, false);
+        }
+      }
     } else if (cred.vendor === 'playht') {
       if (cred.use_for_tts) {
         try {

lib/utils/speech-utils.js

@@ -7,6 +7,7 @@ const bent = require('bent');
 const fs = require('fs');
 const { AssemblyAI } = require('assemblyai');
 const {decrypt, obscureKey} = require('./encrypt-decrypt');
+const { RealtimeSession } = require('speechmatics');
 
 const TtsGoogleLanguagesVoices = require('./speech-data/tts-google');
 const TtsAwsLanguagesVoices = require('./speech-data/tts-aws');
@@ -54,6 +55,61 @@ const testSonioxStt = async(logger, credentials) => {
   });
 };
 
+const testSpeechmaticsStt = async(logger, credentials) => {
+  const {api_key, speechmatics_stt_uri} = credentials;
+  return new Promise(async(resolve, reject) => {
+    try {
+      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);
+    }
+  });
+};
+
 const testNuanceTts = async(logger, getTtsVoices, credentials) => {
   const voices = await getTtsVoices({vendor: 'nuance', credentials});
   return voices;
@@ -532,6 +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;
@@ -1066,5 +1126,6 @@ module.exports = {
   testWhisper,
   testVerbioTts,
   testVerbioStt,
-  getLanguagesAndVoicesForVendor
+  getLanguagesAndVoicesForVendor,
+  testSpeechmaticsStt
 };

package-lock.json

@@ -42,6 +42,7 @@
         "passport-http-bearer": "^1.0.1",
         "pino": "^8.20.0",
         "short-uuid": "^4.2.2",
+        "speechmatics": "^4.0.0",
         "stream-buffers": "^3.0.2",
         "stripe": "^14.24.0",
         "swagger-ui-express": "^5.0.0",
@@ -9046,6 +9047,35 @@
         "node": ">=8"
       }
     },
+    "node_modules/speechmatics": {
+      "version": "4.0.0",
+      "resolved": "https://registry.npmjs.org/speechmatics/-/speechmatics-4.0.0.tgz",
+      "integrity": "sha512-k4za7lTBA8Y3rXXqRBxHWQcwmZ2j4bewSFqKek/FpPNj+fgChtKH4UU1uDUHb9RbZ6R4zTwfZQoaNqHU3dUmEw==",
+      "license": "MIT",
+      "dependencies": {
+        "bufferutil": "^4.0.7",
+        "events": "^3.3.0",
+        "utf-8-validate": "^6.0.3",
+        "ws": "^8.13.0",
+        "zod": "^3.21.4"
+      },
+      "engines": {
+        "node": "^14.15.0 || ^16.10.0 || >=18.0.0"
+      }
+    },
+    "node_modules/speechmatics/node_modules/utf-8-validate": {
+      "version": "6.0.4",
+      "resolved": "https://registry.npmjs.org/utf-8-validate/-/utf-8-validate-6.0.4.tgz",
+      "integrity": "sha512-xu9GQDeFp+eZ6LnCywXN/zBancWvOpUMzgjLPSjy4BRHSmTelvn2E0DG0o1sTiw5hkCKBHo8rwSKncfRfv2EEQ==",
+      "hasInstallScript": true,
+      "license": "MIT",
+      "dependencies": {
+        "node-gyp-build": "^4.3.0"
+      },
+      "engines": {
+        "node": ">=6.14.2"
+      }
+    },
     "node_modules/split2": {
       "version": "4.2.0",
       "resolved": "https://registry.npmjs.org/split2/-/split2-4.2.0.tgz",
@@ -10157,6 +10187,15 @@
       "funding": {
         "url": "https://github.com/sponsors/sindresorhus"
       }
+    },
+    "node_modules/zod": {
+      "version": "3.23.8",
+      "resolved": "https://registry.npmjs.org/zod/-/zod-3.23.8.tgz",
+      "integrity": "sha512-XBx9AXhXktjUqnepgTiE5flcKIYWi/rme0Eaj+5Y0lftuGBq+jyRu/md4WnuxqgP1ubdpNCsYEYPxrzVHD8d6g==",
+      "license": "MIT",
+      "funding": {
+        "url": "https://github.com/sponsors/colinhacks"
+      }
     }
   }
 }

package.json

@@ -52,6 +52,7 @@
     "passport-http-bearer": "^1.0.1",
     "pino": "^8.20.0",
     "short-uuid": "^4.2.2",
+    "speechmatics": "^4.0.0",
     "stream-buffers": "^3.0.2",
     "stripe": "^14.24.0",
     "swagger-ui-express": "^5.0.0",

test/speech-credentials.js

@@ -532,6 +532,39 @@ test('speech credentials tests', async(t) => {
       t.ok(result.statusCode === 204, 'successfully deleted speech credential');
     }
 
+    /* add a credential for Speechmatics */
+    if (process.env.SPEECHMATICS_API_KEY) {
+      result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,
+        body: {
+          vendor: 'speechmatics',
+          use_for_stt: true,
+          api_key: process.env.SPEECHMATICS_API_KEY,
+          speechmatics_stt_uri: 'eu2.rt.speechmatics.com'
+        }
+      });
+      t.ok(result.statusCode === 201, 'successfully added speech credential for speechmatics');
+      const ms_sid = result.body.sid;
+
+      /* test the speech credential */
+      result = await request.get(`/Accounts/${account_sid}/SpeechCredentials/${ms_sid}/test`, {
+        resolveWithFullResponse: true,
+        auth: authUser,
+        json: true,   
+      });
+      console.log(JSON.stringify(result));
+      t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for speechmatics');
+
+      /* delete the credential */
+      result = await request.delete(`/Accounts/${account_sid}/SpeechCredentials/${ms_sid}`, {
+        auth: authUser,
+        resolveWithFullResponse: true,
+      });
+      t.ok(result.statusCode === 204, 'successfully deleted speech credential');
+    }
+
     /* add a credential for nvidia */
     result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
       resolveWithFullResponse: true,