add support for deepgram STT

lib/auth/index.js

@@ -11,7 +11,7 @@ const sql = `
 function makeStrategy(logger, retrieveKey) {
   return new Strategy(
     async function(token, done) {
-      logger.debug(`validating with token ${token}`);
+      //logger.debug(`validating with token ${token}`);
       jwt.verify(token, process.env.JWT_SECRET, async(err, decoded) => {
         if (err) {
           if (err.name === 'TokenExpiredError') {

lib/routes/api/speech-credentials.js

@@ -14,7 +14,8 @@ const {
   testMicrosoftTts,
   testWellSaidTts,
   testNuanceStt,
-  testNuanceTts
+  testNuanceTts,
+  testDeepgramStt
 } = require('../../utils/speech-utils');
 
 const obscureKey = (key) => {
@@ -88,6 +89,11 @@ const encryptCredential = (obj) => {
       const nuanceData = JSON.stringify({client_id, secret});
       return encrypt(nuanceData);
 
+    case 'deepgram':
+      assert(api_key, 'invalid deepgram speech credential: api_key is required');
+      const deepgramData = JSON.stringify({api_key});
+      return encrypt(deepgramData);
+
     default:
       assert(false, `invalid or missing vendor: ${vendor}`);
   }
@@ -175,6 +181,10 @@ router.get('/', async(req, res) => {
         obj.client_id = o.client_id;
         obj.secret = obscureKey(o.secret);
       }
+      else if ('deepgram' === obj.vendor) {
+        const o = JSON.parse(decrypt(credential));
+        obj.api_key = obscureKey(o.api_key);
+      }
       return obj;
     }));
   } catch (err) {
@@ -225,6 +235,10 @@ router.get('/:sid', async(req, res) => {
       obj.client_id = o.client_id;
       obj.secret = obscureKey(o.secret);
     }
+    else if ('deepgram' === obj.vendor) {
+      const o = JSON.parse(decrypt(credential));
+      obj.api_key = obscureKey(o.api_key);
+    }
     res.status(200).json(obj);
   } catch (err) {
     sysError(logger, res, err);
@@ -474,6 +488,19 @@ router.get('/:sid/test', async(req, res) => {
         }
       }
     }
+    else if (cred.vendor === 'deepgram') {
+      const {api_key} = credential;
+      if (cred.use_for_stt) {
+        try {
+          await testDeepgramStt(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);
   } catch (err) {
     sysError(logger, res, err);

lib/utils/speech-utils.js

@@ -2,6 +2,7 @@ const ttsGoogle = require('@google-cloud/text-to-speech');
 const sttGoogle = require('@google-cloud/speech').v1p1beta1;
 const Polly = require('aws-sdk/clients/polly');
 const AWS = require('aws-sdk');
+const { Deepgram } = require('@deepgram/sdk');
 const bent = require('bent');
 const fs = require('fs');
 
@@ -42,6 +43,33 @@ const testGoogleStt = async(logger, credentials) => {
   }
 };
 
+const testDeepgramStt = async(logger, credentials) => {
+  const {api_key} = credentials;
+  const deepgram = new Deepgram(api_key);
+
+  const mimetype = 'audio/wav';
+  const source = {
+    buffer: fs.readFileSync(`${__dirname}/../../data/test_audio.wav`),
+    mimetype: mimetype
+  };
+
+  return new Promise((resolve, reject) => {
+    // Send the audio to Deepgram and get the response
+    deepgram.transcription
+      .preRecorded(source, {punctuate: true})
+      .then((response) => {
+        //logger.debug({response}, 'got transcript');
+        if (response?.results?.channels[0]?.alternatives?.length > 0) resolve(response);
+        else reject(new Error('no transcript returned'));
+        return;
+      })
+      .catch((err) => {
+        logger.info({err}, 'failed to get deepgram transcript');
+        reject(err);
+      });
+  });
+};
+
 const testAwsTts = (logger, credentials) => {
   const polly = new Polly(credentials);
   return new Promise((resolve, reject) => {
@@ -138,5 +166,6 @@ module.exports = {
   testMicrosoftStt,
   testWellSaidStt,
   testNuanceTts,
-  testNuanceStt
+  testNuanceStt,
+  testDeepgramStt
 };