ben.carrasco/jambonz-api-server
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-api-server.git synced 2026-07-04 19:21:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

added testing of microsoft stt credential

Browse Source
This commit is contained in:
Dave Horton
2022-11-10 17:54:40 -05:00
parent ee5e25bb8d
commit 0fc8500361
4 changed files with 86 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/utils/speech-utils.js
+33 -5
View File
@@ -3,6 +3,7 @@ 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 sdk = require('microsoft-cognitiveservices-speech-sdk');
const bent = require('bent');
const fs = require('fs');
@@ -70,6 +71,38 @@ const testDeepgramStt = async(logger, credentials) => {
});
};
const testMicrosoftStt = async(logger, credentials) => {
const {api_key, region} = credentials;
const speechConfig = sdk.SpeechConfig.fromSubscription(api_key, region);
const audioConfig = sdk.AudioConfig.fromWavFileInput(fs.readFileSync(`${__dirname}/../../data/test_audio.wav`));
speechConfig.speechRecognitionLanguage = 'en-US';
const speechRecognizer = new sdk.SpeechRecognizer(speechConfig, audioConfig);
return new Promise((resolve, reject) => {
speechRecognizer.recognizeOnceAsync((result) => {
switch (result.reason) {
case sdk.ResultReason.RecognizedSpeech:
resolve();
break;
case sdk.ResultReason.NoMatch:
reject('Speech could not be recognized.');
break;
case sdk.ResultReason.Canceled:
const cancellation = sdk.CancellationDetails.fromResult(result);
logger.info(`CANCELED: Reason=${cancellation.reason}`);
if (cancellation.reason == sdk.CancellationReason.Error) {
logger.info(`CANCELED: ErrorCode=${cancellation.ErrorCode}`);
logger.info(`CANCELED: ErrorDetails=${cancellation.errorDetails}`);
}
reject(cancellation.reason);
break;
}
speechRecognizer.close();
});
});
};
const testAwsTts = (logger, credentials) => {
const polly = new Polly(credentials);
return new Promise((resolve, reject) => {
@@ -127,11 +160,6 @@ const testMicrosoftTts = async(logger, credentials) => {
}
};
const testMicrosoftStt = async(logger, credentials) => {
//TODO
return true;
};
const testWellSaidTts = async(logger, credentials) => {
const {api_key} = credentials;
try {
package-lock.json
Generated
+7 -6
View File
@@ -27,6 +27,7 @@
"helmet": "^5.1.0",
"jsonwebtoken": "^8.5.1",
"mailgun.js": "^3.7.3",
"microsoft-cognitiveservices-speech-sdk": "^1.24.1",
"mysql2": "^2.3.3",
"passport": "^0.6.0",
"passport-http-bearer": "^1.0.1",
@@ -4308,9 +4309,9 @@
}
},
"node_modules/microsoft-cognitiveservices-speech-sdk": {
"version": "1.24.0",
"resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.24.0.tgz",
"integrity": "sha512-26/lmzsRHgIzP03DSts1pqp+5UINdpDo35jnmi9QRzhYebtLrdEydwPR9TF0wa9TB+vbMeMFWD+gy9VKfWXAvQ==",
"version": "1.24.1",
"resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.24.1.tgz",
"integrity": "sha512-7oAlVge4cPnCeNHeIVUQe4tKZmfGtsriD8rjl7uAoPcwG4hF3BXVVhUEkhlW+B8i5zVAJl3fH4BbAfZPCtrbvg==",
"hasInstallScript": true,
"dependencies": {
"agent-base": "^6.0.1",
@@ -10023,9 +10024,9 @@
"integrity": "sha512-iclAHeNqNm68zFtnZ0e+1L2yUIdvzNoauKU4WBA3VvH/vPFieF7qfRlwUZU+DA9P9bPXIS90ulxoUoCH23sV2w=="
},
"microsoft-cognitiveservices-speech-sdk": {
"version": "1.24.0",
"resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.24.0.tgz",
"integrity": "sha512-26/lmzsRHgIzP03DSts1pqp+5UINdpDo35jnmi9QRzhYebtLrdEydwPR9TF0wa9TB+vbMeMFWD+gy9VKfWXAvQ==",
"version": "1.24.1",
"resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.24.1.tgz",
"integrity": "sha512-7oAlVge4cPnCeNHeIVUQe4tKZmfGtsriD8rjl7uAoPcwG4hF3BXVVhUEkhlW+B8i5zVAJl3fH4BbAfZPCtrbvg==",
"requires": {
"agent-base": "^6.0.1",
"asn1.js-rfc2560": "^5.0.1",
package.json
+1
View File
@@ -36,6 +36,7 @@
"helmet": "^5.1.0",
"jsonwebtoken": "^8.5.1",
"mailgun.js": "^3.7.3",
"microsoft-cognitiveservices-speech-sdk": "^1.24.1",
"mysql2": "^2.3.3",
"passport": "^0.6.0",
"passport-http-bearer": "^1.0.1",
test/speech-credentials.js
+45 -12
View File
@@ -30,7 +30,9 @@ test('speech credentials tests', async(t) => {
json: true,
body: {
vendor: 'google',
service_key: jsonKey
service_key: jsonKey,
use_for_tts: true,
use_for_stt: true
}
});
t.ok(result.statusCode === 201, 'successfully added a speech credential to service provider');
@@ -63,7 +65,9 @@ test('speech credentials tests', async(t) => {
json: true,
body: {
vendor: 'google',
service_key: jsonKey
service_key: jsonKey,
use_for_tts: true,
use_for_stt: true
}
});
t.ok(result.statusCode === 201, 'successfully added speech credential');
@@ -112,20 +116,20 @@ test('speech credentials tests', async(t) => {
});
t.ok(result.statusCode === 204, 'successfully deleted speech credential');
/* add a credential for microsoft */
if (process.env.MICROSOFT_API_KEY && process.env.MICROSOFT_REGION) {
/* add / test a credential for google */
if (process.env.GCP_JSON_KEY) {
result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
resolveWithFullResponse: true,
auth: authUser,
json: true,
body: {
vendor: 'microsoft',
vendor: 'google',
use_for_tts: true,
api_key: process.env.MICROSOFT_API_KEY,
region: process.env.MICROSOFT_REGION
use_for_stt: true,
service_key: process.env.GCP_JSON_KEY
}
});
t.ok(result.statusCode === 201, 'successfully added speech credential');
t.ok(result.statusCode === 201, 'successfully added speech credential for google');
const ms_sid = result.body.sid;
/* test the speech credential */
@@ -134,8 +138,37 @@ test('speech credentials tests', async(t) => {
auth: authUser,
json: true,
});
console.log(JSON.stringify(result));
t.ok(result.statusCode === 200 && result.body.tts.status === 'ok', 'successfully tested speech credential for deepgram');
//console.log(JSON.stringify(result));
t.ok(result.statusCode === 200 && result.body.tts.status === 'ok', 'successfully tested speech credential for google tts');
t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for google stt');
}
/* add / test a credential for microsoft */
if (process.env.MICROSOFT_API_KEY && process.env.MICROSOFT_REGION) {
result = await request.post(`/Accounts/${account_sid}/SpeechCredentials`, {
resolveWithFullResponse: true,
auth: authUser,
json: true,
body: {
vendor: 'microsoft',
use_for_tts: true,
use_for_stt: true,
api_key: process.env.MICROSOFT_API_KEY,
region: process.env.MICROSOFT_REGION
}
});
t.ok(result.statusCode === 201, 'successfully added speech credential for microsoft');
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.tts.status === 'ok', 'successfully tested speech credential for microsoft tts');
t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for microsoft stt');
}
/* add a credential for wellsaid */
@@ -159,7 +192,7 @@ test('speech credentials tests', async(t) => {
auth: authUser,
json: true,
});
console.log(JSON.stringify(result));
//console.log(JSON.stringify(result));
t.ok(result.statusCode === 200 && result.body.tts.status === 'ok', 'successfully tested speech credential for wellsaid');
/* delete the credential */
@@ -191,7 +224,7 @@ test('speech credentials tests', async(t) => {
auth: authUser,
json: true,
});
console.log(JSON.stringify(result));
//console.log(JSON.stringify(result));
t.ok(result.statusCode === 200 && result.body.stt.status === 'ok', 'successfully tested speech credential for deepgram');
/* delete the credential */