Skip to content

Commit b8158aa

Browse files
chughtschughts
committed
processing of 1st set of inputs working
1 parent 70b01f6 commit b8158aa

File tree

1 file changed

+100
-83
lines changed
  • services/speech_to_text

1 file changed

+100
-83
lines changed

‎services/speech_to_text/v1.js‎

Lines changed: 100 additions & 83 deletions
Original file line numberDiff line numberDiff line change
@@ -116,7 +116,10 @@ module.exports = function (RED) {
116116
websocket = null,
117117
socketCreationInProcess = false,
118118
socketListening = false,
119-
contentType = null,
119+
startPacket = { action: 'start',
120+
'content-type' :'audio/wav',
121+
'interim_results': true
122+
},
120123
audioStack =[];
121124
const HOUR = 60 * 60;
122125

@@ -242,13 +245,12 @@ module.exports = function (RED) {
242245

243246
if ("string" === typeof msg.payload) {
244247
msg.payload = JSON.parse(tmp);
245-
if ( msg.payload.mode &&
246-
'start' === msg.payload.mode &&
247-
msg.payload.type ) {
248-
contentType = msg.payload.type;
248+
if ( msg.payload.action &&
249+
'start' === msg.payload.action) {
250+
startPacket = msg.payload;
249251
}
250252
} else {
251-
msg.payload = { 'mode' : 'data', 'data' : tmp };
253+
msg.payload = { 'action' : 'data', 'data' : tmp };
252254
}
253255

254256
return Promise.resolve(msg.payload);
@@ -347,81 +349,86 @@ module.exports = function (RED) {
347349
return p;
348350
}
349351

350-
function connectIfNeeded(dataIn) {
352+
function connectIfNeeded() {
353+
console.log('re-establishing the connect');
354+
websocket = null;
355+
socketCreationInProcess = false;
356+
processSTTSocketStart()
357+
.then(() => {
358+
return Promise.resolve();
359+
//return;
360+
})
351361
}
352362

353-
function connectToSTTSocket(dataIn) {
363+
function processSTTSocketStart() {
354364
var p = new Promise(function resolver(resolve, reject) {
355-
//return connectIfNeeded(audioData);
356-
//resolve();
357-
var audioData = dataIn;
358365
var model = config.lang + '_' + config.band;
359-
var wsURI = 'wss://stream.watsonplatform.net/speech-to-text/api/v1/recognize'
360-
+ '?watson-token=' + token + '&model=' + model;
361-
362-
//if (!audioData) {
363-
// audioData = {mode : 'start'};
364-
//}
365-
366-
if (!websocket && !socketCreationInProcess &&
367-
'start' === audioData.mode) {
368-
socketCreationInProcess = true;
369-
var ws = new WebSocket(wsURI);
370-
ws.on('open', () => {
371-
console.log('******************');
372-
console.log('Web Socket is now open');
373-
374-
var message = {
375-
action: 'start',
376-
'content-type': contentType ? contentType :'audio/wav',
377-
//'content-type': 'audio/webm',
378-
'interim_results': true
379-
//'max-alternatives': 3,
380-
//keywords: ['colorado', 'tornado', 'tornadoes'],
381-
//'keywords_threshold': 0.5
382-
};
383-
console.log('Signalling Start');
384-
ws.send(JSON.stringify(message));
385-
websocket = ws;
386-
socketCreationInProcess = false;
366+
var wsURI = '';
367+
368+
if (endpoint) {
369+
var tmp = endpoint.replace('https', 'wss');
370+
wsURI = tmp + '/v1/recognize'
371+
+ '?watson-token=' + token + '&model=' + model;
372+
// https://stream.watsonplatform.net/speech-to-text/api
373+
} else {
374+
wsURI = 'wss://stream.watsonplatform.net/speech-to-text/api/v1/recognize'
375+
+ '?watson-token=' + token + '&model=' + model;
376+
}
377+
378+
if (!websocket && !socketCreationInProcess) {
379+
socketCreationInProcess = true;
380+
var ws = new WebSocket(wsURI);
381+
ws.on('open', () => {
382+
console.log('******************');
383+
console.log('Web Socket is now open');
384+
385+
console.log('Signalling Start');
386+
ws.send(JSON.stringify(startPacket));
387+
websocket = ws;
388+
socketCreationInProcess = false;
387389
//resolve();
388-
});
389-
(function(dIn) {
390-
ws.on('message', (data) => {
391-
// First message will be 'state': 'listening'
392-
console.log('-----------------------');
393-
console.log('Data Received from Input');
394-
console.log(data);
395-
socketListening = true;
396-
var d = JSON.parse(data)
397-
var newMsg = {payload : JSON.parse(data)};
398-
node.send(newMsg);
399-
if (dIn && d && d.state && 'listening' === d.state){
400-
console.log('We are now listening');
401-
//resolve(true);
402-
}
403-
});
404-
})(dataIn);
405-
406-
ws.on('close', () => {
407-
if (websocket) {
408-
websocket.close();
409-
}
410-
websocket = null;
411-
socketListening = false;
412-
console.log('STT Socket disconnected');
413-
//setTimeout(connectToSTTSocket, 1000);
414390
});
391+
392+
ws.on('message', (data) => {
393+
// First message will be 'state': 'listening'
394+
console.log('-----------------------');
395+
console.log('Data Received from Input');
396+
console.log(data);
397+
var d = JSON.parse(data)
398+
var newMsg = {payload : JSON.parse(data)};
399+
node.send(newMsg);
400+
if (d && d.state && 'listening' === d.state){
401+
socketListening = true;
402+
console.log('We are now listening');
403+
resolve();
404+
}
405+
});
406+
407+
ws.on('close', () => {
408+
//if (websocket) {
409+
// websocket.close();
410+
//}
411+
//websocket = null;
412+
socketListening = false;
413+
console.log('STT Socket disconnected');
414+
setTimeout(connectIfNeeded, 1000);
415+
});
416+
417+
ws.on('error', (err) => {
418+
socketListening = false;
419+
console.log('Error Detected');
420+
reject(err);
421+
});
422+
415423
} else {
416-
if (dataIn) {
417-
//resolve(false);
418-
}
424+
resolve();
419425
}
420-
resolve(socketListening);
426+
421427
});
422428
return p;
423429
}
424430

431+
425432
function stackAudioFile(audioData) {
426433
console.log('Pushing onto the stack');
427434
audioStack.push(audioData);
@@ -433,14 +440,14 @@ module.exports = function (RED) {
433440
var p = new Promise(function resolver(resolve, reject) {
434441
console.log('Sending Audio - outer ');
435442
console.log(audioData);
436-
if (audioData && audioData.mode) {
437-
if ('data' === audioData.mode) {
443+
if (audioData && audioData.action) {
444+
if ('data' === audioData.action) {
438445
console.log('Sending Audio - inner');
439446

440447
// send stack First
441448
if (audioStack && audioStack.length) {
442449
audioStack.forEach((a) => {
443-
if (a && a.mode && 'data' === a.mode) {
450+
if (a && a.action && 'data' === a.action) {
444451
websocket.send(a.data);
445452
}
446453
});
@@ -456,11 +463,11 @@ module.exports = function (RED) {
456463
}
457464
});
458465
} else {
459-
var message = { action: 'stop' };
460-
if (audioData.mode === 'stop') {
466+
//var message = { action: 'stop' };
467+
if (audioData.action === 'stop') {
461468
console.log('Signalling Stop');
462-
websocket.send(JSON.stringify(message));
463-
socketListening = false;
469+
websocket.send(JSON.stringify(audioData));
470+
//socketListening = false;
464471

465472
// Closing as refresh doesn't appear to work
466473
//websocket.close();
@@ -477,13 +484,23 @@ module.exports = function (RED) {
477484
var speech_to_text = determineService();
478485
var p = getToken(speech_to_text)
479486
.then(() => {
480-
return connectToSTTSocket(audioData);
481-
})
482-
.then((listening) => {
483-
if (listening) {
484-
return sendAudioSTTSocket(audioData);
485-
} else {
486-
return stackAudioFile(audioData);
487+
switch (audioData.action) {
488+
case 'start':
489+
//return Promise.reject('Its a start');
490+
return processSTTSocketStart();
491+
case 'stop':
492+
case 'data':
493+
// Add a Delay to allow the listening thread to kick in
494+
setTimeout(() => {
495+
if (socketListening) {
496+
return sendAudioSTTSocket(audioData);
497+
} else {
498+
return stackAudioFile(audioData);
499+
}
500+
}, 1000);
501+
//return Promise.reject('Its a data or stop');
502+
default:
503+
return Promise.resolve();
487504
}
488505
})
489506
.then(() => {

0 commit comments

Comments
 (0)