Send Recorded Twilio Audio To Lex - twilio

Question

Currently I am able to record user input, pass the recording URL to the needed function, and download the audio file locally. What I am trying to do with the audio file is either get a buffer of it to send to Lex or convert it to the format Lex needs.
Per AWS Documentation the following values are accepted for the input stream param value:
var params = {
botAlias: 'STRING_VALUE', /* required */
botName: 'STRING_VALUE', /* required */
contentType: 'STRING_VALUE', /* required */
inputStream: new Buffer('...') || 'STRING_VALUE' || streamObject, /*required */
userId: 'STRING_VALUE', /* required */
accept: 'STRING_VALUE',
requestAttributes: any /* This value will be JSON encoded on your behalf with JSON.stringify() */,
sessionAttributes: any /* This value will be JSON encoded on your behalf with JSON.stringify() */
};
lexruntime.postContent(params, function(err, data) {
if (err) console.log(err, err.stack); // an error occurred
else console.log(data); // successful response
});
Per the twilio documentation it looks like the audio file is pretty flexible...
A request to the RecordingUrl will return a recording in binary WAV audio format by default. To request the recording in MP3 format, append ".mp3" to the RecordingUrl.
What do I need to do to get the twilio recorded audio in the right format for Lex? Is it just a matter of building the correct Lex param set or do I need to do some audio conversion before hand? I am writing this application in node js if that helps and I can add more code if it will help.

Answer 1

I was able to figure this out by downloading the file from Twilio as a PCM and changing my parameters a bit. Also, due to the way that Twilio handles the record verb, I needed to transfer the call to a hold state while waiting for the recordingStatusCallback to POST out. I also send a text to the caller with the final status from Lex.
The code I used to download the file:
app.post('/processRecording', (request, response) => {
var https = require('https');
var fs = require('fs');
let callSID = request.body.CallSid;
let url = request.body.RecordingUrl;
var saveFile = new Promise(function(resolve, reject) {
let fileName = callSID+ ".pcm";
var file = fs.createWriteStream(fileName);
var request = https.get(url, function(response) {
response.pipe(file);
resolve();
});
});
});
const accountSid = 'YOUR ACCOUNT SID';
const authToken = 'YOUR AUTH TOKEN';
const client = require('twilio')(accountSid, authToken);
//Once the file is downloaded, I then fetch the call from the hold state using this code:
saveFile.then(function(){
client.calls(callSID)
.update({method: 'POST', url: '/updateCall'})
.then(call => console.log(call.to))
.done();
});
And my updateCall endpoint looks like this:
app.post('/updateCall', (request, response) => {
let lexruntime = new AWS.LexRuntime();
let recordedFileName = request.body.CallSid + '.pcm';
let toNumber = request.body.To;
let fromNumber = request.body.From;
let twiml = new Twilio.twiml.VoiceResponse();
let lexFileStream = fs.createReadStream(recordedFileName);
let sid = request.body.CallSid;
var params = {
botAlias: 'prod', /* required */
botName: 'OrderFlowers', /* required */
contentType: 'audio/lpcm; sample-rate=8000; sample-size-bits=16; channel-count=1; is-big-endian=false',
accept: 'text/plain; charset=utf-8',
userId: sid /* required */
};
params.inputStream = lexFileStream;
lexruntime.postContent(params, function(err, data) {
if (err) console.log(err, err.stack); // an error occurred
else console.log(data); // successful response
if (data.dialogState == "ElicitSlot" || data.dialogState == "ConfirmIntent" || data.dialogState == "ElicitIntent" ){
twiml.say(data.message);
twiml.redirect({
method: 'POST'
}, '/recordVoice');
response.type('text/xml');
response.send(twiml.toString());
}
else if (data.dialogState == "Fulfilled" ){
twiml.say(data.message);
response.type('text/xml');
response.send(twiml.toString());
client.messages.create({
to: toNumber,
from: fromNumber,
body: data.message
}).then(msg => {
}).catch(err => console.log(err));
}
else{
twiml.say(data.message);
response.type('text/xml');
response.send(twiml.toString());
}
});
});
The recordVoice endpoint is actually a Twilio Serverless Function But I think this is what it would look like as an express endpoint:
app.post('/recordVoice', (request, response) => {
let twiml = new Twilio.twiml.VoiceResponse();
twiml.record({
action: '/deadAir',
recordingStatusCallback: '/processRecording',
trim: true,
maxLength: 10,
finishOnKey: '*'
});
twiml.say('I did not receive a recording');
response.type('text/xml');
response.send(twiml.toString());
});
The /deadAir endpoint is also a Twilio Serverless Function but this is what it would look like:
app.post('/deadAir', (request, response) => {
let twiml = new Twilio.twiml.VoiceResponse();
twiml.pause({
length: 60
});
response.type('text/xml');
response.send(twiml.toString());
});