Turen NPM | npm.io

node-turen

The Node.js client for TurenCore.

Installation

$ npm install turen --save

Get Started

var options = {
  host: 'apigwws.open.rokid.com',
  port: 443,
  key: 'rokid openplatform key',
  secret: 'rokid openplatform secret',
  deviceTypeId: 'rokid device type id',
  deviceId: 'rokid device id',
};
var speech = new TurenSpeech();
speech.on('voice coming', (event) => {
  // voice coming
});
speech.on('voice accept', (event) => {
  // voice accept
});
speech.on('asr end', (asr, event) => {
  // asr
});
speech.on('nlp', (response, event) => {
  // response.asr
  // response.nlp
  // response.action
});
speech.on('disconnect', (socketType/* event or rpc */) => {
  // got if some event is disconnected
});
speech.start(options);

Services

TurenCore provides multiple socket-based services for different functionalities.

RPC

Rpc service is used to call method of TurenCore, which includes:

Restart()
RestartWithoutArgs()
Pickup()
IsPickup()
OpenMic()
SetStack()
SetSkillOption()

CMD

Cmd service is used to call debug method which includes:

openMic()
closeMic()
pickup()
reset()
readyForAsr()
setAngle(deg)

Event

Event service is to be notified for all voice and nlp events, includes:

voice coming returns the speech energy and direction sl when triggered locally.
voice local sleep when sleep locally.
asr begin when cloud speech recognition begins.
asr end returns the final value asr when cloud speech recognition ends.
nlp returns the nlp and action when NLP is done.

A complete events list is at here.

Audio

Audio service is also to debug with TurenCore, we could pull audio streams from every stage like AEC/BF/VAD.

You could use it as a socket handle:

var audioStream = new TurenAudio('mic_in');
audioStream.on('data', (chunk) => {
  // got data
});
audioStream.connect();

Streaming API also could use used:

var writable = fs.createWriteStream('/path/to/your/file');
writable.pipe(audioStream);

Push audio data to TurenCore:

var audioWritable = new TurenAudio('mic_out');
await audioWritable.connect();

// write a buffer
audioWritable.send(new Buffer(1024));
// write a readable stream
audioWritable.send(fs.createReadStream('/path/to/your/pcm/file'));

The available type values of TurenAudio is:

mic_in as a Readable stream to pull the raw data, its format depends on your micphone configuration.
bf_out as a Readable stream to pull the pcm after BF.
bf4_out as a Readable stream to pull the pcm after BF with selected 4 channels.
bf12_out as a Readable stream to pull the pcm after BF with the full 12 channels.
aec_out as a Readable stream to pull the pcm after AEC.
codec_in as a Readable stream to pull the pcm before opu codec.
speech_in as a Readable stream to pull the opu before uploading to cloud speech service.
mic_out as a Writable stream to push raw data.

The following types are available only for our CTC model:

ctc.line_0 as a Readable stream to pull the pcm to the CTC line 0.
ctc.line_1 as a Readable stream to pull the pcm to the CTC line 1.
ctc.line_2 as a Readable stream to pull the pcm to the CTC line 2.
ctc.line_3 as a Readable stream to pull the pcm to the CTC line 3.

Logger

You also could access to TurenCore's logs by TurenLogger:

var TurenLogger = require('turen').client.TurenLogger;
var logger = new TurenLogger();
logger.on('data', (data) => {
  // receives the data
});
logger.connect();