reverb
v0.2.0
Published
Push-based client library for Amazon Echo
Downloads
89
Readme
Reverb
Is there an Echo in here?
What?
Reverb is an open source library for creating custom interactions with the Amazon Echo. Unlike existing solutions, Reverb uses a true "push"-style interaction and does not rely on using a simulated or real web browser behind the scenes. This makes it fast and lightweight on both your network and your machine.
Usage
For now, Reverb requires you to get extract some data by hand. These are the "Cookie" header and the device serial number. These can both be easily acquired using the Developer Tools that come with Chrome. Simply open the Echo web app and open the Network tab of the developer tools. Click on the Filter icon (it looks a bit like a funnel) and filter down to "WebSockets." Then, refresh the page.
There should be a single entry in the list pointing to
dp-gw-na-js.amazon.com
. Select that and go to the "Headers" tab.
The serial number you want will be the x-amz-device-serial
parameter under "Query String Parameters," and the "Cookie"
is the (potentially huge) string next to Cookie:
under
"Request Headers."
Now that you have those, it's simple to get access to the Echo's transcriptions of your requests:
var Reverb = require('reverb');
new Reverb({
cookie: // The Cookie retrieved above
, serial: // The serial retrieved above
}).on('ready', function() {
// this is emitted when the connection is
// established and the hands have been shaked
console.log("Ready!");
}).on('activity', function(activity) {
// parse the transcription and do something
// with it here
console.log("You said:", activity.summary);
}).on('timerStart', function(timer) {
// do something when timers are started
console.log("timer started:", timer.remainingTime);
}).on('timerPause', function(timer) {
// do somethign when timers are paused
console.log("timer paused:", timer.remainingTime);
}).on('timerComplete', function(timer) {
// do something when timers complete
console.log("timer complete");
}).on('alarmSet', function(alarm) {
// do something when alarms are set
console.log("alarm set:", alarm);
}).on('alarmUpdate', function(alarm) {
// do something when alarms are changed
console.log("alarm time updated:", alarm.alarmTime);
}).on('alarmComplete', function(alarm) {
// do something when alarms complete
console.log("alarm complete");
}).on('alarmRemove', function(alarm) {
// do something when alarmas are turned off
console.log("alarm unset");
});
Note that, as with other solutions, you will have to append "cancel" or "stop" to the end of your custom commands to prevent Echo from trying to process it.
Future Work
- Provide an easy mechanism for authenticating
- Perhaps, provide an "Action Dispatcher" to make parsing the activity text easier
- Perhaps, provide a useful
-g
install- Plugins could be dropped into a folder
Why?
The Amazon Echo's voice capabilities are truly amazing, but they have yet to provide a proper API for extending its capabilities. I was not satisfied with the "poll"-style mechanism used in existing solutions and wanted to see if there was something better.