Google’s “smart reply” underline in Android is a flattering neat focus of appurtenance learning. Google’s servers indicate your incoming content messages or emails and writes replies for you. Smart replies hang out at the bottom of an app like Gmail or Google Inbox, and you can collect from several replies formed on the context of the message. Now Google is experimenting with making smart replies even faster by embedding respond options directly into Android notifications.
The examination comes from Google’s new “Area 120″ group, an thought incubator inside the company. Users who sealed up for the group’s early entrance program got an email yesterday announcing the new feature, which is an app the group is just job “Reply.” The app isn’t out yet, but the email shows off two judgment images and gives users a couple to sign up.
The images show a presentation from Hangouts and Android Messages with the approaching content and image, but next them, right in the presentation panel, are a few machine-produced replies. Someone asks “Are you at a restaurant?” and you can fire back a discerning respond with a singular tap.
With the second example, the email claims the complement will go over the common smart respond fare: in response to “When can you be home?” the complement pre-populates a respond that says “13 min,” finish with a automobile emoji. This is something that would be really considerable if it actually works. Google would first have to figure out your plcae (easy with GPS), figure out your home plcae (something you can tell the Assistant and Google Maps), then run a Google Maps query for traffic, calculate the drive time, and write the message. It’s all something Google routinely does with a singular voice query, but now it will do it in response to someone else’s query.
Best of all, Reply won’t just work with Google apps; the proclamation calls out “Hangouts, Allo, WhatsApp, Facebook Messenger, Android Messages, Skype, Twitter DMs, and Slack” as concordant apps.
How this is (probably) going to work
The causal spectator competence be confused about how this can presumably work when Google doesn’t own apps like Twitter, Slack, and Facebook. How can Google change the facilities of apps it doesn’t own?
The answer lies in the pictures, if you demeanour closely. Google shows notifications with the icons and names of “Hangouts” and “Android Messages,” but those notifications don’t come from those apps. As the first colored word in the presentation says, those notifications are actually from the app “Reply.” It’s all a crafty use of Android’s existent APIs.
First, Google will build a presentation listener app. In Android, there exists an API that can concede an app to read, reply, and fundamentally take over the whole presentation panel. It was creatively created with Android Wear in mind, permitting the Wear app to take your presentation information and lamp it to the watch, where it can be repackaged into a smartwatch form factor.
The Reply app will do something along the same lines as Android Wear—it will squeeze all the content and images from your messaging notifications by the notification-listener API and reconstruct those notifications. But while Wear was doing it for a smartwatch format, Reply will do it so it can slipstream the smart respond options into your presentation panels. Just as Android Wear did not need any work from a third party to support notifications, Reply shouldn’t need any additional work, either. That’s how it can work with WhatsApp, Slack, and the other third-party apps.
Reply will need to come up with some solution to not display the strange apps presentation and Reply’s duplicate, cloned presentation alongside any other, presumably by just dismissing the strange app notification.
The presentation listener API still doesn’t concede a third-party app to respond to your messages. For that there’s the “RemoteInput” API, which is another API creatively meant for Android Wear and Android Auto that will substantially be repurposed for Reply. Android’s watch and automobile interface is dictated to be used for voice input; this API allows messaging apps to accept a retard of content from Wear or Auto and send it to the suitable contact. In Android 7.0, Google stretched the use of the API for the “in-line reply” option. Just daub on a presentation and a keyboard would cocktail up, along with a tiny in-line content box. Reply will substantially be an prolongation of this API but with single-tap appurtenance denunciation submit instead of with a voice or keyboard.
Of course, we’ll endorse this is the way this all works when the Reply app comes out, but this is the many apparent way to make the app work. It’s not an ideal setup—ideally you’d wish Android to take caring of this natively, but remember this is just an initial 20 percent project. If people like it, maybe someday it can be combined to a future chronicle of Android, so Google can equivocate this penetrate of a setup.
We’re in the early entrance program and have sealed up for the preview. Hopefully it arrives some time soon.