Apr 28, 2017 — by Eric Brown — 919 views
Google has released a Python-based Google Assistant SDK that’s designed for prototyping voice agent technology on the Raspberry Pi 3.
Google’s developer preview aims to bring Google Assistant voice agent applications to Linux developers. The Google Assistant SDK is initially designed for prototyping voice agent technology on the Raspberry Pi 3 using Python and Raspbian Linux, but it works with most Linux distributions. The SDK lets developers add voice control, natural language understanding, and Google AI services to a variety of devices.
The AI-infused Google Assistant voice agent emerged first on the Linux-based Google Home speaker, and spread to various Android platforms to offer AI voice interactions. Competitors include Amazon’s Alexa, Microsoft’s Cortana, and Apple’s Siri.
The SDK incorporates the Python version of the gRPC API, an open source remote procedure call (RPC) client. Since gPRC allows you to generate bindings to other platforms, developers could also extend the SDK to port to Windows and Mac platforms, as well as other programming languages including C/C++, Node.js, Java, and Ruby.
Raspberry Pi 3
The Google Assistant SDK is free for prototyping use, but requires that you make a deal with Google for commercial deployment. Later this year, “commercial device makers will also be able to use the SDK across a wide range of hardware,” said Google.
The SDK release appears to be linked in part to a survey that Google posted for Raspberry Pi developers in January asking them what kind of Google tools they would find most useful. In the survey announcement, Google said that it planned to bring its AI and machine learning technology to the Raspberry Pi in 2017.
Deeplocal’s RPi-based, Google Assistant infused Mocktail Mixer
(click images to enlarge)
The SDK includes samples and documentation, including instructions on how to make a voice-enhanced Mocktail Mixer developed by Pittsburgh-based Deeplocal. Built around a Raspberry Pi and an Arduino board, the open source bartender device lets you order cocktails by voice command, and dispenses jokes with the drinks.
The Google Assistant SDK lets you add similar Q&A chichat to other devices ranging from a robot to a smart mirror, says Google. Developers can call upon “Actions on Google” to extend the SDK with new commands. For example, an action could turn on an LED light or set a robot into motion. The voice agent can be summoned with the OK Google hotword or a physical red button made available by Google.
Other SDK features include the ability to read transcribed text from user voice input, which can be used to provide additional device functionality. For example, you could scan for happy or sad keywords and change the mood lighting accordingly. Google is currently working on enhancing the SDK with hotword support and companion app integration, among other features, and has established a community development site.
There has already been at least one open source project designed to run Google Assistant on a Raspberry Pi using IFTTT. There are many more RPi projects using Alexa. Amazon has a head start over Google, Microsoft, and Apple when it opened its Alexa SDK to device developers. Over the last year, there has been a growing number of smart home and other mostly Linux-based devices that have integrated Alexa voice agents.
Google Assistant SDK overview
The Google Assistant SDK developer preview is available for download, initially optimized for the Raspberry Pi. More information may be found in Google’s announcement announcement and on the Google Assistant SDK web page.