Make Your Own Virtual Assistant with Mycroft

Virtual assistants such as Alexa, Google Assistant, and Cortana are all the rage. Using devices like Amazon Echo or Google Home gadgets, it’s possible to automate many tasks and have a smart voice assistant. However, using a Raspberry Pi and Mycroft, it’s easy to cobble together an artificial intelligence (AI) voice assistant. Here’s how to make your own virtual assistant with Mycroft!

What is Mycroft?

Mycroft is a platform that’s capable of controlling Internet of Things (IoT) devices and acting as a virtual assistant. Similar to skills for Alexa, Mycroft includes a bevy of skills, from telling the time to controlling media center software. In addition to the Raspberry Pi, Mycroft is available for Android and Linux.

Getting Started

To begin installing Mycroft and creating a virtual assistant, you’ll need several items.

• Raspberry Pi 4 (older models aren’t powerful enough to run Mycroft)

• microSD card (8 Gb or larger)

• Keyboard/mouse

• Speaker (analog)

• USB microphone

• Picroft

For installing Mycroft on a Raspberry Pi 3, the easiest method is using Picroft. It’s an operating system built with Raspbian Jessie Lite as its foundation. As the name suggests, Picroft is specifically engineered as a means to install Mycroft. This project is available as an image which is ready to easily be burned to a microSD card.

However, since there’s Mycroft for Linux, you may run this off of a Raspberry Pi using your favorite Ubuntu/Debian, Arch, or Fedora-based distribution (distro) instead. Alternately, simply run the Docker version of Mycroft on a Raspberry Pi.

Make Your Own Virtual Assistant with Mycroft

With your materials gathered, get started making a Mycroft personal assistant!

Download Picroft

If you’re using Picroft, head over to the GitHub page. From here, download the Picroft image. It comes as a Zip file. As such, you’ll need to extract the image file before you can mount it. The official Raspberry Pi website recommends:

• Windows: 7-Zip

• macOS: The Unarchiver

• Linux: Unzip

Next, you’ll need to mount the Picroft image to your microSD card. There are several software options. Etcher is an excellent cross-platform burner that features installers for Windows, macOS, and Linux. However, a few other choices are:

• Windows: Win32DiskImager

• macOS: ApplePi-Baker

• Linux: LiveUSB Install

Locate the extracted image file in your preferred image mounting software. Typically, that’s your downloads directory or wherever you selected to save your downloaded Picroft image and extracted image.

Using your preferred image writing software, select the extracted image file. Then, choose the microSD card where you’ll mount Picroft. Burn Picroft to your bootable media.

Alternate Install: Mycroft for Linux

Although Picroft is the simplest means of spinning up Mycroft on most Linux distros based on Debian, Fedora, or Arch. Therefore, you may choose your favorite Raspberry Pi-compatible Linux operating system (OS). Raspbian should work perfectly, especially considering Picroft is Raspbian with Mycroft installed and configured. 

cd ~/
git clone https://github.com/MycroftAI/mycroft-core.git
cd mycroft-core
bash dev_setup.sh

Then, to start Mycroft on Linux, simply run:

start-mycroft.sh

As the Mycroft documentation reveals, here’s how to use the Mycroft startup service:

usage: start-mycroft.sh [command] [params]

Services:
 all                      runs core services: bus, audio, skills, voice
 debug                    runs core services, then starts the CLI

Services:
 audio                    the audio playback service
 bus                      the messagebus service
 skills                   the skill service
 voice                    voice capture service
 wifi                     wifi setup service
 enclosure                mark_1 enclosure service

Tools:
 cli                      the Command Line Interface
 unittest                 run mycroft-core unit tests

Utils:
 skill_container  container for running a single skill
 audiotest                attempt simple audio validation
 audioaccuracytest        more complex audio validation
 sdkdoc                   generate sdk documentation

Examples:
 start-mycroft.sh all
 start-mycroft.sh cli
 start-mycroft.sh unittest

You may start every Mycroft service simultaneously with:

$ ./start-mycroft.sh all
Starting all mycroft-core services
Initializing...
Starting background service bus
Starting background service skills
Starting background service audio
Starting background service voice

After installing Mycroft for Linux, you’ll need to pair it with your Mycroft account. To do so, say aloud:

“Hey Mycroft, pair my device.”

Mycroft will reply:

“I am connected to the Internet and need to be paired. Your 6-gigit registration code is XXXXXX.”

Then, enter that registration code in your Mycroft AI account to pair Mycroft with your Linux machine.

However, there is a means to use Mycroft without Mycroft Home. This requires adding your own API keys to configuration files. For instance:

[WeatherSkill] api_key= “YOUR API KEY HERE”

Mycroft Skills

Similar to devices like the Google Home or Amazon Alexa, Mycroft offers skills. These vary from telling the date and time to controlling media center software, or telling jokes. Onboard, Mycroft includes a smattering of skills. As with most AI smart assistants, you’ll need to become acquainted with what queries you can make and how to phrase them. The Mycroft documentation offers loads of examples for each of its specific skills.

• Alarm: “set an alarm for [time, or date and time]” (sets alarm, checks alarm status)

• Audio Record: “record” (play and record audio)

• Configuration: “configuration update” (updates Mycroft configuration)

• Date Time: “what time is it” (tell the date and time)

• Desktop Launcher: “open [application]” (open apps on your desktop such as Firefox)

• DuckDuckGo: “what is Frankenstein” (search DuckDuckGo for basic questions)

• Hello World: “how are you” (Hello World)

• IP: “what is your IP address” (tells IP address of host device)

• Joke: “tell me a joke” (tells a joke)

• Installer: “install [skill],” “uninstall [skill]” (installs or uninstalls skills)

• Mark-1 Demo: Gives a demo of the Mark 1

• Media: “play,” “pause,” “next track” (controls multimedia)

• Naptime: “go to sleep” (puts Mycroft to sleep)

• NPR News: “news” “stop news” (listen to NPR news)

• Pairing: “pair my device” (pair Mycroft with home.mycroft.ai)

• Personal: “what are you” (use this to learn about Mycroft)

• Playback Control: “play” (allows control of audio subsystem)

• Reminder: “remind me to [whatever you want to be reminded of]” (sets a reminder)

• Speak: “say [phrase]” (repeats a phrase or word)

• Singing: “sing a song” (sings songs)

• Stock: “stock price of [desired stock, i.e. Google]” (provides stock prices)

• Stop: “stop” (stops running skills)

• Version: “check version” (tells Mycroft core version)

• Volume: “increase/reduce/mute volume” (controls volume)

• Weather: “what is the weather” (gives the current weather as well as forecast)

• Wiki: “tell me about [topic]” (searches Wikipedia)

• Platform Patch: “Platform patch” (patches official platforms)

In addition to its pre-bundled core skills, Mycroft allows for community developed skills. To add a new skill, merely use git clone on a repo, which will add that repo to your skills folder. Where this lies will depend on the device. On a Raspberry Pi, your skills folder is in the home directory at:

/opt/mycroft/skills

Therefore, to add a new skill, first change directory into your skills folder:

cd /opt/mycroft/skills

Then, git clone the git repo:

git clone https://github.com/kfezer/daily_meditation.git

On Picroft, installation is easier. Merely use the Mycroft Skills Manager (msm) installer:

msm install https://github.com/kfezer/daily_meditation.git

Make Your Own Virtual Assistant with Mycroft: Final Thoughts

Creating a virtual assistant with Mycroft is super simple. While it may not boast the same connectivity and integration as an Alexa or Google Home, a Raspberry Pi running Mycroft offers tons of functionality in what’s an inexpensive, intuitive project. Using Picroft makes this incredibly feasible to spin up a smart assistant on the Raspberry Pi.

Your turn: What smart assistants do you recommend?