Assistant built for the Google Voice AIY that can understand and respond to user voice queries.

Features:

• Hotword detection
• Speech recognition
• Text-to-speech
• Persistent memory between sessions
• Changing button LED colors
• Ability to change device volume
• Ability to store queries that are not understood
• Ability to give device IP address
• Ability to stream or download and play audio from YouTube

More features, such as ability to play music and record audio on demand, are planned for the future

This program is built for the Google AIY Voice Kit. It may not work on other devices.

Downloading:

Run git clone https://github.com/J0hnL0cke/AIY_Assistant.git or download it.

It shouldn't matter where you download it.

I used ~/AIY-projects-python/src/aiy/recognizer because that's where most of the modules I needed were

Dependencies:

Install all necessary dependencies with apt and pip. Modules that are needed have been included.

If you get an error installing with pip3, then try running the commands using pip instead.

sudo apt-get install python-pyaudio python3-pyaudio sox

sudo apt-get install libatlas-base-dev

sudo apt-get install espeak

sudo -H pip3 install --upgrade youtube-dl

pip3 install pyaudio

pip3 install pyttsx3

pip3 install speechrecognition

pip3 install python-vlc

pip3 install bespon

Configure the project

You will need an account on Houndify to use speech recognition.

Create a project and save the project and client ID keys for later.

Next, head over to the Snowboy Dashboard.

You can train a custom voice model or use a universal model.

You will need to download the .pmdl/.umdl file onto the device.

Now run the build.py file and follow the instructions.

When you are asked to enter a path to install to, you can use a relative path like "./files", or an absolute path, like "/~/AIY-projects-python/src/aiy/recognizer/files"

When you are asked to enter a path to your Snowboy model, you can use a relative path from the main.py file, such as "./model.pmdl", or an abslolute path, like "/~/AIY-projects-python/src/aiy/recognizer/model.pmdl"

First run:

cd into the repository if not already in it

Run python3 main.py

You can enter commands by pressing the button and speaking or by entering text in the shell and pressing the enter key.

Press Ctrl+C to exit and close the program

You should be good to go. If you have an issue, report it here

Start at boot

There is a .service file included in this repo if you want to run the assistant at startup. To enable it, run the following commands in a shell:

Create the symlink sudo ln -s ``pwd``/assistant.service /lib/systemd/system

Reload the service files so the system knows about this new one sudo systemctl daemon-reload

Enable the program to run at boot sudo systemctl enable assistant.service

Now, you can start the program manually or reboot, and the code in service_assistant.py will start. Pressing the button on the AIY will then start the assistant.

Start the program manually sudo service assistant start

Show program logs sudo journalctl -u assistant -f

Unfortunately, there is no way to provide keyboard input to the assistant while it is running as a service, so you can only provide commands using voice recognition.

Contribute

Want to help?

Leave a pull request for features that should be added.

Additional commands that should be added are also welcome.