Minion makes it easy to create a personal assistant system, likely controlled by voice that can gather data via sensors, interpret them via commands and react accordingly using actuators. Minion can be always listening depending on your configuration.

It is built to be distributed so the sensors and actuators can be spread over several locations, and to be simple enough to run on single-board computers such as the Raspberry Pi or the Banana Pi

Configuration is handled via a JSON file, which makes it easy to read without requiring much programming knowledge. Minion also has a simple web based dashboard - codename "Big Boss" - to help you manage the various Minion configurations. It is fully functional at this stage, but not fully tested.

  $ git clone https://github.com/matiboy/minion.git

Install basic requirements using the command

  $ pip install -r minimum_requirements.txt

This installs only what is required for Minion and the Big Boss to work. As you install components, the Big Boss will attempt to pip install any extra Python requirement. Please refer to the README.md file in each component for further information.

There are two ways to start a config:

Either copy a config from one of the examples in the "examples" folder.

Or you could simply fire up the Big Boss and go through the 4 sections: Nervous System, Sensors, Actuators and Commands.

This is the very first release of Minion. It is not exactly according to plan, it's clunky, but it goes just above the crossbar so it will do.

Without going into the awesomeness of home automation and voice control, why create another framework? There are great tools out there - a few listed here. None served my purpose, here are a few things that make Minion different:

1. Distributed: other frameworks do not seem to want to communicate between devices. Minion can be installed on a device in the kitchen, another Minion instance in your room, and leave the Minion brain somewhere in the living room. Thanks to the Nervous System, sensors can communicate to the brain, the brain can send commands to actuators, regardless of their location;

2. Modular: it has to be very easy to create new modules and to integrate them into Minion. We hope the architecture selected allows for that. Coming soon is a Yeoman generator to allow contributors to quickly scaffold a sensor, command or actuator;

3. Configurable: while basic configuration is available in all the other tools, more complex combinations seem to be limited. Minion allows for complex configuration so you can have for example 2 microphone listeners with different sensitivity, one for activation commands via Pocketsphinx and one for more complex ones via Api.ai . You can easily have one room with always on listening but also a mobile app on which you need to press to issue commands;

4. Agnostic (mostly): Other tools tend to decide which device and library you need to use. One of the most advanced tool available, Jasper, is specifically designed for the Raspberry Pi. While we are huge fans of the Pi, Minion aims to work on any OS - or rather on several, distributed OSes - using your preferred tools. If you like SoX more than FFmpeg to record, it needs to be possible. Minion is written in Python and its modular architecture means that it is OS and library agnostic.

5. Big Boss: Megalomaniac web based management of Minion instances

Minion requires 4 components to function fully:

• one or more sensors: ears (microphone), eyes (rss reader) or other (GPS location, temperature reader, http API, anything else). Sensors can be enhanced with post processors to e.g. transform audio to a text command etc.
• a brain which understands a list of commands sent by the sensors' signals
• actuators to react when the brain says so: mouth (espeak/say or other TTS), hands (GPIO, shell commands) etc
• and very importantly, a nervous system to pass messages along

Post processors are meant to enhance and modify the output of a sensor before it is passed on through the nervous system. The exact same effect can be achieved using commands, but post processors allow for local processing to occur before going over the wire. Read more about post processors.

Below are a list of post processors. More details can be found in the respective README.md files in the components folders.

Pocketsphinx is one of the choices of speech to text post processors. Its main advantage is that the speech to text process is handled locally on the same device. This means that it is very fast compared to the Google API or API.ai processors.

You may want to use this tool to create your own language model and dictionary.

Tests can be run using Nose. Install Nose in your virtual environment and from the root of Minion's folder, run

  $ nosetests

Feel free to submit issues and enhancement requests.

In general, we follow the "fork-and-pull" Git workflow.

1. Fork the repo on GitHub
2. Commit changes to a branch in your fork
3. Submit pull request

Please make sure to rebase your branch onto the latest before submitting