tutorial.md

I've found mycroft's documentation to be less then ideal, so here I'll document how I made the skill-unitconversion skill, which can do things like tell you how many meters a second 88 miles per hour is.

The first step was installing mycroft. I didn't install the version from the archlinux package repo as it has some permissions issues. Instead I followed the guide here. If you do run into issues, try running chown -R :$USER /opt/mycroft/skills /var/log/mycroft to make it so that your user can edit those two locations.

(It would be better if mycroft ran on a per-user basis without involving any system level logging. Perhaps a systemd user service?)

Then you need to download the mycroft-skills repository using git clone https://github.com/MycroftAI/mycroft-skills.git.

Then create a new skill and register it with mycroft.

cd mycroft-skills
cp 00__skill_template skill-name-of-my-skill #Create a new skill from template
ln -s $(pwd)/skill-name-of-my-skill /opt/mycroft/skills #Make your skill findable by mycroft

You can now start working on your skill in the skill-name-of-my-skill folder. In my example the skill is skill-unit-conversions.

Thankfully we don't need to do the actual unit conversions ourselves, we can load in a python library to do it for us. Unfortunatly since we're not really installing the skill we're not able to automatically install our python dependencies.

From your mycroft-core directory (as per the installation guide I linked to) activate the microft virtual env and install python-pint. We'll also install sympy to eliminate floating point errors and work with fractions like 1/3rd.

source ./venv-activate.sh
pip install pint sympy
deactivate

For deployment you'll need to add your dependencies to manifest.yaml, but we're not going to worry about that until we actually have something production ready.

We're going to use the padatious intent parser instead of adapt, it seems to be what people are moving towards anyway.

First, make sure padatious is available available. In skill-yourskill/__init__.py import the intent_file_handler. from mycroft.skills.core import MycroftSkill, intent_handler, intent_file_handler.

Next we need to create some .entity files for our units. Thankfully pint already has a big list of units, and we're just going to put those in some voc files without worrying too much if they're actually pronouncable or what.

import pint
ureg = pint.UnitRegistry()
ureg._units.keys() #Goes in vocab/en_us/unit.entity
ureg._prefixes.keys() #Goes in vocab/en_us/prefix.entity

We only need to run this once since pint units don't change very often. We might even want to manuelly edit those lists to remove un-pronouncable units, but it's a good starting point.

Prefixes are words like kilo or mili and units are words like meter or gram.