It's possible to create new folders in Finder.app, but
creating files are not supported. I've created a
Keyboard Maestro macro
that lets you spawn new files at will:
If you're not familiar with
Keyboard Maestro
it's a tool similar to
Automator, that allows you to create macros that can be triggered in a number
of ways e.g. when a hot key is pressed. In this case I've mapped
the "New File…" macro to ⌘⌥N.
TLDR: Learn a bit of Python and you can create a
custom skill with
Flask-ask in a breeze.
They yell down on us from their high horses:
Designers shouldn’t code. Designers should pound the drum and developers should dance to the
beat. Well, I've been dancing a secret proverbial tango with
Python lately.
This is my first step outside of the Python Hello World sandbox. I
recently came upon an
Amazon Echo
and I was curious to see if I could create my own skill. As much as I
enjoy receiving flash briefings from NPR, I wanted to do more. I
wanted Alexa to keep track of my daily run:
John Wheeler’s
Flask-Ask framework
makes it easy to create Alexa Skills. If you’re interested in creating
your own skills I would highly recommend that you pony up the $15 and
buy his
Alexa tutorial videos. It’s
an absolute bargain.
Creating the VUI flow
VUI is an acronym I learned last week, it stands for Verbal User
Interface. Here's the VUI flowchart for the Run Log skill:
This is a simplistic VUI, but it will suffice for our proof-of-concept
purposes.
Setting up the development environment
Follow John's tutorial to get up and running with Flask-Ask:
My
IntentSchema
and
SampleUtterances
are very simple for this proof-of-concept skill. I use Amazon's
own
AMAZON.DURATION
slot type that makes it possible for the user to utter a workout
duration and easy for us to parse that duration:
Let me a reveal a secret here: Alexa is not a personal
assistant. Alexa is a voice terminal robot and we have to stand on the
edge of the scene and feed her lines to create the illusion of
dialogue. These lines live in the
Jinja template:
welcome: "How long did your workout take?"
welcome_reprompt: "I did not catch that. How many minutes did your workout take?"
run_logged: "OK, I have logged {{ minutes }} minutes, {{ seconds }} seconds."
personal_best: "{{ minutes }} minutes, {{ seconds }} seconds. That is a new personal best. Congratulations tiger!"
personal_worst: "{{ minutes }} minutes, {{ seconds }} seconds. That is a new personal worst. Please up your game."
This is how the responses map to our VUI flowchart:
Adding some audio flavor
Plain text strings can get you a long way, but for more advanced
speech features
SSML
is required. SSML is a markup language that is tailored to the
creation of synthetic speech, it provides some impressive features
like phonemic pronunciation. Very handy if you want to mention
Arsenal's goal-keeper
Wojciech Szczęsny
in your skill.
I used the audio tag in my skill, to play a sad trombone
when you log a personal worst and an upbeat sax riff if you log a
personal best:
personal_best: <speak><audio src="https://s3-us-west-1.amazonaws.com/run-log/best.mp3" />{{ minutes }} minutes, {{ seconds }} seconds. That is a new personal best. Congratulations tiger!</speak>
personal_worst: <speak><audio src="https://s3-us-west-1.amazonaws.com/run-log/worst.mp3" />{{ minutes }} minutes, {{ seconds }} seconds. That is a new personal worst. Please up your game.</speak>
This proved more difficult than expected, Alexa doesn't take any
old mp3. I had to encode the mp3 in
ffmpeg using these parameters:
ffmpeg -i input.mp3 -b:a 48k -ar 16000 output.mp3
Skill logic
The logic lives in the
alexa-runlog.py
script. The
launch
decorator determines what Alexa should do when the skill is launched:
I often find that I need to add annotation icons to a mockup, to
explain what is going on.
I like Skitch's annotation style. The annotations stand out from
the mockup - you're not in doubt what is the annotation and what
is the actual mockup.
My need for additional color tints grew and so did the length of the
variable names. Say hi to @color-blue-very-very-light!
There was a another problem: if I wanted to change the hue of
@color-blue I would have to change each and every tint.
An unsustainable situation. Today I found out that LESS (and Sass for
that matter) has a lighten and a
darken function. That means that I can have a master
color (e.g. @color-blue) and use lighten and
darken to create the tints:
It's hypnotic, to watch the guitarist move her hands over the
fretboard. It's been a struggle for her, to achieve this
man/machine symbiosis, but it was worth it.
Vim is a text editor, but it's not exactly MS Word. To say that
Vim has a steep learning curve would be an understatement. Learning
Vim is an investment, but it's an investment that pays fine
dividents. Watching a Vim master type is like watching the guitarist
pick chords.
As UX Designers we're all about ease-of-use, but who's going
to invent the complex tools for tomorrows guitar heroes?
Who's going to invent the next guitar, the next Vim?
I'm not arguing that complexity is a goal in itself, but neither
is limiting complexity where it is needed.
In the beginning it's all play. The designers of games are trying
things out - having fun, creating. They invent throw-ins, castling and
chequered flags. Slowly the games solidify, status quo creeps in,
it's business as usual. That's why we can't have a special
rule for diving in the penalty area, where you have to wear a
ballerina outfit for the rest of the match.
The traditionalists are against ballerina outfits in football games.
They want things to be like the last time we played - it was a 1-1
draw and John fell asleep in front of the TV. In another world 1500
years ago, someone woke up, had breakfast and invented chess. For two
weeks chess was soft, like clay - nobody cared about memorizing
opening variants. A five year old chess prodigy is reading up on the
Sicilian right now, but she would have more fun playing Fischer Random
Chess, where the pieces on the first row are shuffled. Pokemon
figurine to B4 - checkmate!
