Kalliope assistant + MM

sispheor

Hello, just a little post to introduce a module that allow you to link Kalliope project with MM.

The module allows you to see on the screen what the assistant say or control your MM by sending notification from Kalliope.

A little demo:
https://youtu.be/QHwctPbJ2ZY!

A couple links:

E3V3A

I really like the look and feel to this, at least from your video. But I think we would need a complete walk-through to install this. In addition, would be good to know the performance for RPi3’s.

Is there any complete installation docs somewhere?

sispheor

The complete installation doc of Kalliope is on the fist link.
You can use the pre compiled image for Rpi.

And about perf, I run both project on the same Pi without any issue.

Give a try. Kalliope wont disappoint you.
And don’t forget to come on the gitter chat if you need direct assistance.

E3V3A

@sispheor Sorry, but I would never use a pre-compiled image of anything. It just doesn’t make sense. Most people already has a bunch of stuff installed that they don’t want to loose.

So from what I understand from your github comment, we need to install at least 3 separate Kalliope things in order to use it. Now, that’s quite a bunch of work, and a lot of separate reading. It would be great to have all steps in one place.

No, I don’t want to sound pessimistic or anything, but I believe in trying to keep basic installation info in one place. I’m so sick of dealing with the lack of documentation and easy to follow instructions for 99% of what you find on github. Those things alone can be total project killers.

KISS principle! (At least to get up and running.)

sispheor

As Kalliope is a complete separate project, yes it has its own installation process. But il you take a look to the manual install doc it’s not a big deal.

I don’t know what you mean by having all info in one place. MM has is installation doc in one place, and modules have their own doc per module.

It’s exactly the same for Kalliope. You need to install the core project and then module (neuron).

Don’t hesitate to tell me if something is missing.

E3V3A

@sispheor

I don’t know what you mean by having all info in one place

What I mean is:

People already has MM installed, so they go to MMM-kalliope in hope to just get it working/installed, but instead just find:

Module to bind Kalliope with your Magic Mirror.
…
Note: On Kalliope, a neuron is available to talk with this module directly.

Huh!? WTF is a neuron!? What is that? Well…ok.

So they go to kalliope_neuron_magic_mirror and find the same screenshot there…but still no installation info to follow. Eh, what does this do, and why is not just part of the MMM? Circular back link to MMM-kalliope, since there is no ref link back to main kalliope.

By this time I’m already annoyed, but try to persist and follow:

Kalliope installation documentation

Then … I have to follow the next link:

Please follow the right link bellow to install requirements depending on your target environment:
Raspbian (Raspberry Pi 2 & 3)

and the next one…and now I’m totally sick of shitty instructions. And whatever was good, I’ve already forgotten.

One page please!?

I think it should be simple enough for you guys to do, especially for simple unified devices like the RPi’s, where everyone is pretty much running the same stuff.

E3V3A

A hot KISS tip for writing any kind of SW installation documentation:

Ask your girlfriend, boyfriend or 9 year old kid, who has not used it before to follow it.

Could they complete it within 1 hour?

No? - Then it’s shit written documentation.

E3V3A

I just want to inform the rest of the forum members, that in case you think I come across as a bit hardass here, I’m in good contact with @sispheor who I think understand now, what the problems are.

What are they?

How to make great project modules more reachable outside the developer expert community!
Everyone want to talk to their computers, Google, Alexa, Siri, Jarvis, AI-this-that-and-what-not!

sispheor

We have added a single line command install for Kalliope.

Tested on a single Raspberry Pi, both project fit perfectly and the whole can be installed in two command line.

Thanks for your feedback !

E3V3A

@sispheor said in Kalliope assistant + MM:

We have added a single line command install for Kalliope.

Is it for the MMM-kalliope?

Fantastic! Sound like music to our ears!

sispheor

No it’s for Kalliope core. MMM-Kalliope need to be installed via only a clone of the module as usual.

And the kalliope neuron can be installed via the kalliope CLI. It’s in the doc of each community neuron.

NoNameRo

@sispheor when trying to install Kalliope on a freshly installed system, I have 2 errors

Searching for SpeechRecognition>=3.7.1
Reading https://pypi.python.org/simple/SpeechRecognition/
No local packages or working download links found for SpeechRecognition>=3.7.1
error: Could not find suitable distribution for Requirement.parse('SpeechRecognition>=3.7.1')

Searching for apscheduler>=3.3.1
Reading https://pypi.python.org/simple/apscheduler/
Downloading https://pypi.python.org/packages/c4/16/8be21944c787ef13cd7581c62d446053d079f36aa54ef20b5b45b378944d/APScheduler-3.5.1.tar.gz#md5=18c2050f24bad3dd82e17dd4b5f182df
Best match: APScheduler 3.5.1
Processing APScheduler-3.5.1.tar.gz
Writing /tmp/easy_install-NmMSMf/APScheduler-3.5.1/setup.cfg
Running APScheduler-3.5.1/setup.py -q bdist_egg --dist-dir /tmp/easy_install-NmMSMf/APScheduler-3.5.1/egg-dist-tmp-px4K5B
  File "build/bdist.linux-armv7l/egg/apscheduler/executors/base_py3.py", line 12
    async def run_coroutine_job(job, jobstore_alias, run_times, logger_name):
            ^
SyntaxError: invalid syntax

creating /usr/local/lib/python2.7/dist-packages/APScheduler-3.5.1-py2.7.egg
Extracting APScheduler-3.5.1-py2.7.egg to /usr/local/lib/python2.7/dist-packages
  File "/usr/local/lib/python2.7/dist-packages/APScheduler-3.5.1-py2.7.egg/apscheduler/executors/base_py3.py", line 12
    async def run_coroutine_job(job, jobstore_alias, run_times, logger_name):
            ^
SyntaxError: invalid syntax

then a little excitement, add the starter configuration to rpi_install_kalliope.sh, and let the user during the installation choose which language he wants to install

sispheor

@NoNameRo said in Kalliope assistant + MM:

SpeechRecognition

Hi,

I had this one time.
Can you try to install the lib manually?

sudo pip install SpeechRecognition

Then run the install again from the kalliope folder (in /home/pi)

sudo python setup.py

It’s weird, sometime pip doesn’t find this lib.

E3V3A

You should not have the pip installation in that script without:

checking if it is already installed
make sure what python version your using. pip2 or pip3?
allow user to cancel that installation, if they know they already have it

Most of the time, if you have both Py2 and Py3 installed, pip would apply to Py2 and pip3 to Py3.

NoNameRo

@sispheor SpeechRecognition was successfully installed, but sudo python setup.py has another error

pi@MagicMirror:~/kalliope $ sudo python setup.py
usage: setup.py [global_opts] cmd1 [cmd1_opts] [cmd2 [cmd2_opts] ...]
   or: setup.py --help [cmd1 cmd2 ...]
   or: setup.py --help-commands
   or: setup.py cmd --help

error: no commands supplied

sispheor

@E3V3A
Actually, we consider here that the user start the installation from a fresh install raspbian.
So he is using Python 2 and pip2.

sispheor

@NoNameRo sorry, forgot one word

sudo python setup.py install

sispheor

@E3V3A but yes you are right, a good script should be idempotent.
I’ll enhance it.

E3V3A

Does MMM-kalliope allow or have any Google Assistant or Alexa functionality?
If, yes, how to install it. I didn’t see anything in the neuron list…

Although Kalliope look very nice and surely better documented now, it will be hard to sell without these features. People wanna talk to AI, not only ask to turn your lights on or off, which you can do by your self. Perhaps a phone plugin so you can call it up and talk?

In addition it is a huge obstacle that you have to make plugins (or neurons or synapses or whatever you call it) for just about everything you want to talk to Kalliope about. This learning should be automated by ML, and integrated to the productivity tools you already have, including future Alexa and GA compatibility.

sispheor

Kalliope is not an IA but a framework to make an assistant. Kalliope is like Ansible actually by design.
From neuron you can cover almost all features provided by current assistant like Google or Alexa.

There is a Android app to control kalliope from your phone and a “Tasker” neuron to control your phone from kalliope.

Users have to make their brain in the language they want by using neuron. We provide “starter kit” to give a basic config and new user to understand the concept.

I Understand that is not what you are looking for. You should give a try to another app.