RPI as Alexa and homeassistant making conversation

I got teased in the last month by video about google home and amazon alexa.
and for sure since the possibilities to connect those devices to HA.

so i have been searching a little, and what the hack, I dont need to buy 1 of those, I have a raspberry!
i followed the instruction on this website:

and now i can talk to my raspberry as if it was an amazon Alexa device.
oke that cool on itselve. but i have also appdaemon which talks.
and i tried it out and they can talk to each other.
so my RPI is having a conversation with itself.

appdeamon: Alexa
appdaemon: how are you?
Alexa: i am fine!
appdaemon: Alexa
appdaemon: tell me a joke
Alexa: what do layers wear to court?
Alexa: lawsuits.

i can now have a 24 hour conversation about everything and nothing going on, with just a RPI, an old webcam and a cheap speaker.

2 Likes

i know that i am not able to let Alexa play music, but it seems that i run into more problems.

could be that i need to figure more things out or that it doesnt work with this Alexa.

emulated hue isnt discovered.
i added my google mail calender, but Alexa doesnt see it.
but i am afraid that the biggest problem is language related.
I am in Germany and som things automaticly are in german, but Alexa is in english.
the Alexa app is also english, but the help in german.

seems like a lot off work has to be done before it works like i want :wink:

maybe there are some Alexa experts out there that know some tips? (yeah i know i can make a new topic, but if no response here i will search on :wink: )

@ReneTode - I played around with the RPI release of the Alexa suite a bit when they dropped the button requirement. My use-case was integrating Alexa voice commands into the smart mirror in my main bathroom. I already had a Pi in there running the mirror, already had speakers hooked up for music, why not add a mic?

After spending many hours on it with limited success, I am forced to conclude that Amazon released the open version of the Alexa suite software as a marketing gimmick to entice us to purchase Echo Dots.

The cost of a Dot is roughly the same as a Pi+Power Supply+Mic. I personally believe their feeling was, “Well, sure you can technically get it to work - but look at how much better it runs on our hardware for the same price! Just buy one of these, it’s easier!”

I DID have voice commands working, but as you mentioned - some parts (music especially) are crippled. Recognition is much slower and not as accurate. It’s neat from a technology standpoint, but unless someone else pipes in with a better experience, I’m afraid that the results you got are “working as intended” from Amazon’s perspective.

i read that music was disabled, so that wasnt what i wanted anyway.
voice command is working and reqognition is absolute great. yeah oke i say a german word once in a while and then she doesnt understand me, duh, off course.
off course its a little slower. the RPI cant keep up.

i dont think they see it as gimmick. when i improve things, they learn also. the more people speak to it the better it gets and even more, if i want to make a product with Alexa on it i would pay them.

you cant compare the price from RPI, power supply and mic to a dot.
if you already have those things (like a lot around here) then it is for free, and if you dont and dont want to use the RPI for other things then you would be stupid to go that way.

my problems are that emulated hue isnt recognized and that languages seem to be off in some parts. and allthough it is set in the app, the calender isnt seen.

i have looked up a bit more and it could also be a problem that i havent set up https. maybe that should be my first goal then.

i have no Alexa device to compare it to, but this makes me smile.

Glad you are seeing some success with it! It’s been some time since I tinkered with the software, so maybe it got better.

Just brainstorming a bit on the issues you’re reporting - have you tried using ha-bridge for control through your AlexaPi? It’s a bit more versatile than emulated_hue, although I suspect they might wall off the smart home stuff to Amazon branded devices only.

I was curious, so I looked into this a bit. Some googling led me to this writeup https://www.hackster.io/awshome/awshome-home-automation-using-rpi-alexa-iot-a3d3dc that offers a work-around. with just a few caveats:

  1. This gets pretty deep into AWS/Lambda, so you’ll need to be comfortable there - (just based on what you’ve written here, I think you’d be okay).
  2. It was written in September, so YMMV on whether it still works
  3. The author concedes it isn’t as quick or accurate as an actual Echo, but I suspect that may be a limitation of the chosen hardware

If you get this going, let us know about it - could make for some interesting possibilities!

thanks for the tips.
the first look at ha-bridge page tells me directly that it doesnt work with my setup.
it says that it only works with the actual echo and dot.
so there lies my first problem and i suspected something like that (but didnt hope for it)

but then i look at your second link and see AWS and lambda. and directly i rmembered that i have seen that here also:

your link makes use of 433 mhz directly on the RPI. so that wouldnt bring me anything.
but it makes me think that with AWS and lambda it must be possible to get HA working.

i dont know if i am comfortable with AWS and lambda. until now i have never seen anything from it. :wink:
only time will tell if i am going to be comfortable with it.

off course i will tell here if it works and what i found out :wink:

oke, i am stuck at the start.
to sign up for an AWS account i need a creditcard that i dont have.
i dont want to give payment information if i want to sign up for a free account anyway, but in this case its even impossible.
so i probably wont get past that point att all and Alexa on the pi is what it now is, just a little speaking toy, not able to use for my automations.

I am waiting the day we can have a server program for the pi3 like Alexa.

AT the end of the day to me the interesting skills are maybe 10 maximum. Maybe the big problem is voice recognition