Sensory repository down

[flos06]

IMPORTANT: Before you create an issue, please take a look at our Issue Reporting Guide.

Briefly summarize your issue:

I am getting this error when trying to install the Alexa SDK on RPI 4. fatal: remote error: Repository not found.

What is the expected behavior?

I'd expect it to find the repository and install

What behavior are you observing?

It is not finding the repository. Which is not surprising seeing as https://github.com/Sensory/alexa-rpi gives a 404 not found

Provide the steps to reproduce the issue, if applicable:

Install sensory

Tell us about your environment:

What version of the AVS Device SDK are you using?

  <x.y.z>

Tell us what hardware you're using:

• [ ] Desktop / Laptop
• [x ] Raspberry Pi
• [ ] Other - tell us more:

Tell us about your OS (Type & version):

• [ ] Linux
• [ ] MacOS
• [ ] Raspbian Stretch
• [ ] Raspbian Jessy
• [ ] Other - tell us more:

[Ddvos]

I am having the same issue. Is there already any sollution?

[flos06]

Not that I know of unfortunately

[Ddvos]

Maybe it could help if there was another repo. We can update de link to https://github.com/Sensory/alexa-rpi in the setup.sh file

[flos06]

I've tried to find one but all the forks are outdated.

[Ddvos]

I think we have to wait on an update from out Amazon....

[0zuzu]

Fuck

[gauthiersornet]

What about these github links : https://github.com/jcakchen/alexa-rpi https://github.com/gonzalo123/alexa_rpi

In the file setup.sh, change the line : TEST_MODEL_DOWNLOAD="https://github.com/Sensory/alexa-rpi/blob/master/models/spot-alexa-rpi-31000.snsr" by this one : TEST_MODEL_DOWNLOAD="https://github.com/jcakchen/alexa_rpi/models/spot-alexa-rpi-31000.snsr"

In the file pi.sh, change the line : git clone git://github.com/Sensory/alexa-rpi.git By this one : git clone git://github.com/gonzalo123/alexa-rpi.git then remove the line : bash ./alexa-rpi/bin/license.sh

[flos06]

Wouldn't you need the license in order for it to work though?

[gauthiersornet]

I'm testing... But the compilation seems to be stuck... I let it and i'll see...

[flos06]

I've tried. Spits out an error at the end that the library isn't licensed

[Ddvos]

I also tried but with no luck. Hope they release an update soon for the Sensory alexa-rpi missing repo

[gauthiersornet]

It is still stuck :/

[flos06]

well unless someone shares their latest license-key.txt it wont matter anyways :(

[sammy4703]

I emailed Sensory about this - here's response:- Hi xxx

Sensory has made a business decision to remove this project from GitHub. We apologize for any inconvenience. If you would like to explore a license of Sensory technology, contact sales@sensory.com.

[flos06]

Well that sucks.

[moskyow]

There is another way to have Alexa on Raspy ?

[flos06]

You can use KITTAI (snowboy) wakeword engine instead of sensory, however you would need the alexa sdk 1.24 as 1.25 removed support for it. Should still be doable though

[moskyow]

You can use KITTAI (snowboy) wakeword engine instead of sensory, however you would need the alexa sdk 1.24 as 1.25 removed support for it. Should still be doable though

Project closed on Dec. 31st, 2020

[Ddvos]

Maybe this one? https://github.com/alexa-pi/AlexaPi

I haven't tried because it looks outdated.

I want to use Alexa for playing Spotify because as far as I know, that isn't possible with Google Assistant. Maybe you guys can use Google Assistant?

[Ddvos]

Here is a temporary solution:

https://github.com/alexa/avs-device-sdk/issues/1980

But as far as I know, you can only use the tab to speak

[flos06]

I've build it using KITT ai and honestly its working fine. FYI Ddvos you can't use spotify on the alexa sdk either :( You have to be whitelisted and they don't seem to do that anymore unless you are a commercial user.

[GitHubUser4234]

I've built it with Kitt.AI as well, but it's triggered way more often by false positives (listening although wake word hasn't been said). That totally sucks.

Amazon, what do you suggest, after all, the steps in the setup guide became invalid too.

Would be great if someone could make the necessary changes to support the superior and free Porcupine Wake Word Engine in AVS.

[flos06]

I've built it with Kitt.AI as well, but it's triggered way more often by false positives (listening although wake word hasn't been said). That totally sucks.

Amazon, what do you suggest, after all, the steps in the setup guide became invalid too.

Would be great if someone could make the necessary changes to support the superior and free Porcupine Wake Word Engine in AVS.

Yes this is really annoying. It even triggers for me without a mic plugged in. -_-

[GitHubUser4234]

Amazon has now removed the Sensory step from the setup guide :disappointed_relieved:

[shivasiddharth]

I have created a hack for triggering Alexa using Picovoice porcupine: https://github.com/shivasiddharth/avs-device-sdk-pi If someone can come up with something more neat, it would be great. If we can have a socket server running along with the UI Manager, then the job is done.

[shivasiddharth]

Here is how it works https://www.youtube.com/watch?v=uYtlX260JOc&t=466s

[fivenote]

I have created a hack for triggering Alexa using Picovoice porcupine: https://github.com/shivasiddharth/avs-device-sdk-pi If someone can come up with something more neat, it would be great. If we can have a socket server running along with the UI Manager, then the job is done.

Excellent workaround! Thanks @shivasiddharth!

I made a few tweaks for my setup. I wanted it to find the SampleApp's window id each time it starts, and keep it minimized when not detecting the watchword. So...

This bash script starts my alexa script (a modified version of the SampleApp start script). It saves the window id to /tmp, then minimizes it. It then starts my version of shivasiddharth's alexa_picovoice_trigger.py that I called wakeAlexa.py

#!/bin/bash

#start alexa and give unique pilexa name for window search
lxterminal -t pilexa -e alexa &

#save alexa window id and minimize
sleep 3
xdotool search --desktop 0 --onlyvisible --name pilexa > /tmp/pilexa
xdotool search --desktop 0 --onlyvisible --name pilexa windowminimize %@
xdotool search --desktop 0 --onlyvisible --name wakeAlexa windowminimize %@

#start porcupine wakeword listener
python3 $HOME/bin/wakeAlexa.py --access_key <my access key> --keywords alexa grapefruit

#cleanup
pkill SampleApp

exit 0

In the wakeAlexa.py script, I modified the command that takes action when the wake word is detected. It unminimizes the window, sends the 't' command, and re-minimizes.

subprocess.call("wmctrl -ia $(cat /tmp/pilexa) && xdotool search --desktop 0 --onlyvisible --name pilexa windowfocus %@ keydown --window %@ t keyup --window %@ t keydown --window %@ Return keyup --window %@ Return windowminimize %@",shell=True)

This is all definitely a workaround. I usually run the SampleApp as a service as I test my apps with it. This workaround puts it all out on the desktop. It would be great if Amazon integrated porcupine into the SampleApp.

Hope this helps us all for now. Thanks again @shivasiddharth!

[shivasiddharth]

@fivenote Brilliant. I attempted to create a socket for communication. I am not well versed in C. So not sure as to why it's not working. Here is the link to the file https://raw.githubusercontent.com/shivasiddharth/avs-device-sdk/master/SampleApp/src/UserInputManager.cpp if someone can make it work, it would be brilliant. We can use pushbuttons, whatever we want.

[GitHubUser4234]

Hi shivasiddharth and fivenote,

Really great to see that there is a workaround for porcupine! But still there is this huge issue of #1993 which makes usage of Alexa very limited. Routines and skills don't work without certification anymore like they did before.

I've currently settled for Android on RPI...no AVS customizations anymore, but at least routines do work...

[j4ckp0t85]

hi guys. i have successfully tried the porcupine method. my problem is that i have a very basic hardware (an orange pi pc) and i mainly don't use windowed terminal (sshsessions only).

i have enabled a vnc server to try it but it's laggy (the wake word detector is good but the tap to talk is not always hitting).

no way without using real desktop terminals?

[kenarsa]

It is possible to use Porcupine as a wake word engine. Picovoice has a free tier that allows using its SDK on up to 3 devices for anyone. That being said, someone (preferably someone on the AVS side) need to do the integration (wrapping Porcupine C class in C++ class that AVS accepts). Disclaimer, I work for Picovoice.