Amazon Echo to HA Controllers

Just got back from my business trip Stateside (I live in Australia) and unboxed the Echo.
 
The setup app doesn't seem to work very well but the Echo somehow got itself on the home WiFi network and is working. Kids and wife having a great time interacting with it although they do think its a bit of a gimmick. FIrst impressions is the speech recognition is better than the Kinect and the cloud based services are rich enough to answer basic questions. I won't get time until the weekend to hack it to interface into my HA system but so far I like what I see/hear, it is a nicely polished product (eg. the voice is better than the standard MS and a little better than the Ivona voice I used previously). Latency for the recognition and response is good, and the ability to hear the 'alexa' keyword over ambient noise & when it is playing music is very good.
 
In summary it has potential.
 
Quick update since CQC integrates the Echo (natively and third party driver).  I am currently using both, and so far it has to be the best control interface for the WAF that I've ever seen.
 
It has basically become how we interact with our lights.  Spent a damned fortune on Taps (IVB, ping me if you want to dump your Hue), and they just sit around.
 
I'd bet 95% accuracy?  Every now and then I will have to repeat a command, but for the most part shes solid.  Awesome addition to the house.
 
FYI, if you ever think about integrating locks, know it is possible to stand outside a window and shout to Alexa.  I tested it locking my door.  :)
 
In the USA for the next few days toying with idea of grabbing an Alexa.

Short on time at the moment to do a ton of research as to how Alexa can be interfaced to a M1 system. Can anyone briefly give me what might be involved?

Live in Australia and also am wondering what limitations might apply using it downunder. I note with interest some Aussie adopter on the forum of Alexa.....be interested in your feedback.

Regards,

Fleetz
 
Spent a bit of time setting up the development environment for the Echo on the weekend and after the usual futzing around I now have Alexa recognising automation voice commands for my home automation system, and voice response. I use node,js as the end point for the Echo 'skills' on Microsoft Azure cloud which then relays commands back to my home server via websockets.
 
Still proof of concept stage, but if anyone is having trouble with setting up skills and using Azure/Node (instead of AWS) check out this project that got me started: https://www.hackster.io/anthony-ngu/amazon-echo-talking-to-node-js-server?offset=2&ref=search&ref_id=echo
 
[edit]
fleetz, I'm in Australia with an Echo and have interfaced it with non Amazon systems. I have no experience with M1 so can't help with that specifically, although from what I can see from the elk web site it seems this security controller can access the network so must have some API/access control. Elk systems are popular on this forum (mostly in the US market??) so I'm sure others can help you with Elk specifics and I can help you with node code that could access your M1 if it has sockets, REST or some form of network API. I also see Dean from CQC has just released a driver for Echo using a similar approach so if you have CQC this would be the simplest approach.
 
Only limitation for Echo in Australia is that the news feed expects you to be in the US with a US zip code. I believe Echo will be released for other markets at some point so the cloud software will be updated to support international I'm sure (but guessing). Apart from that, works fine, and Aussie accents seems OK, in fact the accuracy of voice recognition is very good - best I have seen for an open air recognition setup.
 
deandob said:
Just got back from my business trip Stateside (I live in Australia) and unboxed the Echo.
 
The setup app doesn't seem to work very well but the Echo somehow got itself on the home WiFi network and is working. Kids and wife having a great time interacting with it although they do think its a bit of a gimmick. FIrst impressions is the speech recognition is better than the Kinect and the cloud based services are rich enough to answer basic questions. I won't get time until the weekend to hack it to interface into my HA system but so far I like what I see/hear, it is a nicely polished product (eg. the voice is better than the standard MS and a little better than the Ivona voice I used previously). Latency for the recognition and response is good, and the ability to hear the 'alexa' keyword over ambient noise & when it is playing music is very good.
 
In summary it has potential.
Did you get the setup app to work? I too live in Australia and am currently State side in San Francisco. The Amazon echo arrive today here and I tried to download the iOS app. Won't let me it sees my Apple account in Australia and inhibits me from downloading it. Appreciate any feedback as I am out of here back to Australia on Friday, wondering if there is any work around or do I send it back for credit?

Cheers,
Fleetz
 
Setup a dummy Apple ID account which gets the Echo app downloaded and installed so it looks like I will be able to set the Echo unit up to my WiFi back home.

Might have to setup a dummy Amazon Prime account and connect via VPN to use the Echo's prime features?

Fleetz
 
Apple account won't help - this is an Amazon product and plays inside the Amazon eco-system. I set it up via web client on windows and an Amazon fire my wife uses, both work. Did it in Australia, no issues. Had to setup my wife's account as a developer account to upload the skill definitions to integrate Alexa into my HA system.
 
No need for Amazon Prime account (unless you want to play full songs on Echo, non prime accounts will do previews of songs) and no need for VPN in Australia. It is clear Amazon will push Echo users to be Prime members to get the most out of the echo (which is fair enough). I'm hoping there will be a spotify plugin one day.
 
Good luck with it!
 
deandob said:
Apple account won't help - this is an Amazon product and plays inside the Amazon eco-system. I set it up via web client on windows and an Amazon fire my wife uses, both work. Did it in Australia, no issues. Had to setup my wife's account as a developer account to upload the skill definitions to integrate Alexa into my HA system.
 
No need for Amazon Prime account (unless you want to play full songs on Echo, non prime accounts will do previews of songs) and no need for VPN in Australia. It is clear Amazon will push Echo users to be Prime members to get the most out of the echo (which is fair enough). I'm hoping there will be a spotify plugin one day.
 
Good luck with it!
Cheers Dean....all makes sense. One of the main reason for purchase is looking at integrating in an M1 system. The rest is a bonus. Been playing with "Alexa" here in San Francisco at my friends place where we are staying. Bloody hilarious some of the responses I have been getting using their Prime account. Obviously Alexa doesn't understand all of our Australian! (-:

I will need to see if I can get a developers account so I can do the M1 integration....hope they have not closed them off.

I have some Philips Hue and WeMo devices too so I will have a bit of fun all in all.

Thanks for you input....appreciated.

Regards,

Ian
 
Alexa has no problems with the Australian accent but struggles with Australian phrases (eg. she just beeps at "where is a good place to hit the turps tonight".... :blink:)
 
(eg. she just beeps at "where is a good place to hit the turps tonight".... :blink:)
 
What does it mean; to hit the turps, Dean?
 
One thing that is gonna have to stop is these Amazon Echo commercials where they demonstrate what it can do including the keyword. Every time that comes on my Echo starts playing music...
 
deandob said:
Apple account won't help - this is an Amazon product and plays inside the Amazon eco-system. I set it up via web client on windows and an Amazon fire my wife uses, both work. Did it in Australia, no issues. Had to setup my wife's account as a developer account to upload the skill definitions to integrate Alexa into my HA system.
 
No need for Amazon Prime account (unless you want to play full songs on Echo, non prime accounts will do previews of songs) and no need for VPN in Australia. It is clear Amazon will push Echo users to be Prime members to get the most out of the echo (which is fair enough). I'm hoping there will be a spotify plugin one day.
 
Good luck with it!
Hi Dean,

Cheers for all that.

Back in Australia. Finally got the Amazon Echo iOS app to download and connected no probs to the Echo.

Have Hue and WeMo devices working fine.

Currently have Amazon Prime trial account and that is working fine however I suspect at the end of the 30 days it may well cut me off if it sees me in Australia?

Have set up a Pandora account but that has limitations, setup artist radio stations and it plays one song from them and then other artists. Oh and ads...

Interested in getting it talking to my Ness M1 system....interested in what you have done to get Echo talking to your HA system? What HA system are you running?

Cheers,

Fleetz
 
Hi Fleetz,
 
I have a home grown system developed over the years with  HTML5 for the front end, .NET for the automation services, node.JS for device plugins and comparable to retail / open source solutions available (you can read about it at the start of this thread: http://cocoontech.com/forums/topic/28256-anyone-interested-in-discussing-building-diy-home-automation-software-and-hardware/)
 
Regarding how to get it connected to your Ness system, I followed the approach outlined here to get it to work (in addition I use websockets to communicate from the cloud back to my server), which is specific to Microsoft Azure and Node.JS.
https://www.hackster.io/anthony-ngu/amazon-echo-talking-to-node-js-server?offset=8&ref=search&ref_id=echo
 
I did a few other things but the link above should get you started. Happy to explain more details if you want.
 
After some fiddling about we seem to have a pretty setup going with the Echo and CQC. The Echo is not really optimal for a completely opened ended system like an automation controller. Ultimately it's best for those situations where the things being said are strictly prescribed and well known. Because it can only work with predefined 'utterances', which are spoken sentences into which replacement tokens can be placed, if you really wanted the most absolutely naturalistic, specific speech to do any given thing, it would be very difficult because you'd have a gigantic number of sample utterances and still wouldn't have enough for all situations. 
 
So we went with a fairly small set of generic utterance patterns that can be used in a lot different ways, even if it might not necessarily be the most natural way to say it. That allows the user to get as much use of out of it as possible without us having to try to foresee every possible situation and hard code it in.
 
So we support these types of patterns (with the actual variable bits in [] brackets, and with Control being the keyword to invoke CQC, though you can change that):
 
Alexa, tell control to run [something to run]
Alexa, tell control to run [something] with [some parameters]
Alexa, tell control to set [something]
Alexa, tell control to set [something] to [some value]
Alexa, tell control to turn [something] off
Alexa, tell control to turn [something] on
Alexa, tell control to turn the [something] off
Alexa, tell control to turn the [something] on
Alexa, tell control to turn [something] down
Alexa, tell control to turn [something] up
Alexa, tell control to turn the [something] down
Alexa, tell control to turn the [something] up
Alexa, tell control to turn the [something] down
Alexa, tell control to turn down the [something]
Alexa, tell control to turn up the [something]
lexa, ask control what is [something]
Alexa, ask control what is the [something]
Alexa, ask control is the [something off, running, etc...]
Aleax, ask control if the [something is on, running, etc...]
Aleax, ask control if [something is true, today is tueseday, etc....]
 
On the CQC side you set up a configuration file that tells CQC how to respond to these utterances, though it's more simplified than the above. All of those effectively come down to run, set, or query operations. So there's a map for each of those and you can indicate a handler for each of the somethings that you want to handle. You can have multiple somethings go to a single handler, or have specific handlers for a single something.
 
They can return text to the Echo to be spoken in response, either for confirmation of run/set type operations, or to give back the requested info for query type ops.
 
 
And of course the fact that they want to support multiple target applications, e.g. CQC, means that you have to add that extra verbiage in there to make sure the request gets vectored to the correct target handler. The 'control' bit in the above examples is the key that tells the Amazon server where to vector the speech once it's been converted to text. Their built in command can be like "Alexa, what is the date". But to do that via a third party handler it has to be something like "Alexa, ask control what is the date". So it's not quite as obvious or naturalistic, but not too difficult to get used to.
 
Back
Top