Hi there Stuart
There is indeed room in this world for both mechanisms.
Alright - there are a number of solutions out there, easy to use and without messing with HTTPS which will let you do:
"Alexa, lounge lights off"
to which the response will be "OK" and for many that will be fine and I am certainly not knocking the effort that people have put in to make this work.
Remember X-10 - the original home control protocol (or one of them). You could turn a light on and unless you were there, just hope that it worked. Of course in that situation if you were in the room you KNEW whether or not it had worked and so that would be fine.
Let's take a slightly different example.
"Alexa turn the bedroom heating to 19 degrees".
Trust me when I tell you that when it says "OK" it could quite easily have turned the bathroom heating to 90 degrees.
And that is because of an annoying feature of Alexa in that it accepts the words you speak OR any similar sounding word in the dictionary.
So - would it not be better if Alexa responded "Ok, I've set the bedroom heating to 19 degree".
Indeed with an early background in computer adventures I'm just about finished with a version that can remember the object in question and do something like:
"Alexa tell computer to turn the bathroom heating to 19 degrees and the kitchen light on"
"Alex tell computer to turn it off"
(Clearly referring to the last used object - remind anyone of original Adventure?)
That is what our version does - and if it does not understand it says "Sorry, computer thought you said 'ballroom eating 90 degrees'". You can then very easily pop into Node-Red and add ballroom as an alternative to bedroom if you wish - or start speaking more clearly.
When you are playing with this on your own it's fine - you can adjust your voice accordingly but when you hand the reins over to your married partner - in my case my wife who's accent is very different to mine (I'm UK she's US) the LAST thing you want is "you spent WHAT on this thing?"
And that is the focus of the direction that we went in with this. By being able not only to parse the sentences coming out but return any arbitrary text and variable information back. The only thing we're missing is the ability to constantly interact - it's a case of one sentence in - one sentence back - and as soon as someone improves on that while staying in Node-Red, our code will go in the bin :-)
Regards
Pete.