I love "smart" devices, but hate "devices that needlessly insist on connecting to the Internet".
One of the worst offenders is Dropcam. They have a super camera, easy to set up and use. Great picture quality. Would be an awesome baby monitor or "closed circuit TV replacement". But why the goddamn hell does it need to connect to the Internet? Why is the only option available to needlessly stream video out of my home network to the cloud, only so that I can then stream it back into my home network for viewing??? WTF? That's both a waste of outbound bandwidth and a waste of inbound bandwidth. I should be able to put it on my network, switch off the cable modem, and still be able to view video locally. How hard is that? I could do that with a webcam and a really long USB cable!
Their business model depends on some percentage of their customers using the subscription service.
My guess is: if they offered the version you describe, they'd need to make it much more expensive. Which many consumers would find odd: the one with fewer features would cost much more. Granted, those consumers wouldn't be looking at the big picture...but I find many consumers don't. Up front costs matter a lot to consumers.
As dumb as it sounds, it is probably easier that way. Sometimes in LANs it is easier to get data out then back in. For example, a lot of dorm networks don't support Chromecast devices because chromecasts tries to multicast on the LAN for discovery, but dorms have networking policies that prevent this.
A webcam that sends the data out to the internet then back would avoid the discovery issue by using an external webserver as a rendezvous point.
I don't think people spend a lot of time thinking about their home networking. You could imagine most people just plug in their home routers and it is a crapshoot whether or not the router will support the necessary functionality, whereas a router will always enable communication to the outside world (or people would return it ASAP).
With that said, this seems like a straightforward technical problem that may have technical solutions.
Ease of setup for regular Jane/Joe because they know shit all about router configuration. That's why devices just transfer everything over someone else computer a.k.a. "teh cloud".
If you don't care about recording video or video recognition features, the cheapo chinese cams on amazon actually perform pretty well. For $80 you can get 720p video with IR lights, speakers, microphone & it can move around. Usually it doesn't zoom like a dropcam can.
If you willing configure a NAS server somewhere, you can even record the video locally.
The video quality probably doesn't compare but I've used an old iPhone with iPCamera (i'm sure Android equivalents exist) installed for this purpose, which simply hosts an mjpeg stream at a local IP address. It should be simple to start or stop recording the stream on any device that's connected.
Alexa probably uses forms of machine learning and also queries lots of services to find the answers you need. Also it learns from every user and gets better for every user this way. That would be really hard to do with an offline device.
One of the worst offenders is Dropcam. They have a super camera, easy to set up and use. Great picture quality. Would be an awesome baby monitor or "closed circuit TV replacement". But why the goddamn hell does it need to connect to the Internet? Why is the only option available to needlessly stream video out of my home network to the cloud, only so that I can then stream it back into my home network for viewing??? WTF? That's both a waste of outbound bandwidth and a waste of inbound bandwidth. I should be able to put it on my network, switch off the cable modem, and still be able to view video locally. How hard is that? I could do that with a webcam and a really long USB cable!