Hi!
I’m interested in sharing info on this as well. Had an issue last year and started digging in. Been building some things related since April '24.
Currently have some dexcom cloud data worked out, some fall sensing worked out, and automations as a response. (Around 6-7 Aquara FP2’s in various modes.) An idea on getting Apple Health metrics, not sure where to locate some dashboards, maybe azure and power bi, maybe home assistant. Can send them via json, where the best place to ingest the data is a conundrum. Basically enrich the health picture from multiple sources to cut down on false positives or other people freaking out due to one metric alarming without context. (My 70 year old parents have called in wellness checks twice cause my phone died and dexcom data stopped, nothing like being woken up on a Sat morning by a police officer in your room!)
Also have tts to sonos, and some progress on responses to atom m5’s to an llm. (whole different can of worms when we get into distributed mics that can also be used when walking room to room durring meetings, but very interesting building alexa and google home replacements) Planning to add object detection via cameras to create an alert but maintain privacy. My idea has been to aggregate the dexcom and apple health info (glucose is a 3 hour delay to apple health, need to use dexcom cloud for current), notice “things” and less obtrusively send alerts. Gently signal lights, drop volume on content and have a tts message, the set it back, stream content from an apple tv through obs studio out to tv’s via ndi and swap the scene to the same content but with a transparent graph from graphana showing glucose in the lower right of the sceen as a chyron as needed.
End goal example is detect a fall, check metrics like glucose, if it meets the houston we have a problem level attempt communication via tts to speakers looking for a voice response, if none, start alerting an emergency contact list via txt, email, phone. Also provide health dashboards easily accessible to family, give them the ability to double check if they are worried, maybe trigger a voice message over speakers if say phone is dead. All without violating privacy, ie ya can’t view a camera stream.
Lastly want to miniaturize it on a 3 node kubernetes cluster of low cost hw, like a PI or equivelent, build a design pattern/blueprint others can use to get a jumpstart.
Lots of neat use cases the concepts can expand to, swap a central video stream from one room to another as presence detects the user went from living room to bedroom. Family video intercom based on presence and room. etc.
So…lets compare notes, brainstorm. There’s alot there, and its all kinda fun. If you’ve found some communities that apply would love to hear it.