I’m still trying to navigate the vast breadth of stuff in the google ecosystem. I’m learning and doing pretty good.
I was wondering though, if anyone knew if google cloud or some other ecosystem might have easy access to some kind of API for IoT devices? Is google’s IoT ecosystem too separate from cloud and AI, or is there a way we as developers could utilize both Gemini’s APIs with some of google’s IoT APIs?
I figured this might be a good question to ask and reference. I know google’s ecosystem is huge, so there’s a lot of easy wiggle room to build and combine cool stuff together with Gemini’s API :).
If i had access to Gemini nano, I would absolutely love these . Still though, these are useful docs to have around.
I meant more like, some kind of API endpoint for Google Home. Something where you can send and retrieve data to turn the lights off, or something to that effect.
Google Home for developers is at https://developers.home.google.com/
The services are hosted on cloud. Personally, I would have concerns about a LLM having the capability to turn on/off devices in my home, at least until the probability of hallucinations becomes acceptably tiny. And, as the saying goes, your mileage may vary, so more adventurous souls are welcome to try that out
I currently have +200 sensors and devices around my house that are tied together on single server and connected to AI.
Google home is good place to start, but IoT is a jungle, not all devices are supported, and there’s multiple different protocol’s out there.
Hallucinations are not really a problem, except in case’s where the AI decides to interact with devices that doesn’t exist, only to be confused by a “200” (ok) response from the API because the call was received, in these cases it will just believe that it did what you asked for, but in reality it hasn’t done anything.
Much of this is solved by having an extremely verbose naming convention, as in lamp on my nightstand being called light.bedroom_table_level_light_color_on_off
Fun facts though; I got my start programming things other than video game stuff years back by tinkering with IoT devices. My first project was a home automation project using OpenHAB (which I still recommend more than anything else tbh). So in a weird way I’m actually more comfortable handling IoT stuff in some respects than handling APIs. Obviously I’ve since gained more experience handling APIs, but IoT is where I got my start and caught the developer bug.
Good thing I’ve definitely got the adventurous spirit!
Believe it or not, I made a voice AI control before LLMs even came out. Language models simplify the process so much, it feels like my prior work was made with creaky cogwheels.
Anyways, I think this might be handy to have around. A major problem back when I developed my own home automation system was how scattered everything was, and how much plumbing one had to do to make things function properly. Voice commands were notoriously a pain in the ass from this.
Thankfully, things are a bit different now, and there is a lot of benefit to having everything you need right there in front of you. Plus, as more devs start popping in, I can guarantee you a lot of developers are going to want to implement some of the things Google Assistant can do, but Gemini cannot (yet). It’s the perfect thing to build on with Gemini in my opinion, and I think more people should play around with it and be a lil’ adventurous!
. Still though, these are useful docs to have around.
.