How to get low latency automations?

Zwave, Zigbee, matter over thread and WiFi over thread should all be eligible to run locally, although it also depends on the exact routine you are running, as there are some individual things in a routine that can push it back to cloud operation.

@h0ckeysk8er has been keeping a list of things that can push your routine into the cloud, so he may be able to say more. :thinking:

WiFi which is not using matter is trickier. This requires an open API or a server configuration where you use a different device, like a raspberry pi, or a laptop, as a server on your local LAN and then you communicate with that from an edge driver.

The official integration with Shelly devices is cloud to cloud. But there is a community-created integration which uses an edge Driver and is local. You can read more about it and ask about specific Shelly devices and configurations in the following thread:

Edge Shelly drivers for Gen1 and Gen2 Devices

As @nathancu already mentioned, the easiest answer is stick with devices that use zigbee, Z wave, or matter which communicate communicate directly to a smartthings/Aeotec hub (not to their own third-party hub). Those will definitely be eligible to be local, it just depends on the details of each routine you use.

Wi-Fi devices are trickier. It’s possible for some but not for others, you just have to research model by model.

one more less common situation: if you have more than one smartthings hub on your account, and you combine devices from multiple hubs in a single routine, that will run in the cloud rather than locally, even though you could write a different routine where each individual device could run locally. It’s just the way they currently manage multiple hub communication. It’s possible this will change in the future, but this is the way it is as of the date of this posting.