Well… within max 10 secs (due to the refresh)
I am satisfied with this interval… it works fine.
I haven’t tried to lower it because I don’t want to spam the Unifi Controller with requests. Maybe someone else has trimmed that parameter and can share experiences
It also depends on how fast the Unifi Controller software refreshes its own info… I don’t know the answer to this 
Just tested it out. It took 21 seconds after I arrived at the door. Phone normally has WiFi before I arrive at the door. So, it is not usable for disabling alarm.
It really depends on how fast the Unifi Controller software will update it’s own database.
The reporting to the Binding can only happen after this first step has been completed.
It may take 5-6 secs for Unifi controller to mark you as present and then another 10 secs for the binding to pull this updated info.
You can try to lower the refresh= on the bridge down to 1 sec
If I wrote it how I should of wrote it, the binding should keep the HTTP connection open so theoretically this shouldn’t cause any sort of bottleneck.
My dev env is all whack right now as I just had to reinstall my OS due to a SSD failure 
Just to be clear: I am not blaming the Binding. My hopes were not that high anyway, that this could work for my use-case because timing would be really critical.
another “workaround” would be to increase a bit the entry delay on your alarm system (let the PIRs trigger but not fire the alarm within a 35 secs window till disarming)
by the way: I don’t know if it’s such a great idea to automate this aspect…
think about: theft of mobile phone device prior to the burglary attempt…
I also have my alarm panel integrated in OH2 but I don’t auto-disarm. I do it manually.
I do auto-arm based on several criteria (one of which is presence)
Actually, I would like to use your suggestion and I would not even call it “workaround”. I was using that with my old alarm system.
But my new alarm system (Homematic IP) currently does not offer delayed alarms. It only offers delayed arming and delayed firing of door open events.
But the latter is not usable in this case, because it would not fire the alarm if the burglar quickly enough closes the door.
So, I will have to buy a remote control and always take it with me or use the phone app to disarm.
But this is really off-topic now, I guess.
This x 1000.
Arming automatically with no code required = OK
Disarming without some sort of user input (passcode required) = BAD IDEA
Out of curiosity, if you are using this binding for presence detection, why is this better than the network binding ?
Thomas asked the same question …
Here was my answer…
I can only confirm that this binding works absolutely great as presence detection, I’ve tried the network/ping method before but imo there is no comparison.
Unifi doesn’t impact battery but also no glitches no missed pings or any of thos, it works flawless.
I fully agree and I use a mix of different Android and iPhone devices hopping over four UAP-AC-PRO. I’m more than impressed, how reliable it is. No false detection!
Regarding presence of wired devices - couldn’t this be done through the snmp binding and checking ifOperStatus of the port you’re interested in? Or are we talking about clients connected to cascaded switches?
Of course it can be done with snmp
But I believe that Matthew is planning to implement this feature in the binding as well
I haven’t tested if it works now as is.
From what I remember, the Unifi Controller reports wired clients as well
This should be fixed in the next ESH stable release (probably build 318).
This can’t be done if it is a remote network
why not?
if there is no direct connection to the remote network I do not see how the snmp binding will work… How do I get from the controller subnet to the remote subnet if there is no routing?
remote network does not mean no routing maybe you wanted to write “isolated network” or something similar
of course you need routing in place for the snmp to work
if using cascaded switches (only Layer 2 links) you could face a problem
I guess its down to my English 
I have a controller at home and about 15 different sites scattered around Europe. I would call them remote, because they are not really isolated. Anyway there is not direct routing from controller subnet to the “isolated” subnets.