Ok, I see two potential approaches I could take.
filter.approximate
andfilter.accuracy
From a cursory look at my history, those cell tower pings are showing accuracy of 0, so I would guess I cannot use this approach as it would filter out my GPS pings which are 1.3-7, etc.
<entry key='filter.accuracy'>0.1</entry>
Is filter.approximate a boolean, and it omits all "cell and wifi locations that are coming from geolocation provider"? So:
<entry key='filter.approximate'>true</entry>
Would this be the correct approach to filter the WiFi/Cell locations? I imagine that's a server wide setting too, not something that could be set per device?
The filter.approximate won't work here. It only works when Cell/WiFi location resolution was done on the server side. So you have to use filter.accuracy.
I see...
Considering that the cell pings are showing "0" I would guess 1-3 would be considered less accurate. Am I understanding that correctly? Wouldn't setting the following filter out all pings, or am I thinking about this backwards?
Sorry to be redundant.
<entry key='filter.accuracy'>0.1</entry>
Cell locations shouldn't have zero accuracy.
Anton, these errant locations continue to plague my records, and more significantly some automations that get triggered when they shouldn't.
To preface this, I have a tiny bit of coding experience, but I am not proficient. A while back I spent some time prototyping a Boolean that would filter location reports with accuracy=0 using a LLM. I know, not ideal, but realistically it's the only way I could approach creating a filter like this.
My question for you is, do you have any caveats or guidance that you're willing to offer for this endeavor?
The objective being that the server filters the reports with 0 accuracy and permits any non zero report. Of course the current filters could be applied as well.
Thanks Anton.
Of course I'm open to other suggestions, but currently I get numerous location reports throughout the day that are in a consistent region that is a handful of miles away from the device. They all have accuracy=0, and whether they are cell based, or other I don't know. They are in any case, problematic and undesired. (If useful for diagnosis, I don't believe they happened with the old app, so maybe something Play Services is doing).
Are you sure you're using the latest server version?
Thanks for the tip. Looks like I'm a few versions back on 6.8.1. Did something change related to this?
I'll update and keep an eye on things.
So I'm on 6.10.0 now and after a few hours I have an example.
The colored lines along the bottom and coming up the left side are correct. The position that I have open (1) is incorrect. It should be over near (2) which was inside a building that has a metal roof. There are a few other nearby and incorrect positions during that same time period while I was in the building.

Ah, that's perfect. I'll check into that. Thank you.