I had an AI running and working on this for almost 20h - just a few interruptions.

All-in-all it was pretty successful, but only because I had stictly forbidden to make any edits from its own "knowledge" and use a docs MCP server which I had previously fed with the Electron documentation for exactly one version (39.2). This allowed me to check occasionally, whether it had actually queried the information from there. Anyway, my indoctrination has apprently been extreme but successful. It affirmed in every 2nd sentence that all information iis solely taken from the docs server.

I had checked back with the original docs and didn't see any mistake. What could have happened that it missed some new members in the classes (due to the methodology via querying MCP for docs), but I would consider the total result as pretty solid.

Finally, I have reviewed every single changed file and made a number of editis (which also led to the additional commits), also reverted some unnecessary changes in favor of compatibility.

I had also give special instructions regarding numeric types (doing an assessment whether integral or fractional when the docs to not state anything and build a list of properties where it's unclea).

The ultimate criteria for me were as to whether it would properly identify and change the type of the property which caused the issue that@markatosi had run into (decimal displayFrequecy but int in our definitions):#944

And it did.

Summary

With the extended matrix tests, 10 new test failures have surfaced:https://github.com/ElectronNET/Electron.NET/actions/runs/19586773630/attempts/1#summary-56097317718

Two of them had turrned out to be a timeout issue on newer Electron versions on Windows (takes 1.5s).
It's fixed in this PR (commit "getPrinters: Increase timeout").

The remaining 8 failures were automatically fixed by the model update in this PR (without having been aiming at that).

Now these are the results:

But I'm going to remove the matrix-test commit because it belongs intothe other PR.