It looks like Kee is behaving as designed so it’s not yet clear what issue we should create on GitHub - Google has changed their sign-in page so that the sign-in page contains several form fields and therefore no Kee entry can match the single form field accurately enough for it to be safe to auto-fill the field. If we change the accuracy thresholds to make them work here, Kee will start auto-filling all kinds of forms that it shouldn’t.
I suspect we can improve things by considering the newer metadata that sites like Google are including in their form fields (e.g. the specific hints it provides to indicate a field is a username or a current password field). That could also break everything but is less likely too than the quick hack of changing the threshold. I wonder if we could create a new form detection algorithm which places full trust in the information provided by each web site? That would be a huge change from the current situation because we historically found that websites are not trustworthy in that regard - e.g. frequently supplying false information to try to trick password managers into failing.
I’d like to think site designers are now less hostile to password managers but it will need very thorough testing and a fair bit more planning before we know what to implement.
This looks like a great example to test with when we can come up with a plan though so I have made a note to refer back to this in future.
In the mean time, there might be some way to trick Kee into thinking the entry is a good enough match - maybe creating some fake form fields for example, but I don’t have time to investigate exact steps right now.