I have had some issues with Ente Photos for several months on a Pixel 7, always up to date with latest GrapheneOS and Ente release (installed via Google Play Store).
Their machine learning is done by indexing the photos on device, then uploading those vectors e2ee to their servers. However, I've had 2 issues with this.
One, the indexing has been extremely slow. About 10-100 per hour, instead of the expected 18,000 or so. I have tried giving Unrestricted background battery, leaving the app open in the foreground, disabling hardened memory allocator, all with little or no effect.
Two, per a discussion I had with Ente developer on their forum, they had trouble getting their ML framework (ONNX) to work on GrapheneOS and instead had to use a different one (GGML). No further details were given on why.
Has anyone had this issue of slow indexing? Or any insight into why that particular ML framework wouldn't work on Graphene?
The last issue I've been having is that notifications for when something is added to a shared album happen intermittently and very infrequently. Experimented, again no impact from unrestricted background battery being allowed. This is installed via Google Play Store.
For what it's worth, I still recommend Ente despite these issues. It's fairly feature complete compared to Google Photos, plus is e2ee, offers easy export sync, and lets people upload to a shared album without an account or app.