Hey, one of the folks building Ente here.
We have to pick our battles, and we currently aren't looking at serving customers on low-end devices. But if they access Ente on a laptop, the indexing will run there, and the computed indexes will sync to their low-end devices, e2ee.
Also, given that compute on smart phones is getting better and cheaper with every iteration, we believe it's best to bet on Edge ML for the long term.