If I would be doing this then

I would probably start with simple smooth planar surfaces from a low friction material.

If it would not be enough I would probably add an air pump which would create an air cushion between the surfaces (one of the surfaces would have a grid of small holes and air would be pumped between the surfaces). Check whether it is acceptable despite possible noise, air flow disturbing hands.

If this would not be good enough then I may play with magnets whether some suitable arrangement can be created. Maybe an simulation to search through all the possible magnetic field arrangements.

The cutout in my drawing is cylindrical, and supposed to hold split spherical bearing seats...
This way the ball wont ride on the base material, and we wont need super precise spherical cutouts.
From what I remember, these split seats also allow adjustment of the friction as they are set in a thread.
But it was a long ago I used these the last time, so I could be wrong here. Wouldn't that allow for smooth motion?

Probably not. There will be friction between the seat and the ball. I doubt this is better than just two simple planar surfaces from some low friction material (e.g. teflon). No need to complicate it with balls when you are not using them to eliminate friction on the load bearing surfaces.
My guess is that the balls only make the forces more concentrated to a smaller area leading to more wear.  Better to keep pressure (forces per unit area) smaller overall by two simple planar surfaces from a low friction material.

lmao