There's nothing really systematic, but there is a fair bit of literature already on what you might call discrete noncommutative Riemannian geometry à la Connes, dealing with noncommutative spectral triples that are discrete in some sense.
On the one hand, there's been a great deal of work on so called "finite" or "discrete" spectral triples, spectral triples corresponding to noncommutative spaces with metric dimension $0$ but possibly non-zero $K$-theoretic dimension ("$KO$-dimension"); these have been used as $0$-dimensional "internal spaces" in noncommutative-geometric particle physics models. The seminal papers on finite spectral triples and their classification are by T. Krajewski and by M. Paschke and A. Sitarz, and the most up-to-date survey on their standard applications within noncommutative-geometric mathematical physics, including a great deal of introductory material, is by K. van den Dungen and W. van Suijlekom. For a particularly interesting recent development, take a look at M. Marcolli and W. van Suijlekom on connections with spin networks and lattice gauge fields.
On the other hand, there's been a great deal of work on constructing interesting spectral triples for graphs and fractals, though I must confess to knowing little about this line of work. A literature trawl brings up potentially interesting papers by M. Requardt, E. Christensen, C. Ivan, and M. Lapidus, and J. W. de Jong, for instance. There's also a fair bit of work on semi-finite spectral triples for graph $C^\ast$-algebras, by A. Rennie and collaborators, but this I know even less about.