Currently, a significant barrier to building predictive models of cell-based self-assembly processes is that molecular models cannot capture minutes-long cellular dynamics that couple distinct components with active processes, while reaction-diffusion models lack sufficient detail for capturing assembly structures. Here we introduce the Non-Equilibrium Reaction-Diffusion Selfassembly Simulator (NERDSS), which addresses this gap by integrating a structure-resolved reaction-diffusion algorithm with rule-based model construction. By representing proteins as rigid, multi-site molecules that adopt well-defined orientations upon binding, NERDSS simulates formation of large reversible structures with sites that can be acted on by reaction rules. We show how NERDSS allows for directly comparing and optimizing models of multi-component assembly against time-dependent experimental data. Applying NERDSS to assembly steps in clathrin-mediated endocytosis, we capture how the formation of clathrin caged structures can be driven by modulating the strength of clathrin-clathrin interactions, by adding cooperativity, or by localizing clathrin to the membrane. NERDSS further predicts how clathrin lattice disassembly can be driven by enzymes that irreversibly change lipid populations on the membrane. By modeling viral lattice assembly and recapitulating oscillations in protein expression levels for a circadian clock model, we illustrate the wide usability and adaptability of NERDSS. NERDSS simulates user-defined assembly models that were previously inaccessible to existing software tools, with broad applications to predicting self-assembly in vivo and designing high-yield assemblies in vitro.