Solid composite propellants (SCPs) are ubiquitous in the field of propulsion. In order to design and control solid SCP rocket motors, it is critical to understand and accurately predict SCP regression. Regression of the burn surface is a complex process resulting from thermo-chemical-mechanical interactions, often exhibitingextreme morphological changes and topological transitions. Diffuse interface methods, such as phase field (PF), are well-suited for modeling processes of this type, and offer some distinct numerical advantages over their sharp-interface counterparts. In this work, we present a phase-field framework for modelingthe regression of SCPs with varying species and geometry. We construct the model from a thermodynamic perspective, leaving the base formulation general. A diffuse-species-interface field is employed as a mechanism for capturing complex burn chemistry in a reduced-order fashion, making it possible to model regressionfrom the solid phase only. The computational implementation, which uses block-structured adaptive mesh refinement and temporal substepping for increased performance, is briefly discussed. The model is then applied to four test cases: (i) pure AP monopropellant, (ii) AP/PBAN sandwich, (iii) AP/HTPB sandwich,and (iv) spherical AP particles packed in HTPB matrix. In all cases, reasonable quantitative agreement is observed, even when the model is applied predictively (i.e., no parameter adjustment), as in the case of (iv). The validation of the proposed PF model demonstrates its efficacy as a numerical design tool for future SCP investigation.