The realm of music composition with artificial intelligence stands as a pertinent and evolving field, attracting increasing interest and exploration in contemporary research and practice. This paper presents a constraint-programming based approach to generating four-voice diatonic chord progressions according to established rules of tonal harmony. It uses the strength of constraint programming as a formal logic to rigorously model musical rules and to offer complete control over the set of rules that are enforced. This allows composers to iteratively interact with the model, adding and removing constraints, allowing them to shape the solutions according to their preferences. We define a constraint model of basic tonal harmony, called Diatony. We show that our implementation using the Gecode solver finds optimal solutions in reasonable time and we show how it can be used by a composer to aid in their composition process.