Subgroup discovery is a data mining task that aims at finding patterns describing unusual statistical deviations concerning a specific target variable. This task has many applications, such as searching for risk groups for infectious diseases. One issue in this task is redundancy, which is the extraction of two or more patterns that provide similar knowledge. DINOS is an evolutive algorithm for subgroup discovery that outperforms state of the art solutions of the issue with a novel redundancy reduction method. It is composed of various steps that can be performed in several ways (e.g: population construction, best solution selection, class selection) to adapt to specific situations. Adequate software architecture is essential to maintain the flexibility needed to use this algorithm in real applications and new investigations, which is not the case of its original implementation. This work introduces DAF, an architectural design for the DINOS algorithm oriented to the extensibility of its components. It defines a framework based on its logic, but not restricted to the techniques originally used. The proposal specifies a base structure with extension points that allow DINOS to function regardless of the specific implementation. The solution has high extensibility in terms of adding new alternatives for each component and algorithm configuration capabilities. Finally, it does not introduce undesired changes to DINOS’ original logic and there is no reduction in the subgroups’ quality.