AlgoPython is a platform designed to teach Python programming. Currently, it only provides feedback on syntax errors and task validation. This article studies the elements necessary for implementing a system that provides adaptive epistemic feedback. After presenting the feedback decision problem, we analyze a reinforcement learning algorithm to address it. Next, we define two fundamental elements: the elements to be considered for decision (elements relating to the learner and their output) and feedback modeling and production. Finally, we propose an initial proof of concept to evaluate feedback production and reinforcement learning algorithm selection for feedback decision.