Using the constraint-based modeling approach, we have developed a diagnostic component, which is able to identify errors made by learners of a logic programming language when implementing a given task specification. It uses patterns to hypothesize the intention of a learner and programming techniques to model conditions on the semantic well-formedness of the program code. These conditions are expressed by means of constraints, which are evaluated on the student solution. Guiding feedback can be derived from constraint violations and is presented to the student with different degrees of informativity. The component has been integrated into a web-based tutoring system and tested on a number of exercises by the participants of an introductory course in logic programming.