Users are finding themselves interacting with increasingly complex software systems and expanding information resources. However many of these systems have little to no awareness of the personally-understood user context which expresses why they are being used. In this paper we propose a framework for modelling and proactively retrieving previously accessed and created information objects and resources that are within the context of a user's current situation. We first consider theories of context to understand the discrete aspects of context that may delineate a user's composite situations. With this we develop a framework for modelling user interaction in context along with a re-configurable algorithm for making personal recommendations for desired information objects based upon the environmental, content-based and task sequence contextual similarity of the current situation to past situations. To measure the effectiveness of our approach we use a two week activity log from four real users in a preliminary lab-based evaluation methodology. Initial results suggest the framework as a static personal recommendation algorithm is effective to varying degrees during periods of interaction for users of various characteristics.