Recommenders are used with increasing frequency in a wide variety of domains ranging from e-commerce to healthcare. However, in some domains, item information can exist at different heterogeneous data sources covering different aspects of the data, which makes it impossible or not preferable to build one single recommendation engine. Moreover, in such systems, different users of the recommender can be interested in different aspects of the items. Motivated from these, we build a framework which can orchestrate different recommendation algorithms based on users feedback in an interactive and online manner. Initial experiments on a real-world application demonstrate the value of our solution.