Daffodil is a digital library system targeted at strategic support of advanced users during the information search process. It provides user-customizable "stratagems" for exploring and managing digital library objects with meta data annotations over a federation of heterogeneous digital libraries. Bingo! is a focused crawler that learns how to gather thematically relevant documents from the Web and Deep-Web sources. This paper presents a coupling architecture for Daffodil and Bingo that allows advanced users to explore digital libraries and Web sources in a comprehensive and coherent way. Starting from a user's interest profile in Daffodil, Bingo! is instructed to find thematically similar documents on the Web, leading to high-quality recommendations that reach beyond the information that can be directly found in digital libraries. Our experimental studies demonstrate that this coupling does indeed lead to a powerful tool suite that improves the precision and recall compared to conventional library and Web search mechanisms