As librarians, we recognize that users seldom approach a research project with a clear search strategy and thus we develop means to help them clarify their research objectives. However, our traditional methods do not translate well to the web environment, while at the same time demand for web access increases. Addressing this challenge, the Oregon State University Libraries and the Department of Computer Science are collaborating on the development of two test beds to investigate interfaces that will lead users to web resources that are valuable and useful by using collaborative filtering recommendation technology. Collaborative filtering (CF) allows an information portal to continually learn what resources are likely to be useful to which users (and which questions) by observing search patterns as well as explicit recommedations from users. The first test bed, the Tsunami Digital Library (TDL), is a portal to website with relevant and authoritative information on tsunamis with an interface aimed at both researchers and the general public. The second test bed is a search and recommentation portal interface to the OSU Libraries' web resources. In this paper, we explain CF and its potential relevancy to libraries, describe the TDL project, and discuss future plans.