Sci-tech novelty retrieval service for the National Science and Technology Major Project (NSTMP) Water Project is discussed. First, its project aims and technique features are described. Then, major implement measurements and quality control strategy of sci-tech novelty retrieval are proposed, which include foreword control and active styles, 4 research topics-oriented classifications (e.g., technology, demonstration, policy and standard groups), 3 difficulty scales and corresponding measurements. Finally, we analyze the major information service characteristics of NSTMP in a whole life cycle. In particular, based on our present service models at National Science Library, CAS, future embedded cloud service concepts for the NSTMP are proposed and discussed.