The paper describes a global framework enabling competency-based search of learning resources making a heavy use of semantic technologies. First it presents the generic components. Then it exemplifies how this framework can be adapted to a given environment, i.e. which knowledge representations, Web Services descriptions and other components have to be tailored or added to the framework. Then it shows step by step how a query is processed, Finally, users' feedback, lessons learnt and future trends are provided as well as comparisons with other approaches already published about annotating and retrieving learning objects in Learning Objects Repositories.