The interest in reasoning over stream data is growing as quickly as the amount of data generated. Our intention is to change the way stream data is analyzed. This is an important problem because we constantly have new sensors collecting information, new events from electronic devices and/or from customers and we want to reason about this information. For example, information about traffic jams and customer orders could be used to define a deliverer route. When there is a new order or a new traffic jam, we usually restart from scratch in order to recompute the route. However, if we have several deliveries and we analyze the information from thousands of sensors, we would like to reduce the computation requirements, e.g. reusing results from the previous computation. Nowadays, most of the applications that analyze stream data are specialized for specific problems (using complex algorithms and heuristics) and combine a computation language with a query language. As a result, when the problems become more complex (in e.g. reasoning requirements), in order to modify the application complex and error prone coding is required. We propose a framework based on a high-level language rooted in logic and constraints that will be able to provide customized services to different problems. The framework will discard wrong solutions in early stages and will reuse previous results that are still consistent with the current data set. The use of a constraint logic programming language will make it easier to translate the problem requirements into the code and will minimize the amount of re-engineering needed to comply with the requirements when they change.