Interactive visualization of high-performance computations is important area in supercomputing. Interactivity assumes that visualization of results of computation is generated during computation process. However there is a problem: due to overwhelming size of data to visualize, a visualization program should be itself parallel and executed on supercomputer. Beside that, such program should allow to be changed dynamically, because visualization pipeline may change due to user steering of interactive visualization. Current mainstream frameworks for interaction with supercomputer programs assume usage of external parallel programming methods. In current paper, an original parallel programming model is suggested that have built-in capabilities for online interactive visualization. At basic level, it is based on messages and reactions. At higher level it uses promises for inter-operation of computation and visualization parts.