Alert monitoring of psychological health for web text is to evaluate and forecast the unhealthy psychological status of web users by the technology of web text mining. Based on the idea of “Service Distributed and Management Centralized”, this paper proposes a cloud architecture for psychological health analysis and describes the service component representation, workflow customization and workflow execution. With the excellent semantic representation ability of resource description frame (RDF), a service component metadata representation method is presented. Through querying the metadata of service component, end-users could choose the service components and customize the logic workflow on demand. In the workflow execution, the workflow case database is constructed. Through analyzing the similarity between history case and user workflow, a series of workflow reuse strategies are proposed. To validate the performance of the cloud architecture for psychological health analysis, a set of experiments are prepared to test the viability, the results demonstrate that the cloud architecture has higher precision, reuse and scalability for psychological health analysis of web text.