CamServer is a Python package for data stream processing developed at Paul Scherrer Institute (PSI). It is a
key component of SwissFEL's data acquisition, where it is deployed on a cluster of servers and used for displaying and processing images from all cameras. It scales linearly with the number of servers and is capable of handling multiple high-resolution cameras at 100 Hz, as well as a variety of data types and sources. The processing unit, called a pipeline, runs in a private process that can be either permanent or spawned on demand. Pipelines consume and produce ZMQ streams, but input data can be arbitrary using an adapter layer (e.g., EPICS). A proxy server handles requests and creates pipelines on the cluster's worker nodes
according to rules. Some processing scripts are available out of the box (e.g., calculation of standard beam metrics) but users can upload custom ones. The system is managed via its REST API, using a client library or a GUI application. CamServer's output data streams are consumed by a variety of client types such as data storage, image visualization, monitoring and DAQ applications. This work describes the use of CamServer, the status of the SwissFEL's cluster and the development roadmap with plans for SLS 2.0.