LUME-EPICS is a dedicated Python API and related server that provides an EPICS interface for machine learning and physics-based simulation models using variable representations defined in the LUME-Model package, which enforces minimal data requirements for serving model variables. The server is configurable for use with either Channel Access, pvAccess, or a synchronized dual server, accepting inputs and outputs in the form of scalars, images (e.g. areaDetector for Channel Access and NTNDarrays with attributes for pvAccess), and arrays. Features include templated variable and display generation, Bokeh client widgets, and minimal model abstraction. To illustrate LUME-EPIC's utility, we will showcase a neural network representation of the LCLS injector and a live Bmad simulation model using live process variable values from the LCLS.
|Talk duration||10 minutes + questions|