Speaker
Description
Osprey is a production-ready agentic AI framework designed for safety-critical control systems at large scientific facilities. It provides a plan-first approach to task execution, semantic channel finding, connector-based access to existing control stacks, and isolated environments for running generated python code. The framework offers a transparent and auditable way to translate natural-language operator requests into control-system actions. A central focus of Osprey is usability and rapid onboarding: self-contained tutorials, reproducible deployment patterns, and a facility-agnostic interface allow laboratories to adopt agentic workflows with minimal customization. Over the past year, the framework has matured into a stable platform with a growing capability library and consistent safety mechanisms. It has been successfully deployed at several U.S. Department of Energy accelerator facilities within the Genesis AI initiative, and the same architecture is being applied in non-DOE environments such as the ITER fusion experiment and the Extremely Large Telescope. This contribution summarizes the current status of the framework, recent feature developments, and lessons learned from multi-facility adoption.
| In which format do you inted to submit your paper? | LaTeX |
|---|