Speaker
Description
Developing Tango Device Servers still requires a significant amount of repetitive engineering work: defining attributes, commands and properties is relatively structured, but implementing device-specific protocol logic from manuals, datasheets and vendor documentation remains time-consuming and error-prone. This contribution presents S2Innovation’s AI-assisted Tango Device Server generator, a web-based tool that combines structured device modelling with retrieval-augmented generation over device documentation.
The application allows users to define a device interface through a graphical form, upload device manuals or datasheets, and generate Python/PyTango server code tailored to the selected attributes, commands and properties. The backend uses FastAPI and LangChain to connect with multiple LLM providers, while the retrieval layer extracts relevant documentation fragments to improve code accuracy. Initial tests with devices such as power supplies and teslameters showed that generated code often requires only limited manual corrections before becoming functional.
The goal is not to replace Tango developers, but to reduce boilerplate, shorten integration cycles and make device-server development more accessible to beamline engineers, controls engineers and scientific software teams.
| Tags | AI, device server, web |
|---|
