Speaker
Description
This project explores methods for automating the creation of device servers in the TANGO Controls environment using Large Language Models (LLMs). The primary goal is to streamline and accelerate the coding process for device servers, reducing the time and effort required by developers. The application features a web-based user interface where users specify device attributes, commands, and properties, while the backend generates functional server code using LLMs such as GPT, Claude, or Gemini.
A key aspect of this project is using technical documentation as a contextual knowledge source for the LLMs. By employing the Retrieval-Augmented Generation (RAG) technique, the application automatically selects relevant documentation fragments to enhance code accuracy, particularly for implementing device-specific features.
This poster presents the system architecture, underlying technologies, and real-world testing outcomes, illustrating the potential of LLMs to improve device server development in distributed control environments.
The project concept was initially presented at the last TANGO Community Meeting during the “lightning talks” and is part of a master’s thesis conducted at Jagiellonian University in Poland, in cooperation with the commercial company S2Innovation, which specializes in TANGO Controls software solutions.
