Speaker
Dr
Frank Mayet
(MPY1 (MPY Fachgruppe 1))
Description
In this talk a work-in-progress implementation of an accelerator operations assistant is presented. The assistant is based on the open Mixtral:8x7b-instruct LLM and can tap into the high-level control system toolkit 'doocs_generic_experiment', written for and used at the dedicated R&D accelerator ARES at DESY. Furthermore, it has access to the electronic logbook, as well as machine-specific documentation. In order to achieve this, advanced prompt engineering techniques, such as chain of thought (CoT), ReAct prompting, and retrieval augmented generation (RAG) are leveraged.
Primary author
Dr
Frank Mayet
(MPY1 (MPY Fachgruppe 1))
