Speaker
Description
Operational logbooks are essential for documenting accelerator performance, interventions, and operator experience, but their content is often inconsistent, unstructured, and difficult to search. This limits both human retrieval and the use of AI systems that rely on high-quality historical data. Project ARIEL (Agentic Retrieval Interface for Electronic Logbooks) introduces a modular, facility-agnostic framework that standardizes how logbook information is ingested, enriched, and searched across accelerator laboratories. Each participating site hosts its own ARIEL database while adopting a shared schema, data-enhancement modules, and interoperable search components. Enhancement modules provide semantic metadata, text and figure embeddings, and optional machine-state snapshots at ingestion time. On the retrieval side, ARIEL supports keyword, embedding-based, multimodal, and machine-state search, forming a unified foundation for an agentic retrieval layer capable of orchestrating multiple search strategies. This contribution presents the architecture, schema design, and early cross-facility prototypes, and describes how ARIEL fits into the broader DOE Genesis AI mission to establish shared, interoperable AI infrastructure for accelerator facilities.
| In which format do you inted to submit your paper? | LaTeX |
|---|