FRAGATA: Semantic Retrieval of HPC Support Tickets via Hybrid RAG over 20 Years of Request Tracker History

A team from the Galician Supercomputing Center (CESGA) has developed FRAGATA, a novel semantic search system designed to unlock two decades of institutional knowledge trapped in a legacy ticketing system. For over 20 years, CESGA's technical support team used Request Tracker (RT), whose basic search engine struggled with typos, different languages, and varied wording, making past solutions hard to find. FRAGATA addresses this by implementing a hybrid RAG (retrieval-augmented generation) architecture that combines modern information retrieval techniques to understand the intent behind a query, not just keyword matches.

The system is fully deployed on CESGA's own infrastructure, featuring a clever design that supports incremental updates without service downtime. To handle the computational load of processing over 20 years of ticket history, the most expensive stages of the retrieval pipeline are offloaded to the center's FinisTerrae III supercomputer. Preliminary results indicate a "substantial qualitative improvement" over the native RT search, meaning support engineers can now efficiently locate relevant past incidents and solutions, dramatically accelerating problem resolution and preserving critical operational knowledge that was previously difficult to access.