GoKawiilJev-driven pipeline diagnoses 76% of SREGym-Lite faults without an LLM agent
The research team built a diagnosis pipeline that uses Jev, a lightweight decision-making tool, to analyze Kubernetes cluster evidence without any LLM agent involved. The pipeline programmatically collects cluster objects, events, logs, and resource data, then has Jev select likely root causes and supporting evidence to assemble a diagnosis report. Tested across 21 SREGym-Lite faults, the pipeline passed 80 of 105 diagnoses (76.2%) with a median diagnosis time of 14.6 seconds.
GoKawiil's interpretation of the reporting above, not reported fact.
This suggests that structured, programmatic evidence collection paired with a fast decision-making model could replace more expensive LLM-agent-based approaches for certain diagnostic tasks, potentially cutting costs and latency. The approach still has gaps—nearly a quarter of diagnoses failed—indicating it may need further refinement or hybrid use with agents for complex incidents. The open-sourcing of the pipeline on GitHub could let other teams test and extend this method for site reliability engineering workflows.
- A Jev-only pipeline, without an LLM agent, diagnosed 80 of 105 cases (76.2%) across 21 test faults.
- Median diagnosis time was 14.6 seconds, indicating a fast decision process.
- The pipeline collects cluster evidence programmatically and has Jev choose among preset options rather than generate free-form analysis or reports.
Source: sregym.com — Yiming Su, 2026-10-07
Published there as: “Jev-Driven SRE Diagnosis: What Worked and What Failed”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.