Llm-Agents
Llm-Agents
- slurm-rca-bench
partial
An incident-diagnosis benchmark for the Slurm control plane: 10 scenarios (6 runnable, 2 causal chains measured on a Docker test cluster), closed-vocabulary scoring, and degenerate baselines. No agent scored yet.
- cluster-sre-agent
partial
An LLM agent for Slurm incident diagnosis, built as five ablatable configurations. The dependency graph and the read-only tool guard are built; the agent configurations are not, and no accuracy has been measured.