Docs — GPU (Tier-3) · Energy Governance

Entropy-aware scheduling and NVML-audited energy accounting for multi-GPU/MIG clusters. Initial endpoints will expose per-request power/energy audit and entropy-gated rate control.

Planned Endpoints

• POST /govern/audit → log per-request {elapsed_ms, avg_W, energy_Wh}
• POST /govern/rate → apply Φ-based throttling or burst control

Containerized side-car and Kubernetes/Slurm plugins will be documented here upon release.