Module gpu_metrics 

GPU metrics collection for observability.

Collects per-GPU utilization, memory, temperature, and power metrics using vendor-specific interfaces:

• NVIDIA: nvidia-smi CLI (avoids hard NVML dependency)
• AMD: sysfs under /sys/class/drm/card{N}/device/
• Intel: sysfs under /sys/class/drm/card{N}/device/
• Apple: IOKit via powermetrics (macOS only)

Structs

GpuHealthReport

Per-GPU health check result

GpuUtilizationReport

Per-GPU utilization snapshot

Enums

GpuHealthStatus

GPU health status

Functions

check_gpu_health

Collect GPU health reports.

collect_gpu_metrics

Collect GPU utilization metrics for all detected GPUs.