Changelog¶
See CHANGELOG.md for the full release history.
v0.1.0 (2026-05-19)¶
Initial release.
- KEDA External Scaler gRPC server implementing
externalscaler.ExternalScalerServer - Direct NVML GPU metrics via
go-nvmlC-bindings - 6 GPU metrics: utilization, memory utilization, memory used (MiB and %), temperature, power draw
- Pre-built profiles:
vllm-inference,triton-inference,training,batch - Multi-GPU aggregation:
max,min,avg,sum - Scale-to-zero via activation thresholds
- Per-GPU index targeting
- Mock collector for testing without hardware
- DaemonSet manifests and Helm chart
- CI: build, test, lint, Helm lint, Docker push