Skip to content

2.1.779

Choose a tag to compare

View all tags
@arthur-louisa arthur-louisa released this 07 Aug 04:29
· 268 commits to dev since this release
2.1.779
4e1834b
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

🚀 Arthur Engine Release

August 7, 2026

This release brings GPU-aware autoscaling and throughput improvements to GenAI Engine deployments, alongside clearer SSL configuration for platform connections.

GPU Deployment & Scaling

Autoscaling

  • Introduced a new standalone GPU-metric autoscaling add-on chart for GenAI Engine on EKS Auto Mode + Karpenter, driving pod autoscaling based on GPU utilization and memory usage via DCGM metrics through a bundled dcgm-exporter → Prometheus → prometheus-adapter → HPA pipeline (#1992)
  • Added a backward-compatible flag in the core genai-engine chart to allow external HPA management without conflicts during Helm upgrades (#1992)

With this add-on, deployments can now scale pod replicas dynamically as GPU load rises and falls, with Karpenter independently provisioning and consolidating GPU nodes to match demand—removing the previous fixed-replica limitation.

CloudFormation Defaults

  • Aligned the GenaiEngineServerWorkerCount default to 3 workers across all GPU CloudFormation templates, matching recent Helm chart guidance (#2108)

Benchmarks on g4dn.2xlarge instances show 3 workers deliver roughly 44% more throughput (29.8 req/s vs 20.7 req/s) with optimal GPU utilization and well within container memory constraints—improving efficiency with no additional configuration required.

Configuration & Platform

Environment Variables

  • Renamed the KEYCLOAK_SSL_VERIFY environment variable to more accurately reflect its role in SSL verification for platform connections, with full backward compatibility for existing configurations (#1763)

Existing deployments continue to work unchanged while adopting the clearer naming convention is encouraged.

Release Management

Version Increments

  • Incremented the arthur-engine version and merged accumulated dev branch changes into main (#2107)
  • Incremented the arthur-engine version to 2.1.779, bundling the GPU worker alignment, SSL configuration, and autoscaling add-on changes (#2112)

These internal version increments coordinate the production build and deployment pipeline and carry no direct user impact.

Release notes generated by Louisa

Assets 2
Loading