install-zymtrace-profiler

What problem does it solve?

This Skill installs the zymtrace profiler agent so your cluster can collect continuous CPU and (optionally) CUDA GPU profiling data and forward it to the existing zymtrace backend gateway.

Core Features & Use Cases

• Helm DaemonSet / Kubernetes deployment: Installs the profiler across all nodes (GPU-capable scheduling optional) and connects it to your backend gateway endpoint.
• Multi-platform install options: Supports Kubernetes without Helm (manifest), Docker single-host, and bare-metal binary installs with systemd.
• CPU-only vs CUDA GPU profiling: Enforces the CUDA 12.x+ requirement for GPU profiling and configures GPU metrics and framework metrics for vLLM, SGLang, and Dynamo-Triton where available.
• Air-gapped / private registry support: Mirrors and configures the profiler image for restricted environments.
• Verification and guardrails: Runs a post-install verification workflow to confirm license validity and backend connectivity, and routes GPU/workload setup needs clearly.

Quick Start

Tell the AI to install the profiler with GPU support on your EKS cluster and connect it to your backend gateway, then follow the verification prompts until the agent confirms backend streaming.