Thank you for helping make private edge inference reproducible, observable, and safe. Start by choosing the repository that owns the change.
| Change | Owning repository |
|---|---|
| Unified install, status, validation, logs, or orchestration | edge-cli |
| Ubuntu, k3s, NVIDIA runtime, GPU Operator, device plugin, or DCGM substrate | k3s-nvidia-edge |
| Ollama, GGUF, WebUI, OpenTelemetry, Prometheus, Grafana, Helm profiles, or benchmarks | llm-observability-stack |
| Read-only GGUF model runtime contract or evidence schema | gguf-observability |
| Cross-repository validation and sanitized test reports | edge-llm-tests |
| Organization profile or shared community metadata | .github |
- Read the target repository's README,
AGENTS.md, contributing guide, and architecture or language-boundary documentation. - Search existing issues and pull requests for overlapping work.
- Keep the change focused and preserve repository ownership boundaries.
- Use an imperative commit subject and explain why the change is necessary.
- Run every repository-specific test listed in its contributing instructions.
For Go changes, the normal minimum is formatting, unit tests, race tests, vet, and a build. Helm changes require linting and rendering each affected values profile. Runtime validation must distinguish read-only diagnostics from commands that modify a host or cluster.
Describe what changed, why, how it was validated, and any security or operational impact. Include concise sanitized output for failures or live checks. Never attach credentials, Kubernetes Secret data, kubeconfig, private IP addresses, pod logs containing user data, private prompts, model responses, or model weights.
By participating, you agree to the Code of Conduct. Report security concerns using SECURITY.md, not a public issue.