Documentation
If you are completely new, the smoothest path is:
Core mental model used across all pages:
- operator decides cluster-level node states,
- agent discovers node hardware and capability,
- agent enforces node-level controls,
- scheduler consumes node supply labels + workload constraints.
Section guide
- Getting Started
- concepts, install, runtime modes, workload compatibility
- Architecture
- operator/agent roles, CRDs, policy model, telemetry/control interfaces
- Hardware
- CPU and GPU support model, heterogeneity strategy, runtime caveats
- Simulator
- digital-twin behavior, algorithms, runtime flow
- Experiments
- benchmark design and measured outcomes
What to expect
- Clear control-loop model: operator decides, agent enforces.
- Kubernetes-native contracts: CRDs + scheduling constraints as intent/supply language.
- Practical path to adoption: quickstart first, then progressive deep dives.