New
Your scheduler. Your notebooks. No YAML.
How it works
From a YAML file to a running job.
Four steps, all visible on the job itself. No separate control plane.
Describe the work.
Submit a GryviaAIJob with kubectl, the CLI, the dashboard or the API. GPUs, image, priority and an optional queue are plain fields on a custom resource.
Admit it, or hold it.
An opt-in gate checks quotas and hard budgets first. Gang admission and queues come from Kueue, or from the alpha native queue gate: all pods fit together, or the job waits.
Place it where it fits.
The operator filters and scores nodes on GPU type, RDMA, NVLink, free GPUs and memory, and records the choice and the reason in the job status.
Run it. Watch it.
It becomes an Indexed Job or StatefulSet with the distributed environment set. Metrics, checkpoint guards and network telemetry (experimental) show how it is doing.
What you get
One platform. Six operators.
Everything is a CRD with a controller behind it. Turn on what you need, leave the rest off.
Native queue scheduler · alpha
Share a cluster fairly.
Queue guarantees and ceilings, borrowing, all-or-nothing gangs and priority preemption, driven by one ConfigMap. Opt-in, unit-tested, and not combinable with Kueue.
Two jobs fit inside the cluster. Both are admitted, each counted against its queue.
Illustration of the documented behaviour. It does not reorder waiting jobs and is not yet verified on GPUs. Read the guide
Install
Up in minutes.
Start on a laptop with no GPUs, or install on the cluster you already have.
git clone https://github.com/zyvorai/zyvor-gryvia && cd zyvor-gryvia ./scripts/kind-demo.sh kubectl -n gryvia-system port-forward svc/gryvia-ui 8443:443
Creates a kind cluster with fictional demo data so the dashboard is populated. Open https://localhost:8443 and accept the self-signed certificate.
Apple Silicon companion
Your Macs, scheduled.
A separate control plane for Mac fleets: native Swift worker and app, unified-memory admission, owner-aware idle harvesting and signed provenance. It lives in its own repository and does not use the Kubernetes path.
- Native workers
- Swift agent and SwiftUI app for macOS 15+ on Apple Silicon.
- Memory-aware
- Jobs are admitted against each Mac's unified-memory budget.
- Owner first
- Work yields when the Mac's owner returns.
Where it stands
Every claim, with its test.
Gryvia is alpha: the API may change before a stable release. GPU paths are tested on kind with simulated GPUs, fake clients and CPU-only clusters, not yet on GPU hardware, and performance figures are targets until published benchmarks back them. The README's “What works today” lists what was tested and where.