Advanced Features Guide
Overview of Gryvia's advanced capability areas, with what is implemented and what is only a CRD or a design.
Authentication and TLS
Signing in
Gryvia CLI Guide
The Gryvia CLI (Rust, in cli/) provides a command-line interface for managing GPU clusters, submitting jobs, monitoring quotas, and analyzing costs.
Gryvia Platform Setup Guide
This guide covers everything after the Kubernetes cluster exists: installing the operators, dashboard and
Gryvia Bare Metal Deployment Guide
Hardware and network preparation for bare metal GPU clusters. The Terraform and Ansible material in the repository is
Frequently Asked Questions
Answers about what Gryvia does today. Where a feature has a CRD but no controller, or has not been verified on real hardware, the answer says so. See the roadmap and the CRD reference (which lists which kinds have a controller).
GPU as a Service
Run Gryvia as a multi-tenant GPU cloud: you (the provider) onboard tenants, publish a catalog of GPU
GPU nodes: drivers, CUDA and the device plugin
Gryvia does not install NVIDIA drivers itself. It uses NVIDIA's own GPU Operator
INTEGRATIONS
External System Integrations
ML Workflows Guide
Gryvia's machine learning workflow kinds: hyperparameter tuning, DAG pipelines, model registry, inference serving and interactive workspaces, plus the training-side kinds that do have a running controller.
Network Intelligence Guide
Guide to Gryvia's eBPF-based network intelligence: the kernel programs, the per-node collector, the
Operational Playbooks
Standard operating procedures and runbooks for Gryvia operations.
Operations: upgrade, uninstall, backup, troubleshooting
Upgrade
Gryvia Roadmap
Where Gryvia stands and what is still open. This page replaces an earlier month-by-month plan that carried invented dates, effort estimates, targets and budgets. There are no committed dates here. Anything under "Open work" is aspirational, not a promise.
Scheduling Guide
What Gryvia does today when a GryviaAIJob is submitted, and which of the more advanced scheduling ideas (gang scheduling, DRF fair share, backfill, elastic training, preemption) exist only as library code or design.
Storage and Network Operators
Overview of the optional Gryvia Storage and Network operators.