RTX AI Platform

GitOps-first single-node Kubernetes platform for local AI workloads on NVIDIA RTX GPUs.

Includes:

• vLLM inference serving
• MLflow tracking
• GPU monitoring
• Gateway API networking
• FluxCD GitOps management
• NVIDIA GPU runtime integration

---

Prerequisites

• GPU visible (nvidia-smi)
• NVIDIA container runtime configured (nvidia-container-toolkit)
• Kubernetes cluster with a default StorageClass

---

Quickstart

GitOps-ready via FluxCD, but optimized for quick local setup:

make

Then open:

• MLflow - http://<node-ip>:<node-port>/mlflow
• Grafana - http://<node-ip>:<node-port>/grafana
• Prometheus - http://<node-ip>:<node-port>/prometheus
• vLLM - http://<node-ip>:<node-port>/llm/v1/models

---

License

MIT — see LICENSE