RunAI on NVIDIA Brev

setup_prereqs.sh runs automatically when your instance is created. It installs Helm, Ingress, Prometheus, Knative, TLS certs, and CoreDNS.

Prerequisites

• Brev instance with at least 24 CPUs
• Port 443 open on the instance

Deploy RunAI

1. Set your JFrog token (get this from Run:AI support):

export RUNAI_JFROG_TOKEN="your-token-here"

2. Run the deploy script:

./deploy_runai.sh

3. On your local machine, add the hosts entry printed at the end:

sudo bash -c 'echo "YOUR_PUBLIC_IP runai.brev.cloud" >> /etc/hosts'

4. Open https://runai.brev.cloud and accept the self-signed cert warning.

5. Login: test@run.ai / Abcd!234

Re-deploy

Both scripts are idempotent. To tear down and redeploy:

./reset_runai.sh
./deploy_runai.sh

Configuration

Variable	Default
RUNAI_DOMAIN	runai.brev.cloud
RUNAI_VERSION	2.24.37
RUNAI_CLUSTER_NAME	brev-cluster

Troubleshooting

kubectl get pods -n runai-backend    # control plane pods
kubectl get pods -n runai            # cluster pods