Activate AI Studio

To use AI Studio, enable Object Storage, Model Gallery, and Agent Atelier.

Getting startedGetting started

1. Make sure the Model Gallery and Agent Atelier images are available in the registry used by the cluster. If the required images are not in the local registry, upload them.

2. Make sure the cluster meets the infrastructure requirements for the selected models.

3. Make sure NVIDIA® GPU support is enabled for the GPU worker nodes when deploying the cluster. Wait until NVIDIA® GPU support is ready.

4. Make sure AI Studio dependencies are enabled. If Object Storage is disabled, enable it in the existing StorageConfig resource:

   kubectl patch storageconfig main --type=merge -p '{"spec":{"enabled":true}}'
   

   Wait for the storage to get ready. Check the installation status using the kubectl get componentinstallation storage-main command: the PHASE field should turn Ready.

Enabling Model GalleryEnabling Model Gallery

1. Save this configuration example to the modelgalleryconfig.yaml file.

2. Configure the models and resources. The example includes Embeddings, Ethics, and YandexGPT 5 Lite, which require one A100 80 GB and two H100 80 GB GPUs. Before enabling other models, make sure additional GPUs, CPUs, RAM, and disk space are available.

3. Compare the nodeSelector values with the cluster node labels:

   kubectl get nodes -L nvidia.com/gpu.product
   

4. Apply the manifest:

   kubectl apply -f modelgalleryconfig.yaml
   

5. Check the installation status:

   kubectl get componentinstallation model-gallery-main
   

   Wait for the PHASE field to turn Ready. If an error occurs, check the MESSAGE field in the command output.

Enabling Agent AtelierEnabling Agent Atelier

1. Save this configuration example to the agentatelierconfig.yaml file.

2. If you do not need Code Interpreter, set spec.settings.codeInterpreter.enabled: false. To use it, make sure Certificate Manager is on and allocate resources for sessions.

3. Apply the manifest:

   kubectl apply -f agentatelierconfig.yaml
   

4. Check the installation status:

   kubectl get componentinstallation agent-atelier-main
   

   Wait for the PHASE installation field to turn Ready.

Opening the web UIOpening the web UI

Once the installation is complete, the AI Studio web UI will be available at https://aistudio.sys.<cluster_domain>. Use the Identity and Access Management credentials to log in.

Updating settingsUpdating settings

To change the models, resources, or Code Interpreter settings, edit the relevant resource:

kubectl edit modelgalleryconfig main
kubectl edit agentatelierconfig main

For descriptions of settings, see Configuration.

What's nextWhat's next

• Disabling AI Studio
• AI Studio