Ollama Helm Chart

Ollama Helm Chart packages a self-hosted LLM service for developers and infrastructure teams that already use Kubernetes. It brings Ollama into a cluster so applications can access models running on the cluster's hardware.

• CPU and GPU deployments Run Ollama on CPUs or use NVIDIA and AMD GPUs, subject to Ollama's hardware compatibility.
• Model preparation Download models, load selected models into memory at startup or create custom models as part of deployment.
• Application connections Applications can use the Ollama API, the official ollama-js and ollama-python clients, or LangChain's JavaScript and Python integrations.
• Knative Serving Use Knative in place of a standard Kubernetes deployment, with a separate model preparation job sharing persistent model storage with the serving application.

The community chart is open source under the MIT license. GPU support requires a newer Kubernetes baseline than CPU-only operation, and Ollama doesn't support every GPU, particularly some AMD cards. In Knative mode, model preparation requires persistent storage, and the cluster must permit the relevant persistent-volume features. Deployment doesn't wait for the model preparation job to finish.