Connect an external model provider

Before you start

Obtain the provider's exact model ID, API endpoint and credentials. Confirm its privacy, cost and model-use terms. Prompts sent to this model leave the appliance. Local GPU and model-cache capacity do not size the remote provider.

Connect

1. Open Models → Create and select an external model.
2. Enter a unique local model name, provider/model identifier and the required endpoint fields offered by the form.
3. Provide the API key privately. Existing Secret references and provider settings must belong to the deployment; never paste them into public examples.
4. Save, wait for catalog publication and send a small API request.

Operate

Use Edit to change supported parameters, and Stop/Start to remove or restore the local route. Stop does not shut down the provider's server or cancel its billing. Remote provider logs are not Kubernetes Pod logs; inspect returned API errors and the provider's own diagnostics.

See external model fields for advanced configuration.