Quickstart
This path creates a single model that forwards requests to a built-in managed model. You do not need an upstream provider account or API key.
1. Create a workspace
Section titled “1. Create a workspace”Create or select a workspace in the dashboard. The Overview page becomes the place to find its generated runtime URL, API key, deploy status, and custom domain controls.
2. Create a namespace
Section titled “2. Create a namespace”Create the assistant namespace in Organization → Namespaces.
3. Create a model
Section titled “3. Create a model”Create a support model in the assistant namespace. Paste the first-model
source from the next guide and save it. Its public model ID is
assistant/support; the source routes it to default/gpt-oss-20b.
4. Deploy and test
Section titled “4. Deploy and test”Click Deploy from the model editor after saving. The embedded test console calls the deployed runtime with the selected model. The source must be saved before deployment.
5. Call the runtime
Section titled “5. Call the runtime”Copy the generated runtime URL and default API key from Overview, then follow
Make your first API call. Use
assistant/support as the request model.
To route a model through an API key you own, complete this quickstart first, then connect an upstream provider.