Skip to main content

​
Before you begin

Install the CLI, authenticate with rig login, and use an isolated test workspace. These examples require the v0.13 CLI. The manifests have been checked with the v0.13 parser; verify application health after deployment. Open the ai-chat source. These manifests explicitly select incremental deployment.

​
Deploy from your machine

From a terminal on your machine, clone the repository and enter the example directory:
ai.managed: true enables managed routing. The model parameter allows rigbox/default, rigbox/fast, or rigbox/free; these are aliases, not guarantees of a particular model or price.
Record the workspace and app IDs printed by deployment. Use those exact IDs wherever this guide shows WORKSPACE_ID or APP_ID; do not guess a public URL from an app name.

​
Verify the result

Open the app and send a short message. A healthy HTTP process does not prove AI authorization: check a completed response, then inspect logs if the proxy rejects it.

​
Access and recovery

The example uses the authenticated route by default. Managed AI requires your account/workspace to have access and available credits. If deployment fails, read the terminal error and installation output before retrying. rig app logs --app APP_ID --install shows install logs when an app exists. Check required credentials, resource limits, the start command, and the configured health path. An image deployment may require explicit root-filesystem replacement consent; do not add --reimage to a workspace containing data you need.

​
Clean up

After saving any data you need, delete only the isolated example workspace. This deletes its applications and workspace data; inspect persistent volumes and snapshots separately before removing them.
Continue with deployment guides or the manifest reference.