Guides
The library docs say what each piece does. These guides say which piece to reach for and why, from what the lab proved. Each links to the lab scenarios that show it working.
| Guide | Read it when |
|---|---|
| Choosing a generative UI tier | You're deciding how an answer should look, and where it goes |
| Canvases, the workspace and state ownership | A canvas should follow the agents, and the agents should see what the person sees |
| Questions, approvals and sub-agents in the UI | An agent needs the person, or delegates to agents the person should see |
| MCP Apps as host and as viewer | A partner's app should show in your chat, or your canvas in someone else's host |
| Testing with scripted models and recorded streams | You want tests that don't depend on a model, and numbers that do |