generative-ui

To scaffold a React generative UI app, you provide an app idea and framework choice to trigger a single prompting flow that initializes the project, writes API keys, and starts the dev server. This handles project setup and initial component wiring automatically.

Yes, you can choose Vite or Next.js as your framework when scaffolding a generative UI app. The tool collects your framework preference alongside your project name and app type during the initial prompt to configure the environment accordingly.

Streaming props require Zod schema constraints that prohibit using z.record, z.map, or z.set, while enforcing schema labeling via .describe(). These guardrails improve runtime validation and prop reliability for agent-driven rendering.

The CLI initializes authentication non-interactively by running the tambo create-app and tambo init flow, which authenticates via browser and automatically writes the correct API key to your environment without pausing for extra confirmations.

You add custom React components by defining 1-2 app-specific components with strict Zod schemas and registering them in the existing component registry. This allows the agent to render them dynamically from natural language inputs.

Zod schema limitations for generative UI component props include the inability to use z.record, z.map, or z.set structures. You must use .describe() for schema labeling to ensure runtime validation and prop reliability during agent-driven streaming.