Inference Agent UI

Build production-ready agent interfaces using the ui.inference.sh component registry and the AI SDK stream protocol.

Treat this skill as integration guidance, not a design system. Reuse existing app design tokens/layout unless the user asks for a visual redesign.

Use This Workflow

1. Confirm the integration mode:

• generic-stream: Use any backend that returns AI SDK UI message streams.
• inference-runtime: Use inference.sh runtime and add uiTransform.

2. Install registry components with shadcn:

• Add core primitives first:
  • npx shadcn@latest add "https://ui.inference.sh/r/agent.json"
• Add widgets only when needed:
  • npx shadcn@latest add "https://ui.inference.sh/r/widgets.json"
• Add optional extras only if the task requires them:
  • actions.json, citations.json, sources.json