Level, Draw, or Narrate
The up to date Design Mode presents a number of enter strategies. Customers can click on on a component to immediate an agent for edits associated to that particular element. For modifications involving a number of parts, multi-select performance permits customers to reference a number of parts concurrently, enabling duties like matching types or adjusting teams.
Drawing capabilities allow customers to focus on particular areas or areas on the interface for the agent’s consideration. This characteristic captures a frozen body of the viewport, guaranteeing the agent receives exact visible context. The newest iteration additionally introduces extra correct focusing on and sooner efficiency for voice enter, making visible interactions really feel extra built-in into the event workflow.
Beneath the hood, deciding on a component feeds the agent two key information streams: the ingredient’s technical id (like its XPath, element kind, attributes, and Fiber tree props) and a screenshot for spatial context. This twin method equips the agent to effectively find and modify the related code.