Let your agents prompt users during workflows — CHI provides the UI to collect the structured input they need. No code is generated — the UI adapts on the fly based on user responses.
CHI works by defining a UI Definition Language (UIDL) that a Large Language Model (LLM) uses to describe the UI Component Tree. User interactions are captured as events and sent back to the LLM, which dynamically updates the Web UI component tree in response.
This creates an iterative interaction loop that continues until sufficient information has been gathered to generate a result conforming to the target schema.
CHI builds on the Vaadin platform, where high-level UI components ensure the interface remains coherent and accessible.