Hakisoft Universal AI Assistant

Hakisoft Universal AI Assistant redefines the AI assistant paradigm by replacing static chat bubbles with a morphic UI system — an interface that structurally transforms based on detected user intent.

When a user sends a message, the agent (powered by the Groq API via CopilotKit) doesn’t just reply with text. It classifies the context — travel, fitness, development, gastronomy, cultural — and triggers useCopilotAction hooks that mount specific widget compositions: live weather panels, interactive maps, currency converters, and fitness dashboards, all populated with real API data in the same response turn.

What makes it genuinely interactive: every widget is functional, not decorative. The weather panel pulls 7-day forecasts from Open-Meteo. The map queries OpenStreetMap/Overpass for actual POIs. The currency widget fetches live rates from Frankfurter. The agent orchestrates which MCPs to call, handles failures gracefully, and composes the layout — no human designer involved.

The /app-builder route pushes this further: users describe an app in natural language and receive a fully functional React + TypeScript + Tailwind application with live preview via CodeSandbox Sandpack, refined iteratively through conversation.

Originality: most generative UI projects render pre-built components with AI-filled slots. This system lets the agent decide the interface architecture itself — layout, widget combination, color context, and data sources — dynamically per request, powered entirely by real public APIs with zero mock data.

Previously we had tried adding CopiloKit to the development, but unfortunately due to a lack of knowledge it caused us conflicts, but during the test day we managed to make great progress.