Hey everyone, I am the maker of Smart Components. I originally started building this because replicating that Copilot-style ghost text natively in React felt like a really fun engineering puzzle. When building AI features, I realized how much time is wasted bridging raw LLM streams to the DOM. You end up wrestling with fetch streams, race conditions, debouncing, and trying to get the math right for ghost-text positioning (especially on mobile). So, I built an open-source, completely headless library to handle all those complex browser and state behaviors natively. What it does out of the box: : Drop-in predictive ghost text that perfectly aligns with user input, including built-in fixes for iOS Safari tap delays. useSmartState: A hook that snaps chaotic, mid-stream LLM outputs directly into clean React state. Headless & Agnostic: You bring your own styles (Tailwind, Shadcn) and your own models (OpenAI, Anthropic, Local). I am open-sourcing this today to help other developers skip the messy boilerplate and get straight to building. I would love your feedback, and I will be hanging out in the comments all day to answer any questions or talk about the technical implementation.

Smart React Components

Headless React components for building AI interfaces.

Headless React components for building AI interfaces.