Thesysdev Opens ChatGPT's Intelligent UI Pattern as a Free Next.js Tool
A new MIT-licensed repo recreates ChatGPT's Intelligent UI, streaming interactive maps, charts, forms, and mini apps instead of plain text replies.
- Open-source clone of ChatGPT's Intelligent UI, streaming interactive answers instead of plain text, now at 620+ stars on GitHub.
- Built on OpenUI Lang 1.0: model emits a DSL that renders into stateful React primitives with maps, charts, inputs, SVG, and sandboxed HTML.
- Ships 59 helper functions so answers compute client-side, enabling calculators, planners, and dashboards without round-trips to the model.
- Works with OpenUI Cloud, any OpenAI-compatible API via
LLM_BASE_URL, or local Ollama models like qwen3 and gpt-oss. - Includes a benchmark harness that parses outputs and screenshots them across widths and themes for model evaluation.
- MIT licensed Next.js and TypeScript project with a live hosted demo you can try first.
Thesysdev has released Open Intelligent UI, an MIT-licensed Next.js implementation of ChatGPT’s Intelligent UI pattern. A prompt such as “Plan a day in San Francisco” can produce an interactive map, photos, and an editable itinerary. Developers can inspect and modify the language, renderer, state system, API route, and evaluation harness, making the repository a practical reference for building generative interfaces.
Written in TypeScript, the repository has 630 GitHub stars. A hosted demo runs without installation and includes starter prompts for city exploration, calculators, and a playable drum pad.
| Area | Implementation |
|---|---|
| License | MIT |
| Application stack | Next.js and TypeScript |
| Generated format | OpenUI Lang 1.0 |
| Default backend | OpenUI Cloud through the Thesys-hosted Gateway |
| Other backends | OpenAI-compatible endpoints and Ollama |
| Thread storage | Browser localStorage |
From prompt to component tree
OpenUI Lang 1.0 gives the model a compact domain-specific language for describing components, state variables, and function calls. The client parses that language as tokens arrive and builds a React tree incrementally. Layout remains under the model’s control while each generated element maps to a component registered by the application.
A broad primitive set
- Layout and content: text, containers, lists, expandable sections, and other structural elements.
- Inputs:
Slider,DatePicker,Timer, and category filters. - Maps and media: MapLibre GL maps,
MapMarker, street routes, and image search results. - Data and learning tools: charts, KaTeX math, dashboards, and a flipping
Flashcard. - Embedded experiences:
HtmlViewcan run generated HTML, canvas projects, or small games inside a sandboxed iframe.
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
