wandb/openui

OpenUI is a tool that lets you describe a web interface component in plain English, like "a login form with a blue submit button", and then instantly see a working, rendered version of it in your browser. Instead of manually writing HTML, CSS, and layout code, you type what you want and an AI generates the visual result live. The way it works is straightforward: you give OpenUI a description of a UI element (a button, a form, a card layout, etc.) and it uses a large language model (an AI service that understands and generates text and code) to produce the HTML. You can then ask for modifications in the same conversational way, "make the button larger" or "change the font color", and the AI updates the result. You can also have it convert the generated HTML into other formats used by different frontend frameworks, such as React, Svelte, or Web Components. OpenUI supports connecting to multiple AI providers: OpenAI, Groq, Anthropic (Claude), Gemini, Cohere, Mistral, and locally-running models via Ollama. You supply whichever API key you have, and OpenUI routes to that service. You would use this when you want to prototype a UI quickly without doing the manual coding yourself, or when you're experimenting with what a component should look like before investing time in a polished implementation. It's useful for developers who want to skip the tedium of writing repetitive UI code and for teams testing ideas visually. The backend is written in Python, with TypeScript on the frontend. You can run it via Docker (the recommended approach) or by cloning the repository and running it directly with Python. A development container and Gitpod configuration are also included for contributors.