simon-he95/vue-tui

Vue TUI lets you build terminal-style interfaces using Vue components, and those same components can render either in a web browser or as a real command-line program. The core pitch is "write your UI once, run it in the browser or the terminal." You get building blocks like boxes, text, inputs, lists, and tables, all styled and laid out as if they lived in a terminal screen. At a high level, you wrap your app in a terminal provider that manages a grid of text cells. Components like TBox and TInput draw into that grid. From there, one renderer can turn the grid into HTML for a browser page, and another renderer can output ANSI escape codes to a real terminal via stdout. The same Vue component tree drives both. The project also ships streaming markdown rendering, virtualized lists for large datasets, and log viewers designed to handle high volumes of output without choking. The most compelling use case is building an AI agent console, a terminal surface that streams an agent's text output, shows tool-call status, renders markdown, and accepts user input, all in one cohesive view. Beyond that, anyone who wants to build a dashboard that looks like a terminal but runs in a browser, or a genuine CLI tool with a richer interface than plain print statements, would find this useful. It also supports headless testing, so you can verify your terminal UI behaves correctly without spinning up a real terminal. The project is careful about separating browser-safe code from Node-only code. Browser imports will never pull in native dependencies, while CLI-specific functionality lives behind a dedicated entry point. Some experimental features push further than you might expect, there is support for rendering 3D scenes, video frames, charts, and mermaid diagrams inside a terminal, though the more advanced graphics require specific terminal emulators like kitty or iTerm2 and degrade to ASCII art otherwise. The stable surface covers the terminal core, DOM and CLI rendering, Vue components, and markdown, while high-throughput and agent-specific APIs are still marked experimental.