Vercel Streamdown – Markdown for AI Streaming (streamdown.ai)

Vercel announced Streamdown, a drop-in replacement for react-markdown built specifically for AI-powered streaming. Packaged with the AI Elements Response component (and available standalone via npm), Streamdown addresses the practical problems of tokenized, incremental model output: it can parse and intelligently fix incomplete Markdown (unclosed code fences, headings, lists), stream rendered content smoothly, and provide polished defaults for common UI needs like Tailwind-styled headings, lists and code blocks. It also adds ergonomic features such as Shiki-based syntax highlighting with a hover-to-copy button, Mermaid diagram support with configurable themes, and KaTeX-rendered LaTeX math via remark/rehype plugins. For engineers, Streamdown is significant because it combines streaming-friendly parsing with security and extensibility familiar to the remark/rehype ecosystem. Default rehype plugins include rehype-harden (to limit untrusted image/link origins), rehype-raw and rehype-katex; default remark plugins include GFM and remark-math. The component extends react-markdown props with streaming and security options (parseIncompleteMarkdown true by default, shikiTheme, mermaidConfig, controls for copy/download), making it easy to replace react-markdown in chat or assistant UIs while getting safer, prettier, and more robust handling of partially generated Markdown, code, math, and diagrams.