Streamdown

What is Streamdown?

Streamdown is a standalone library that facilitates the streaming and rendering of Markdown content in AI applications, addressing the challenges of handling incomplete or tokenized inputs. Its mission is to provide developers with a secure, easy-to-use tool that ensures perfectly formatted outputs without manual intervention, making it ideal for real-time AI interfaces. By incorporating plugins for advanced syntax like GFM, math expressions, and diagrams, it enhances the functionality of AI-driven content display. Streamdown powers the AI Elements Response component but can be installed independently for broader use. It solves problems related to prompt injection and unterminated Markdown blocks, promoting safer and more efficient development.

Streamdown's Core Features

• Offers built-in Tailwind classes for styling common Markdown elements like headings, lists, and code blocks, providing consistent and visually appealing formatting out of the box.
• Supports GitHub Flavored Markdown (GFM) including task lists and tables, enabling users to render complex Markdown syntax seamlessly in streaming environments.
• Provides interactive code blocks with Shiki syntax highlighting and a copy button, improving code readability and user interaction in technical content.
• Enables rendering of LaTeX mathematical expressions using remark-math and KaTeX, which is particularly useful for educational, scientific, or technical AI applications.
• Supports interactive Mermaid diagrams that stream as code blocks with a render button, allowing for dynamic visualization of charts and flows.
• Parses unterminated Markdown blocks such as headings, inline code, bold, italic, and links, ensuring smooth handling of incomplete streaming inputs.
• Includes security hardening features to prevent prompt injection by restricting image and link origins, enhancing safety in AI-powered applications.
• Allows customizable props for streaming behavior, security settings like allowed URL prefixes, and theme options for code blocks, giving developers flexibility in implementation.