Technical Whitepaper & User Guide
AI Markdown Live Viewer & Safe Formatter Technical Architecture Guide
LLM responses formatted in Markdown can sometimes contain unsanitized HTML or XSS injection vectors. SafeAI Markdown Viewer provides a zero-server sandbox environment that safely parses CommonMark and GitHub Flavored Markdown (GFM) into clean, styled HTML directly within browser memory.
1. AST-Based Parsing & XSS-Resistant Rendering Pipeline
Our rendering pipeline transforms raw markdown into an Abstract Syntax Tree (AST), ensuring all raw HTML tokens pass through an uncompromising DOMPurify-grade sanitizer. Code blocks, responsive data tables, task lists, and blockquotes render with elegant slate typography without ever touching an external server.
2. GitHub Flavored Markdown (GFM) Specification Support
| Markdown Syntax | Supported Format | SafeAI Rendering Enhancements |
|---|---|---|
| Fenced Code Blocks | ```lang ... ``` | Syntax highlighting with 1-click clipboard copy |
| Data Tables | | Col A | Col B | | Responsive horizontal scrolling and zebra styling |
| Interactive Task Lists | - [x] Done Task | Modern checkbox state visualization |
| Inlines & Blockquotes | > Quote, **Bold**, ~~Strike~~ | High-contrast readable typography |
3. Markdown Authoring & Rendering Troubleshooting
-
✓
Line Break Rendering: Standard Markdown requires two trailing spaces for hard linebreaks. SafeAI supports soft line breaks for chat LLM outputs.
-
✓
Broken Table Alignment: Ensure the header separator row (`|---|---|`) contains matching column delimiters.
-
✓
Sanitized Elements: Dangerous tags such as `