What's New in v0.4.1
This is a maintenance release focused on performance improvements and bug fixes. Here are the main highlights:
- 93% reduction in diagram update calls during streaming (prevents browser crashes)
- Simplified LLM XML format - less tokens = faster diagram generation
- XML auto-fix improved from 58.7% to 99% success rate
- Minimal style mode for faster black/white diagram generation
Performance Improvements
- Debounced streaming diagram updates to prevent browser crashes during long sessions (#262)
- Simplified LLM XML format to output bare mxCells only - reduces token usage (#254)
- Added minimal style mode toggle for faster diagram generation (#260)
- XML auto-fix with 99% fix rate for malformed LLM output (#247)
- Optimized Vercel costs by removing analytics and tuning function settings (#238)
- Added append_diagram tool for better truncation handling (#252)
- Added configurable MAX_OUTPUT_TOKENS to prevent Bedrock truncation (#251)
Bug Fixes
- Fixed browser crash during long streaming sessions with slow models (#262)
- Fixed recovery from invalid XML in localStorage on startup (#261)
- Fixed malformed XML handling from DeepSeek gracefully (#235)
- Fixed UI revert and user notification when feedback submission fails (#237)
- Added user feedback for clipboard copy failures (#236)
- Fixed custom model setting bug (#227)
- Improved Azure provider auto-detection and validation (#223, #225)
- Fixed clipboard issues (#189)
- Persisted processed tool calls to prevent replay after chat restore (#224)
- Limited auto-retry to 3 attempts and enforced quota checks (#219)
- Disabled history XML replacement by default (#217)
Contributors
Thanks to all contributors who made this release possible!
Full Changelog: v0.4.0...v0.4.1