What's new

March 10, 2026

• Added English-only output for all personas — reduces token cost ~40%
• Added Anti-pattern rules for each expert persona
• Added SearchProvider interface for future AI search integration
• Improved Expert maxTokens reduced 4096→2048 for faster responses
• Improved Context-aware prompts: personas know output goes to AI IDE agent

March 9, 2026

• Added Council of Architects — 4 expert personas + Lead Architect
• Added MCP tool: upgrade_architecture_blueprint
• Added Support for custom Lead Architect model selection
• Added Parallel expert execution with Promise.allSettled
• Added Graceful degradation when expert calls fail