Released: Sep 02, 2025
This alpha release represents a significant milestone in MAESTRO's development, with comprehensive improvements across the backend, frontend, and documentation. The focus has been on performance optimization, stability improvements, and making the platform more user-friendly.
Performance Improvements
Complete Async Migration
- Migrated entire backend to async/await architecture for better concurrency
- Implemented async database operations with asyncpg
- Fixed event loop blocking issues that were causing UI freezes
- Added per-mission concurrency control with intelligent rate limiting
Parallel Processing
- Implemented parallel web searches for faster information gathering
- Added concurrent document retrieval with semaphore controls
- Increased MAX_WORKER_THREADS to 20 for better performance
- Optimized LLM request handling with queue management
Backend Enhancements
Mission Management
- Fixed mission resume functionality with proper context restoration
- Added recovery capabilities for resuming from specific research rounds
- Implemented outline revision and validation mechanisms
- Fixed citation processing and report generation phases
- Added intelligent chunk sufficiency evaluation for document searches
Agent Improvements
- Simplified planning and messenger agents for better results
- Enhanced outline generation with validation and retry logic
- Improved research agent with better error handling
- Fixed agent activity logging and statistics display
- Added DEBUG_REFLECTION environment variable for troubleshooting
Reference & Citation System
- Improved reference citation system with Unicode bracket support for llms using these
- Support for multiple citations in single brackets in writing mode (e.g., [1,2,3])
- Fixed reference filtering and renumbering based on actual usage
- Added inline reference citations for writing mode
- Enhanced citation processing during report generation
UI/UX Enhancements
Chat Interface
- Fixed chat history search and filtering functionality
- Improved chat sidebar ordering (now uses created_at timestamp)
- Enhanced message handling and role alternation
- Fixed draft editing and saving functionality
- Added better WebSocket stability and reconnection handling
Mission Interface
- Added comprehensive Settings tab to display mission configuration
- Fixed mission status updates and WebSocket notifications
- Implemented UI recovery buttons and modal for outline revision
- Fixed outline history to show all versions from each round
- Improved Start button to use existing approved questions
Writing Mode
- Added LaTeX formula rendering support throughout
- Fixed LaTeX scrolling issues with scoped KaTeX CSS
- Implemented proper draft saving and editing
- Fixed reference display and formatting
- Enhanced writing context management
General UI
- Moved toast notifications to bottom-right corner
- Fixed duplicate toast notification issues
- Improved error messages and user feedback
- Fixed various scrolling and display issues
Settings & Configuration
Mission Settings
- Comprehensive mission settings capture and storage
- Auto-save functionality for chat settings
- PostgreSQL migration system with chat settings support
- Updated default research parameters for optimal performance
- Improved settings persistence with proper state management
Database
- Added proper PostgreSQL migration system
- Implemented chat settings table and migrations
- Fixed database connection pooling issues
- Added support for JSONB fields for better performance
Documentation Overhaul
Complete Restructuring
- Simplified troubleshooting documentation
- Removed outdated and incorrect information
- Fixed all references to non-existent CLI commands
- Updated environment variables documentation
New Sections
- Added comprehensive example reports from various models
- Created dedicated prompts page showing actual prompts used
- Added hardware requirements and deployment configurations
- Included model-specific known issues and workarounds
Visual Updates
- Updated color scheme to match MAESTRO logo
- Improved navigation structure focusing on user needs
- Added card-based layouts for better readability
- Enhanced code block and admonition styling
Upgrade Guide
For users upgrading from v0.1.4:
-
Update and Rebuild
git pull docker compose down docker compose up -d --build
-
Clear Browser Cache
- Required due to UI updates and WebSocket changes
Known Issues
- Qwen 3 30B-A3B has difficulties with structured output generation
- GPT-OSS 20B may have instruction following issues with non-traditional reports
- Some models may produce formatting inconsistencies in long documents
Acknowledgments
Thank you to all users who reported issues and provided feedback. This release addresses many of the stability and performance concerns raised by the community.
What's Next
- Further UI polish and responsiveness improvements
- Additional model support and optimization
- Enhanced collaboration features
- Advanced document processing capabilities
- Expanded web search provider options
For detailed technical changes, please refer to the Git commit history.
Report issues: GitHub Issues