- Create docs/MACOS_APP_MANUAL.md with full documentation:
- Installation and build instructions
- Menu bar interface guide with keyboard shortcuts
- Preferences configuration (General, LLM, Network, Notifications)
- Real-time updates via WebSocket
- Network modes (localhost vs LAN)
- Troubleshooting section
- Development guide with project structure
- FAQ section
- Update README.md:
- Add native macOS app to "What You Get" section
- Update architecture diagram to show menu bar app
- Add macOS app manual to documentation table
- Update project structure to include macos-app directory
Implement a comprehensive debate framework for improved trading decisions:
- Add 6 debate personas: Optimist, Pessimist, Fundamentalist, Technician,
Devil's Advocate, and Neutral
- Implement structured debate rounds with position-taking and rebuttals
- Add weighted consensus mechanism for aggregating agent opinions
- Add confidence calibration tracking to improve accuracy over time
- Include comprehensive test suite (38 tests)
New components:
- src/agents/debate.py - Complete debate system implementation
- DebateAgent: Individual agent with persona-specific prompts
- DebateOrchestrator: Coordinates multi-agent debates
- ConfidenceCalibrator: Tracks prediction accuracy
- tests/test_debate.py - Full test coverage
The debate system helps avoid groupthink by having agents with different
analytical biases challenge each other's positions before reaching consensus.
A comprehensive autonomous trading system for Polymarket prediction markets
featuring multi-LLM provider support, a native macOS menu bar app, and a
web-based control dashboard.
Key features:
- Multi-agent trading system (Research, Risk, Execution, Reflection agents)
- LLM provider flexibility (Anthropic, OpenAI, Google, xAI, Local models)
- Automatic provider fallback chain for resilience
- Native Swift/SwiftUI macOS menu bar application
- FastAPI web dashboard with real-time WebSocket updates
- Risk management with kill switch
- Technical indicators and market analysis
