[Show and Tell] Lumina Capital — Production-Grade Multi-Agent Hedge Fund Engine (LangGraph, 8-Legend Council, TimesFM & Qlib)

[ramuza0113-sketch]

Hi TradingAgents community! 👋

Inspired by multi-agent trading research and architectures like TradingAgents, we built and open-sourced an institutional-grade engine designed for real-world market execution:

Lumina Capital (AI Hedge Fund Engine v6.0)
🐙 GitHub: https://github.com/ramuza0113-sketch/lumina-ai-hedge-fund
🚀 Product Hunt: https://www.producthunt.com/posts/lumina-capital

---

🤖 What We Built Upon the Multi-Agent Paradigm

1. 8-Legend Parallel Deliberation (LangGraph):
   Independent parallel reasoning across 8 distinct investor personas (Warren Buffett, Michael Burry, George Soros, Cathie Wood, William O'Neil, Benjamin Graham, cis, and a CRO risk officer).
2. Fact-Linked Cross-Debate:
   Direct head-to-head clash between the strongest Bull and Bear agents backed by primary financial data (SEC Form 4 insider trades, FRED liquidity, Options Gamma Flip).
3. Hard CRO Circuit-Breaker Veto:
   The Chief Risk Officer agent has absolute veto power (forcing allocation to 0.0%) during negative gamma regimes or market turbulence, overriding bullish personas.
4. Google TimesFM & Qlib Alpha158:
   Ensemble of zero-shot foundation time-series forecasting and 158 mathematical quantitative factors.
5. Real-time 3D Causal Knowledge Graph:
   Interactive 3D graph representation (Graphiti × Neo4j) visualizing real-time market relationships and narrative contagion.

Huge respect to the TradingAgents team and community for pioneering this space! We would love your feedback and collaboration on agent debate dynamics and execution safeguards.