Regime-Aware AI Options Agent

Traditional algorithmic trading bots rely on rigid static heuristics, while naive LLM trading wrappers suffer from severe hallucination and unconstrained drawdown risks. This project bridges this gap by introducing an institutional-grade, autonomous quantitative trading system. It strictly decouples probabilistic LLM strategy reasoning (Google Gemini 3.6 Flash) from a 100% deterministic Python Hard Risk Gate. The system operates through an isolated, multi-step pipeline natively integrated with the Alpaca Trading API and FastMCP Server. First, it calculates technical indicators (SMA, RSI, ATR, Realized Volatility) to classify the market into distinct regimes (e.g., High Volatility or Bullish). Next, the AI synthesizes this data into strict JSON proposals, tailoring strategies to the regime (e.g., Iron Condors for high volatility, Bull Call Spreads for uptrends). Before execution, the Deterministic Hard Risk Gate evaluates the AI's intent. It enforces a strict maximum 5.0% account equity allocation per trade and absolutely bans naked short options, ensuring every strategy is defined-risk. Finally, approved orders are formatted to standard OCC symbology and dispatched to Alpaca, while a continuous monitor handles automated Take-Profit (+50%) and Stop-Loss (-40%). The entire system is fully operable via an interactive CLI daemon and a Streamlit dashboard featuring an Explainable AI audit log.