English | 中文 | हिंदी | Español | Français | العربية | বাংলা | Русский | Português | Bahasa Indonesia

📈 Hybrid AI Trading System 📈

An expert decision-support system for NASDAQ and Oil (WTI) ETF trading, leveraging a 14-model hybrid artificial intelligence powered by unified cloud LLMs for robust and nuanced trading signals.

---

📚 Table of Contents

• 🌟 About the Project
  • 🚀 Dual-Ticker Strategy (Analysis vs. Trading)
  • 🧠 Hybrid AI Engine
  • 🧘 Decision Philosophy: "Cognitive Prudence"
  • ✨ Key Features
  • 💻 Tech Stack
  • 🧠 AI & LLM Architecture (NexusAI-Client Cloud Multi-Provider)
  • 🧠 FinAcumen (Financial Memory)
• 📂 Project Structure
• 🚀 Quick Start
  • ✅ Prerequisites
  • ⚙️ Installation
• 🛠️ Usage
  • Simulation Mode (Paper Trading)
  • Real Execution (Trading 212)
• 🧪 Production Backtesting
• 🤝 Contributing
• 📜 License

---

🌟 About the Project

This project is an expert decision-support system for ETF trading, using a 14-model hybrid AI approach. It is designed to provide a comprehensive and robust analysis by combining several AI perspectives.

🚀 Dual-Ticker Strategy (Analysis vs. Trading)

The system uses an innovative approach to maximize model accuracy:

• High-Fidelity Analysis: AI models analyze global reference indices (^NDX for Nasdaq, CL=F for WTI Crude Oil). These indices offer longer history and "purer" trends, without the noise related to trading hours or ETF fees.
• ETF Execution: Real orders are placed on the corresponding tickers on Trading 212 (SXRV.DE, CRUDP.PA), using T212 live prices (via positions API) for position sizing. Portfolio state is synchronized directly from T212 (sync_state_from_t212()), and live prices are injected into the analysis pipeline (_inject_t212_live_price() in src/data.py).

🧠 Hybrid AI Engine

The system merges thirteen distinct signals (plus a meta-model):

1. Classic Quantitative Model: RandomForest/GradientBoosting/LogisticRegression ensemble trained on technical and macroeconomic indicators.
2. TimesFM 2.5 (Google Research): State-of-the-art foundation model for time-series forecasting.
3. TensorTrade / PPO (Reinforcement Learning): RL agent (stable-baselines3) training a PPO policy in a custom Gymnasium trading environment with persistence across cycles.
4. Oil-Bench Model: Energy-specialized model merging EIA fundamental data (Stocks, Imports, Refinery utilization) and sentiment for WTI trading.
5. Textual LLM (NexusAI Cloud Multi-Provider): Contextual analysis of raw data, real-time news via the AlphaEar skill, and integration of dynamic macro-economic web research. Powered by NexusAI-Client with automatic failover across multiple free and paid frontier cloud providers.
6. Visual LLM (NexusAI Multimodal Vision): Direct chart pattern analysis (enhanced_trading_chart.png).
7. Sentiment Analysis: Hybrid analysis combining Alpha Vantage and "hot" trends from AlphaEar (Weibo, WallstreetCN).
8. Decentralized Data (Hyperliquid): Analysis of speculative sentiment on Oil (WTI) via Funding Rate and Open Interest.
9. Vincent Ganne Model: Geopolitical and cross-asset analysis (WTI, Brent, Gas, DXY, MA200) for detecting macroeconomic bottoms.
10. Grebenkov Model: Trend-Following mathematical model calibrated for cross-asset analysis using Agnostic Risk Parity.
11. Hidden Markov Model (HMM): Probabilistic model for market regime detection (bullish/bearish) based on historical price variations.
12. FinAcumen (Experience Memory Engine): An intelligent ReAct agent loop that evaluates market conditions by writing and executing Python queries, equipped with a vector "Financial Memory".
13. 🏛️ Weekend Council (Strategic Retrospective): A weekly, async, multi-persona LLM deliberation running every Saturday and Sunday at 09:00. Six personas (Stratège / Risk Manager / Quant / Sceptique / Tacticien / Comportementaliste) each execute on distinct cloud LLM providers (Groq, Cerebras, Mistral, Cohere, Nvidia NIM / OpenRouter / OrcaRouter, Gemini Free & Pro) for genuine reasoning diversity. The Judge emits a per-ticker stance that becomes the 11th weighted vote in the real-time consensus (9.5% weight, decaying linearly over 7 days).
14. Hybrid Fusion Engine: The meta-model orchestrating dynamic weighting and cognitive consensus across all sub-models.

The goal is to produce a final decision (BUY, SELL, HOLD) with an absolute priority on Accuracy First.

🧘 Decision Philosophy: "Cognitive Prudence"

Unlike classic trading algorithms that panic as soon as volatility explodes, this system applies an informed investor approach:

• Strong Consensus Required: A quantitative model (Classic) may cry wolf (SELL), but if cognitive models (Text LLM, Vision, TimesFM) remain neutral, the system will prefer HOLD.
• Confidence Filter: A movement decision (Buy or Sell) is only validated if the global confidence exceeds a safety threshold (generally 40%). Below this, the system considers the signal as "noise" and remains on standby.
• Capital Protection: In VERY_HIGH risk mode, HOLD serves as a shield. It prevents entering an unstable market and avoids exiting prematurely on a simple technical correction.

✨ Key Features

• Unified Cloud LLM Architecture via NexusAI-Client: Complete removal of heavy local Ollama and GGUF dependencies. Direct high-speed API calls with zero-downtime automatic failover across 9+ cloud providers (Gemini Free/Pro, Groq, Cerebras, Mistral, Cohere, Nvidia NIM, OpenRouter, OrcaRouter, DeepSeek).
• Sub-Minute Cycle Execution: High-speed parallelized cloud inference brings the full multi-model analysis cycle down from 15 minutes to ~30-45 seconds.
• Dual-Ticker Approach: Analyze the index, trade the ETF.
• T212 Live Prices: Real-time recovery of EUR prices via the Trading 212 API (<1s), with yfinance fallback and parquet cache.
• Dated Brent Spread: Monitoring of physical market tension via the spread between Brent Spot (Dated) and Brent Futures.
• Network Resilience: Circuit breakers and timeouts across all external network calls.
• Cache Auto-Invalidation: Parquet cache auto-detects staleness (> 1 day) and forces a refresh.
• Autonomous Morning Brief Agent: An overnight analytical workflow (morning_brief/morning_brief.py) running daily via schedule.py. Synthesizes market reports and injects fundamental awareness into daily trading cycles.
• News & Blockchain Sentiment: Integration of AlphaEar and Hyperliquid to capture social and speculative sentiment.
• Centralized Risk Management: The AdvancedRiskManager centralizes Anti-Loss (Stop-Loss) and Trailing Stop logic.

💻 Tech Stack

• Language: Python 3.12+
• Calculations & Data: pandas, numpy, yfinance, pyarrow, pandas_datareader, hyperliquid-python-sdk
• Machine Learning: scikit-learn, shap, stable-baselines3, gymnasium
• AI & LLM: nexusai-client, google-genai
• Web Scraping & Search: beautifulsoup4, duckduckgo_search, crawl4ai
• Visualization: matplotlib, seaborn, mplfinance
• Utilities: tqdm, rich, python-dotenv, schedule

---

🧠 AI & LLM Architecture (NexusAI-Client Cloud Multi-Provider)

The system leverages NexusAI-Client to unify all cloud AI model calls into a resilient, zero-maintenance gateway:

• Zero Local Footprint: No local LLMs to download or run.
• Automatic Fallback Chain: Free tiers (Gemini Free, Groq, Cerebras, Nvidia NIM, OrcaRouter, Cohere) are queried first, seamlessly cascading to paid models or alternative providers on 429/503 errors.
• Multimodal Vision: Seamless technical chart analysis via multimodal frontier models.
• Strict JSON Parsing: Automatic extraction and validation of trading decisions, search queries, and oil allocations.

---

📂 Project Structure

Trading-AI/
├── morning_brief/                   # Overnight autonomous agent for fundamental analysis
│   ├── morning_brief.py             # Brief orchestrator via NexusAI-Client
│   └── output/                      # Generated daily markdown reports
├── src/                             # Core modules
│   ├── adaptive_weight_manager.py   # Dynamic model weighting based on performance
│   ├── advanced_risk_manager.py     # Trend-Aware risk management and sizing
│   ├── bootstrap.py                 # Core initialization logic
│   ├── chart_generator.py           # Generates technical charts for visual LLM
│   ├── classic_model.py             # Scikit-learn quantitative models ensemble
│   ├── config_weights.py            # Base weights configuration for the hybrid engine
│   ├── data.py                      # Data fetching, caching, and preprocessing
│   ├── database.py                  # SQLite database management for metrics
│   ├── eia_client.py                # Energy Information Administration API client
│   ├── enhanced_decision_engine.py  # Hybrid fusion engine orchestrating all models
│   ├── enhanced_trading_example.py  # Pipeline execution and orchestration
│   ├── features.py                  # Technical and macroeconomic feature engineering
│   ├── grebenkov_model.py           # Trend-Following math model
│   ├── hmm_model.py                 # Hidden Markov Model for regime detection
│   ├── llm_client.py                # Unified LLM inference via NexusAI-Client
│   ├── news_fetcher.py              # Financial news crawling and parsing
│   ├── oil_bench_model.py           # Energy-specialized WTI trading model
│   ├── performance_monitor.py       # P&L and risk metrics monitoring
│   ├── sentiment_analysis.py        # Sentiment analysis engine
│   ├── t212_executor.py             # Trading 212 order execution & state sync
│   ├── tensortrade_model.py         # Reinforcement learning model
│   ├── timesfm_model.py             # Google TimesFM foundation model wrapper
│   ├── vincent_ganne_model.py       # Geopolitical bottom-detection model
│   ├── web_researcher.py            # Dynamic web research query generator
│   ├── council/                     # Weekend AI Council deliberation suite
│   │   ├── weekend_council.py       # 3-round multi-provider debate orchestrator
│   │   └── council_prompts.py       # Personas and debate templates
│   └── agents/                      # FinAcumen cognitive ReAct agent
├── tests/                           # Comprehensive unit and integration test suite
├── main.py                          # Pipeline entry point
├── schedule.py                      # Production scheduler
└── scheduler_config.json            # Centralized parameter configuration

---

🚀 Quick Start

✅ Prerequisites

• Python 3.12+ (via uv)
• API keys in .env (Gemini, Groq, Mistral, Nvidia, Alpha Vantage, EIA, etc. — see .env.example)

⚙️ Installation

1. Install uv: astral.sh/uv
2. Setup TimesFM 2.5 & Dependencies:

   python setup_timesfm.py
   uv sync
   uv run python -m playwright install chromium

3. Configure Environment: Copy .env.example to .env and fill in your API keys.

---

🛠️ Usage

# Paper trading simulation (NASDAQ)
uv run main.py --simul

# Paper trading simulation (Oil)
uv run main.py --simul --ticker CRUDP.PA

# Live/Demo execution on Trading 212
uv run main.py --t212

# Run full automated scheduler (8:30 AM - 6:00 PM)
uv run schedule.py

# Run Weekend AI Council on demand
uv run python -m src.council.weekend_council --days 7

# Run Morning Market Brief on demand
uv run python morning_brief/morning_brief.py

---

🤝 Contributing

Contributions are welcome! Please open an issue or submit a PR.

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.