💹 AI Financial Research & Advisory Agent

An end-to-end LangGraph-powered financial research and advisory system that autonomously collects, analyzes, and audits stock market data. It integrates machine learning, LLMs, and compliance-driven auditing to simulate real-world investment decision-making with built-in human oversight.

🚀 Features

• Real-time data ingestion from Yahoo Finance, NewsAPI, and SEC Edgar
• Technical indicator computation: RSI, MACD, Moving Averages, Volatility
• ML-driven BUY / SELL / HOLD signal generation using Random Forest
• GPT-4–powered financial advisory with compliance-safe recommendations
• Automated compliance checks, restricted stock detection, and risk scoring
• MongoDB-based audit trail ensuring traceability and explainability
• LangGraph workflow orchestration and checkpointing for reproducibility

🧠 System Architecture

[Data Ingestion] → [Feature Engineering] → [Signal Generation] → [Compliance Check] → [Financial Advisory/Human Review] → [Audit & Logging] → END

LangGraph Workflow Diagram

🛠️ Tech Stack

• LangGraph
• LangChain
• OpenAI GPT-4
• Scikit-learn
• YFinance
• NewsAPI
• SEC-API
• MongoDB
• Streamlit

⚙️ Installation & Setup

git clone https://github.com/<your-username>/ai-financial-research-agent.git
cd ai-financial-research-agent
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Create a .env file and include:

NEWS_API_KEY=your_newsapi_key  
SEC_EDGAR_API_KEY=your_sec_api_key  
MONGO_URI=your_mongodb_uri  
OPENAI_API_KEY=your_openai_key  

Then execute:

python workflow.py

(Optional UI):

streamlit run main.py

📊 Example Output

Audit log recorded for GOOGL | Decision: Automated Financial Advisory  
{
  "symbol": "GOOGL",
  "signal": "BUY",
  "confidence": 87.42,
  "accuracy": 81.65,
  "risk": 0.34,
  "compliance_status": "PASS",
  "decision_source": "Automated Financial Advisory",
  "final_advice": "Based on RSI and MACD trends, GOOGL indicates short-term bullish strength with controlled risk."
}

🧩 Future Enhancements

• Interactive Streamlit dashboard for visualization
• Multi-asset portfolio simulation & optimization
• LLM-based personalized risk narratives
• Real-time trading integration via broker APIs

👨‍💻 Author

Sayam Kumar
📧 sayamk565@gmail.com | 📞 +1 (437) 876-4544
🔗 LinkedIn | GitHub

🪪 License

Licensed under the MIT License - free to use, modify, and distribute with proper attribution.