📈 FinSight AI: Autonomous Financial Analyst

FinSight AI is a production-grade multi-agent system designed to perform comprehensive investment analysis. It combines Quantitative Finance (Technical Analysis), Machine Learning (XGBoost Forecasting), and Generative AI (Llama-3 via Groq) to deliver actionable insights.

Unlike simple chatbots, FinSight triangulates data from three sources—Price Action, News Sentiment, and Company Filings—to generate professional-grade investment memos.

---

🚀 Key Features

• 🤖 Multi-Agent Architecture:
  • Quant Agent: Calculates RSI, MACD, Bollinger Bands, and runs XGBoost price predictions.
  • Sentiment Agent: Analyzes market psychology using NewsAPI & VADER (Social Sentiment).
  • Research (RAG) Agent: Retrieves fundamental risks from 10-K filings using Vector Search.
  • Manager Agent: Synthesizes all data into a cohesive investment report using Llama-3.
• 📊 Interactive Dashboard: Built with Streamlit & Plotly. Includes Candlestick charts, moving averages, and live metrics.
• 📈 Backtesting Engine: Simulates historical performance of strategies (e.g., Golden Cross) vs. Buy & Hold.
• 💾 Robust Data Engineering:
  • SQLite Database: Caches stock data for offline access and speed.
  • Hybrid Fallback: Tries Database -> Fails to API -> Updates Database.
• 🐳 Production Ready: Includes Docker support, Pytest suite, and a FastAPI backend.

---

🏗️ System Architecture

The system uses LangGraph to orchestrate a directed acyclic graph (DAG) of agents:

1. User Input: Ticker Symbol (e.g., AAPL, BTC-USD).
2. Parallel Execution: Quant, Sentiment, and RAG agents run simultaneously.
3. Synthesis: The Manager Agent (LLM) receives structured JSON from all sub-agents.
4. Output: A structured Markdown report + Interactive Visualizations.

---

🛠️ Tech Stack

• LLM & Orchestration: LangChain, LangGraph, Groq (Llama-3-70b).
• Machine Learning: XGBoost, Scikit-Learn.
• Data & Database: yfinance, SQLite, SQLAlchemy.
• Sentiment: NewsAPI, VADER Sentiment.
• Vector DB: ChromaDB / HuggingFace Embeddings.
• Frontend: Streamlit, Plotly.
• Backend API: FastAPI.

---

⚡ Quick Start

1. Clone the Repository

git clone [https://github.com/yourusername/finsight-ai.git](https://github.com/yourusername/finsight-ai.git)

cd finsight-ai

2. Set Up Environment Variables

Create a .env file in the root directory:

GROQ_API_KEY=gsk_your_key_here
NEWS_API_KEY=your_newsapi_key

3. Install Dependencies

pip install -r requirements.txt

4. Initialize the Database

Run the ingestor to populate your local SQLite database with initial data:

python -m src.data_engine.ingestor_quant

5. Run the Application

Option A: Streamlit Dashboard (UI)

streamlit run src/ui/app.py

Option B: FastAPI Backend

uvicorn src.api.main:app --reload

Option C: Docker

docker build -t finsight .
docker run -p 8501:8501 --env-file .env finsight

---

📂 Project Structure

FinSight/
├── docs/                # Documentation & Architecture diagrams
├── logs/                # Application logs
├── notebooks/           # Jupyter notebooks for research & experiments
├── src/
│   ├── agents/          # LangGraph Agent logic (Quant, RAG, Sentiment)
│   ├── api/             # FastAPI backend endpoints
│   ├── data_engine/     # Database (SQLite) & Ingestion scripts
│   ├── ml_engine/       # XGBoost training, Backtesting, & Feature Engineering
│   └── ui/              # Streamlit Frontend
├── tests/               # Pytest suite
├── Dockerfile           # Containerization setup
└── requirements.txt     # Python dependencies

🧪 Testing

Run the automated test suite to ensure signal logic and API endpoints are working:

pytest tests/

---

🤝 Contributing

Contributions are welcome! Please open an issue or submit a pull request for any features, bug fixes, or documentation improvements.

📄 License

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