• Try QuantIntelli+ Live
• View Prediction Analysis

🧠 Dual-Engine Analysis:
Combines a fast XGBoost model for baseline statistical predictions with a deep Google Gemini LLM for contextual analysis.

🌐 Advanced RAG Pipeline:
Dynamically searches the web using Tavily, Google, and DuckDuckGo to gather real-time, relevant information like team news, injuries, form, and H2H stats.

📄 Content Enrichment:
Goes beyond snippets by fetching and parsing full-text articles from top-ranking search results to provide deeper context to the LLM.

📊 Structured Analytical Output:
Generates a detailed, professional-grade report with sections for Dual Recommendation, Conflict Resolution, Market Efficiency, and Risk Analysis.

💾 Persistent Session Logging:
Uses Supabase to log every prediction and its subsequent analysis, creating a traceable record of the agent's reasoning.

🕹️ Interactive UI:
Built with Gradio for an intuitive, easy-to-use interface that guides the user through the two-stage analysis process.

QuantIntelli+ operates a unique two-stage workflow to deliver its insights.

• A user inputs match odds (e.g., Liverpool vs Chelsea 2.1 3.4 3.8).
• The pre-trained XGBoost model instantly processes the odds and outputs a baseline prediction (Home Win, Draw, or Away Win) with probabilities.
• This initial session is logged to Supabase, creating a unique ID for the match.

• The user toggles "Analysis Mode" and asks for a deeper dive.
• The RAG pipeline activates:
  • It generates targeted queries ("Liverpool injury news", "Chelsea recent form", etc.).
  • It dispatches these queries across multiple search providers (Tavily, Google) for comprehensive coverage.
  • Top results are "enriched" by fetching the full webpage content.
• The LLM synthesizes the data:
  • A meticulously crafted prompt is sent to Google Gemini, containing the statistical prediction, market odds, and all the enriched web context.
• The final report is generated:
  • Gemini returns a structured, multi-part analysis.
  • This analysis, including the extracted contextual outcome, is logged back to the original Supabase record.

• Python 3.9+
• Access to Google Gemini, Tavily, and Google Custom Search APIs
• A Supabase project

git clone https://github.com/your-username/quantintelli.git 
cd quantintelli

pip install -r requirements.txt

Create a .env file in the root directory and populate it with your API keys and credentials. Use .env.example as a template:

# Google Gemini API
GOOGLE_API_KEY="your_gemini_api_key"

# Supabase
SUPABASE_URL="your_supabase_project_url"
SUPABASE_SERVICE_KEY="your_supabase_service_key"
SUPABASE_PREDICTION_TABLE_NAME="your_table_name" # e.g., 'predictions'

# Web Search APIs (Optional, but recommended for full functionality)
TAVILY_API_KEY="your_tavily_api_key"
GOOGLE_API_KEY_CS="your_google_cloud_platform_api_key"
GOOGLE_CSE_ID="your_google_custom_search_engine_id"

Ensure your trained model and scaler files are placed in the model/ directory:

model/xgboost_model.pkl
model/scaler.pkl

python app.py

Navigate to the local URL provided by Gradio (e.g., http://127.0.0.1:7860) to start interacting with QuantIntelli+.

This tool is for educational and research purposes only. It is an exploration of hybrid AI systems and should not be used for actual financial betting. The predictions and analyses generated are not financial advice. Always gamble responsibly.

Contributions are welcome! If you'd like to contribute, please open an issue or submit a pull request. For questions or feedback, feel free to reach out via GitHub Discussions.