Nifty50GPT-Final: Offline Financial SQL LLM

Nifty50GPT-Final, developed by Student One Private Limited, is a 1.1 billion parameter model built on the TinyLLaMA architecture. It is uniquely designed to translate natural language queries into SQL, specifically for analyzing Indian stock market data. This model operates entirely offline, bundling a DuckDB database (student_data.duckdb) with over 10 years of historical data for NIFTY stocks, Indian indices, and global indices.

Key Capabilities

SQL Generation: Converts prompts into SQL queries executable on the included DuckDB database.
Offline Operation: Functions without internet access or external APIs, ensuring privacy and low latency.
Financial Data Access: Provides structured data for 50+ NIFTY stocks, various Indian indices (e.g., NIFTY50, SENSEX), and global indices.
Diverse Query Support: Handles fundamental metric lookups, year-over-year growth, CAGR calculations, OHLCV data, rolling averages, and NIFTY-wide aggregates.

Ideal Use Cases

Research & Experimentation: Perfect for academic or personal research into Indian financial markets.
Educational Purposes: A valuable tool for learning about financial data analysis and SQL generation.
Local Financial Analysis: Enables quick, private analysis of historical stock data without cloud dependencies.

Important Considerations

This model is provided for research and educational purposes only and does not offer financial advice. Users must verify all outputs, as the data may not be current, and the model is not certified by any financial authority. Use exact stock symbols and supported metric names to avoid hallucinations.