231116/talk2data repository overview

⁠Overview

This project enables users to interact with databases using natural language queries, transforming them into executable SQL statements through a Generative AI model. It leverages Groq's LLM (deepseek-r1-distill-llama-70b) for SQL generation, integrated with a backend database and a frontend interface built with Streamlit.

⁠Key Features

• Natural Language Interface: Users can input plain English questions.
• SQL Generation: Converts natural language queries into SQL using Groq's LLM.
• SQL Validation: Ensures generated SQL is valid before execution.
• Token Usage Tracking (NOT WORKING): Monitors token usage per query for cost and performance insights.
• Real-Time Execution: Executes validated SQL queries on the connected database.
• Dynamic Schema Awareness: Adapts to the database schema for accurate query generation.

⁠Technologies Used

• Backend: Python, SQLAlchemy
• Frontend: Streamlit
• Database: SQLite (or any other supported by SQLAlchemy)
• LLM Integration: Groq's ChatGroq class
• Tokenizer: Hugging Face's transformers library for token counting