Text2SQL

Overview

Text2SQL is an interactive Streamlit application that allows users to query CSV data using natural language. The application converts user questions into SQL queries, executes them against a SQLite database, and displays the results.

Features

• CSV file upload and automatic SQLite database creation
• Natural language to SQL query conversion using the Ollama LLM
• Question caching for improved performance on similar queries
• Automatic query execution and result display
• Support for various data types, including dates
• Interactive Streamlit interface

Requirements

• Python 3.7+
• Streamlit
• Pandas
• SQLite3
• Langchain
• Ollama
• Scikit-learn

Installation

1. Clone the repository:

   git clone https://github.com/devbm7/text2sql.git
   cd text2sql
   

2. Install the required Python packages:

   pip install streamlit pandas langchain-community requests scikit-learn
   

3. Install Ollama following the instructions at Ollama's official website.

Setup

1. Ensure Ollama is installed and the llama3.1:8b model is available.

2. Start the Ollama server with the llama3.1:8b model:

   ollama run llama3.1:8b
   

Usage

1. Run the Streamlit application:

   streamlit run main.py
   

2. Open your web browser and navigate to the URL displayed in the terminal (usually http://localhost:8501).

3. Use the application:

   • Upload a CSV file using the file uploader.
   • Enter your question about the data in natural language.
   • View the generated SQL query and the query results.

How It Works

1. The user uploads a CSV file, which is converted into a SQLite database.
2. The user enters a question in natural language.
3. The application checks if a similar question has been asked before and uses the cached query if available.
4. If no similar question is found, the application uses the Ollama LLM to generate a SQL query from the natural language question.
5. The generated SQL query is executed against the SQLite database.
6. The results are displayed in a tabular format.

Limitations

• The accuracy of the SQL query generation depends on the capabilities of the Ollama LLM model.
• Large CSV files may take some time to process and load into the SQLite database.
• The application currently supports a single table per session.

Contributing

Contributions to improve Text2SQL are welcome. Please feel free to submit pull requests or open issues to suggest improvements or report bugs.

License

MIT License