/NaturalSQL

A demo to use natural language to query SQLite DB

Primary LanguageTypeScriptMIT LicenseMIT

NaturalSQL

A demo to using natural language to query SQLite DB

About

This repository contains code that combines SQLite querying with OpenAI's GPT-3.5-turbo language model. The code allows users to interact with a SQLite database using natural language queries through a chat-like interface.

Features

  • Connects to a SQLite database (chinook.db) using the sqlite3 library.
  • Utilizes OpenAI's GPT-3.5-turbo language model for natural language processing.
  • Supports querying the database by generating SQL queries based on natural language.
  • Retrieves table and column information from the database to assist with query generation.
  • Allows users to ask questions about the music database and receive SQL query responses.

Installation

  1. Clone this repository to your local machine:

    git clone git@github.com:JackZong/NaturalSQL.git

  2. Install the necessary dependencies:

    yarn install

  3. Configure the API key and base path by editing the config.json file:

    {
  "apiKey": "<your_open_ai_api_key>",
  "basePath": "<base_path>",
  // if you are using rapid open ai APIs, this demo based on Rapid API's open ai APIs
  "rapidApiKey": "<your_rapid_api_key>",
  "rapidHost": "<rapid_host>"
}

Usage

  1. Ensure that the SQLite database (chinook.db) is located in the ./db directory.

  2. Start the application:

    yarn start

  3. The application will connect to the database and retrieve the table and column information.

  4. The application will use OpenAI's GPT-3.5-turbo language model to generate an SQL query based on the hard-code natural language.

  5. The SQL query will be executed against the SQLite database, and the results will be displayed in the console.

Example

example

TODO

  • Supports to send message based on user inputs.

License

This project is licensed under the MIT License.

Acknowledgments