SQL-Agent with LangGraph, LangChain and LLM

This repository contains a SQL Agent that converts natural language questions into SQL queries, executes the queries on a MySQL database (can be used on any database), and provides human-readable responses. It utilizes OpenAI’s ChatGPT, LangChain, and SQLAlchemy to interpret natural language inputs and interact with a database. Any LLM can be used instead of ChatGPT.

Features

• Converts natural language questions to SQL.
• Executes SQL queries on a database with pre-defined tables.
• Returns formatted, user-friendly query results.
• Checks if a question is relevant to the database schema.
• Provides error handling, retries, and generates alternative responses if needed.

Prerequisites

• Python 3.8+
• OpenAI API key

Installation

1. Clone the Repository

   git clone https://github.com/AjayKrishna76/SQL-Agent.git

2. Install Dependencies

   pip install -r requirements.txt

3. Set Up Environment Variables Create a .env file in the root directory to securely store sensitive information, like your OpenAI API key:

   OPENAI_API_KEY=your_openai_api_key

4. Database Configuration Configure your MySQL database connection by modifying the db_config dictionary in the code:

   db_config = {
   'host': 'localhost',
   'user': 'root',
   'password': 'password',
   'database': 'database'
   }

   Any database can be used here.

5. Create the workflow

This workflow diagram outlines the main steps of the SQL Agent:

1. check_relevance: Determines if the question is relevant to the database schema.
2. convert_to_sql: Converts a relevant question into an SQL query.
3. execute_sql: Executes the generated SQL query on the database.
4. generate_human_readable_answer: Translates the SQL query result into a user-friendly response.
5. regenerate_query: If the query fails or is incorrect, attempts to refine and regenerate it.
6. end_max_iterations: Stops the workflow if maximum retries are reached.
7. generate_funny_response: Provides a playful response if the question is not relevant.

• Each node represents a function, with arrows indicating the flow between stages. The workflow ends successfully at generate_human_readable_answer or end_max_iterations if retries are exhausted.