SQL Agent with LangGraph and Streamlit

Overview

This project implements a conversational SQL agent using LangGraph and Streamlit. The agent can:

• List available database tables
• Retrieve table schemas
• Generate SQL queries from natural-language questions with detailed reasoning
• Check generated SQL for common mistakes and rewrite if needed
• Execute queries against a SQLite database and stream results back to the user
• Support configurable filters to refine query results

Demo Video

Check out the SQL agent in action:

▶️ Watch Demo

Features

• Agentic workflow built as a LangGraph state graph
• Natural-language reasoning step separated from SQL generation
• Error-checking phase to catch SQL pitfalls (NULL handling, quoting, joins, etc.)
• Configurable loop between generate → check → run phases
• Streamlit UI for interactive chat and result display

Prerequisites

• Python 3.10
• Streamlit
• LangGraph (or your installed package)
• sqlite3 (standard Python library)

Installation

1. Clone the repository:

   git clone https://github.com/animesh1012/data_analyst_agent.git
   cd sql-langgraph-agent

2. Create a virtual environment and install dependencies:

   python -m venv .venv
   source .venv/bin/activate   # on macOS/Linux
   .\.venv\Scripts\activate  # on Windows
   pip install -r requirements.txt

Running the Agent

Launch the Streamlit app:

streamlit run SQL_AGENT_APP.py

Open the displayed URL (usually http://localhost:8501) in your browser. Ask your SQL questions in the chat—e.g. "Show me the top 5 countries based on covid cases"—and watch as the agent:

1. Lists tables
2. Fetches schemas
3. Plans its SQL in English (no code)
4. Generates and checks the SQL
5. Executes it and returns the results

Code Structure

├── SQL_AGENT_APP.py                             # Streamlit application entrypoint
├── sql_agentic_flow_cookbook.ipynb              # Experiment Jupyter Notebook
├── requirements.txt                             # Python dependencies
├── README.md                                    # Project documentation (this file)
└── covid_19_report.db                           # Sample SQLite database (You can use any db)
└── worldometer_data.csv                         # Sample data used in demo (You can use any db)