LangGraph-Based Intelligent SQL Agent

LangGraph-orchestrated intelligent SQL agent that converts natural language questions into executable SQL queries, executes them on a relational database, and returns structured results via a FastAPI backend. This is a stateful SQL Agent that:

• Accepts natural language questions
• Translates them into SQL
• Executes queries on a real database
• Handles errors and retries using LangGraph control flow
• Exposes the agent via a FastAPI REST API

High level Flow

Why LangGraph?

Because:

• You need state
• You need conditional routing
• You may need retry loops
• You want deterministic control over agent behavior

Tech Stack

Syntax	Description
Layer	Technology
Language	Python 3.10+
Agent Framework	LangGraph
LLM	OpenAI / Azure OpenAI (configurable)
Database	SQLite (start simple)
Backend API	FastAPI
ORM / SQL	SQLAlchemy
Environment	.env

Repository Structure

langgraph-sql-agent/langgraph-sql-agent/
│
├── app/
│   ├── main.py              # FastAPI entry
│   ├── agent/
│   │   ├── graph.py         # LangGraph definition
│   │   ├── state.py         # Graph state schema
│   │   ├── nodes.py         # Agent nodes
│   │   └── tools.py         # SQL execution tools
│   │
│   ├── db/
│   │   ├── models.py        # SQLAlchemy models
│   │   ├── init_db.py       # Sample data
│   │   └── session.py
│   │
│   └── config.py
│
├── tests/
├── README.md
├── requirements.txt
└── .env.example