	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > LangChain > LangChain SQL Agent - Natural Language Database Queries

LangChain SQL Agent - Natural Language Database Queries

Author: Venkata Sudhakar

ShopMax India's analytics team wants to query the orders database using plain English instead of writing SQL manually. LangChain's SQL Agent lets non-technical staff ask questions like 'How many orders from Delhi were placed last month?' and automatically generates, executes, and interprets the SQL query against the actual database.

LangChain's create_sql_agent connects an LLM to a SQLDatabase object. The agent inspects the database schema, generates a SQL query for the user question, executes it, and interprets the result. It uses tools like sql_db_list_tables, sql_db_schema, sql_db_query, and sql_db_query_checker internally. The agent iterates if the first query fails, making it robust against schema variations and common SQL mistakes.

The example below creates an in-memory SQLite database with ShopMax India order data and uses the SQL agent to answer business questions in natural language.

from langchain_openai import ChatOpenAI
from langchain_community.utilities import SQLDatabase
from langchain_community.agent_toolkits import create_sql_agent
import sqlalchemy

engine = sqlalchemy.create_engine("sqlite:///:memory:")
with engine.connect() as conn:
    conn.execute(sqlalchemy.text("""
        CREATE TABLE orders (
            order_id TEXT, customer TEXT, city TEXT,
            product TEXT, amount INTEGER, order_date TEXT
        )
    """))
    conn.execute(sqlalchemy.text("""
        INSERT INTO orders VALUES
        ('ORD-1001', 'Priya Sharma', 'Delhi', 'Laptop', 65000, '2026-04-15'),
        ('ORD-1002', 'Amit Kumar', 'Mumbai', 'Headphones', 29990, '2026-04-20'),
        ('ORD-1003', 'Sneha Patel', 'Delhi', 'Smartphone', 45000, '2026-04-22'),
        ('ORD-1004', 'Ravi Nair', 'Bangalore', 'Tablet', 32000, '2026-04-25'),
        ('ORD-1005', 'Divya Menon', 'Delhi', 'Speaker', 8500, '2026-04-28')
    """))
    conn.commit()

db = SQLDatabase(engine)
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
agent = create_sql_agent(llm, db=db, verbose=False, agent_type="openai-tools")

result = agent.invoke("How many orders came from Delhi and what was the total revenue?")
print(result["output"])

It gives the following output,

There were 3 orders from Delhi with a total revenue of Rs 118,500. The orders were for a Laptop (Rs 65,000), Smartphone (Rs 45,000), and Speaker (Rs 8,500).

In production, restrict the agent to read-only database access by granting SELECT-only permissions - never connect the SQL agent to a write-enabled connection in production. Use include_tables to limit which tables the agent can see, preventing accidental access to sensitive tables like user credentials or payment details. For ShopMax India, create a separate read replica or a reporting schema for the SQL agent to query safely without impacting transactional performance.

Send your comments, suggestions or queries regarding this site to [email protected].