How to beginner · 3 min read

How to use AI to write SQL queries

Q: How to use AI to write SQL queries

Use an AI language model like gpt-4o to convert natural language descriptions into SQL queries by sending prompts to the chat.completions.create endpoint. Provide clear instructions and database schema context in the prompt to get accurate SQL code.

Quick answer

Use an AI language model like gpt-4o to convert natural language descriptions into SQL queries by sending prompts to the chat.completions.create endpoint. Provide clear instructions and database schema context in the prompt to get accurate SQL code.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install openai>=1.0

Setup

Install the OpenAI Python SDK and set your API key as an environment variable to authenticate requests.

bash

pip install openai>=1.0

Step by step

This example shows how to send a natural language request to generate a SQL query using the gpt-4o model. It includes the database schema in the prompt for context.

python

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

prompt = (
    "You are an expert SQL assistant. Given the following database schema:\n"
    "Tables:\n"
    "- users(id INT, name TEXT, email TEXT)\n"
    "- orders(id INT, user_id INT, amount FLOAT, order_date DATE)\n"
    "Write a SQL query to find the total order amount per user in 2025."
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": prompt}]
)

sql_query = response.choices[0].message.content
print("Generated SQL query:\n", sql_query)

output

Generated SQL query:
SELECT u.name, SUM(o.amount) AS total_amount
FROM users u
JOIN orders o ON u.id = o.user_id
WHERE o.order_date BETWEEN '2025-01-01' AND '2025-12-31'
GROUP BY u.name;

Common variations

You can use asynchronous calls, switch to other models like claude-3-5-sonnet-20241022, or add streaming to get partial results as they generate. Adjust the prompt to include more schema details or constraints for complex queries.

python

import asyncio
import os
from openai import OpenAI

async def generate_sql_async():
    client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
    prompt = (
        "Given tables users(id, name) and orders(user_id, amount), "
        "write a SQL query to get the average order amount per user."
    )
    response = await client.chat.completions.acreate(
        model="gpt-4o",
        messages=[{"role": "user", "content": prompt}]
    )
    print("Async generated SQL:\n", response.choices[0].message.content)

asyncio.run(generate_sql_async())

output

Async generated SQL:
SELECT u.name, AVG(o.amount) AS average_amount
FROM users u
JOIN orders o ON u.id = o.user_id
GROUP BY u.name;

Troubleshooting

If the AI generates incorrect or incomplete SQL, refine your prompt by adding more schema details or examples. Ensure your API key is set correctly in os.environ["OPENAI_API_KEY"]. For rate limits or errors, check your OpenAI dashboard and handle exceptions in code.

✅

Key Takeaways

Provide clear database schema context in prompts for accurate SQL generation.
Use gpt-4o with the OpenAI Python SDK for straightforward SQL query generation.
Async and streaming calls enable more flexible and responsive query generation workflows.

Verified 2026-04 · gpt-4o, claude-3-5-sonnet-20241022

Verify ↗