LangChain SQL Agent: Apne Database se Baat Karein 📊

🔧 Prerequisites

Humein kuch libraries install karni hongi:

pip install langchain langchain-openai openai sqlalchemy

SQLAlchemy ek popular library hai jo Python ko SQL databases se connect karne me madad karti hai.

Step 1: Database Setup (SQLite) 🗃️

Hum ek simple in-memory SQLite database banayenge aur usme kuch data daalenge.

from langchain_community.utilities import SQLDatabase

# In-memory SQLite database banayein
db = SQLDatabase.from_uri("sqlite:///Chinook.db")

# Check karein ki connection hua ya nahi
print(db.dialect)
print(db.get_usable_table_names())

Explanation:

• Humne ek sample 'Chinook' database use kiya hai, jise aap online download kar sakte hain. Yeh music store ka data rakhta hai.
• SQLDatabase.from_uri() database se connection banata hai.
• get_usable_table_names() humein batata hai ki database me kaun-kaun si tables hain.

Step 2: LLM aur Agent Banayein 🤖

Ab hum ek Language Model (LLM) initialize karenge aur use database ke saath jodkar ek agent banayenge.

from langchain_openai import ChatOpenAI
from langchain_community.agent_toolkits import create_sql_agent

# OpenAI ka LLM initialize karein
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)

# SQL Agent banayein
agent_executor = create_sql_agent(
    llm=llm,
    db=db,
    agent_type="openai-tools",
    verbose=True
)

Explanation:

• ChatOpenAI se hum GPT model ko access kar rahe hain.
• create_sql_agent ek magic function hai jo LLM aur Database ko ek saath kaam karne ke liye taiyar karta hai.
• verbose=True se hum dekh sakte hain ki agent background me kya soch raha hai aur kaun si SQL query bana raha hai.

Step 3: Database se Sawal Puche! 💬

Ab hamara agent taiyar hai. Chaliye isse kuch natural language me sawal puchte hain.

# Agent se sawal puchein
response = agent_executor.invoke({
    "input": "List the total sales per country. Which country has the highest sales?"
})

print(response["output"])

Yeh agent aapke sawal ko samjhega, uske liye SQL query generate karega, database par chalayega, aur result ko aasan bhasha me aapko dega. 🎯

# Ek aur sawal
agent_executor.invoke({"input": "Show me all artists from the 'Rock' genre."})