Connect LLM to SQL database with LangChain SQLChain

How to Tutorial for using LangChain SQLChain

  1. Beginning the Process

First, ensure that PostgreSQL is installed on your machine and that you have an OpenAI account. If necessary, establish a new Python virtual environment before proceeding.

make sure to install requirements.txt from :

you will find the .json file on the repo as well

run : 

!pip install -r requirements.txt

if you wanna create some data on your postgres server

import psycopg2
import json


import psycopg2
import json

# Load the JSON data from the file
with open("./data/my_database.json", "r") as f:
    data = json.load(f)

# Connect to the PostgreSQL database
conn = psycopg2.connect(
    host="localhost",
    database="postgres",
    user="user",
    password="password",
    port="5432"
)

# Create a cursor object
cur = conn.cursor()

# Create the table
cur.execute("""
    CREATE TABLE IF NOT EXISTS person (
        id SERIAL PRIMARY KEY,
        name VARCHAR(100),
        address TEXT,
        email VARCHAR(100),
        phone_number VARCHAR(30),
        birthdate DATE,
        job VARCHAR(100)
    )
""")

# Insert the data into the table
for person_id, person_data in data["_default"].items():
    name = person_data.get("name", "")
    address = person_data.get("address", "")
    email = person_data.get("email", "")
    phone_number = person_data.get("phone_number", "")
    birthdate = person_data.get("birthdate", None)
    job = person_data.get("job", "")

    cur.execute(
        "INSERT INTO person (name, address, email, phone_number, birthdate, job) VALUES (%s, %s, %s, %s, %s, %s)",
        (name, address, email, phone_number, birthdate, job)
    )


cur.execute("SELECT table_name FROM information_schema.tables WHERE table_schema = 'public'")
table_names = [row[0] for row in cur.fetchall()]

# Print the table names
print("Table names:")
for table_name in table_names:
    print(table_name)
# Commit the changes and close the connection
conn.commit()

and then run : 

from dotenv import load_dotenv
env_content = """
GROQ_API_KEY=your groq api key without ''
"""

# Write the content to a file named .env
with open('.env', 'w') as f:
    f.write(env_content)
# Load environment variables
load_dotenv(".env")

if you do not have groq api just make one here:

Set OpenAI API environment variables
import os
# Set OpenAI API environment variables (if needed)
os.environ["OPENAI_API_BASE"] = "https://api.groq.com/openai/v1"
os.environ["OPENAI_MODEL_NAME"] = "llama3-8b-8192"
os.environ["OPENAI_API_KEY"] =os.getenv("GROQ_API_KEY")

initialise your server again

from langchain_community.utilities import SQLDatabase

username="postgres"
password="password" # plain (unescaped) text
host="localhost"  # Hostname or IP address of the PostgreSQL server
port="5432"       # Port number
mydatabase="postgres"

pg_uri = f"postgresql+psycopg2://{username}:{password}@{host}:{port}/{mydatabase}"
db = SQLDatabase.from_uri(pg_uri)

then create your ChatOpenAI object

import os
from langchain.prompts.prompt import PromptTemplate

OPENAI_API_KEY = os.getenv("GROQ_API_KEY")
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(
    openai_api_key=OPENAI_API_KEY,
    model_name="llama3-8b-8192"
)

import this libs

from langchain_community.llms import OpenAI
from langchain_experimental.sql import SQLDatabaseChain,SQLDatabaseSequentialChain
from langchain.chains.llm import LLMChain
from langchain.prompts import PromptTemplate 
from langchain.chains.sql_database.prompt import DECIDER_PROMPT, PROMPT, SQL_PROMPTS

create a prompt

sql_chain = SQLDatabaseChain.from_llm(llm, db)

PROMPT = """ 
Given an input question, first create a syntactically correct postgresql query to run,  
then look at the results of the query and return the answer.  
The question: {question}
"""

Run a query

decider_chain = LLMChain(llm=llm, prompt=DECIDER_PROMPT, output_key="table_names")

class CustomSQLDatabaseSequentialChain(SQLDatabaseSequentialChain):
    def run(self, question):
        result = super().run(question)
        print(f"Query result: {result}")
        return result

db_chain = CustomSQLDatabaseSequentialChain(llm=llm, database=db, verbose=True, top_k=3,
                                            decider_chain=decider_chain, sql_chain=sql_chain, prompt=PROMPT)


question = "Whom the person has a name starting with 'S'?"
db_chain.run(question)

you will get result like

>Entering new CustomSQLDatabaseSequentialChain chain...

Table names to use: ['employees', 'person']

Finished chain.

Query result: Question: Whom the person has a name starting with 'S'? SQLQuery:SELECT "name" FROM employees WHERE "name" LIKE 'S%';

run this as well if you wanna display the results : 

from sqlalchemy import create_engine, text

# Assuming db_chain is your CustomSQLDatabaseSequentialChain instance
query = db_chain.run(question)

# Extract the SQL query from the combined string
sql_query = query.split('SQLQuery:')[1].strip()

# Establish a connection to your database
engine = create_engine(pg_uri)

# Execute the SQL query
with engine.connect() as connection:
    result = connection.execute(text(sql_query))

# Print the results
for row in result:
    print(row)

by jojo @ig : suho0066

Last updated