LangChain SQL Agent for Retail Data Analysis

This project demonstrates how to use a Large Language Model (LLM) as an intelligent agent to query a SQL database using natural language. The agent can understand questions, write and execute SQL queries, and return the results in a user-friendly format, including CSV files and plots.

The repository contains two main components:

1. fake_data_generator.py: A script to create a realistic SQLite database (lt_walmart_data.db) containing mock retail data (sales, products, inventory, etc.).
2. agent_with_sql.py: A script that runs a LangChain SQL agent which connects to the generated database to answer questions.

Features

• Natural Language to SQL: Ask complex questions about your data in plain English.
• Data Generation: Quickly bootstrap a sample database with realistic, interconnected data.
• CSV Export: Automatically saves query results with headers into a CSV file (outputs/result.csv).
• Automatic Charting: Generates a plot (outputs/chart.png) for time-series or categorical data.

Prerequisites

• Python 3.10+
• A Google Cloud Project with the Vertex AI API enabled.
• The gcloud CLI installed and authenticated on your local machine.

Setup & Installation

1. Clone the Repository

   # git clone <repository_url>
   cd agentic_ai_demo

2. Set Up Environment and Install Dependencies

   We recommend using uv, an extremely fast Python package installer.

   Option A: Using uv (Recommended)

   First, install uv if you don't have it:

   # On macOS / Linux
   curl -LsSf https://astral.sh/uv/install.sh | sh
   # On Windows
   powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

   Then, create the virtual environment (it will be named .venv) and install the dependencies:

   uv venv
   source .venv/bin/activate # On Windows, use: .venv\Scripts\activate
   uv pip install -r requirements.txt

   Option B: Using venv and pip

   If you prefer to use Python's built-in tools:

   python -m venv venv
   source venv/bin/activate
   # On Windows, use: venv\Scripts\activate
   pip install -r requirements.txt

3. Authenticate with Google Cloud Log in with your Google account to allow the application to use the Vertex AI services.

   gcloud auth application-default login

How to Use

Step 1: Generate the Database

First, run the data generator script. This will create a data/ directory and place the SQLite database file (lt_walmart_data.db) inside it.

python fake_data_generator.py

You will see output indicating the progress of data generation for each table. This only needs to be done once.

Step 2: Query the Data with the Agent

Now you can ask the agent questions about the data. Run the agent_with_sql.py script and pass your question as a command-line argument.

If you don't provide a question, it will use a default example: "Show revenue trend for the last 30 days".

python agent_with_sql.py "Your question in quotes"

Example:

python agent_with_sql.py "What are the top 5 best-selling products by total revenue?"

The agent will then:

1. Think about the question and construct a SQL query.
2. Execute the query against the lt_walmart_data.db database.
3. Parse the results.
4. Save the data to outputs/result.csv.
5. If applicable, create a chart and save it to outputs/chart.png.
6. Print a summary message to the console.

Example Questions to Try

• "Which 5 stores have the highest total revenue?"
• "Show me the daily sales revenue for the last 14 days."
• "What are the top 10 products with the lowest stock levels across all stores?"
• "Which store location has the most negative feedback comments?"