Data Cleaning Agent: Workshop Template

A hands-on workshop template where participants will build a Jupyter-based agent that leverages LangChain and OpenAI to perform data cleaning tasks. This template provides the foundation - we'll write the code together during the workshop!

What We'll Build

During this workshop, you'll learn to create an AI-powered data cleaning agent that can:

• Generate python (pandas) code to clean datasets
• Handle missing values intelligently
• Create automated data summaries
• Route user queries to appropriate cleaning functions

Prerequisites

• Anaconda or Miniconda installed
• Git (to clone this repo)
• An OpenAI API key set in your environment (OPENAI_API_KEY)

Setup

Step 1: Clone the Repository

git clone https://github.com/Tchanwangsa/Data-Cleaning-Agent_Workshop-Version.git
cd Data-Cleaning-Agent_Workshop-Version

Step 2: Create Conda Environment

Use Anaconda Prompt (Windows) or Terminal (macOS/Linux):

# Create and activate environment
conda create -n data-cleaning-agent python=3.11 -y
conda activate data-cleaning-agent

# Install Jupyter and kernel support
conda install jupyter ipykernel -y
python -m ipykernel install --user --name data-cleaning-agent --display-name "Data Cleaning Agent"

macOS/Linux: If you get "conda: command not found", run conda init and restart your terminal.

Step 3: Configure API Key

Copy the .env.example file or run:

- Windows: copy .env.example .env
- macOS/Linux: cp .env.example .env

Edit the .env file and add your OpenAI API key:

OPENAI_API_KEY=your_actual_api_key_here

Step 4: Launch Environment

Jupyter Notebook:

jupyter notebook

VS Code:

code .

Important: Select the "Data Cleaning Agent" kernel when opening notebooks. (This is the environment we created in step 2)

Step 5: Install Dependencies

Open main.ipynb and run the first cell:

# Install necessary packages
%pip install langchain openai pandas numpy matplotlib seaborn python-dotenv import-ipynb

Workshop Structure

During the workshop, we'll work through: 2. Setup & Introduction - Getting familiar with the tools we'll be working with 3. LLM Integration - Connecting to OpenAI for code generation 4. Feature Development - Building specific cleaning modules 5. Building Helper Functions - Creating data analysis helpers 6. Query Routing - Creating an intelligent dispatcher 7. Testing & Deployment - Putting it all together

Additional Resources

• Setup LLM call tracing with LangSmith
• Check available models and costs on OpenAI Pricing