Data Analysis Agent

An AI agent that can load datasets, perform analysis, and create visualizations using Claude and LangChain.

What This Agent Can Do

This agent is like having a data analyst assistant that can:

• Load data from CSV and Excel files
• Analyze data using pandas (filtering, grouping, statistics)
• Execute Python code for complex analysis
• Create visualizations (scatter plots, histograms, bar charts, etc.)
• Answer questions about your data in natural language

How It Works

The agent has access to 5 specialized tools:

1. load_data: Loads CSV/Excel files into memory
2. list_datasets: Shows what datasets are currently loaded
3. get_dataset_info: Gets detailed info about a dataset
4. python_repl: Executes Python code for complex analysis
5. create_visualization: Creates plots and charts

The agent decides which tools to use based on your questions!

Setup Instructions

1. Create Conda Environment

conda create -n data-analysis-agent python=3.11
conda activate data-analysis-agent

2. Install Dependencies

pip install -r requirements.txt

3. Configure API Key

cp .env.example .env

Edit .env and add your Anthropic API key:

ANTHROPIC_API_KEY=sk-ant-api03-xxxxx

4. Run the Agent

python data_analysis_agent.py

Example Usage

Basic Analysis

You: Load the file sample_sales_data.csv

Agent: [Loads data and shows summary]

You: What are the top 3 products by total revenue?

Agent: [Uses python_repl to calculate]
The top 3 products by total revenue are:
1. Laptop: $489,200
2. Monitor: $234,900
3. Chair: $122,400

You: Show me a bar chart of revenue by region

Agent: [Creates visualization]
Visualization saved to: plot_sales_data_bar.png

Advanced Analysis

You: Calculate the average quantity sold per product category

Agent: [Executes analysis]
Average quantity sold by category:
- Electronics: 156.3 units
- Furniture: 58.7 units

You: Create a scatter plot showing price vs quantity sold

Agent: [Creates scatter plot]
Visualization saved to: plot_sales_data_scatter.png

Understanding the Code

Key Components

Datasets Dictionary:

datasets: Dict[str, pd.DataFrame] = {}

This global dictionary stores all loaded datasets so they're accessible to all tools.

Tools:

• Each tool is a Python function wrapped in a Tool object
• The agent reads the tool descriptions to decide when to use them
• Tools can call each other (e.g., python_repl can access loaded datasets)

Python REPL Tool: This is the most powerful tool - it lets the agent write and execute Python code for complex analysis. The agent has access to pandas, numpy, matplotlib, and all loaded datasets.

Sample Queries to Try

Data Loading

• "Load the file sample_sales_data.csv"
• "What datasets do I have loaded?"
• "Tell me about the sales_data dataset"

Analysis Questions

• "What's the average price by category?"
• "Which region has the highest total revenue?"
• "Show me products with quantity sold above 100"
• "Calculate correlation between price and quantity sold"

Visualizations

• "Create a histogram of prices"
• "Make a scatter plot of price vs quantity"
• "Show me a bar chart of revenue by month"
• "Create a box plot of quantity by region"

Complex Analysis

• "Find products where revenue exceeds $50,000"
• "Calculate the monthly growth rate for each region"
• "What percentage of total revenue comes from Electronics?"
• "Show me the top 5 products by revenue in the West region"

Troubleshooting

"Dataset not found": Make sure you've loaded the dataset first using load_data

"Column not found": Check the column names with get_dataset_info

Code execution errors: The python_repl tool has access to pandas, numpy, matplotlib, and seaborn - make sure your code uses these libraries

Plots not showing: Plots are saved as PNG files in the current directory - look for files like plot_*.png

Resources

• Pandas Documentation: https://pandas.pydata.org/docs/
• Matplotlib Gallery: https://matplotlib.org/stable/gallery/
• LangChain Agents: https://python.langchain.com/docs/modules/agents/
• Python REPL Tool: https://python.langchain.com/docs/integrations/tools/python