Skip to content

README.md

Latest commit

 

History

History
137 lines (104 loc) · 3.48 KB

README.md

File metadata and controls

137 lines (104 loc) · 3.48 KB

Data Analysis Agent

An AI-powered data analysis agent built with PydanticAI that allows you to analyze datasets using natural language queries. Ask questions about your data in plain English and the agent automatically generates and executes pandas code to extract insights.

How It Works

Natural Language Query
        |
        v
+------------------+
| PydanticAI Agent |  Receives dataset schema + question
+------------------+
        |
        v  generated pandas code
+------------------+
| Safe Executor    |  Runs code against DataFrame
+------------------+
        |
        v
+------------------+
| Result Formatter |  Produces readable text + tables
+------------------+
        |
        v
  AnalysisResult (code, answer, table)

The agent receives the dataset schema (column names, types, sample values) as context, generates a pandas expression or code block, executes it in a sandboxed environment, and returns a formatted result.

Project Structure

Data-Analysis-Agent/
├── main.py          # Entry point (CLI and interactive mode)
├── agent.py         # PydanticAI agent definition
├── executor.py      # Safe pandas code execution
├── data_loader.py   # CSV loading and schema extraction
├── schemas.py       # Pydantic models for input/output
├── config.py        # Configuration and constants
├── sample_data.csv  # Sample employee dataset
├── requirements.txt
├── .env.example
└── .gitignore

Setup

  1. Install dependencies:
pip install -r requirements.txt
  1. Set your OpenAI API key:
cp .env.example .env
# Edit .env and add your API key

Usage

Interactive mode

python main.py --data sample_data.csv

Single query

python main.py --data sample_data.csv --query "What is the average salary by department?"

With your own data

python main.py --data your_dataset.csv

Example Queries

Using the included sample_data.csv (employee dataset):

Ask: What is the average salary by department?
Code: df.groupby('department')['salary'].mean()

Ask: Who has the highest rating?
Code: df.loc[df['rating'].idxmax()]

Ask: How many employees are in each city?
Code: df['city'].value_counts()

Ask: What is the total salary of employees with more than 5 years experience?
Code: df[df['experience_years'] > 5]['salary'].sum()

Ask: Show the top 3 highest paid employees
Code: df.nlargest(3, 'salary')[['name', 'department', 'salary']]

Sample output

============================================================
DATA ANALYSIS RESULT
============================================================
Query:  What is the average salary by department?
Code:   df.groupby('department')['salary'].mean()
------------------------------------------------------------
Result: Series with 3 entries.

+----+--------------+---------+
|    | department   |  salary |
+====+==============+=========+
|  0 | Engineering  |  102400 |
+----+--------------+---------+
|  1 | Marketing    |   77000 |
+----+--------------+---------+
|  2 | Sales        |   69200 |
+----+--------------+---------+
============================================================

Configuration

Key settings in config.py:

Setting Default Description
MODEL_NAME openai:gpt-4o-mini PydanticAI model identifier
MAX_RETRIES 3 Retry attempts for LLM calls

License

MIT