🧪 AI Unit Test Generator

A command-line tool that automatically generates Python unit tests for your functions using a Large Language Model (LLM) via the Hugging Face Inference API.

---

📋 Table of Contents

• Overview
• Features
• Prerequisites
• Installation
• Configuration
• How to Run
• Usage Examples
• Limitations

---

Overview

This tool takes a single Python function as input, validates it, and sends it to an LLM (mistralai/Mistral-7B-Instruct-v0.2) to generate a complete, runnable unittest test file — including imports, edge cases, exception handling, and branch coverage.

---

Features

• ✅ Accepts input via file path, manual paste, or stdin pipe
• ✅ Validates that the input contains exactly one non-empty function
• ✅ Automatically strips comments before sending to the AI
• ✅ Generates tests covering all branches, edge cases, and exceptions
• ✅ Outputs a clean, runnable Python test file with no markdown

---

Prerequisites

Before running this tool, make sure you have the following installed:

• Python 3.8+
• A Hugging Face account with an API token → huggingface.co/settings/tokens

---

Installation

1. Clone the repository:

git clone https://github.com/your-username/ai-unit-test-generator.git
cd ai-unit-test-generator

2. Install dependencies:

pip install huggingface-hub python-dotenv

---

Configuration

1. Create a .env file in the root of the project:

touch .env

2. Add your Hugging Face token to the .env file:

HF_TOKEN=your_huggingface_token_here

---

How to Run

The tool supports three input methods:

Option 1: Provide a file path (recommended)

python app.py

Then select option 1 and enter the path to your .py file:

--- AI Unit Test Generator ---
1. Provide a file path
2. Input source code manually

Select an option (1 or 2): 1
Enter the path to your .py file: my_function.py

Option 2: Paste source code manually

python app.py

Then select option 2 and paste your function, then press:

• Ctrl+D on Mac/Linux
• Ctrl+Z then Enter on Windows

Option 3: Pipe input via stdin

cat my_function.py | python app.py

Saving the output to a file

cat my_function.py | python app.py > test_my_function.py

Then run the generated tests:

python test_my_function.py

---

Usage Examples

Example input (my_function.py):

import math

def calculate_circle_area(radius):
    if radius < 0:
        raise ValueError("Radius cannot be negative.")
    if radius == 0:
        return 0.0
    return math.pi * (radius ** 2)

Run the tool:

cat my_function.py | python app.py > test_circle.py
python test_circle.py

Expected output:

......
----------------------------------------------------------------------
Ran 6 tests in 0.001s

OK

---

Limitations

• Only supports one function per input — multiple functions will be rejected
• Empty functions (only pass, ..., or docstrings) are rejected
• Generated expected values for math-heavy functions should be manually verified, as LLMs can occasionally produce incorrect calculations
• Requires an active internet connection to reach the Hugging Face Inference API
• Free Hugging Face accounts may experience rate limiting on heavy usage