Image2Test: Using ChatGPT to Build Manual Tests from Screenshots

This project provides tools and resources for generating, analyzing, and comparing manual test cases using similarity metrics. It includes a Jupyter Notebook for generating test cases, a Streamlit application for analyzing test similarity, and CSV files containing human-generated test cases.

Project Structure

1. Notebook

• Path: notebook/img2test.ipynb
• Purpose:
  • Generate manual test cases using GPT.
  • Save the generated test cases in Markdown format.
• Key Features:
  • Accepts a Google Drive link for an image.
  • Generates test cases based on a system and user prompt.
  • Saves the output along with metadata (system prompt, user prompt, and image link).

2. Streamlit Application

• Path: main.py
• Purpose:
  • Analyze the similarity of test cases using TF-IDF, BERT, and ROUGE-L metrics.
  • Provide both overall and row-by-row analysis of test cases.
• Key Features:
  • Upload HTML files containing test data.
  • Perform similarity analysis and visualize results.
  • Download analysis results as CSV files.

3. Human-Generated Test Cases

• Path: human_tests_csv/
• Purpose:
  • Store manually created test cases for various scenarios.
• Files:
  • human - SEARCH.csv: Test cases for search functionalities.
  • human - MAPS.csv: Test cases for map-related functionalities.
  • human - EMAIL.csv: Test cases for email functionalities.
  • human - ECOMMERCE.csv: Test cases for e-commerce functionalities.

How to Use

1. Generate Test Cases

• Open the Jupyter Notebook: notebook/generation_of_manual_tests_with_gpt.ipynb.
• Provide a Google Drive link to an image and customize the prompts.
• Run the notebook to generate test cases and save them as Markdown files.

2. Analyze Test Cases

• Run the Streamlit application: main.py.
• Upload an HTML file containing test data.
• Use the "Overall Analysis" tab for summary metrics and visualizations.
• Use the "Row-by-Row Analysis" tab for detailed comparisons of individual test steps.

3. Compare with Human-Generated Tests

• Use the CSV files in human_tests_csv/ as a reference for manually created test cases.
• Compare the generated test cases with these files to evaluate quality and coverage.

Requirements

Python Libraries

• streamlit
• pandas
• numpy
• matplotlib
• beautifulsoup4
• sklearn
• sentence-transformers
• nltk
• rouge

Additional Tools

• Jupyter Notebook for running the test generation notebook.
• Streamlit for running the similarity analysis application.

Future Enhancements

• Add support for more similarity metrics.
• Improve the visualization of analysis results.
• Automate the comparison of generated and human-created test cases.