BCFT-AI

AI Testing Agent

Python app created in Google Colab

The link to the app is as follows: https://colab.research.google.com/drive/1tBnudPYg9Ze7uriQwbwQESd5zl2UL4Y1

The AI Testing Agent streamlines software testing by using AI to compare expected vs. actual results through uploaded documents and screenshots.
Originally Designed for evaluating my AI Document search solutions, it delivers clear evaluations and concise, ready-to-paste summaries for reporting.
The AI agent helps developers make testing faster, smarter, and more consistent — all within the simple Google Colab interface. By taking the manual guesswork out of testing analysis, the agent allows developers to focus more on building and improving software rather than sifting through results.
It uses Gemini’s powerful language capabilities to offer thoughtful insights based on both context and content — making it especially helpful for testing apps with complex logic or AI-driven features. Whether you’re working solo or presenting your findings in a larger QA report, the ready-made summary feature ensures your conclusions are clear and shareable in seconds.
Designed with ease of use in mind, the workflow fits smoothly into any developer’s routine, and it’s fully adaptable for future testing needs.

Technical Details

📌 Project Summary

AI Testing Agent is an automated quality assurance tool designed to evaluate software test results using AI-powered comparison and summarization.
Originally built to test document search solutions developed with Vertex AI on Google Cloud, this tool enables intuitive evaluation using uploaded documents and screenshots.

🚀 Key Features


🧰 Technologies Used

Component Tech/Tool Used
AI Model Gemini API (via Google Cloud Workspace)
Environment Google Colab Notebook
UI Elements Colab Widgets (File Uploads)
Text Processing Python, Gemini Language Model
Image Handling PIL / OpenCV

📦 Setup Instructions

  1. Open the Colab Notebook
    Clone or open the project via Google Colab.

  2. Install Requirements (Optional)
    Colab includes most dependencies, but custom ones can be installed via standard !pip install commands.

  3. Upload Test Files
    Use the provided widgets to upload:
    • A PDF containing expected results
    • A screenshot of actual output (PNG or JPG)
  4. Run Evaluation
    Enter the query used during testing and execute the cell to generate both full and summary evaluations.

🧪 Example Use Case

Testing the accuracy and relevance of a document search query for a Vertex AI-powered app: