🍎 Teacher's AI Pre-Marking Agent

In this day and age, learners will use AI to their best of ability to complete assignments. As a fellow teacher myself, I would not say NO to using AI for homework, but I would love to know how they used it to help in their assignments and advise accordingly.

This web application helps teachers pre-mark assignments in Google Classroom by comparing student worksheets with their AI chat transcripts. It uses Google Gemini Pro to detect AI usage patterns and evaluate work against a specific rubric.

Disclaimer : As the author of this project (Assisted by Gemini-CLI), I don't recommend turning this into a fully automated Assignment Marking AI Agent. I believe that doing so would be greatly unfair to learners and doesn't not help on the process of nuturing future human minds that will need to work with AI. Every learner should have a chance to defend themselves in the spirit of fairness and transparency. Also, you may note that the AI can get things wrong, always practice discretion and empathy in learner assessments.

🚀 How to Start

1. Login: Open the app and click the "Log in with Google" button.
2. Select Course: Choose the course you want to grade from the dropdown.
3. Select Assignment: Choose the specific assignment.
4. Fetch Submissions: Click the button to load all "Turned In" work.
5. Review AI Pre-Marking:
   • Click "Analyze AI Usage" for each student.
   • The AI will read the PDFs and propose a grade and feedback.
   • You can adjust the grade and edit the feedback directly in the app.
6. Post Grades: Click "Post to Classroom" to send your approved grade back to the Google Classroom gradebook.

🖼️ Screenshots

User Interface

Features

AI Usage Analysis

☁️ Deploy to Google Cloud Run

To make this available to all your teachers via a URL:

1. Google Cloud Console Setup

1. Go to the Google Cloud Console.
2. OAuth Consent Screen: Ensure it's configured for "Internal" (if using Google Workspace) or "External".
3. Credentials:
   • Create an OAuth 2.0 Client ID.
   • Select "Web application" as the type.
   • Add https://your-app-url.a.run.app (your final Cloud Run URL) to Authorized redirect URIs.
   • Download the JSON and rename it to credentials.json in this project.

2. Deployment Command

Run this in your terminal:

gcloud run deploy teacher-helper --source . --set-env-vars "GOOGLE_API_KEY=your_key_here"

• Select a region (e.g., us-central1).
• Allow unauthenticated invocations if you want teachers to access it directly (they will still need to log in with their Google account inside the app).

3. Usage on Cloud Run

When the app starts, it will ask for a Redirect URI. Enter your public Cloud Run URL (e.g., https://teacher-helper-xyz.a.run.app).

📜 Rubric Applied

• UVP (20%)
• Product Description (15%)
• Benefits vs Features (15%)
• Demographics (10%)
• Customer Persona (15%)
• Evidence of AI Usage (25%) (Analyzes if AI was used strategically or just copy-pasted).

Special Thanks

This project was made during the Build with AI 2026 x IWD: Gemini CLI Hackathon ONLINE.

The Organizing Team at GDG Münster & GDG on Campus Universiti Teknologi Malaysia - Johor Bahru, Malaysia

Team Members Maria Eugenia Marquez Jimenez & Punitha Ramesh who have stayed strong for the entire effort, inclusive bugs, breakdowns, fails and all "429 RESOURCE_EXHAUSTED". You guys are the best and have helped me overcome the "WTF" moments and almost close to giving up.


Wan will never forget that final cheer when that screen managed to function properly.