GenAI Feedback Tool

An AI-powered grading pipeline that automates the evaluation of student blog posts using OpenAI's GPT models. Built to scale personalized, rubric-aligned feedback for around 100 students participating in Computing Talent Initiatives (CTI) open-source programs (Summer Open-Source Experience (SOSE) or Micro-Internship Experience).

---

Overview

The GenAI Feedback Tool is a system designed to solve the challenge of delivering timely, high-quality feedback to a large student population. It automates the full feedback cycle:

• 🧾 Extracts student IDs from Google Sheets
• 📥 Fetches submission data from Canvas LMS via paginated APIs
• 📝 Parses blog content from Google Docs
• 🤖 Evaluates responses using fine-tuned GPT models with rubric prompts
• 📤 Writes scores and feedback to Google Sheets
• 🎯 Post grades and comments to Canvas Gradebook
• Draft and send emails to students who received feedback.

---

Key Features

• Sheet-Aware ID Filtering
  Filters submissions using only student IDs listed in selected sheets (e.g., SOSE or Micro), improving performance with more students.

• Canvas LMS Integration
  Uses the Canvas Submissions API with pagination to collect metadata and submission history in bulk.

• Google Docs Parsing
  Grabs .txt content from Docs.

• OpenAI GPT Evaluation
  Uses fine-tuned GPT models and prompt engineering to produce rubric-aligned scores and feedback per section.

• Google Sheets Automation
  Writes all feedback and scores back to the source roster sheet for transparency and review.

• Email Drafting & Approval
  Auto-generates email drafts when feedback is ready and logs them in the sheet. Emails are only sent after instructor approval.

• Canvas Grade Push (In Progress)
  Push scores/comments back to Canvas submission.

• Cron-Job Ready
  Can be run on a schedule (every 30–60 minutes) for near real-time grading.

---

⚙️ Technologies Used

• Google Apps Script (JavaScript)
• OpenAI API (GPT-3.5 / fine-tuned model)
• Canvas LMS API (REST, with pagination)
• Google Docs API
• Google Sheets API

---

📁 Project Structure

File	Description
Grading.gs	Orchestrates grading per checkpoint
GenAIRating.gs	Handles GPT scoring logic using structured prompts
GetSubmissions.gs	Interfaces with Canvas & Google Docs
MicroRubric.gs	Rubric definitions for Micro-Internship
SOSERubric.gs	Rubric definitions for SOSE
EmailDrafter.gs	Email drafting, approval, and status tracking
pushToCanvas.gs	Pushes scores/comments to Canvas LMS
README.md	Project documentation and setup

---

🧠 How It Works

1. ✅ Select a Checkpoint + Sheet (e.g., SOSE and CH1, Micro-Int A and CH3)
2. 📬 GetSubmissions.gs collects all Canvas submissions and parses Google Docs content.
3. 🤖 GenAI_Rating.gs loads the correct rubric, calls the fine-tuned GPT model, and returns JSON feedback.
4. 📊 Grading.gs writes structured feedback and scores to the right columns in the sheet.
5. ✉️ EmailDrafter.gs checks for feedback completion and drafts personalized emails for students.
6. ✅ Instructor can review/edit and mark email for send-out.

---

🖼️ System Diagram