DPAssist

AI-Powered Portfolio Evaluation for Teachers

Built for educators who shouldn’t have to choose between providing meaningful feedback and having time to teach.

Live Production System

🌐 Teacher Dashboard: https://dpassist-edu.streamlit.app/

Dashboard Preview

Teacher Dashboard Submissions Dashboard
Teacher Dashboard Submissions Dashboard

📖 Open Teacher Setup Guide

🎥 Watch Summary Video

The cloud-hosted version of DPAssist is currently deployed and processing real student portfolio submissions.

The Problem

Digital portfolio assignments are among the most valuable — and most time-consuming — assignments a teacher can give. A single class of 25 students, each submitting a multi-page portfolio, can easily require 10–15 hours of reading, grading, and writing individual feedback.

Teachers who want to give students the chance to revise and improve before a deadline face an impossible tradeoff:

DPAssist solves this by automating the feedback loop while keeping the teacher in control of every final grade.

What It Does

DPAssist is a cloud-hosted portfolio evaluation system consisting of:

Teachers may also run the system locally if desired.

From the moment a student submits a portfolio link via Google Forms, the system takes over:

  1. Detects the new submission from Google Sheets
  2. Scrapes the portfolio website or repository
  3. Evaluates the work against the teacher’s rubric using Google Gemini
  4. Sends actionable feedback to the student before the deadline
  5. Creates a scored Gmail draft for the teacher after the deadline

The teacher uploads a rubric PDF once. After that, the workflow is automated.

Key Design Decisions

Evidence-First AI Evaluation

Most AI grading tools produce vague feedback that sounds encouraging but provides little instructional value.

DPAssist enforces strict evaluation rules:

If validation fails, the submission is flagged rather than producing inaccurate feedback.

Two-Phase Feedback Model

Phase Timing Recipient Contains
Student Feedback Before deadline Student Evidence-based improvement feedback without scores
Teacher Draft After deadline Teacher Full rubric scores, summary comments, and grading recommendations

By withholding scores before the deadline, students focus on improving their work rather than negotiating points.

Rubric Parsing

Teachers upload existing PDF rubrics with no required reformatting.

Gemini extracts:

A regex-based fallback parser provides additional resilience when AI extraction is unavailable.

School Network Compatibility

Many schools restrict IMAP access.

DPAssist automatically falls back from Gmail draft creation to direct teacher email delivery when necessary, preventing silent failures.

Architecture

Student submits Google Form Portfolio Link (via Google Form)
        │
        ▼
┌─────────────────────┐
│   Google Sheets     │
└────────┬────────────┘
         │
         ▼
┌────────────────────────────┐
│ Render Background Worker   │
│ background_service.py      │
└──────────┬─────────────────┘
           │
      ┌────┴─────┐
      │          │
      ▼          ▼
┌──────────┐ ┌──────────┐
│ Scraper  │ │ Rubrics  │
└────┬─────┘ └────┬─────┘
     │            │
     └─────┬──────┘
           ▼
┌─────────────────────┐
│ Gemini Evaluator    │
│ + Validator         │
└─────────┬───────────┘
          │
     ┌────┴─────┐
     │          │
     ▼          ▼
 Student      Teacher
 Feedback      Draft

Components

File Purpose
background_service.py Submission monitoring and processing
teacher_ui.py Streamlit teacher dashboard
evaluator.py Gemini AI evaluation and validation
rubric_parser.py PDF rubric extraction
rubric_manager.py Rubric storage and retrieval
scraper.py Portfolio content extraction
email_service.py Student feedback email delivery
gmail_drafts.py Teacher draft creation
google_sheets.py Submission and rubric storage
config.py Configuration management
orchestrator/ Evaluation pipeline
schemas/ Structured validation models

Technology Stack

Layer Technology
AI Evaluation Google Gemini
Teacher Dashboard Streamlit Cloud
Background Worker Render
Storage Google Sheets
Email Delivery Gmail SMTP / Gmail Drafts
Scraping BeautifulSoup4, Requests
Runtime Python 3.9+

Deployment Options

Production deployment uses:

No local installation is required for teachers.

Local Deployment

Advanced users may run DPAssist locally using Python.

Setup

See TEACHER_SETUP.md for complete installation and deployment instructions.

Requirements

Developer Quick Start

pip install -r requirements.txt
cp .env.example .env
python teacher_ui.py

Real-World Usage

DPAssist has been tested with actual student portfolio submissions at Charlotte Latin School.

Sample units successfully processed include:

The system tracks processed submissions to prevent duplicate evaluations and duplicate email delivery.

Production Status

DPAssist is currently deployed and processing live student submissions.

Production environment:

The production system consists of a Streamlit Cloud teacher dashboard and a Render-hosted background worker connected through Google Sheets.

Successfully verified in production:

Evaluation Rules

The AI evaluation layer enforces the following constraints:

MEETS / PARTIALLY_MEETS
    → Must include evidence quote(s)

NOT_YET
    → Must include zero evidence quotes

MEETS + LOW CONFIDENCE
    → Rejected by validator

HIGH CONFIDENCE
    → Allowed only when evidence is complete

All rubric criteria must be evaluated exactly once.

Validation failures stop the workflow rather than allowing incomplete or inaccurate feedback to be sent.

License

Proprietary — Educational Use Only

DPAssist — because teachers’ time is better spent teaching.