Under the Hood

How Kriterion Works

From assignment creation to final grades - here's every step, plus the architecture powering it.

The Workflow

STEP 01

Professor Creates Assignment

Define rubrics, upload test cases, set deadlines, and choose allowed languages. Starter code and attachments are stored on S3.

STEP 02

Student Writes Code

An in-browser VS Code-like editor with syntax highlighting, file explorer, and multi-file support. Or just upload files directly.

STEP 03

Submission & Compilation

Code is sent to a Celery worker, compiled inside a sandboxed Docker container. Instant feedback: compiled successfully or full error output.

STEP 04

Test Case Execution

Each test case runs against the student's code with time limits. Outputs are compared line-by-line. Partial credit is calculated automatically.

STEP 05

Plagiarism Detection

JPlag's Greedy String Tiling algorithm runs across all submissions. Matching code regions are identified with line numbers and similarity scores.

STEP 06

Review & Final Grade

Faculty review auto-grades, run student code interactively, override scores, and publish final grades - all from one unified grading interface.

Architecture

How Everything Connects

A multi-tier architecture with async task processing, sandboxed execution, and cloud storage.

Browser

Next.js App

FastAPI

REST API

Auth

JWT + RBAC

Celery

Task Queue

PostgreSQL

Primary DB

AWS S3

File Storage

Docker Sandbox

Code Execution

JPlag Engine

Plagiarism Detection

Submit → Queue

Student submits code via the API. FastAPI dispatches a Celery task with submission metadata.

Queue → Sandbox

Celery worker pulls code from S3, writes it to a temp directory, and runs it inside a Docker container.

Sandbox → Grade

Test results are captured, scores calculated, plagiarism checked, and the submission record is updated in PostgreSQL.

Tech Stack

Tools & Technologies

Every layer of Kriterion is built with production-grade open-source tools.

Frontend

Next.js 14

React framework with App Router, SSR, and file-based routing

TypeScript

Type safety across all components and API calls

Tailwind CSS

Utility-first CSS with shadcn/ui component library

React Query

Server-state management, caching, and automatic refetching

Backend

FastAPI

Async Python framework with OpenAPI docs and dependency injection

SQLAlchemy

ORM with relationship loading, migrations, and PostgreSQL support

Pydantic v2

Data validation and serialization for all request/response models

Celery + Redis

Distributed task queue for async grading and plagiarism jobs

Infrastructure

Docker

Containerized services and sandboxed student code execution

PostgreSQL

Primary database for courses, assignments, submissions, and grades

AWS S3

Object storage for starter code, solution files, and submission artifacts

JPlag

Java-based plagiarism engine using Greedy String Tiling (CLI subprocess)

See it for yourself.

Create an account and set up your first assignment in under five minutes.

Get Started