{% extends "base.html" %} {% block title %}About - AnZar{% endblock %} {% block content %}

AnZar

Version 2.5 ยท October 2026

AnZar is a personal project, not a product. It was written for one lecturer's own courses and is shared with colleagues who want to try what AI-assisted grading does in practice. There is no company behind it, there is nothing to buy and no account is needed.

The tool grades the quizzes and the file submissions of a Canvas course against criteria written by the teacher. The model proposes points and a short reason per criterion; the teacher accepts, adjusts or rejects them on a review page, and only the accepted scores go back to Canvas. It runs on local models (LM Studio, Ollama), on university-hosted endpoints, and on any endpoint that speaks the OpenAI API format.

Before using it in a course

A model touches student grades here, and student answers leave the machine when the endpoint is not local. Institutions set their own rules for automated assessment, and many require approval before it is used. So check the rules of your institution first, and make all parties aware that AI assessment is used and to what degree: students, co-lecturers and the exam board should know which parts a model grades, against which criteria, where the answers are sent and who reviews the scores before they become final. AnZar keeps the criteria and the grading runs on disk for exactly this kind of conversation. The responsibility for the grades remains with the teacher of the course.

Under the hood: Python with a pywebview window, the Canvas REST API and OpenAI-format endpoints. The frontend uses Alpine.js, MathJax and Phosphor icons.

{% endblock %} {% block extra_css %} {% endblock %} {% block scripts %}{% endblock %}