ShellNewHandler VS AutoLabSuite

AutoLabSuite's answer:

• The rubric is versioned with the course, so an old grade stays explainable after the criteria change.
• Editing a rubric never silently recalculates work that was already graded.
• It runs on the university's own servers and against its own GitLab.
• Reports are laid out by criterion, so a reviewer is not hunting through a PDF.
• The interface is in English, Russian, Polish and Belarusian.

AutoLabSuite's answer:

The work stays where the department already keeps it. Students submit through a merge request in the university's own GitLab, or on GitHub. We do not bring our own code hosting.

The criteria are readable before the work starts. They live as a file in the course repository, so a student sees the same rubric the grader applies, and any change to it shows up in the history.

A confirmed grade does not move on its own. Edit the rubric mid-semester if you need to; work that was already graded is not recalculated. Changing a grade is a deliberate act by the teacher, it requires a reason, and it is logged.

AutoLabSuite's answer:

• Brest State Technical University, Brest, Belarus

AutoLabSuite's answer:

Departments that teach programming and collect code from students. Most of the load falls on the lecturer grading lab work every week, so that is who the interface is built around. A department already keeping coursework on GitHub or its own GitLab does not have to move anything.

AutoLabSuite's answer:

I teach at Brest State Technical University: web technologies, distributed systems and cloud technologies, language processes of intelligent systems.

My problem was never that grading is dull. It was that a better lab costs more to grade. Every time I made a task harder or more interesting, the time to check one submission went up with it. Twenty minutes was the good case. An hour was normal. With around 500 lab submissions and 50 coursework projects in a semester, that arithmetic quietly decides what you are allowed to assign, and I was giving out easier labs than I wanted to.

In the autumn of 2025 the courses moved to GitHub. Not to build a product. Students have to learn pull requests anyway, the mechanism already exists, and it is better than anything I would invent in its place. A submission became a pull request, and the history of a work finally sat in one place.

The automation came after that. A lab now takes one to three minutes to check. That did not buy me free evenings. It let me raise the difficulty of the labs without paying for it in grading time, which was the point from the very beginning.

AutoLabSuite's answer:

Python, FastAPI, PostgreSQL, Alembic, Next.js, TypeScript, React, Playwright, Docker. Sign-in through OIDC, tested against Keycloak.