jupyterlab-nitro-ai-judge 1.1.0

JupyterLab extension for submitting Nitro AI Judge notebook solutions

JupyterLab Nitro AI Judge

JupyterLab extension that adds a notebook toolbar button, Submit to Nitro AI Judge, and guides the user through login, contest selection, task selection, CSV submission, source attachment, and feedback review.

Features

• notebook toolbar button for direct submission flow
• login prompt backed by nitro-ai-judge-cli
• contest and task loading from Nitro AI Judge
• task statement and data download into the notebook workspace
• CSV file picker rooted at the current notebook directory
• source selection from either a file or the current notebook exported as Python
• automatic polling until scoring feedback is available
• results view with total score plus per-subtask score and metric values

Requirements

• Python 3.10+
• JupyterLab 4.x
• nitro-ai-judge-cli>=1.1.0 available in the same Python environment as JupyterLab

Install

python -m pip install jupyterlab-nitro-ai-judge

For local development in this repo:

npm install
python -m pip install -U pip build hatchling hatch-jupyter-builder jupyterlab
python -m pip install -e .

Development

Build the frontend bundle:

python -m pip install -U jupyterlab
npm run build:prod

Build the Python package:

python -m build

Usage

1. Open a notebook in JupyterLab.
2. Click Submit to Nitro AI Judge in the notebook toolbar.
3. Log in if prompted.
4. Select the contest and task.
5. Download the task data you need into the notebook workspace.
6. Pick the output CSV file.
7. Choose either a source file or the current notebook as the source attachment.
8. Submit and wait for feedback.
9. Review the total score and per-subtask metrics in the Nitro panel.

Publishing

Before publishing, replace the placeholder GitHub URLs in package.json and pyproject.toml.

PyPI release flow:

python -m pip install -U build twine
python -m build
python -m twine check dist/*
git tag v1.1.0
git push origin main --tags

The included GitHub workflows build on pushes and publish to PyPI on version tags.