An automated pipeline that clones any GitHub repository, executes its Jupyter notebooks in isolated environments, and scores reproducibility using the ReproScore framework (RRS, ROS, RCS).
Automated workflow from repository input to reproducibility scoring
Provide any public GitHub repository URL โ from any research field.
Clone the repository with shallow depth for efficiency.
Static analysis across 5 categories: Environment, Data, Documentation, Portability, Signals.
Detect required Python version from runtime.txt, .python-version, or setup files.
Create a per-repo pyenv + venv with all declared dependencies installed.
Run all notebooks via nbconvert, capturing outputs and errors per cell.
Diff original vs re-executed notebook outputs at the cell level.
Compute Reproducibility Outcome Score and Composite Score from execution evidence.
Live data from the latest pipeline runs โ updated on each push
Works on any machine with Python 3.10+ and pyenv
# 1. Clone the repo
git clone https://github.com/VasundharaShaw/ReproScore-Any
cd ReproScore-Any
# 2. (Optional) Set GitHub token for higher API rate limits
export GITHUB_API_TOKEN=your_token_here
# 3. Check environment
python3 -m pipeline setup
# 4. Score a repository interactively
python3 -m pipeline run --interactive
# 5. Or score multiple repos from the DB
python3 -m pipeline run --count 5
Score any GitHub repository directly in your browser
Explore reproducibility results across all scored repositories
The notebook covers score distributions, execution results, reproducibility metrics, and error analysis. To run with your own data, point it at your local output/db/db.sqlite.