The local evaluation loop
The Python runner keeps reference answers with the grader and passes only inputs to a candidate. Matched reports show what changed. Local tuning profiles and saved experiments stay on your computer.
What runs on this website?
GitHub Pages serves documentation and recorded JSON/HTML reports. The explorer loads those static files and renders the same result components as the local lab. To execute candidates, edit cases, or save experiments, start the Python lab locally.
The hosted viewer has no execution backend, model credentials, browser storage, analytics, or cloud integrations. AWS services are not part of this architecture.
Read the data contracts · How the site is built and deployed