Quibble documentation
Quibble compares tabular prediction methods on the same held-out rows and reports quality, disagreements, execution costs, and replacement checks. Version 0.0.1 is an experimental pre-release of a local tool, not a hosted service.
Start with About Quibble for its purpose and limits, then choose a guide:
| Goal | Guide |
|---|---|
| Read the project story and engineering lessons | Predicting from Tables blog series |
| Install from source and run the offline demo | Getting started |
| Choose data, design a variant, and run in the browser | Experiment workbench |
| Compare your own CSV or Parquet | Testing your data |
| Automate runs | CLI reference |
| Interpret saved evidence | Reports, Swap checks |
| Find all guides and current validation boundaries | Documentation index |
| Diagnose a problem | Troubleshooting |
These Markdown guides can be read directly in a checkout or browsed as a GitHub documentation folder. Installation commands assume you are in the source checkout's root, not inside this folder. Source: ushakrishnan/quibble. The repository includes a GitHub Pages deployment workflow. A site URL is live only after deployment succeeds. The v0.0.1 GitHub pre-release is separate from publishing the Python package to PyPI.
In the running workbench, choose Docs for a site-style reader with grouped sidebar navigation, current-page highlighting, an on-page section index, breadcrumbs, and Previous/Next links. Blog articles have their own series navigation, separate from the documentation sidebar. Navigation collapses on smaller screens; the original Markdown remains directly browsable on GitHub.