1. Home
  2. Guide
  3. Documentation

Deepchecks Documentation Guide: Checks, Suites and Examples

Updated November 16, 20265 min readDocumentation
Deepchecks documentation topics for data validation, model testing and monitoring

Deepchecks documentation is the official reference for every check, suite and object in the library, published at docs.deepchecks.com. Knowing how it is laid out helps you jump straight to a quickstart, an API page or a worked example instead of scrolling through pages that do not match your task.

The sections below walk through the main areas of the documentation, the core concepts it relies on, practical reading habits and what to do when a page does not answer your question.

How the Documentation Is Organized

The documentation is grouped by data type and by product. Tabular, NLP and computer vision each have their own quickstarts, usage guides and check galleries, while monitoring and LLM evaluation products keep separate documentation sections with their own setup steps.

Inside each area you will usually find a getting-started path, conceptual explanations, a gallery of checks with runnable examples, and an API reference generated from the source code. Version selectors let you read pages that match the release you have installed.

SectionWhat You Find ThereWhen to Use It
QuickstartsShort end-to-end notebooksFirst day with a new data type
Usage guidesDataset objects, conditions, customizingBuilding real validation pipelines
Checks galleryOne example per checkUnderstanding a specific result
API referenceClasses, parameters, return typesWriting or debugging code

Core Concepts Explained in the Docs

A handful of ideas appear on almost every page, and understanding them makes the rest of the reference much easier to read. The documentation introduces them early and links back to them from individual check pages.

Checks

A check inspects one potential problem, such as feature drift, label imbalance, duplicate samples or weak model segments. Each check returns a value you can process in code plus a visual display for notebooks and reports.

Conditions

Conditions turn a check result into a pass, warning or fail decision. Adding conditions is how you convert exploratory checks into automated tests that can stop a CI pipeline when a threshold is crossed.

Suites

A suite is an ordered list of checks, often with conditions attached, that runs together and produces one concluding report. Built-in suites cover common stages such as data integrity, train-test validation and model evaluation.

Finding the Right Check Quickly

The checks gallery is grouped by purpose, which is the fastest way to locate the right tool. Start from the question you need answered, then open the group that matches it and read the short description at the top of each example.

  • Data integrity: mixed types, null values, duplicates, conflicting labels.
  • Train-test validation: feature drift, label drift, new categories, leakage signals.
  • Model evaluation: performance reports, weak segments, calibration and overfitting signs.

Reading API Reference Pages

API pages list every constructor parameter with its default value. When a check behaves differently than expected, the reference usually explains why, for example a default sample size, a minimum category frequency or a specific drift method chosen for numeric columns.

Match the documentation version to your installed release. Parameter names occasionally change between versions, and reading the latest page while running an older package is a common source of confusing errors.

Running Documentation Examples Locally

Most gallery pages are generated from runnable scripts or notebooks that load small public datasets. Copying an example into your own environment is a fast way to see real output before pointing the same check at your private data.

from deepchecks.tabular.suites import data_integrity

result = data_integrity().run(my_dataset)
result.show()          # inside Jupyter
result.save_as_html("integrity.html")

Documentation for CI and Monitoring

Separate sections describe how to run suites inside continuous integration and how to deploy the open-source monitoring service. The monitoring quickstart covers installing the deployment helper, starting the system and onboarding a model so check results can be tracked over time.

These pages also explain where open-source features end and managed or enterprise features begin. Reading that boundary early helps teams plan hosting, licensing and access control before a production rollout.

Customizing Checks with the Docs

Usage guides explain how to adjust built-in checks and create custom ones. You can change parameters, attach your own conditions or write a new check class that plugs into the same suite framework, so custom logic still appears in the familiar report format.

Start by modifying an existing check before writing a new one. Small changes to thresholds and parameters cover many needs, and the example pages show the exact methods for each check.

Tips for Reading the Docs Efficiently

A few habits make the reference far quicker to use, especially when you are debugging under time pressure or evaluating whether a check fits your data.

  • Use the site search with exact check names copied from your code.
  • Open the gallery example before the API page to see real output.
  • Read the “conditions” part of each check page for testing ideas.
  • Bookmark the quickstart for each data type you use.

When the Documentation Is Not Enough

If a page does not answer your question, search the GitHub issue tracker for similar reports, then ask in the community channel mentioned in the README. Include your version and a short reproducible example to get faster, more accurate help.

Sharing the Docs Across a Team

Teams benefit from agreeing on a small set of documentation pages everyone uses. Linking the relevant quickstart and check pages in your internal README keeps new members on the same version and avoids repeated questions.

Frequently Asked Questions

It is published at docs.deepchecks.com, with separate sections for the testing library, monitoring and LLM evaluation.

Read the version that matches the package installed in your environment. You can check it with deepchecks.__version__.

Yes. Quickstarts and the checks gallery include examples built on public datasets that you can copy and run locally.

The usage guides cover suites, conditions and customization, and individual check pages show how conditions apply to each check.

Conclusion

Deepchecks documentation becomes far more useful once you know its structure: quickstarts for the first run, usage guides for pipelines, the gallery for individual checks and the API reference for exact parameters. Keep the version aligned with your install and you will find answers in minutes rather than hours.

Treat the documentation as a working companion rather than a one-time read, returning to the gallery whenever a new check appears in your reports.