Examples

This folder contains examples of how to use Evidently to evaluate, test, and monitor different types of AI systems: agentic workflows, LLM applications, and traditional ML models.

Start here

If you are new to Evidently, start with these three top-level tutorials:

1. Agentic Systems: Tracing and Evaluation

Learn how to trace multi-step agentic workflows, inspect intermediate steps, and evaluate agent behavior end to end.

2. LLM Validation: Input and Output Quality

Learn how to evaluate the quality of LLM application inputs and outputs, inspect results, and build validation workflows for LLM-powered systems.

3. Classic ML Validation: Data Quality and Drift

Learn how to validate datasets for traditional ML workflows, detect data quality issues, and analyze data drift between reference and current data.

Folder structure

cookbook Short, focused examples that demonstrate specific Evidently functionality or implementation patterns.

tutorials Longer end-to-end examples that combine multiple Evidently capabilities for realistic use cases.

service An example of how to run and serve the full Evidently system locally.

grafana Examples of using Evidently with Grafana dashboards. These examples show how to export Evidently metrics and visualize them in Grafana using a local setup.

datasets Sample datasets used in the examples.

How to use these examples

• Start with the three top-level tutorials for the quickest introduction.
• Use the cookbook for focused implementation snippets and feature-specific examples.
• Explore tutorials for more complete workflows and extended use cases.
• Use service if you want to run Evidently locally as a full system.
• Use grafana if you want to visualize Evidently metrics in external dashboards.

Notes

Some examples may depend on local datasets, external services, or optional Python dependencies. Please check the instructions inside each notebook or folder README for setup details.