> ## Documentation Index
> Fetch the complete documentation index at: https://docs.oumi.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# MANAGE EVALUATORS

> Save, edit, version, and reuse evaluators

An evaluator is saved once and reused across evaluations. Oumi keeps every version of it, so you can change an evaluator without losing track of how earlier evaluations were scored.

## SAVE AN EVALUATOR

In the evaluator form, click the `Save` icon, enter an **Evaluator name**, and click `Save Evaluator`. The evaluator appears on the **Evaluators** page.

To save a copy as a separate evaluator, open the menu next to the `Save` icon and choose `Save as…`. Use it to try a variation, such as a stricter prompt, without changing the original.

## EDIT AN EVALUATOR

Open the evaluator from the **Evaluators** page and click `Edit`. Each time you save your changes, Oumi creates a new version of the evaluator, so evaluations that already ran keep the version they used. You can't change an evaluator's type. To switch from an LLM evaluator to a rubric, for example, create a new evaluator.

The **Versions** tab lists every version of the evaluator. When you run an evaluation, the evaluator picker offers the latest version.

![Versions tab of the banking77-accuracy evaluator listing Version 2 and Version 1](https://mintcdn.com/oumi/5eryyJDD07V8QVa7/images/evaluators-versions.jpg?fit=max\&auto=format\&n=5eryyJDD07V8QVa7\&q=85\&s=11ff850fb429f0cf62f227a7275382d0)

## THE EVALUATOR PAGE

Open an evaluator to see its configuration and use it:

![Evaluator page for banking77-accuracy with the Edit, Preview, and Run Evaluation buttons, the evaluator's settings, and its prompt](https://mintcdn.com/oumi/5eryyJDD07V8QVa7/images/evaluators-evaluator-page.jpg?fit=max\&auto=format\&n=5eryyJDD07V8QVa7\&q=85\&s=c685b4ed7329d42af64efbab0201a519)

* **Overview:** The evaluator's type, judge model, prompt or criteria, data fields, and the evaluations that used it.
* **Preview:** For LLM and rubric evaluators, runs the judge on a few rows of a dataset. See [Test with Preview](https://docs.oumi.ai/guides/evaluators/judge.md#test-with-preview).
* **Run Evaluation:** Opens the evaluation form with this evaluator selected. It also fills in the evaluator's judge model as the **Assistant Model**, so change it to the model you want to evaluate.

## WITH THE API

Evaluators are also available through the API and the CLI. For the configuration of each type, see the [evaluator schema](https://docs.oumi.ai/reference/schema/evaluators.md) and the [evaluator recipe examples](https://docs.oumi.ai/reference/recipes/evaluators.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.