> For the complete documentation index, see [llms.txt](https://docs.radar.citadel-ai.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.radar.citadel-ai.com/en/application-owner-guide/evaluation-library/evaluation-packages.md).

# Create Evaluation Packages

An Evaluation Package is a set that groups multiple Evaluation Metrics and each metric's runtime parameters.

When you create an Evaluation Package, you can preconfigure not only the metrics to use, but also each metric's action, borderline handling, and LLM model.

Created Evaluation Packages can be selected when applying metrics to an application. By saving frequently used metric configurations as Evaluation Packages, you can assign the same evaluation settings to multiple applications at once.

### What you can configure in an Evaluation Package

In an evaluation package, you can select which metrics to include and also configure their runtime parameters.

<table><thead><tr><th width="147.640625">Parameter</th><th>Overview</th></tr></thead><tbody><tr><td>Action</td><td>Select <code>Block</code> to control the AI system's behavior if an issue is detected in the input or output.<br>Select <code>Flag</code> to let the action proceed while marking the issue for later review.</td></tr><tr><td>Borderline handling</td><td>For borderline results, selecting <code>Pass</code> treats the case as acceptable, while <code>Fail</code> treats it as an issue.<br>Choose <code>Pass</code> to minimize false positives, or <code>Fail</code> if you want to prevent false negatives for this metric.</td></tr><tr><td>Remote LLM</td><td>Select <code>Fast Model</code> to prioritize response time and cost, or <code>High-Accuracy Model</code> to prioritize evaluation precision. This option is disabled if the selected metric is a <code>Local</code> type.</td></tr></tbody></table>

Defining these parameters in the evaluation package allows you to seamlessly apply the combination of metrics and their specific evaluation behaviors simultaneously to the target application.

### Creation procedure

To create an Evaluation Package, select `Evaluation Packages` from `Evaluation Library` in the sidebar.

The  procedure is the same as [/pages/BdoYnqjkdffClBJPWARx#id-1-b.-add-evaluation-metrics-individually](https://docs.radar.citadel-ai.com/en/application-owner-guide/evaluation-library/pages/BdoYnqjkdffClBJPWARx#id-1-b.-add-evaluation-metrics-individually "mention").

<figure><img src="/files/vk3u5std8Erl3dYxg0V2" alt=""><figcaption></figcaption></figure>

#### Example: Creating an Evaluation Package

You can create evaluation packages based on how and where they will be used. Common ways to group them include creating packages for:

* Specific use cases, such as internal or external applications.
* Different departments, like Development or Customer Support.
* Specific targets, such as input or output logs.
* Different environments, such as production or testing.
* Specific operational policies, like strict blocking versus flag-based monitoring.