> For the complete documentation index, see [llms.txt](https://docs.karini.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.karini.ai/guardrails/guardrail-cost-dashboards.md).

# Guardrail Cost Dashboards

### Guardrail costing

Guardrail Costing provides a structured framework to track and manage expenses generated by **Amazon Bedrock Guardrails** when they are used for content moderation and filtering. It surfaces real-time cost visibility across configured filtering mechanisms and helps users understand how spend is distributed across guardrail policy types.

Guardrails can be invoked from:

* **Prompt Playground**
* **Recipes**
* **Copilots**

Cost is calculated based on the guardrail evaluations performed through these experiences within the selected time range.

**Cost categories included in Guardrail Costing**

* **Content Filters**: Cost for applying content safety filters.
* **Denied Topics**: Cost incurred when configured topics are restricted or denied.
* **Word Filters**: Cost for checking content against configured word/phrase filters.
* **Sensitive Information**: Cost for filtering or blocking sensitive information.
* **Contextual Grounding**: Cost for grounding responses within allowed context.

### Costing and Billing Details

Amazon Bedrock Guardrails charges are incurred **only for the policy types configured in the guardrail**. 

Charging depends on whether content is blocked and where enforcement happens:

* **If a guardrail blocks the input prompt**
  * Charged for **guardrail evaluation of the input prompt**
  * **No charges for foundation model inference** (model call does not occur)
* **If a guardrail blocks the model response**
  * Charged for guardrail evaluation of:
    * **input prompt**, and
    * **model response**
  * Charged for **foundation model inference** because a response is generated prior to guardrail evaluation
  * Charged for the **model response that was generated prior to the guardrail’s evaluation**
* **If a guardrail does not block the input prompt or the model response**
  * Charged for guardrail evaluation of:
    * **input prompt**, and
    * **model response**
  * Charged for **foundation model inference**

### **Guardrail Dashboards**

The Guardrail Costing dashboard provides a consolidated view of guardrail spend and activity across the selected time range. It includes time controls and two primary metrics: 

1. **Total Cost** 
2. **Total API Requests**

<figure><img src="https://415930246-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F7ZrVuiAUMyuYVvrK5KaB%2Fuploads%2FgXFeGBmn2f8UThziIkmH%2Fimage.png?alt=media&amp;token=f8451bca-7473-4890-9dca-9a1ca8af82bc" alt=""><figcaption></figcaption></figure>

**Dashboard controls**

* **Date range (From/To)**: Defines the analysis window; all metrics update to reflect the selected period.
* **Granularity**
  * **Hourly**: Automatically displays data for the **previous day and the current day**
  * **Daily**: Displays data for the **last month up to the current date**
* **Clear filters**: Resets applied filters and restores the default view for the selected granularity.

**Total Cost**

* Displays the total guardrail-related cost incurred within the selected time range.
* Provides a breakdown of total spend by policy category:
  * Content Filters
  * Denied Topics
  * Word Filters
  * Sensitive Information
  * Contextual Grounding
* Reflects guardrail usage across Prompt Playground, Recipes, and Copilots.

**Total API Requests**

* Displays the total number of API requests processed by guardrails within the selected time range.
* Visualizes request volume over time to show usage patterns based on hourly or daily granularity.