# Amazon GenAI

{% hint style="info" %}
See the changelog of this Action type [here](/actions/amazon-genai.md).
{% endhint %}

{% hint style="warning" %}
Note that this Action is only available in Tenants with access to Amazon Bedrock. [Get in touch with us](/support/support.md) if you don't see it and want to access it.
{% endhint %}

## Overview <a href="#overview" id="overview"></a>

The **Amazon GenAI** Action allows users to enrich events by generating structured outputs using models hosted on **Amazon Bedrock**, such as Claude, Titan, or Jurassic.

{% hint style="warning" %}
In order to configure this action, you must first link it to a Listener. Go to [Building a Pipeline ](/the-workspace/pipelines/building-a-pipeline.md)to learn how this works.
{% endhint %}

<figure><picture><source srcset="/files/FSk5jqz6sVWOgtMe2OWU" media="(prefers-color-scheme: dark)"><img src="/files/FAdpM95SilPU0LATlV2r" alt=""></picture><figcaption></figcaption></figure>

## Ports

These are the input and output ports of this Action:

<details>

<summary>Input ports</summary>

* **Default port** - All the events to be processed by this Action enter through this port.

</details>

<details>

<summary>Output ports</summary>

* **Default port** - Events are sent through this port if no error occurs while processing them.
* **Error port** - Events are sent through this port if an error occurs while processing them.

</details>

## Configuration

{% stepper %}
{% step %}
Find **Amazon GenAI** in the **Actions** tab (under the **AI** group) and drag it onto the canvas.
{% endstep %}

{% step %}
To open the configuration, click the Action in the canvas and select **Configuration**.
{% endstep %}

{% step %}
Enter the required parameters:

<table><thead><tr><th width="231">Parameter</th><th>Description</th><th data-hidden></th></tr></thead><tbody><tr><td><strong>Region</strong><mark style="color:red;"><strong>*</strong></mark></td><td>Choose the Google Cloud location for AWS (e.g., <code>eu-central-1</code>).  Your region is displayed in the top right-hand corner of your AWS console.</td><td></td></tr><tr><td><strong>Model</strong><mark style="color:red;"><strong>*</strong></mark></td><td><p>Enter your Model ID or Model Inference Profile (arn) e.g. e.g., <code>arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-v2</code> </p><ul><li>Go to the <a href="https://console.aws.amazon.com/bedrock/home"><strong>Amazon Bedrock console</strong></a>.</li><li>Go to <strong>Model Access</strong> in the left sidebar.</li><li>You’ll see a list of available foundation models (FMs) like <strong>Anthropic Claude</strong>, <strong>AI21</strong>, <strong>Amazon Titan</strong>, <strong>Meta Llama</strong>, etc.</li><li>Click on a model to view its <strong>Model ID</strong> (e.g., <code>anthropic.claude-v2</code>) and <strong>ARN</strong> (e.g., <code>arn:aws:bedrock:us-east-1::foundation-model/anthropic.claude-v2</code>).</li></ul></td><td></td></tr><tr><td><strong>System Instructions</strong></td><td>Optional instructions to influence the behavior of the model (e.g., "You are a security analyst...").</td><td></td></tr><tr><td><strong>Prompt Field</strong><mark style="color:red;"><strong>*</strong></mark></td><td><p>Select the field in the event containing the prompt to send to the model. Must be <code>string</code>. This field will be sent as-is to the model. </p><p></p><p>Amazon Bedrock models support both English and multilingual prompts, depending on the model selected.</p></td><td></td></tr><tr><td><strong>Temperature</strong></td><td>Adjusts randomness of outputs: greater than <code>1</code> is random, <code>0</code> is deterministic, and <code>0.75</code> is a good starting value. Default value is <code>0.1</code></td><td></td></tr><tr><td><strong>Max Tokens</strong></td><td>Maximum number of tokens to generate. A word is generally 2-3 tokens. The default value is 128 (min 1, max 8892).</td><td></td></tr><tr><td><strong>Top P</strong></td><td><strong>Top P</strong> sets a probability threshold to limit the pool of possible next words. Whereas <code>temperature</code> controls <strong>how random</strong> the selection is,<code>top_p</code> controls <strong>how many options</strong> are considered. Range: <code>0–1</code>. Default is <code>1.0</code>.</td><td></td></tr><tr><td><strong>JSON credentials</strong><mark style="color:red;"><strong>*</strong></mark></td><td>Provide the secret JSON credentials used to authenticate against Amazon Bedrock.</td><td></td></tr><tr><td><strong>Output Field</strong><mark style="color:red;"><strong>*</strong></mark></td><td>Give a name to the output field that will return the evaluation.</td><td></td></tr></tbody></table>
{% endstep %}

{% step %}
Click **Save** to complete.
{% endstep %}
{% endstepper %}

{% hint style="info" %}
Use conditional logic upstream to prevent sending unstructured or non-informative prompts to the model, helping to optimize costs and relevance.
{% endhint %}

## Example

Read our use case to learn how to use this Action in a real cybersecurity scenario.

<table data-view="cards"><thead><tr><th></th><th data-hidden data-card-cover data-type="files"></th></tr></thead><tbody><tr><td><a href="/spaces/lMswUMhL1LeEvusY1XNC/pages/YmdTjvTNc3LFCyJt16xV">Using Amazon GenAI to classify HTTP logs</a></td><td><a href="/files/j5qSI8DALlfqOTdTDFBk">/files/j5qSI8DALlfqOTdTDFBk</a></td></tr></tbody></table>


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.onum.com/the-workspace/pipelines/actions/ai/amazon-genai.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.