# AI Agents

## What is an AI Agent?

An **AI Agent** is a software program or system designed to operate autonomously or semi-autonomously by perceiving its environment, processing information, and taking actions to achieve defined goals. It can interact with humans, other systems, or the physical world.

## AI Agents in Data Pipeline

Within a data pipeline, an AI agent operates as a sophisticated component that automates key processes such as data ingestion, processing, and transformation, as well as complex decision-making. By leveraging AI and machine learning methodologies, it augments the efficiency of the pipeline, enabling the extraction of critical insights, optimization of workflows, and the initiation of autonomous actions.

AI Agents act as intelligent orchestration layers—automating tasks such as ingestion, preprocessing, anomaly detection, and triggering downstream actions.

## Role of an AI Agent in a Data Pipeline

### Data Ingestion & Preprocessing

* Collects data from multiple sources such as APIs, databases, and sensors.
* Cleans, normalizes, and transforms raw data before storage or further processing.

### Intelligent Data Processing

* Detects anomalies, missing values, or inconsistencies.
* Applies feature engineering or automated data labeling.

### Automated Decision-Making

* Executes ML model predictions for classification, clustering, or forecasting.
* Triggers alerts or initiates downstream processes based on real-time insights.

### Optimization & Orchestration

* Dynamically adjusts pipeline parameters based on real-time conditions.
* Improves data flow efficiency through intelligent resource allocation.

### Self-Learning & Adaptation

* Continuously improves accuracy and efficiency by learning from new data patterns.
* Refines decisions through iterative feedback loops.

## Agentic Execution Loop

AI Agents operate in a **loop-based execution cycle**:

1. **Perception** – Capture input data or environment signals.
2. **Reasoning** – Process information using AI/ML or LLM outputs.
3. **Action** – Call a function, trigger a tool, or run a workflow step.
4. **Observation** – Evaluate the results of the action.
5. **Iteration** – Repeat the loop until a satisfactory outcome is achieved.

This iterative cycle ensures adaptability and continuous improvement of pipeline workflows.

{% hint style="info" %}
*<mark style="color:green;">**Please note:**</mark> Agents are useful when you need an LLM to determine the workflow of an app. But they often overkill.*
{% endhint %}

## AI Agent Use Case

{% hint style="success" %}
*Check the illustration on how to use an AI Agent component in the pipeline workflow.*
{% endhint %}

{% embed url="<https://files.gitbook.com/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYaQlnE9b5yeAn1yiuRyq%2Fuploads%2Fp1y0TlNcq6VZBY5yTQ1A%2FAgent%20Component%20(1).mp4?alt=media&token=8e96c42a-8071-452c-890a-2568f2bb1c86>" %}

{% hint style="info" %}
**Please note:** This use case **demonstrates how to consume an exported Agent as a tool inside a pipeline** workflow. Users can also configure an  AI Agent component using the Data Pipeline interface.
{% endhint %}

## Exporting an AI Agent as a Tool

The **Export as Tool** feature allows users to register an AI Agent script as a reusable tool within the **Data Science Lab (DSL)**. This enables integration of agentic logic into other workflows and pipelines.

{% hint style="warning" %} <mark style="color:orange;">**Prerequisites**</mark>

Before exporting an AI Agent as a tool, ensure the following:

* You have permission to access the **DSL module**.
* At least one AI Agent is created under your user account.
  {% endhint %}

### Steps to Export an AI Agent as a Tool

{% stepper %}
{% step %}

### Access & Open the Data Science Lab

* Log in to the platform and navigate to the **Data Science Lab** module using the Navigation panel.
  {% endstep %}

{% step %}

### Access Agentic Tools

* Navigate to the **Agentic Tools** list page.
* Select an **active agentic tool** from the displayed list.

  <figure><img src="/files/hvmGwP2u1sHbAXI6tCYb" alt=""><figcaption></figcaption></figure>

{% endstep %}

{% step %}

### Open Workspace

* You will be redirected to the **Workspace interface** of the selected agentic tool.
  {% endstep %}

{% step %}

### Select Notebook

* From the **Repo folder**, select the notebook containing the agentic script.

{% hint style="info" %}
**Please note:** For newly created agentic projects, you must first create an agentic script in a notebook before exporting.
{% endhint %}
{% endstep %}

{% step %}

### Open Context Menu

* Click the **ellipsis (⋮) icon** next to the notebook.
* From the context menu, select **Register**.

  <figure><img src="/files/ho7vxjm4LvVxo6Z2ichr" alt=""><figcaption></figcaption></figure>

{% endstep %}

{% step %}

### Register Notebook

* The **Register** dialog box appears.
* Select the **Export as a Tool** option.
* Review the script preview displayed in the **Preview** section.
* Click **Finish** to complete the export process.

  <figure><img src="/files/k67cjR36TUgABTIxDrvu" alt=""><figcaption></figcaption></figure>

{% endstep %}

{% step %}

### Confirmation

* A success message confirms that the tool has been created and registered successfully.&#x20;

  <figure><img src="/files/2nMQrL39mWr2S8C55nBW" alt=""><figcaption></figcaption></figure>

- [x] Once registered, the exported tool becomes available in the Agentic Tools list for reuse across other projects and pipelines.
  {% endstep %}
  {% endstepper %}

## Consuming an Exported AI Agent

Once an AI Agent script is successfully registered as a tool, it can be integrated into a **Data Pipeline Workflow** using the **AI Agent component**. This allows the exported tool to participate in automated data flows alongside other pipeline components.

### Steps to Consume an Exported AI Agent

{% stepper %}
{% step %}

### Access the Data Engineering Module

* Navigate to the **Data Engineering** module.
* Open the **Pipelines list** page.
* Select a pipeline designed for agentic workflows.
* Example: The `mail_agent` pipeline containing an email listener and out-event components.

{% hint style="info" %}
**Please note:** If no pipeline exists, create a new pipeline before proceeding.
{% endhint %}
{% endstep %}

{% step %}

### Add the AI Agent Component

* Click the **Add Component/Event** icon.
* In the **Components/Events** panel, open the **Components tab**.
* Expand the **AI Agents** component category.
* Drag and drop the **AI Agent** component onto the canvas.
  * The component will auto-connect to the existing event component in the workflow.
    {% endstep %}

{% step %}

### Configure the AI Agent Component

* **Basic Information**
  * The **Basic Information** tab opens by default when the component is selected.
* **Meta Information**
  * Open the **Meta Information** tab.
  * Set the **Number of Outputs** and define the **Output Node** for the agentic component.
  * Create or add an event to the agent component.
    {% endstep %}

{% step %}

### More Configuration for Agent Component&#x20;

* **Agent Configuration**
* Configure the following Meta Information for the dragged Agent Component:
  * **Role:** Define the role of the agent.
  * **Description:** Provide details clarifying the agent's task.&#x20;
  * **Select Agentic Project:** Choose a registered agentic project from the dropdown list.
* **Task Details**
  * Navigate to the **Task Details** section and configure:
    * **Out Event**: Select the out event from the dropdown.
    * **Description**: Enter clear task instructions for the agent to execute.
    * **JSON Schema**: Upload a JSON schema that defines the expected output format.
  * Click **Save Component in Storage**.
    * A success message confirms the component properties have been saved.
      {% endstep %}

{% step %}

### Update and Run the Pipeline

* Click the **Update Pipeline** icon to apply changes.
  * A success message confirms the pipeline workflow has been updated.
* Run or **activate** the pipeline.
  {% endstep %}

{% step %}

### Monitor Execution

* **Component Status**
  * Open the **Component Status** tab to check the execution status of the agentic tool.

* **Log Panel**
  * Click the **Logs** icon in the toolbar to open the Log Panel.
  * Use the **Refresh** icon to update the logs.
    * A confirmation message indicates that the latest logs have been added.
    * Review the detailed runtime logs for the **AI Agent** in the **Logs** tab.
      {% endstep %}
      {% endstepper %}

* [x] By following these steps, users can successfully integrate an exported AI Agent tool within a Data Pipeline, configure its behavior, and monitor its execution in real-time.


---

# 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.bdb.ai/bdb-user-documentation/platform-modules/11.0/data-engineering/data-pipelines/pipeline-editor/pipeline-components/ai-agents.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.