# AI Agents (Private Beta)

{% hint style="warning" %}
AI Agents is currently available in Private Beta.

If you’d like to enable it for your organization, or if you have any questions or feedback, please reach out to Select Star Support or email <support@getselectstar.com>.
{% endhint %}

### Overview

AI Agents help you keep metadata consistent and organized by automatically applying actions to tables, dashboards, and columns. Using natural language, administrators can define automation rules that run daily to ensure metadata stays accurate and up to date.

<figure><img src="/files/zQGhiqfo4LiuDaHBwPd7" alt=""><figcaption></figcaption></figure>

Key Features

* Create tagging rules using natural language
* Intelligent filter validation with preview
* Tags, Owners, Collections management
* Bulk operations on tables, dashboards, and columns
* Advanced hierarchical filtering (e.g., "tag tables that have a column that has a Tag Y")
* Daily automatic execution

### Getting Started

#### Prerequisites

* **Private Beta access**: Contact Select Star Support
* **Admin role**: Required to create, view, enable/disable agents
* **AI features enabled**: In Admin Panel

#### Basic Workflow

1. Enable AI features in the Admin Panel.
2. Go to the AI Agents section in the Ask AI page.
3. Create a new agent by writing tagging instructions in natural language.
4. Review the preview to confirm the agent’s interpretation.
5. Type **"save"**, **"confirm"**, or a similar instruction to save the agent and schedule daily execution.
6. Once created, the agent is enabled by default and added to a queue for execution. The first execution may take some time if other agents are ahead in the queue. After the initial run, the agent will execute daily.
7. On the AI Agents page, you can enable, disable, or delete agents.

#### Available Actions

| Action                      | Description                      |
| --------------------------- | -------------------------------- |
| **Add tags**                | Apply tags to matching assets    |
| **Remove tags**             | Remove specific tags             |
| **Assign owners**           | Set business or technical owners |
| **Add to collections**      | Add assets to collections        |
| **Remove from collections** | Remove from collections          |

#### Filter Examples

You can define rules based on different criteria, such as names, dates, lineage, popularity, or existing tags:

<table><thead><tr><th width="196.03125">Filter Type</th><th>Example Command</th></tr></thead><tbody><tr><td><strong>By name</strong></td><td>Find columns named "customer_email"</td></tr><tr><td><strong>By date</strong></td><td>List tables created in the last 30 days</td></tr><tr><td><strong>By lineage</strong></td><td>Find tables with downstream lineage</td></tr><tr><td><strong>By popularity</strong></td><td>List dashboards with popularity > 80</td></tr><tr><td><strong>By data source</strong></td><td>Tag all snowflake tables as "Tested</td></tr><tr><td><strong>Hierarchical</strong></td><td>Find tables that have a column with "email" in name</td></tr></tbody></table>

#### Notifications

Asset owners are receive notification when AI Agents update assets they own. We also maintain the history of updates on the object overview page.

<figure><img src="/files/imgTBaWbWvfm33rbE6rG" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/40wkJv4uW94uphLFIWWa" alt=""><figcaption></figcaption></figure>

#### Limitations

* **Supported asset types**: tables, columns, and dashboards
* **Supported filter attributes**: name, data source type, popularity, created on, updated on, lineage (existence only), and hierarchical filters
* **Date filters**: `updated on` and `created on` are available only for tables and columns. Dashboards do not support these filters.
* **Lineage filters**: Can only check if upstream or downstream lineage exists or doesn't exist. Specific lineage questions (e.g., "tables connected to specific dashboard X") are not supported.
* **Filter validation**: The system validates filters and provides helpful error messages when unsupported filters are used. If a filter isn't recognized, the AI will suggest corrections or alternatives.

### Step-by-Step Tutorial (Tag Agent)

#### 1. Navigate to AI Agents

* Open the **Ask AI** page.
* Click on the **AI Agents** item.

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

#### 2. Create a New Agent

* Click **Create Agent**.
* Enter a natural language instruction describing your tagging rule.

> List all tables with the name “Log”

<figure><img src="/files/cauooVFyuJ1CCtr9s087" alt=""><figcaption></figcaption></figure>

#### 3. If needed, update the filter.

> Update the filter to also include tables the name includes “analytics”.

**Note**: using “update” is important, otherwise the prompt will create a new filter with only the new instruction.

<figure><img src="/files/HXMGqV7eVuQSssiJmaLk" alt=""><figcaption></figcaption></figure>

#### 4. Add the tag

> Add "analytics" tag

<figure><img src="/files/0CCSR4IZFHjcNfMz7bQi" alt=""><figcaption></figcaption></figure>

#### 5. Save the agent

Type "save", "confirm", or a similar instruction to save the agent.

> Confirm

<figure><img src="/files/I01SnYTyAsGeol9XSItb" alt=""><figcaption></figcaption></figure>

Once the agent is running, you can track its progress in the list view and visit the chat to update the prompt.

<figure><img src="/files/KWJSwKcJGlp7H2BGmpsl" alt=""><figcaption></figcaption></figure>


---

# 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.selectstar.com/features/ai-tagging-agent.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.