---
title: Functions writer AI assistant
description: Use the functions writer AI assistant to generate, edit, and debug JavaScript code for data transformation, event, and visitor functions.
url: https://docs.tealium.com/server-side/functions/ai-assistant/
---

<blockquote>
This feature is only available to select customers. If you are interested in trying this feature, [contact support](https://docs.tealium.com/support/).
</blockquote>


Functions writer is a capability of the AI assistant built specifically for the functions code editor.

The functions AI assistant can:

* Write and edit function code.
* Generate test payloads.
* Debug and explain existing function code.
* Advise on best practices for performance, personally identifiable information (PII) handling, and error handling.

To open the AI assistant, click **Ask AI** in the functions code editor.

![](https://docs.tealium.com/images/server-side/functions-ai-assistant.png)

## Requirements

This feature requires the following:

* An account admin must enable the **Functions AI assistant** at the account level and profile level in **AI Settings**. For more information, see [Tealium AI features](https://docs.tealium.com/ai-features/).

## How it works

The functions AI assistant provides a dedicated chat session about the currently open function. Open the AI assistant and describe what you need in plain English. The functions AI assistant responds and generates JavaScript code, trigger rules, or test payloads for the current function.

The AI assistant always confirms its actions before making changes to your function configuration.

![](https://docs.tealium.com/images/server-side/functions-ai-assistant-response.png)

### Session context

The AI assistant automatically includes the context of the current function. This context includes the function code, active trigger rule, test payload, account and profile details, and defined attributes. While the function is open, the AI assistant persists the context, even if you close and re-open the chat window. 

Sessions expire after 30 minutes of inactivity or if you close the function window. Each function you open starts a new AI chat session. 

### Data transformation functions

The AI assistant reads the active runtime and generates code accordingly. When a data transformation function is open, the following constraints apply:

* The AI assistant only generates synchronous code. It excludes calls to `fetch`, `async`, `await`, Promises, and network I/O.
* Generated code modifies `event.data.udo` directly and does not reassign `event`.
* Generated code cannot write to attributes with the `tealium_` prefix.

For more information, see [About data transformation functions](https://docs.tealium.com/about-data-transformation-functions/).

### Event and visitor functions

Event and visitor functions execute in a runtime that supports async operations and outbound HTTP calls. When an event or visitor function is open, the AI assistant can generate code that:

* Makes outbound HTTP calls.
* Accesses stored auth credentials using `helper.getAuth()`.
* Re-sends events using `track()`.

For more information, see [Event and visitor functions](https://docs.tealium.com/event-visitor-functions-v3/).

## Use the functions AI assistant

To use the functions AI assistant:

1. Go to **Server-Side > Functions**.
1. Open a function or create one.
1. In the code editor, click **Ask AI** to open the AI assistant panel.
1. Describe what you want the function to do or ask questions about the function. The more specific your description, the more accurate the result. For examples, see [example prompts](#example-prompts).
1. Review the response and apply the changes, or continue chatting to make adjustments.

### Apply changes

When the AI assistant generates code or a test payload, it displays a button to confirm that you want to replace the current function or payload with the generated one.

Click **Use This Code In Function** or **Use This Test Payload** to replace all existing code or test payload.

## Example prompts

The following examples show how to describe tasks for each function type.

**Data transformation functions**

```wrap
Trim whitespace and lowercase the customer_email field.
```

```wrap
Hash customer_email with SHA-256 and store the result in customer_email_hashed. Lowercase the email before hashing. Keep the original field.
```

```wrap
Set customer_segment to high_value if order_total is greater than 500, otherwise set it to standard.
```

**Event and visitor functions**

```wrap
When tealium_event equals purchase, POST order_id, order_total, and customer_email from event.data.udo to https://api.example.com/orders as JSON. Use a Bearer token from auth key "orders_token". Skip the call if order_id is missing.
```

```wrap
Using customer_id from event.data.udo, call GET https://api.example.com/customers/{customer_id} with a Bearer token from auth key "crm_token". Extract loyalty_tier from the response and re-track it with track() alongside the original tealium_event.
```

**Test payloads**

```wrap
Generate a test payload for a purchase event with two products.
```

```wrap
Create a test payload where customer_email is missing, to test the null guard.
```

```wrap
Give me a test payload that satisfies the trigger rule you just created.
```

**Debug and explain existing code**

```wrap
Review my current code and identify any edge cases I have not handled.
```

```wrap
Explain how my current trigger rule works.
```

## Rate limits

Tealium tracks token usage at the account level across multiple AI features. If you exceed the account-level limit, you cannot use AI features until the limit resets. The AI assistant displays the reset time when you reach the limit.