> ## Documentation Index
> Fetch the complete documentation index at: https://www.adaline.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Use roles in prompts

> Structure prompts using role-based messages with system, user, assistant, and tool roles

Every prompt in Adaline is built from messages. The Editor provides a structured, role-based interface for composing these messages — giving you precise control over how the model interprets your instructions.

## Roles and content blocks

Each message has two components:

1. **Role** — Specifies who is speaking in the conversation.
2. **Content block** — Contains the actual text, images, or other content.

<img src="https://mintcdn.com/adaline/-Bsa4BuK4fSdmy67/images/iterate/message-structure.png?fit=max&auto=format&n=-Bsa4BuK4fSdmy67&q=85&s=f5800ecad056a38de2e2d59ca0366b41" alt="Message structure showing roles and content blocks" title="Message structure showing roles and content blocks in Adaline" style={{ width: "100%" }} width="677" height="166" data-path="images/iterate/message-structure.png" />

Choose from the following roles:

| Role          | Purpose                                                                                                                               |
| ------------- | ------------------------------------------------------------------------------------------------------------------------------------- |
| **System**    | Sets the overall instructions, context, and behavioral guidelines for the model.                                                      |
| **User**      | Represents end-user inputs — the questions or requests sent to the model.                                                             |
| **Assistant** | Defines the model's responses. Use this in [multi-shot prompts](/docs/iterate/multi-shot-prompting) to demonstrate expected output format. |
| **Tool**      | Contains tool responses to tool calls. Used when building [prompts with tool interactions](/docs/iterate/use-tools-in-prompt).             |

The **role** determines how the model treats the message, while the **content block** holds what is actually sent — whether that is [text](/docs/iterate/use-text-in-prompt), [images](/docs/iterate/use-images-in-prompt), [PDFs](/docs/iterate/use-pdfs-in-prompt), or a combination.

## Next steps

<CardGroup cols={2}>
  <Card title="Use Text in Prompts" icon="text" href="/docs/iterate/use-text-in-prompt">
    Compose text content with variables, comments, and structured formatting.
  </Card>

  <Card title="Use Images in Prompts" icon="image" href="/docs/iterate/use-images-in-prompt">
    Add image inputs for multi-modal prompts.
  </Card>
</CardGroup>