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

# Authoring – AI Assistant

> Use the in-editor AI Assistant to build, modify, and test decision models in natural language — it can edit the diagram, generate test inputs, and answer questions about the model.

The in-editor AI Assistant helps you build, modify, and test decision models in natural language — it can edit the diagram, generate test inputs, and answer questions about the model.

<img src="https://mintcdn.com/aletyx-3353d50c/-SprOR0k3X7EKq06/images/decision-control/how-to/ai-assistant.png?fit=max&auto=format&n=-SprOR0k3X7EKq06&q=85&s=f293a101cb1de309e22f5a11a9c758b7" alt="AI Assistant panel open in the editor" width="1920" height="1037" data-path="images/decision-control/how-to/ai-assistant.png" />

<Info>
  **Required to use the assistant**

  All how-tos on this page require:

  * **Access:** Authoring
  * **Deployment:** the AI Assistant must be enabled by your administrator
  * **Credential:** a valid AI token configured in **Settings → AI Assistant**
</Info>

<Note>
  **History is per workspace**

  The assistant's chat history and pending queue are stored per workspace — switching to a different model shows that model's own conversation, not a single global one.
</Note>

## How to open the AI Assistant panel

Open the in-editor assistant to get help building or analyzing a model.

**Steps:**

1. With a model open, click the floating **Aletyx AI** trigger button (bottom-right of the editor).

**Verify:** The assistant panel opens with a message box; if no token is set, a "no token" prompt appears instead.

## How to ask the AI Assistant to create or modify a model

Describe a change in natural language and let the assistant draft DMN edits.

**Prerequisites:** AI Assistant panel open and a valid AI token (**Settings → AI Assistant**).

**Steps:**

1. Type your request (for example, "Add a decision node called Eligibility") in **Describe the changes you want to make…**.
2. Press <kbd>Enter</kbd> or click **Send**.
3. Wait for the response (the editor locks while the model is being changed).

**Verify:** The assistant replies; for model changes, a diff is offered for review.

<Tip>
  **Be specific**

  Mention the node names, data types, and conditions you want. "Add an `Eligibility` decision using `Applicant Age` and `Credit Score`" gets a much better result than "add eligibility logic".
</Tip>

## How to attach a file to the AI Assistant

Give the assistant supporting documents (PDF, CSV, images, spreadsheets, etc.).

**Prerequisites:** AI Assistant open; file attachments enabled by your administrator; file ≤ 2 MB.

**Steps:**

1. Click the paperclip (attach) button, or drag a file onto the open panel.
2. Confirm the file badge appears, then send your message.

**Verify:** The file badge shows under the input and the assistant's reply reflects the attachment.

## How to review and apply AI-suggested model changes

Approve or discard the model edits the assistant proposes.

**Prerequisites:** An assistant reply that includes model changes (diff opened).

**Steps:**

1. Inspect the diff overlay in the editor.
2. Accept or revert the proposed changes.

**Verify:** Accepted changes appear on the canvas; reverting leaves the model unchanged.

## How to load AI-generated test inputs into the DMN Runner

Move sample inputs the assistant produced straight into the runner.

**Prerequisites:** An assistant message containing JSON test inputs; DMN Runner available.

**Steps:**

1. On that message, click the **Test** button (play icon).

**Verify:** A "Test inputs loaded into DMN Runner." toast appears and the runner is populated.

## How to manage AI Assistant requests (queue, cancel, resend)

Control in-flight and failed assistant requests.

| You want to…                                      | Do this                                                                 | Result                                                       |
| ------------------------------------------------- | ----------------------------------------------------------------------- | ------------------------------------------------------------ |
| Send a follow-up while a reply is still streaming | Type and send another message — it's marked **Queue** ("Next in queue") | Messages run in order, one after the other                   |
| Remove a queued message before it runs            | In the **Next in queue** preview, click the **✕** on the queued item    | The message is dropped from the queue and not sent           |
| Abort the current request                         | Click **Cancel** next to "Thinking…"                                    | The current request stops; nothing is committed to the model |
| Retry a failed message                            | Click the resend (↻) icon on the failed user message                    | The same prompt is sent again as a new request               |

## How to clear the AI Assistant chat history

Wipe the assistant conversation for the current workspace.

**Prerequisites:** AI Assistant open with at least one message.

**Steps:**

1. Click the trash icon in the assistant panel header.
2. Confirm **Clear History** in the dialog.

**Verify:** The conversation empties.

## How to configure the AI Assistant token

Store the credential the in-editor assistant needs.

**Prerequisites:** AI SaaS integration enabled.

**Steps:**

1. Open **Settings** (masthead cog) → **AI Assistant** (or click **Open Settings** in the no-token prompt).

   <img src="https://mintcdn.com/aletyx-3353d50c/-SprOR0k3X7EKq06/images/decision-control/how-to/ai-token-settings.png?fit=max&auto=format&n=-SprOR0k3X7EKq06&q=85&s=0d6f203c05b27a9f7c24e53540f4572e" alt="AI Assistant token settings" width="1895" height="922" data-path="images/decision-control/how-to/ai-token-settings.png" />

2. Paste your token and validate it.

**Verify:** The token validates successfully and the assistant becomes usable.