Xerte Community Documentation

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Overview ======

The **AI Assistant** is available for a large number of Xerte pages. In most cases, the AI Assistant contains inputs specific to that page and as such each AI Assistant is different. However, several settings //are// identical across all AI Assistants. We refer to these as **Base AI Generation Settings** and are explained on this page.

If you're interested in learning more about the AI settings of a specific page, as well as tips on how to configure the Base AI Generation Settings for these pages, please visit the following page: [[":Assistants:AI Assistant Per Page|AI Assistant Per Page]].

====== Base AI Generation Settings ======

Many interactivities that support automatic content generation use the same core settings. Their default values can be set at the **learning object** level by opening the **Base AI Generation Settings** assistant property.

{{:playground:ai_assistant_base_highlight.jpg?200}}

{{playground:ai_assistant_base_settings.jpg?400&direct |Base AI settings.}}


===== Reading Level =====

**Reading Level** helps the model decide how complex the generated language should be.

You can use:
  * simple descriptors, such as ''basic English'', ''beginner English'', or ''Advanced Dutch''
  * level-based values, such as ''C1 Level English''
  * combined values, such as ''Basic English A2 level''

You can also add extra descriptors for more control. For example:
  * ''A2 Level Beginner English, encountering words on this subject for the first time''
  * ''A2 Level Beginner English with advanced knowledge of [subject] terminology''

This setting is available at the **learning object** level, where you can choose a dropdown value or enter a custom one.

It can also be set per page. To do this, enable **Show Advanced Options** and enter a custom value there.

===== Education Level =====

**Education Level** helps the model judge the difficulty of the generated content.

You can use:
  * age ranges
  * school levels
  * professional or practical backgrounds

Examples:
  * ''University, advanced knowledge of subjects relating to maths'' produces more challenging content
  * ''Middle school'' produces less challenging content
  * ''Middle school, no knowledge of subject'' lowers the difficulty further

This setting is available at the **learning object** level, where you can choose a dropdown value or enter a custom one.

It can also be set per page. To do this, enable **Show Advanced Options** and enter a custom value there.

===== Tone and Style =====

**Tone and Style** tells the model what kind of writing style to use.

Examples:
  * ''Formal'' suits professional contexts
  * ''Semi-formal, active'' may feel more approachable for students
  * ''Semi-formal, active, excited'' produces more enthusiastic content

You can also request more specific styles, although results may be less consistent. For example:
  * ''as if spoken by a radio host''
  * ''Shakespearean''

This setting is available at the **learning object** level, where you can choose a dropdown value or enter a custom one.

It can also be set per page. To do this, enable **Show Advanced Options** and enter a custom value there.

===== Additional Instructions =====

**Additional Instructions** is an optional catch-all field available in most interactivities.

Use it when the standard fields are not enough to describe what you want. For example, when creating flashcards about birds, you could specify that the generated content should focus on:
  * bird anatomy
  * colours
  * physical characteristics of bird species

There is no fixed limit to this field, so you can be as detailed as needed.

Because this field accepts almost any kind of instruction, not every possible input combination can be tested. If generation gives unexpected results, try changing this field or temporarily leaving it empty.

===== Information Source =====

**Information Source** lets you choose what the model should use as context for generation.

Available options include:
  * Single File (Limited Scope) -- upload a file directly, and use that file as context for this generation
  * using all context resources -- use every file which has already been processed as context for this generation
  * using the learning object itself -- use the information in your latest learning object preview as context for this generation
==== Single File (Limited Scope) ====

When you upload a file directly in **Information Source**, the model will try to use only that file as its source.

This means the scope is limited to the specific file selected for that interactivity.

++++Exhaustive list of all supported file upload types|
The **Single File (Limited Scope)** option supports the following file types and link sources.

**Video Files**
  * MP4 (.mp4)
  * AVI (.avi)
  * MPEG (.mpeg)
  * QuickTime (.mov)
  * Other binary video files

**Audio Files**
  * MP3 (.mp3)
  * WAV (.wav)
  * OGG (.ogg)
  * Other MPEG audio files

**Document and Text Files**
  * PDF (.pdf)
  * Plain Text (.txt)
  * Microsoft Word (.doc, .docx)
  * Rich Text Format (.rtf)
  * OpenDocument Text (.odt)
  * Microsoft Excel (.xls, .xlsx)
  * CSV (.csv)
  * Microsoft PowerPoint (.ppt, .pptx)

**Video Links / Domains**
  * Vimeo
  * YouTube
  * PeerTube

With PeerTube, YouTube, or Vimeo, you can paste the link directly into the upload field.
++++
==== Learning Object ====

Using the **learning object itself** as context is useful when you want to generate content from material already added to the learning object, such as making a quiz based on text pages.

When selected, the system saves the current state of the learning object as if you had clicked **Preview**. This does **not** publish the learning object. That saved state is then used as context for generation.

==== All Context Resources ====

If you want to use multiple files, select **All Context Resources**.

This option uses all files synced at the learning object level in the **AI Context Resources/Links** table, found in the **AI Context Management** assistant which can be accessed on the learning object level. This table contains all of the files which have already been synced and are ready to use as AI context.

See the **AI Context Management** section for more information.

====== AI Context Management ======

**AI Context Management** is a learning object-level feature for organising files used as context for generated interactivities.

{{:playground:ai_assistant_context_highlight.jpg?200}}

{{:playground:ai_assistant_context_settings.jpg }}


===== Adding Files =====

To add a file, create a new entry in the grid using the **plus** icon.

Entries can be:

  * YouTube links
  * Vimeo links
  * PeerTube links
  * uploaded files for AI use (see [[ai_assistant_landing_page#single_file_(limited_scope)|single file upload section]] for supported sites and file types)

To upload a file, use **Import Media**.

These AI files are separate from normal media files.

You can also add:

  * a name (Title)
  * a note (Description)

These are only for your own organisation. They are **not** used by the model during generation.

After submitting an entry, the file may take a few minutes to process.


===== Syncing and Refreshing =====

Each time you open the page or change the table, it refreshes and shows all successfully synced files.

If a file does not appear after syncing, try adding it again.

===== Deleting Processed Files =====

To remove processed files linked to an entry, delete the associated row by selecting it, then using the **trash** icon.

You may need to wait briefly while the system removes the files.

Once an entry has been successfully processed and synced, you can delete the original uploaded file to save storage space, as long as you do **not** delete the entry itself from the table.

This is especially useful for large files such as videos.


===== Deleting Original Uploaded Files =====

To delete the original uploaded files, open the **Import Media** window while adding a new row and delete them there.

A dedicated overview for corpus files in the project properties window is planned for a future update.

===== Using Context Files in Generation =====

To make the model use these files as context, go to **Information Source** in the AI settings of each page and select one of the following:

  * **Use Context Resources** to use all synced context files
  * **Single File (Limited Scope)** to use one specific file

This will ensure that all of the synced files in the **AI Context Resources/Links** table are considered for that particular generation.