Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 12 additions & 9 deletions docs/ai_actions/work_with_ai_actions.md
Original file line number Diff line number Diff line change
Expand Up @@ -5,8 +5,14 @@
# Work with AI actions

AI actions define what results are available to editors in AI-enabled areas, such as, for example, the AI Assistant.
If AI Actions have been configured in your application, and you have the required [permissions]([[= developer_doc =]]/permissions/policies/#ai-actions), including `Action configuration/Edit` and `Action configuration/Create`, you can reconfigure the existing AI actions, and create new ones.
If AI Actions, including at least one service connector, have been configured in your application, and you have the required [permissions]([[= developer_doc =]]/permissions/policies/#ai-actions), including `Action configuration/Edit` and `Action configuration/Create`, you can reconfigure the existing AI actions, and create new ones.

!!! note "Variations in AI action settings"

The following description outlines the UI options displayed when either the default OpenAI connector or the Anthropic Claude [LTS update]([[= developer_doc =]]/ibexa_products/editions/#lts-updates) is installed and configured.

Response generation settings may vary depending on the AI service, model implementation, and action type, so the settings shown below might differ from those available in your installation.

Check warning on line 15 in docs/ai_actions/work_with_ai_actions.md

View workflow job for this annotation

GitHub Actions / vale

[vale] docs/ai_actions/work_with_ai_actions.md#L15

[Ibexa.EOLWhitespace] Remove whitespace characters from the end of the line.
Raw output
{"message": "[Ibexa.EOLWhitespace] Remove whitespace characters from the end of the line.", "location": {"path": "docs/ai_actions/work_with_ai_actions.md", "range": {"start": {"line": 15, "column": 1}}}, "severity": "WARNING"}
## View AI actions

With the right permissions, you can view all AI actions configured in the application by navigating to the Admin Panel and selecting **AI actions**.
Expand All @@ -19,7 +25,7 @@
- **Refine text** - used by default in [online editor](create_edit_content_items.md#ai-assistant) for refining text, for example: "Rewrite text in formal tone"
- **Generate alternative text** - used by default in the [image asset editing screen](upload_images.md#ai) to generate alternative text, for example: "Generate short alternative description of an image"

It may happen that a set of sample AI actions has been [installed with the AI actions package]([[= developer_doc =]]/ai_actions/install_ai_actions/#install-sample-ai-action-configurations-optional), and there is a number of existing AI actions that you can modify and clone.
It may happen that a set of sample AI actions has been [installed with the AI actions package]([[= developer_doc =]]/ai_actions/install_ai_actions/#install-sample-ai-action-configurations-optional), and there is already a number of existing AI actions that you can modify and clone.

!!! note "Custom action types"

Expand Down Expand Up @@ -52,22 +58,19 @@

The default request can be seen at the top of the settings area, on a light blue background.

- **Max tokens** - sets a maximum number of "[words](https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them)" or tokens that can be used in a single request by both the request and the response
- **Model** - decides what AI service model is used to generate the response

- **Max tokens** - sets a maximum number of "[words](https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them)" or [tokens](https://docs.claude.com/en/docs/about-claude/glossary#tokens) that can be used in a single request by both the request and the response

- **Length of prompt output** - sets a maximum number of words of the generated result

- **Temperature** - controls the randomness of the response.
Takes a value between 0 and 2, but the usual range is between 0 and 1.
The output is more random at higher temperatures.
For more information, see the parameter description in [OpenAI API reference](https://platform.openai.com/docs/api-reference/chat/create#chat-create-temperature)
For more information, see the parameter's description in [OpenAI API reference](https://platform.openai.com/docs/api-reference/chat/create#chat-create-temperature) or [Anthropic's glossary](https://docs.claude.com/en/docs/about-claude/glossary#temperature)

![AI action options](img/ai_action_options.png)

!!! note "Action settings availability"

Action settings differ depending on the AI service used, model implementation and actual action type.
Therefore the settings visible in your installation may vary from the ones presented above.

5\. Click **Save and close** to apply the changes or **Discard** to discard them and close the window.

## Create new AI actions
Expand Down