Using AI in Documentation

Learn how to use AI in the service editor.

Accessing the interface

Inside any documentation item's description, you can press / and select Ask AI from the list of commands that appear.

Tip: Write "AI" after typing slash for a quicker selection.

States of the AI

The AI can take any input and will aim to provide an as-meaningfull-as-possible response. Once an input is given, the AI will process the information received while formulating an answer. This is dependent on the nature of your request and the data available in your specification.

While the AI is thinking, you can cancel the request if it is taking too long. This will cancel only the last request in the conversation. But it will not close the conversation with the AI

Or discard the entire conversation, which will close the channel with the AI initiated there.

The answer UI

Once a reply is provided, it will come in the following format:

1. The reply of the AI

2. Input to continue the conversation based on the answer provided

3. Actions you can do for the answer provided by the AI

   1. You can replace your selected section with the AI answer

   2. You can insert it after your selection

   3. You can ask the AI to formulate a longer answer

   4. You can ask the AI to retry formulating an answer

   5. You can discard the conversation.

Additionally, you can review the entire conversation with the AI by clicking on the "clock" icon and previewing it in a "chat-like" functionality.

Writing new text

Inside any documentation item description, you can press / and select Ask AI from the list of commands that appear.

The AI can take any input and will aim to provide an as-meaningfull-as-possible response. Because the input was done on a new line, the AI will take the rest of the description as a bounded context for how to answer best your question or request.

Improving based on existing text

Another option to work with AI is by selecting with your cursor an existing text and then instantiating the AI upon it. This will make the bounded context of the answer more specific to your selection.