Skip to main content

Managing Pega GenAI prompts for Pega Knowledge

A user with the Prompt Manager role creates Pega GenAI™ prompts that authors can use to obtain assistance as needed in Pega Knowledge™. This feature enables authors, with the help of Pega GenAI, to create high-quality content that meets organizational standards.

If the Prompt Manager does not create a prompt that accurately conveys the actions and results you want, Pega GenAI does not create entirely relevant results. Because the prompts created by a Prompt Manager can significantly impact a high number of authors and the content they create, you want those prompts to be as accurate as possible.

補足: Assign the Prompt Manager role only to users who are skilled at creating AI prompts.

Creating prompts

  1. In the navigation pane of the Authoring portal, click Configurations > Generative AI.
    補足: This menu option is available in the portal only if you have the Prompt Manager role.
  1. On the Generative AI landing page, click Add new prompt.
    The landing page is empty until you create the first prompt.
  2. In the Create prompt data dialog box, in the Article field section, click the radio button that corresponds to the type of prompt that you want to create.
  3. In the Prompt Label field, enter a label for the prompt.
    The label displays in the dropdown box when the author clicks AI options next to the title, abstract or content field during article creation.
  4. In the Prompt data field, enter the Pega GenAI prompt instructions.
  5. Click Submit.

Your prompt is now visible on the Generative AI landing page. You can click the prompt label to edit the prompt.

Testing prompts

  1. In the navigation pane of the authoring portal, click Knowledge > Content.
  2. On the Content landing page, click Create new content.
  3. Click AI options next to the field for which you created the prompt
  4. Select the prompt that you are testing from the dropdown list.
    Pega GenAI processes the request and shows the results in the Knowledge helper dialog box.
  5. Edit the title or prompt data, and then click Regenerate.
    This action helps you ensure that the prompt generates the correct output.
  6. Optional: Repeat step 5 as many times as necessary to obtain the desired output.
  7. When you are satisfied with the result, publish the prompt by clicking Accept.

The prompt is visible in the dropdown box next to the article element when an author clicks GenAI options. You can edit this prompt again later by clicking the Edit icon next to the prompt label on the Generative AI landing page.

Deleting prompts

  1. In the navigation pane of the Knowledge Portal, click Configurations > Generative AI.
  2. On the Generative AI landing page, click the Delete icon to the right of the prompt label that you want to delete.
  3. In the Delete prompt data dialog box, click Submit.

Best practices for creating prompts

The following list describes best practices that Prompt Managers can follow when creating prompts. Supplement this list with further best practices applicable to the specific conditions of your organization or business.

  • All prompt data must include a reference to the type of prompt that you are creating enclosed in curly brackets. For example, {TITLE}, {ABSTRACT} or {CONTENT_SOURCE}. Your prompt data is invalid without this reference.
  • The {TITLE}, {ABSTRACT}, and {CONTENT_SOURCE} references must be uppercase text.
  • {CONTENT_SOURCE} is the content in HTML format, which includes standard HTML tags and Pega-specific fields.
補足: Pega GenAI uses the HTML source when reformatting text, not the formatted text visible in the Editor.
  • Pega GenAI cannot process Pega-specific formatting for snippets, internal text, and article links. Because these fields might be available in the source text, it is important to configure the {CONTENT_SOURCE} prompts so that these fields are not modified (unless you want to change them). Specifically, do not change the following classes: km-snippet-ref, internalKC, and clickableKC.
  • Pega GenAI in Pega Knowledge has a limit on the number of tokens it can accept and return. When reformatting text, process smaller chunks at a time, as large amounts of text are unsuitable for Pega GenAI manipulation.
補足: The current version of Pega GenAI supports up to 8000 tokens.
  • The quality of the Pega GenAI output depends on the clarity and specificity of the instructions. Well-crafted prompts that effectively convey the desired task or context tend to produce more accurate and relevant results.

このトピックは、下記のモジュールにも含まれています。

トレーニングを実施中に問題が発生した場合は、Pega Academy Support FAQsをご確認ください。

このコンテンツは役に立ちましたか?

改善できるところはありますか?

We'd prefer it if you saw us at our best.

Pega Academy has detected you are using a browser which may prevent you from experiencing the site as intended. To improve your experience, please update your browser.

Close Deprecation Notice