Known Limitations and Workarounds

Review the following guidelines to ensure you don't encounter known issues with your configuration.

1. Filters aren't passed automatically into prompts.

   Because canvas and dashboard filters aren't automatically transposed into the prompt, consumer users must explicitly include these in the question or prompt. Auto transposing is under development for the Oracle Analytics AI Assistant, but not yet part of the product.

   For example, if a consumer user is watching a dashboard with sales data and the top filters for the canvas provide a default for visualizations to a given organization "XYZ" and a given region "ABC" he is responsible for.

   As this user asks questions of Oracle Analytics AI Assistant, the dashboard or canvas filters won't automatically be sent to their prompts. They may have to explicitly specify the value of the filters in the prompt, such as: "For organization XYZ and for region ABC".

2. Custom calendar behavior is based on calendar type.

   If your workbook source is leveraging custom calendars where a fiscal year doesn't match the Gregorian calendar (January through December), then you need to cautiously include time level objects in your local subject area for indexing. To do this, remove the date datatype column from your local subject area and add only the time hierarchy level columns (month, quarter, year) that are calculated for your custom calendar. For example, in the case of the Oracle Fiscal Calendar (June – May), quarter Q1 includes the months of June, July, and August.

3. Portability of AI Assistant metadata is limited.

   Moving indexes and synonyms across workbooks and subject areas is planned but may not be present in the release you're using.