- Version: 2.67
- Release date: 11th August 2025
New Features and Enhancements
Ability to Modify the AI Content Filter
Due to customer feedback and the type of knowledge some of our customers manage and share with their users, we were asked if it was possible to modify the content filter and associated safe guards so that more sensitive data could be used within their document and LLM (Large Language Model) searches.
All customers are now able to modify the content filter within both the AI Document Search flow action and LLM assistant skill. By default, Tenjin uses a safe policy, but in specific circumstances some of our customers legitimately require a less safe option, which can now be enabled through the new Content Filter option; options are default (safe) and unsafe.
Updated Flow Actions
- Query Documents Using AI - New content filter drop-down within the advanced section of the flow editor.
- Chat Completion - New content filter drop-down within the advanced section of the flow editor.
Minor Updates / Bug Fixes
- Update Tenjin to use a new assistant registration type as recommended by Microsoft.
- Security update for 'predictable value range from previous values'.
- Resolve overlapping text at small screen widths when using 'act as' functionality.
- Remove unnecessary vertical scrollbar when viewing flow snapshots.
- Stop text bleed within flow debug when viewing long values with no space character or line break.
- Reduce the maximum number of QnA metadata values to 50.
- Make sure the QnA question field is highlighted when it has a validation error, not the QnA answer input.
- Update the Chat Completion help files in response to recent flow action name changes so that it is clear what flow action they refer to.
Share this article
Comments
0 comments
Article is closed for comments.