Release Notes

Threads

This release enables Vision capabilities for Groups using Azure as their AI provider through Image Descriptions provided by another model. (752)
Access to this feature is controlled by a new option in the Create/Edit Assistants page visible only in Groups using Azure as their AI provider and when selecting a model that ordinarily supports Vision. (752)
The feature is tagged as experimental and is the only feature that can be toggled in locked assistants, besides the assistant's published status. (752)
When Vision capabilities through Image Descriptions are enabled, PingPong will generate an image description of each uploaded image and append it to the user's message before adding it to the Thread. (752)
Image descriptions are filtered out of the user message before displaying in the Thread UI. Images are then displayed in the file attachments section below each user message, like other files. (752)
When the feature is enabled, a new “Experimental Vision Support” banner replaces the current “No Vision Capabilities” banner in Thread File Uploads. A user-facing Modal accessible by clicking on the banner provides more details about the feature and a link to a feedback form.

Fixes an issue where after navigating to a new Group page using the Select Group dropdown, the code ensuring that an assistant is linked in the URL wouldn't run. (762)

Schema Upgrade	Migration Script	Permissions Update	Task Definition Update	Configuration Update
YES	No	No	No	No

Fix: Main Page Assistant won't run URL-dependent code after navigation from Group Selector by @ekassos in #762
Vision capabilities through image descriptions for Azure by @ekassos in #761

Full Changelog: 603-srv229-web142...605-srv230-web144