Release Notes v0.8.3

What’s changed?

We are excited to announce the release of version 0.8.3, which includes significant new features such as the ability to upload a file from your OneDrive account, image generation with DALL·E 3, and response streaming, which are also available in the mobile experience. Additionally, this release includes accessibility improvements, bug fixes and enhancements designed to improve the overall user experience. 

New Features 

• Clarity now gives users the ability to access, browse, and upload a file from their Yale affiliated Microsoft OneDrive for Work or School account to their chats in Clarity AI chatbot. Currently, the file upload is limited to one file at a time. Once the issue with multiple file uploads is resolved, you will be able to select and upload multiple files from your OneDrive account. 
• Clarity now allows you to generate images right within the Clarity AI chatbot using the DALL·E 3 AI model and download the generated images. When you submit a request that includes “Create an image of….”, an image will be generated in the response and available for download. DALL·E 3 images expire after 24 hours so users will need to save their images if they want to keep them. This functionality is an additional feature that complements the existing “GPT-4o” agent, so you don’t need to select another agent in the dropdown menu. 
• Clarity now supports simulated response streaming, enabling a seamless experience when it comes to long running responses without requiring further action by the user.

Accessibility Improvements 

• Chat Input element aria-label has been updated to “What would you like to ask?”  
• Aria-described by attribute has been added to the Chat Input element that points to hidden text saying “The agent can make mistakes. Check important info.”   
• Fonts are now referenced and stored locally and references to Google’s CDN have been removed.   
• Sidebar, Navbar and ChatThread components have been updated to use semantic HTML that clearly describes their meaning in a human- and machine-readable way.   
• Aria-label of “New chat name” has been added to Rename Chat element.  
• Aria-labels have been added to Info Icon and Remove File elements.  
• Aria-activedescendant has been added to Agent Dropdown Selection element.  
• Chat Input element is now keyboard focusable and navigable; when a user opens a dialog, the keyboard focus is automatically on the Chat Input element. Dialog in the Chat Input element can be submitted by pressing Enter when in focus.  
• Button colors have been updated to have a higher contrast ratio of 3 to 1.   
• The alt text of the logo has been updated to match the logo text set in FoundationaLLM:Branding:LogoText branding configuration.   
• An element has been added for “Loading” Icon elements that wrap them with the role “status.”  
• The ability to close all Tooltips with the Escape Button has been added.  
• All Tooltip triggers are now focusable by keyboard.  

Bug Fixes and Improvements 

• Until a multi-file upload can process without issue, file uploads will be limited to 1 file per prompt but users will be able to upload multiple files within a conversation.
• Navigation issue with tabindex that was causing Tab navigation to behave inconsistently has been fixed.   
• Tooltip behavior has been fixed so that users can now select text within the Tooltip without it disappearing.  
• File reference storage has been moved to Cosmos DB to enable better tracking of files through the upload, vectorization, and response process.