• Added ability to select individual files from a folder when configuring assistant tools (search, include, etc.)
• No longer required to select an entire folder when using folder content in tools

• Default number of steps per assistant run increased from 3 to 8
• Maximum allowed steps increased from 8 to 12

• Messages now adapt their width based on content length
• Better readability across all screen sizes
• Optimized layout for both mobile and desktop views

• Now using a more restrictive permission scope (drive.meet.readonly) to access Google Meet transcripts
• Dust only requests access to Google Meet-generated files instead of all Google Drive files

• Added feedback functionality for custom assistant responses
• Users can now rate assistant responses and optionally share conversations
• Introduced feedback statistics dashboard in the assistant manager
• Feedback metrics now available through workspace usage API

Conversations are now automatically sorted by most recent activity:
• When posting a message in a conversation, it moves to the top of the list
• Most recently used conversations appear first

• Added DeepSeek V3, a new large language model, to the platform
• Model can be used in Dust Apps, as a global agent, and in custom assistants
• Known for high performance on benchmarks, fast response times, and efficient processing
• Enabling DeepSeek on a Workspace requires reaching out to Dust

• Added ability to disable storage of execution logs for custom Dust apps
• When enabled, the app will not store inputs and outputs of blocks in the database
• Only affects custom Dust apps (conversation agents are not impacted)

• Improved assistant loading time for quicker auto-complete availability
• Streamlined input bar animation for smoother transitions

• Added comprehensive documentation to help deploy Dust within organizations
• Introduced resources including webinar registration and community access
• Created step-by-step guidelines for successful enterprise implementation