6.114.0

May 28, 2025

What's new 💡

Data Studio

Review Extension — Added document status to help users easily track the progress during the review process.
Bounding Box Labeling:
- Improved the interaction so users can draw a new bounding box without clicking outside to deactivate resize handles on another box when the "Show resize handles on hover" setting is enabled.
- Added a loading indicator during bounding box data fetching to provide better feedback.
Notification — Updated the notification behavior so the popover no longer opens automatically when redirected to the Project Report Dashboard. Instead, a blue dot will appear next to the bell icon when there are new notifications.

LLM Labs

Custom models — Improved the error message when users make a request to an incorrectly formatted vLLM endpoint URL.
Sandbox — Improved our Sandbox UI by hiding the corresponding chunk dropdown when there is no knowledge base selected.
Sandbox — Improved our Sandbox UI by removing the "cheapest" and "fastest" tags when users are only using one application.
Knowledge base — Adjusted our file upload process; users can now edit the chunk directly after uploading the file.
Knowledge base — Added a loading indicator to block user interaction while the edit chunk process is still in progress.
Direct Access LLMs — Removed retired models from VertexAI.
Direct Access LLMs — Added Claude Sonnet 4 as our new model.

Bug fixes 🐞

Data Studio

Fixed an issue in Span Labeling where the “Consensus,” “Accepted,” and “Rejected” legends were missing from the status bar in Labeler mode when “Rapid Labeling Feedback” was enabled.
Fixed an issue in Span Labeling where the “Accepted by” and “Rejected by” statuses, shown in the tooltip when hovering over an applied label, did not update immediately after rejecting, deleting, or editing labels, requiring a page refresh.

LLM Labs

Fixed an issue where evaluation fails when using GPT-o3 mini model as evaluator.
Fixed an issue where deleted deployed or saved applications are duplicated in the evaluation table and filter.
Fixed an issue where the snack bar error appears when editing the chunk using embedding model from Azure OpenAI.
Fixed an issue where the blue color in tables is different from Data Studio.
Fixed an issue where free trial plan is not started for newly created teams.

Last updated 22 days ago

What's new 💡

Data Studio

Review Extension — Added document status to help users easily track the progress during the review process.

Bounding Box Labeling:

Improved the interaction so users can draw a new bounding box without clicking outside to deactivate resize handles on another box when the "Show resize handles on hover" setting is enabled.
Added a loading indicator during bounding box data fetching to provide better feedback.

Notification — Updated the notification behavior so the popover no longer opens automatically when redirected to the Project Report Dashboard. Instead, a blue dot will appear next to the bell icon when there are new notifications.

LLM Labs

Custom models — Improved the error message when users make a request to an incorrectly formatted vLLM endpoint URL.

Sandbox — Improved our Sandbox UI by hiding the corresponding chunk dropdown when there is no knowledge base selected.

Sandbox — Improved our Sandbox UI by removing the "cheapest" and "fastest" tags when users are only using one application.

Knowledge base — Adjusted our file upload process; users can now edit the chunk directly after uploading the file.

Knowledge base — Added a loading indicator to block user interaction while the edit chunk process is still in progress.

Direct Access LLMs — Removed retired models from VertexAI.

Direct Access LLMs — Added Claude Sonnet 4 as our new model.

Bug fixes 🐞

Data Studio

Fixed an issue in Span Labeling where the “Consensus,” “Accepted,” and “Rejected” legends were missing from the status bar in Labeler mode when “Rapid Labeling Feedback” was enabled.

Fixed an issue in Span Labeling where the “Accepted by” and “Rejected by” statuses, shown in the tooltip when hovering over an applied label, did not update immediately after rejecting, deleting, or editing labels, requiring a page refresh.

LLM Labs

Fixed an issue where evaluation fails when using GPT-o3 mini model as evaluator.

Fixed an issue where deleted deployed or saved applications are duplicated in the evaluation table and filter.

Fixed an issue where the snack bar error appears when editing the chunk using embedding model from Azure OpenAI.

Fixed an issue where the blue color in tables is different from Data Studio.

Fixed an issue where free trial plan is not started for newly created teams.