New Features

• Added AI Provider performance dimensions (connection latency / time-to-first-token) and displayed them in the admin.

Experience Improvements

• Front-end alerts upgraded to in-page toast controls that don't interrupt the flow.

Bug Fixes

• Fixed BYOK save alert blocking and streamed-content loss, and xmlContent extraction issues.
• Fixed the model-scheduling dashboard's blank performance column and free-tier quota unification issues.