The release notes summarize the changes introduced over the previous week.
New Features & Improvements
- Login, signup, SSO, and social login pages have a redesigned look with light and dark theme support
- Marketo connector now supports multi-workspace instances, letting customers browse, select, or manually specify the workspace it uses
- Quality Evaluator evaluation jobs now run roughly 3 times faster thanks to increased processing concurrency
- Portal now shows a "Translated via" column in the translation history table indicating which MT/AI engine handled each translation
- Batch-assigning translation memories in TMS project templates is now clearer and more reliable, letting you choose whether to add to or replace existing assignments, and correctly reflecting current selections even when settings differ across languages or workflow steps
- Users invited to Strings and TMS from the Dashboard now get the same invitation experience as being invited directly from those products
Bug Fixes
- Fixed: Translation memory creation in Phrase TMS hung indefinitely for all users; TM creation now completes normally
- Fixed: An unrecognized locale code from DeepL was breaking all DeepL language pairs for every customer using DeepL
- Fixed: In-context preview for PPTX files showed no slide images
- Fixed: Multilingual XLSX in-context preview showed target text rendered into the wrong language column
- Fixed: Buyer-restricted project managers could access projects and jobs outside their allowed buyers via direct links or the API
- Fixed: Project list failed to load for project managers with buyer-access restrictions enabled
- Fixed: A job could be dropped from its batch's pre-translation completion event, leaving it out of automatic pre-analysis
- Fixed: Job editing page loaded slowly when many relevant providers were available
- Fixed: Downloading a target file with an already-used request ID returned a misleading success response with no file
- Fixed: Word count reporting in Phrase Data was incorrect for a retroactively-added review step on jobs with translation-memory-leveraged segments
- Fixed: Rules and AI Checks table now wraps long rule text across multiple lines instead of truncating it