COMMITS
July 26, 2026
T
chore(mcp): announce skill baseline 2026-07-26.1
tangjinzhou committed
T
feat(studio): framing vocabulary follows canvas orientation
tangjinzhou committed
T
feat(studio): render the reasoning stream in chat — no more dead air
tangjinzhou committed
T
feat(studio): chat agent treats its underlying model as internal infrastructure
tangjinzhou committed
T
feat(agent): review_visuals — delegated eyes for the in-app chat agent
tangjinzhou committed
T
Merge sentence-level bilingual translation rework (upstream main)
tangjinzhou committed
T
feat(captions): checkerboard swatch for the no-color state
tangjinzhou committed
T
feat(captions): instant line editing, no auto-retranslate
tangjinzhou committed
T
feat(captions): distribute sentence translations across cues
tangjinzhou committed
T
fix(agent): offline executor desegments cue-split transcripts — same rows as the browser
tangjinzhou committed
T
feat(captions): sentence-level bilingual translation
tangjinzhou committed
T
Merge caption cue rework (upstream main) into the agent-channel line
tangjinzhou committed
July 25, 2026
T
feat(studio): canvas follows the source aspect; cues are geometry-sized single lines
tangjinzhou committed
T
fix(studio): EN cue budget counts word spaces — no more third display line
tangjinzhou committed
T
fix(studio): language-aware cue size — EN cues hold a clause, not 3-4 words
tangjinzhou committed
T
T
fix(studio): retire caption edit force-show — it painted two captions after pause
tangjinzhou committed
T
feat(studio): captions render REGULAR by default — weight leaves the preset
tangjinzhou committed
T
feat(studio): caption base 48px, bolder translation line, bold toggle
tangjinzhou committed
T
refactor(studio): caption presets carry look only — font size leaves the preset
tangjinzhou committed
T
refactor(studio): captionStyle persists SPARSELY — defaults live in the resolver only
tangjinzhou committed
T
revert(studio): default caption width is aspect-invariant; single cue budget
tangjinzhou committed
T
feat(studio): aspect-aware default caption width
tangjinzhou committed
T
T
fix(studio): multi-line captions share ONE plate (main stack + translation)
tangjinzhou committed
T
T
T
refactor(studio): captions become derived state — transcript is the only store
tangjinzhou committed
T
feat(studio): chat syncs to cloud independently of canvas content
tangjinzhou committed
T
feat(studio): fixed caption cues + ICU word segmentation
tangjinzhou committed