This release introduces significant upgrades to the AI Canvas, adds inline metadata editing for assets, polling-based update notifications, and transitions our coding agent adapter to Antigravity 2.0.

🎨 AI Canvas Features & UX Polish

AI Canvas Plan Mode: Added a dedicated planning mode to guide AI canvas generation tasks.
Vector Shapes & AI Tools: Integrated drawing cards with vector shape capabilities and responsive AI tools.
Text Cards & Transformations: Support for text cards with right-click options for mirroring and rotation.
Architecture Cleanups: Refactored canvas side-effects (keyboard controls, canvas state storage, event dispatchers) to enhance stability and UI responsiveness.

📝 Asset Drawer Inline Editing

Inline AI Description & OCR Editing: You can now edit AI-generated descriptions and OCR text directly from the asset drawer.
State Management & Fixes:
- Reset edit state when switching between assets.
- Correctly preserve OCR empty-text status and disable tooltip when idle.
- Added i18n support for all editing triggers and validation messages.

🤖 Agent CLI Upgrade (Antigravity 2.0)

Branding Refactor: Renamed the Gemini coding agent integration to Antigravity 2.0 throughout settings, backend detection, and configuration validation.
Model Defaults: Updated recommended default coding models to gemini-3.5-flash in English, Japanese, Korean, Simplified Chinese, and Traditional Chinese translation files.