- Popup no longer flashes "Loading AI model…" unnecessarily on cold starts when the bookmark index is already cached and up to date. The embedding pipeline now loads lazily on first search rather than eagerly on every service worker respawn. (#7, #9)
- Initial release improvements.
- Test release.