Features

Enable multithreaded preview generation by default with 4 worker threads when enough logical processors are available. Existing saved settings are preserved. Loading time reduced by about 50%.
Add a conservative memory scheduling option for multithreaded previews. It is disabled by default; enabling it reduces concurrent mesh work and queued upload batches, which may slow decoding.
Show the combined host and decoder private working sets during loading. This figure excludes WebView2 and GPU memory.

Improvements

Start uploading complete geometry batches while later chunks are still being generated. Keep mesh work, host batches, and upload preparation bounded, and discard staged results on cancellation or failure.
Pack preview data into shared arenas for fewer IPC reads and GPU buffer allocations without changing draw order.
Keep the optional decoder process-memory limit effective with either scheduling mode.

There are two options for you to install, pick the one you prefer:

A one-click installer (LitematicaPreview-<version>-win-x64-setup.exe): Installs per-user without requiring administrator privileges.
A portable .zip package (LitematicaPreview-<version>-win-x64-portable.zip): Download and extract the portable archive and run LitematicaPreview.exe.

Full Changelog: v0.2.0...v0.3.0