Windows-first file explorer rebooted from scratch as a Rust application with a thin Tauri renderer.
- Beat
Fileson perceived speed, not just binary size. - Keep all file-system logic in Rust.
- Use Tauri and SvelteKit only as the UI renderer.
- Optimize for Windows first while keeping Rust interfaces portable enough for future macOS/Linux backends.
- Navigation must be incremental, cancelable, and benchmarked.
- The UI never owns sorting, searching, metadata hydration, or file operations.
- Expensive enrichments such as thumbnails, Git state, hashes, preview generation, and cloud metadata stay off the hot path.
- Large lists must be virtualized from day one.
This repo has been reset and reinitialized on the latest stable Tauri scaffold with:
Tauri 2SvelteKitTypeScriptbunRust
ROADMAP.md- product and architecture roadmapTODOS.md- implementation backlog for the next build phases
Svelte is a good fit for this architecture because the frontend is intentionally thin. It keeps renderer code smaller and simpler than a typical Vue setup while still being perfectly capable of hosting virtualized views, panes, tabs, and command surfaces.
Vue would also work, but for a renderer-only desktop shell I would pick Svelte first unless your team already has strong Vue momentum.
bun install
bun run tauri devTo artificially slow navigation for UI testing, set VITE_EXPLORER_NAV_DELAY_MS before starting dev.
PowerShell example:
$env:VITE_EXPLORER_NAV_DELAY_MS = "500"
bun run tauri devBuild a single-pane details view backed by a Rust directory engine that streams incremental results into a virtualized Svelte list.