Your disk is an ocean — Diskhoji is the seeker that sounds it. One standalone native Rust app — a single ~6 MB binary — charts millions of files a second into a cushion treemap and a heatmap of a thousand watching cells. No Electron. Nothing leaves your ship.
# one line — Linux & macOS, prebuilt binary $ curl -fsSL https://diskhoji.org/install.sh | sh # windows — powershell > irm https://diskhoji.org/install.ps1 | iex $ curl -fsSL https://diskhoji.org/install.sh | sh # or by hand: prebuilt binary $ tar xzf diskhoji-v0.2.0-linux-x86_64.tar.gz && cd diskhoji-v0.2.0 $ ./diskhoji # opens the native app; ./diskhoji ~/ scans at once # or build from source — Linux & macOS, needs rustup.rs $ git clone https://github.com/singhpratech/diskhoji && cd diskhoji $ cargo build --release && ./target/release/diskhoji ~/
A native window by default — --web serves the same dashboard on localhost if you prefer the browser. macOS binaries are built on CI for Apple Silicon and Intel. cargo install diskhoji is charted for the next release.
A little offline diversion, right here on deck: leap the corrupt reefs, gather the amber bytes, see how many fathoms you can log. Space, ↑, or tap to jump. Works with the rigging cut — no connection needed.
A treemap makes the giant files impossible to miss — but a thousand small files vanish into slivers. So Diskhoji carries a second chart: a heatmap where every item holds an equal cell and color carries the size, like watch-lights on dark water. One click to change charts; zoom, select, and right-click hold their course in both.
Cushion-shaded, area = size. Double-click to dive into any folder, breadcrumbs to surface again.
One cell per item, ring = folder, color = size on a log scale. The small-but-many stay in sight.
Scanner, chart engines, and a true native window compile into a single ~6 MB executable. No installer, no webview, no 200 MB of Electron, no runtime to fetch. Stow it anywhere and run it.
A work-stealing scanner crews every core: 1.85 million files catalogued in 2.7 seconds cold, 0.4 warm. Rescans cost nothing, so you actually make them.
The squarified treemap is computed in Rust — about 15 ms for a two-million-node tree. The window only paints pre-computed rectangles; million-file folders never becalm the UI.
Explorer tree ⟷ chart ⟷ file types ⟷ largest files. Spotlight a file type across the map; click a leviathan to find its berth in the tree.
Right-click → delete permanently, behind a confirmation you can trust. The in-memory model updates in place — every total and chart stays true at once.
Binds to 127.0.0.1 only. No telemetry, no updater, no network at all. Keeps to one filesystem, never follows symlinks, refuses to scuttle the scan root.
Before the sea had charts, it had a keeper.
In the oldest Vedic hymns, Varuna is lord of the waters, robed in the splendor of lapis lazuli, and keeper of ṛta — the cosmic order. Nothing moves on his ocean unrecorded: the scriptures say he knows every blink of every eye, and the stars are his thousand spies keeping watch through the night.
Diskhoji — khoji, खोजी, the seeker — sails in that tradition, across seas of data instead of water.
Inspired by the classics. Copied from none. The great treemap analyzers of the 2000s taught everyone what a disk looks like — then stayed in harbor: one OS, one thread, one view. Diskhoji is a ground-up rebuild of the idea in Rust — multi-core, two charts, one small native binary, and at home on Linux, macOS, and Windows.