DedupCommando documentation

Beta — v0.9.2 . DedupCommando performs destructive operations (delete, hardlink, reflink) on real files. Read the safety guide before applying anything, and keep backups.

DedupCommando is a Linux terminal tool (CLI and TUI) that finds byte-for-byte identical files and whole duplicate folders, and reclaims the space they waste — built for ZFS pools, including storage hosted on Proxmox VE systems. Data safety comes first: every destructive batch runs under a ZFS snapshot, "deleted" files are moved to a quarantine instead of being unlinked, and content is re-validated immediately before each action.

Install

APT (Debian 13 / Proxmox VE 9+, recommended) — install from the signed APT repository (auto-updates via apt upgrade):

# as root (Proxmox default); on non-root Debian run: sudo -i
curl -fsSL https://dedupcommando.github.io/apt/dedcom-archive-keyring.gpg \
  -o /usr/share/keyrings/dedcom-archive-keyring.gpg
echo "deb [signed-by=/usr/share/keyrings/dedcom-archive-keyring.gpg] https://dedupcommando.github.io/apt stable main" \
  | tee /etc/apt/sources.list.d/dedcom.list
apt update && apt install dedcom

Release binary (any Linux, glibc ≥ 2.39) — download the tarball for your architecture (amd64 / arm64) from GitHub Releases, verify it, then:
```
tar xzf dedcom-<version>-<triple>.tar.gz
install -m 755 dedcom /usr/local/bin/dedcom
```
From source — a Docker-based build, no local Rust toolchain required (see CONTRIBUTING).

Cargo (cargo install dedcom) is planned — not yet available; for now use APT, the release binary, or build from source. The installation chapter of the manual covers every option in detail.

First scan

dedcom              # multi-panel commander (default)
dedcom --classic    # classic stepwise wizard
dedcom --read-only  # read-only observer
dedcom -h           # all options

Pick scan roots, choose an intensity profile (Idle on a busy host), and start. After the scan, mark a keeper and an action per group, then review and apply — or save the plan as a shell script. The quickstart chapter walks through a typical session.

Read the docs

User manual — install, data safety, scanning, actions, the Commando and Classic interfaces, headless/cron, maintenance, troubleshooting, hotkeys.
Safety, recovery and limitations.
Verifying releases.
Release notes — what changed in each version.
On GitHub: README · Contributing (DCO) · Security policy · Trademarks.

Safety first

A ZFS snapshot of every dataset the batch touches is taken before the first action — if any snapshot fails, the whole batch is aborted.
"Delete" moves files to a quarantine, not unlink — reversible until you explicitly purge it.
Content is re-validated (re-hash / re-stat) right before each action; a mismatch aborts that action.
Files are published atomically with renameat2(RENAME_NOREPLACE) — no check-then-rename race.
A single-instance lock prevents concurrent writers; cross-dataset moves are refused, never a silent copy-and-delete.

Three ways to reclaim space

Delete to quarantine — remove a duplicate, keep it recoverable until purged.
Hardlink — point duplicates at one shared inode (within a single dataset).
Reflink — copy-on-write block clone on ZFS with block_cloning (within one dataset, like a hardlink); independent metadata, blocks shared until a file changes.

One file in each group is the keeper; the rest become links or go to quarantine.

How it works

Scan — walk your chosen roots, hash candidates with BLAKE3 (with an optional byte-for-byte re-compare), and group identical files. Scans are resumable and cached for near-instant re-runs.
Review — browse duplicate groups in the multi-panel commander (default) or a classic stepwise wizard (--classic); mark a keeper and the action for each group. It also finds "twin folders" — directory trees whose scanned contents are identical.
Apply — review the plan and apply interactively, or save it as a shell script. A resource governor (Turbo / Balanced / Idle) keeps a scan from starving VMs or backups on a busy host.

Built for ZFS, runs on Proxmox VE

DedupCommando is designed for ZFS: snapshots, dataset-aware boundaries, and reflink all build on it. It is tested on Proxmox VE 9.1 (OpenZFS 2.3), where ZFS is available out of the box. This is file-level deduplication — finding and removing duplicate files — not ZFS's built-in block-level dedup (zfs set dedup), and not compression.

On non-ZFS filesystems scanning still works, but there is no snapshot safety, so applying actions is not recommended there.

Requirements

Linux, kernel ≥ 3.15, x86_64 or aarch64; the pre-built packages need glibc ≥ 2.39 (Debian 13, Ubuntu 24.04, Proxmox VE 9).
ZFS strongly recommended (snapshot safety, dataset detection, reflink); zfs in PATH, typically run as root.
A UTF-8, 256-color terminal.

Verify your download

Each release ships SHA-256 checksums, minisign signatures, a CycloneDX SBOM and a SLSA build-provenance attestation. See Verifying releases.