SPB Git

Repository Public License

README.md 34.7 KB

ORIX Clean icon: a storage ring with one quarter swept clean

# ORIX Clean

# Reclaim Space. Keep Control.

A native macOS storage analyzer and cleanup utility that measures everything it shows,
explains everything it proposes, and removes nothing without your confirmation.

Author: Simon-Pierre Boucher · part of the ORIX family of Apple utilities

macOS Apple Silicon Swift SwiftUI Storage Privacy Offline Accounts

Version Minimum macOS Architectures Build Core tests UI tests Coverage Scan Treemap Signing Notarization DMG Release Network Dependencies Languages

Commits Last commit Language Size Tags

Download ORIX Clean 1.0.0 (DMG, 5.1 MB) · ZIP · SHA-256 · All releases


ORIX Clean storage map: a treemap of a home folder with a list of its folders beside it

The storage map on the sample home folder. Every block is a real file or folder, sized by its measured size on disk.

About the screenshots. They are captures of the real application (window captured by its window ID) running on a generated sample home folder of 182 real files (1.89 GB) written by orix-bench fixture, so that no personal file name appears here. Sizes, duplicate groups, similar-image groups and cleanup results are what the engines measured on those files. The History screenshot shows real free-space samples of the development Mac.


# Contents

  1. Overview · 2. Product Philosophy · 3. Application Preview · 4. Storage Analysis · 5. Interactive Storage Map · 6. Large Files · 7. Duplicate Detection · 8. Similar Images · 9. Cache Cleanup · 10. Uninstall Assistant · 11. Leftover Files · 12. Developer Cleanup · 13. Downloads Cleanup · 14. Trash Analysis · 15. Cleanup Preview · 16. Safe Deletion · 17. Storage Intelligence · 18. Storage History · 19. ORIX Clean Actions · 20. Reports · 21. Filesystem Architecture · 22. APFS Considerations · 23. Technology Stack · 24. Installation · 25. Build Instructions · 26. Testing · 27. Benchmarks · 28. Privacy · 29. Accessibility · 30. Localization · 31. Roadmap · 32. Release Status · 33. Known Limitations · 34. ORIX Ecosystem · 35. License

# 1. Overview

ORIX Clean answers three questions about a Mac's storage, in this order:

  1. Where did the space go? A parallel scanner reads names and sizes straight from the filesystem and draws them as an interactive treemap.
  2. What could be removed, and what happens if it is? Duplicates, similar pictures, caches, developer data, old downloads, applications and their files — each candidate comes with a reason, a risk level and a consequence.
  3. Do you really want that? Every removal passes through a preview that lists the exact items. They go to the Trash, per-item outcomes are recorded, and they can be put back.
Platform macOS 15.0 or later, universal binary (built and run on macOS 27.0, Apple M5 Max)
Language / UI Swift 6 (strict concurrency), SwiftUI + AppKit, Swift Charts, Canvas
Core ORIXCleanCore, a UI-free Swift package: 19 files, 3,573 lines, 56 tests
App 18 Swift files, 3,352 lines; 684 localized strings (English, Canadian French)
Dependencies None. No third-party package, no network code, no helper tool
Bundle identifier ai.spboucher.orix.clean
Distribution Developer ID signed, Hardened Runtime, Apple notarization, DMG + ZIP

# 2. Product Philosophy

Principle What it means in the code
Every number is measured Sizes come from getattrlistbulk(2) during the scan. Nothing is derived from file names, and there is no simulated progress: counters shown during a scan are the scanner's own counters.
Honest savings Recoverable space is a range when APFS reports shared blocks: from the sum of private sizes to the sum of allocated sizes. Hard links are reported as freeing nothing.
No automatic cleanup Nothing is preselected in duplicates or similar images. A CleanupAuthorization can only be built from the plan the user confirmed; the engine skips anything else.
Reasons, not fear Each candidate states why it is listed, what happens if it is removed and what it costs to get back. There are no urgency banners and no "your Mac is at risk" language.
Nothing leaves the Mac No account, no analytics, no update check. Scripts/check.sh fails the build if networking APIs appear in the product sources.
Supported APIs only No private API, no privileged helper, no administrator prompt, no attempt to get around System Integrity Protection or privacy controls. What cannot be read is reported as not read.

# 3. Application Preview

Overview
Overview. Volumes as macOS reports them, the measured breakdown by type, and recommendations that quote their evidence.
Overview in light mode
Light mode. The same screen with the light palette; appearance follows the system or the setting.
Storage map, dark
Storage map, dark. Folders are frames, files are blocks coloured by type.
Storage map, light
Storage map, light.

# 4. Storage Analysis

The scanner walks a folder or a volume with a small pool of threads. Each directory is read in one or a few system calls that return, for every entry, its name, type, device, modification date, link count, content size, allocated size and — on APFS — private size, clone identifier and sharing flags.

  • Three sizes, never mixed. Content size is the length of the files. Size on disk is what is allocated. Private size is what APFS says would come back if that file alone were removed.
  • Accuracy check. On a 31,259-object tree the allocated total equals du -sk to the byte (2,412,920,832); a test repeats this comparison on every run.
  • Never follows symbolic links. Mount points and other devices are boundaries; a volume mounted inside the scanned folder is detected from the opened directory itself.
  • Hard links are counted once. Other names of the same file are kept in the tree but add nothing to totals.
  • Cloud-only files are never opened. Items whose content is not on the Mac (SF_DATALESS) are flagged; hashing or previewing them would download them.
  • Cancellable and pausable, with live counters. Unreadable folders are listed as issues, and the totals say what is missing.
  • Startup volume. Scanning / enters the Data volume through its firmlinks and skips /System/Volumes, so nothing is counted twice.

# 5. Interactive Storage Map

Storage map in Canadian French The map with the interface in Canadian French. Number and unit formats follow the locale.
  • Squarified treemap drawn in a SwiftUI Canvas: area is proportional to measured size within each folder.
  • Level of detail. Only regions large enough to see are laid out; siblings that would be under a few square points are merged into one cell labelled "N smaller items" with their real combined size. Laying out a 467,907-node tree takes 0.16 ms.
  • Interactions. Double-click to drill down, breadcrumb to go back up, pinch to zoom (up to 12×), drag to pan, hover for details, right-click for Reveal in Finder, Quick Look, Copy Path and Add to Cleanup Selection.
  • Search dims everything that does not match and lists the matches.
  • Two size modes (on disk / content) and two palettes (ORIX, and the Okabe–Ito colour-blind-safe set).
  • A list beside the map shows the same data sorted by size; it is the keyboard and VoiceOver path through the tree.

# 6. Large Files

Large files list with filters

Largest files or largest folders, filtered by minimum size, age, type, name and whether to look inside packages. The age filter uses the modification date and says so: not modified is not not used. When Spotlight has recorded a last-opened date for a file, it is shown next to it as additional evidence; ORIX Clean never claims a file is unused.

# 7. Duplicate Detection

Duplicate groups with one copy kept per group Four groups found in the sample. In the second, both files are APFS clones: the app says they share storage, and the selection bar gives the recoverable space as a range.
flowchart LR
  A[Regular files in the scan] --> B[Group by content size]
  B --> C[Sampled SHA-256: head, middle, tail]
  C --> D[Full SHA-256, streamed]
  D --> E[Groups of identical digests]
  E --> F[You choose the copy to keep]
  F --> G[Byte-for-byte check against the kept copy]
  G --> H[Move to Trash]
  • The sampled hash only narrows candidates; the full hash decides. A test forces every sampled hash to collide and checks that no false group appears.
  • A file whose size no longer matches the scan is dropped rather than compared.
  • Hard links are one file under several names: not offered, because removing a name frees nothing.
  • APFS clones are real duplicates that share blocks: flagged, with a recoverable range instead of a single figure.
  • Files inside applications, photo libraries and other packages are skipped unless asked for.
  • One copy always stays. The UI refuses to tick the last copy, and the deletion engine independently refuses to remove a duplicate whose kept copy is missing, selected, different, or the same inode.

# 8. Similar Images

Groups of similar images with confidence

Two measurements are combined, both computed on the Mac: a 64-bit difference hash (structure) and a Vision feature print (appearance). Images whose structure is nearly identical (4 bits or fewer of 64) are grouped on structure alone; beyond that, Vision must agree. A fast mode compares structure first; a thorough mode compares every pair and also finds crops (limited to 3,000 images). Groups are built around a representative image, so a chain of "A resembles B, B resembles C" does not put A and C together. The confidence shown is the weakest link of the group. Nothing is preselected, and at least one image per group must stay.

# 9. Cache Cleanup

Caches and logs with risk badges and a selection bar

A documented catalogue of 35 locations (12 for caches and logs, 23 for developer tools): application caches in ~/Library/Caches, web caches of eight browsers (the cache folder only — never the profile, history, cookies or passwords), logs, crash reports and old temporary items.

  • Not listed, on purpose: Apple system caches (com.apple.*) and caches known to hold state (CloudKit, HomeKit, and others in an explicit exclusion list).
  • Every row gives why it is listed, what happens if you remove it, how you get it back, and the estimated recoverable space.
  • If the owning application is running, the row says so and the engine leaves the item in place.
  • No performance promise. The screen states that removing caches does not make a Mac faster and that applications will use part of the space again.

# 10. Uninstall Assistant

Application detail with signature information and associated files

For each installed application: size, version, bundle identifier, who signed it (read from the static code signature, without running it), team identifier and the last-used date Spotlight recorded. Associated files are searched in twelve Library locations and labelled by certainty:

Certainty Rule Risk
Matches the bundle identifier Named exactly after the identifier (or a helper identifier that extends it) Review required
Shared by the developer's apps Group container keyed by the team identifier High — never preselected
Name match only Named after the application's display name High — never preselected

Documents are never part of an uninstall. Applications that ship with macOS are shown but cannot be selected.

# 11. Leftover Files

Library items named after an application identifier for which no installed application is found (by the scan of the application folders and by Launch Services). The screen says plainly that this is an inference from names: command-line tools, plug-ins and applications on other volumes can look like leftovers. Items sharing a vendor prefix with an installed application are not flagged. Everything is review required.

# 12. Developer Cleanup

Developer cleanup lab grouped by tool

Xcode DerivedData (one row per project), archives (marked review: they hold the dSYMs of shipped builds), device support files, simulator caches, and the caches of Swift Package Manager, CocoaPods, Carthage, Homebrew, npm, Yarn, pnpm, pip, uv, Gradle, Maven, Cargo, Go and Playwright. After a scan, build folders (node_modules, .build, target…) are listed only when they sit next to their project manifest.

Reported but never removed: simulator devices, simulator runtimes and the Docker disk image, each with the supported command or menu to clean it.

# 13. Downloads Cleanup

Downloads grouped into installers, disk images, archives and incomplete downloads

Installers, disk images, archives and incomplete browser downloads, with the date each arrived in the folder. An archive that sits next to an item of the same name is noted as "may be already extracted (not verified)". Everything here is review required.

# 14. Trash Analysis

Trash content with sizes

Size and content of the Trash of the home folder and of each mounted volume. Reading the Trash requires Full Disk Access; without it the screen says the content is unknown instead of showing zero. Emptying is the only permanent deletion in the application: it has its own confirmation that requires typing DELETE.

# 15. Cleanup Preview

Cleanup preview listing every selected item with reason and consequence

No removal can start anywhere else. The preview shows every path, grouped by risk, with its reason and consequence, the content size, the estimated recoverable space and the recovery option. Items can be unticked there. High-risk items require a separate acknowledgement.

# 16. Safe Deletion

Cleanup results with per-item outcome and Put Back buttons Result of a real cleanup of the sample caches. The app states that the space is still occupied until the Trash is emptied.
flowchart TD
  P[Confirmed plan + authorization] --> A{Path in the confirmed set?}
  A -- no --> S1[Skipped]
  A -- yes --> B{Safety policy allows it?}
  B -- no --> S2[Skipped: protected]
  B -- yes --> C{Still there, same file, same size?}
  C -- no --> S3[Skipped: changed since the scan]
  C -- yes --> D{Owning app running?}
  D -- yes --> M[Requires manual action]
  D -- no --> E{Duplicate: identical to the kept copy?}
  E -- no --> S4[Skipped]
  E -- yes --> T[Move to Trash]
  T -- ok --> R1[Moved to Trash, location recorded]
  T -- no Trash on volume --> M
  T -- error --> F[Failed, with the system's reason]
  • Five outcomes, never merged: Moved to Trash, Permanently deleted, Skipped, Failed, Requires manual action.
  • A last line of defence (SafetyPolicy) refuses system areas, standard folders themselves, volume roots, the inside of application bundles and of photo, music and video library packages — whatever asked.
  • Put Back restores an item while it is still in the Trash, and refuses to overwrite anything that now occupies the original place.
  • A volume without a Trash yields Requires manual action; the engine never falls back to permanent deletion.

# 17. Storage Intelligence

Recommendations are computed only from data that has actually been analysed; an analysis that has not run produces none. Each one carries its evidence (counts, measured sizes, paths), a risk level, the consequence and a link to the relevant screen. Thresholds are constants in RecommendationEngine. A growth projection is offered only with at least three samples over six hours, and is labelled a projection.

# 18. Storage History

Used space over time from recorded samples

Free-space samples of each volume (every 15 minutes while the application is open, and after each scan and cleanup), sizes of watched folders, and the log of every cleanup with its per-item outcome. The chart starts at the first recorded sample and the screen says that nothing exists before it.

# 19. ORIX Clean Actions

Reusable cleanup workflows

Reusable sequences of real operations: scan, find duplicates, collect caches, developer data or old downloads, review, move to Trash, report. A workflow is rejected at save time and at run time unless every removal step comes directly after a review step, and that review is the same preview sheet. Declining it ends the run with nothing removed.

# 20. Reports

Report generator with a Markdown preview

Nine report types (storage overview, largest files, duplicates, similar images, caches, developer data, application uninstall, cleanup results, storage trends) exported as PDF, CSV, JSON or Markdown. Each includes the scan date and scope, what could not be read, measured sizes, estimated recoverable space and actual results. CSV cells that start with a formula character are neutralised.

# 21. Filesystem Architecture

flowchart LR
  subgraph App[SwiftUI application]
    V[Views] --> M[AppModel]
  end
  subgraph Core[ORIXCleanCore - no UI]
    SC[FileScanner] --> T[(ScanTree arena)]
    T --> Q[Queries + Treemap]
    T --> D[DuplicateFinder]
    T --> I[SimilarImageFinder]
    CAT[CleanupCatalog] --> PL[CleanupPlan]
    D --> PL
    I --> PL
    Q --> PL
    PL --> AU[CleanupAuthorization]
    AU --> DE[DeletionEngine]
    SP[SafetyPolicy] --> DE
    DE --> RP[DeletionReport]
    RP --> H[(HistoryStore SQLite)]
    RP --> R[Report: PDF CSV JSON MD]
  end
  M --> SC
  M --> CAT
  M --> DE
  FS[(Filesystem)] --> SC
  DE --> TR[(Trash)]

The scan result is an arena of fixed 64-byte records (parent, children range, flags, three sizes, date, file id) plus one name buffer — about 125–160 bytes per object in practice. Children always follow their parent, so directory totals are one reverse pass. Details: docs/architecture/.

sequenceDiagram
  participant U as User
  participant A as App
  participant S as Scanner threads
  participant K as Kernel
  U->>A: Scan folder
  A->>S: start(root)
  loop each directory
    S->>K: open(O_NOFOLLOW) + fstat
    K-->>S: device (boundary check)
    S->>K: getattrlistbulk
    K-->>S: names, types, sizes, sharing flags
    S->>S: append records under lock
  end
  A-->>U: live counters (10 per second)
  S->>A: tree + issues

# 22. APFS Considerations

Situation What ORIX Clean does
Clones (cp -c, Finder duplicates) Both files show their full allocated size; the private size shows what removal would really free. Recoverable space is given as a range.
Hard links Counted once in totals; never offered as duplicates.
Sparse files Content size and size on disk are both shown; the map can use either.
Transparent compression Allocated size reflects the compressed size; flagged in the tree.
Local snapshots (Time Machine, updates) Not enumerated. Data removed today can stay on disk until the snapshot expires — stated in every report.
Purgeable space The figure macOS reports is shown separately on the Overview; ORIX Clean does not claim it.
Cloud-only files Counted as zero on disk, never read.
Trash Moving to the Trash frees nothing until the Trash is emptied — stated in the preview and in the results.

# 23. Technology Stack

Area Technology
Language Swift 6, strict concurrency, MainActor default isolation in the app
UI SwiftUI, AppKit (panels, Quick Look, workspace), Swift Charts, Canvas
Filesystem getattrlistbulk, getattrlist, clonefile (tests), FileManager.trashItem
Hashing CryptoKit SHA-256, F_NOCACHE streaming reads
Images ImageIO thumbnails, Vision feature prints, Accelerate
Signatures Security framework (SecStaticCode)
Persistence SQLite (C API), WAL
Reports Core Text / Core Graphics PDF, Codable JSON
Project XcodeGen, Swift Package Manager, Swift Testing, XCUITest

Decisions and their reasons (Canvas rather than Metal, SQLite rather than SwiftData, no updater, no helper) are recorded in docs/architecture/overview.md.

# 24. Installation

  1. Download ORIXClean-1.0.0.dmg.
  2. Check it: shasum -a 256 ORIXClean-1.0.0.dmg must match SHA256SUMS.txt.
  3. Open the disk image and drag ORIX Clean to Applications.
  4. Optional: System Settings › Privacy & Security › Full Disk Access › enable ORIX Clean, to include protected folders and the Trash.

# 25. Build Instructions

bash
brew install xcodegen create-dmg          # project generator, DMG builder
Scripts/build.sh generate                 # strings catalogue + Xcode project
Scripts/build.sh core-test                # 56 engine tests (swift test)
Scripts/build.sh debug                    # Debug build
Scripts/build.sh uitest                   # 8 UI scenarios on a disposable disk image
Scripts/sign.sh                           # Release, universal, Developer ID, Hardened Runtime
Scripts/notarize.sh                       # Apple notarization + stapling (existing keychain profile)
Scripts/package.sh && Scripts/notarize.sh dist/ORIXClean-1.0.0.dmg dmg && Scripts/finalize.sh

Requirements: Xcode 27 (Swift 6.2 or later for default actor isolation), macOS 15 SDK or later. More in docs/development/setup.md.

# 26. Testing

Suite Tests What it proves
Scanner 16 Totals equal the kernel's and du; Unicode and 1,500-byte paths; symlink loops; hard links; APFS clones; sparse files; permission denied; files changing during 25 scans; 32,000 files; packages; cancel and pause; mounted volume boundary and removal
Treemap 1 Areas proportional, inside the parent, no overlap, honest aggregate cell, viewport culling
Duplicates 9 Exact groups and digest equal to shasum; a one-byte difference outside the sampled windows; forced sample collisions; hard links; clones; 48 MB files; unreadable and changed files; packages; cancellation
Similar images 3 Resized, recompressed and PNG copies grouped; unrelated images not grouped; crops in thorough mode; corrupt images counted
Safety policy 2 39 protected paths refused, 13 ordinary paths allowed
Deletion 11 No removal without authorization; high-risk acknowledgement; real Trash and Put Back; no overwrite on restore; vanished, replaced and changed items; duplicate verification; a group is never emptied; partial failure; running applications; read-only and full volumes
Catalogue, downloads, Trash 6 Only documented locations; exclusions; every rule documented; build folders need their manifest
Applications 3 Discovery, associated files by certainty, system apps never offered, conservative leftovers
History, reports 3 SQLite round trip, trend, recommendations only from evidence, four renderers, CSV formula neutralisation
Workflows 2 Removal without review is invalid; a declined review removes nothing
UI (XCUITest) 8 First launch; scan, map and search; large files; duplicates; similar images; caches, developer, downloads, Trash; uninstall preview → cancel → real cleanup → Put Back; report export; declined action; light mode, settings, window sizes

Destructive tests run only on temporary directories and on disk images created for the test. Core line coverage: 86.7 % (2,662 of 3,070 lines). Details: docs/development/testing.md.

# 27. Benchmarks

Measured on 2026-11-21, MacBook Pro (Mac17,6, Apple M5 Max, 48 GB), macOS 27.0, internal APFS SSD, release build, warm file cache. Reproduce with orix-bench.

Measurement Result
Scan ~/Library/Developer — 985,558 files, 699,525 folders 17.4 s · 96,757 objects/s · 202 MB peak memory
Scan /Applications — 448,727 files, 64,278 folders, 3 threads 3.04 s · 168,644 objects/s · 8.9 CPU-s · 77 MB
Same, 1 thread / 8 threads 3.45 s, 3.4 CPU-s / 4.07 s, 23.4 CPU-s
Treemap layout, 467,907 nodes at 1600×1000 0.16 ms → 1,798 cells
Duplicate search, sample folder (685 MB hashed) 0.07 s (warm cache; cold reads are bound by the disk)
Similar images, sample folder (31 images, fast mode) 0.21 s
Signed application size 13 MB (universal)

Directory enumeration is bound by the filesystem rather than by cores, which is why the default is three threads (one in Low Power Mode or under thermal pressure). Method and raw numbers: docs/development/performance.md.

# 28. Privacy

  • No account, no network code, no analytics, no advertising, no update check.
  • A scan reads names and metadata. File contents are read only for the analyses you start (hashing, image fingerprints), and stay in memory.
  • Stored locally in ~/Library/Application Support/ORIX Clean/history.sqlite: volume samples, watched-folder sizes, cleanup logs, saved actions. Settings › Privacy & Access › Erase History removes them.
  • macOS asks before the app reads Desktop, Documents, Downloads, removable or network volumes. Full Disk Access is optional and never requested by a prompt of ours.
Settings window

More: docs/PRIVACY.md.

# 29. Accessibility

VoiceOver labels on controls and rows, full keyboard navigation (⌘1–⌘9 sections, ⌘O scan, ⌘↩ review), a list equivalent of the treemap, a colour-blind-safe map palette, risk levels that differ by shape as well as colour, respect for Reduce Motion, and a compact density for the Large Files list. Not verified: no session with a VoiceOver user and no instrumented contrast measurement — see docs/ACCESSIBILITY.md.

# 30. Localization

English and Canadian French (684 strings). Sizes, numbers and dates are formatted for the locale. Known gap: strings have no plural variants yet, so "1 items" / « 1 éléments » can appear.

# 31. Roadmap

  • Plural-aware strings.
  • A Finder-sync or Services entry point ("Analyze with ORIX Clean").
  • Background sampling (login item) so history does not depend on the app being open.
  • Signed automatic updates, if a way is found that keeps the "no network code by default" promise.
  • Sunburst as an alternative view of the map.
  • An Icon Composer layered icon for macOS 26 and later.

# 32. Release Status

Item Status (2026-11-21)
Version 1.0.0, build 1
Source repository Public since 2026-11-21 at the owner's request: https://git.spboucher.ai/orix-clean (gitsrv:srv/git/orix-clean.git)
Developer ID signature Valid — Developer ID Application: Simon-Pierre Boucher (3YM54G49SN), Hardened Runtime, secure timestamp, no entitlements
Apple notarization — app Accepted, submission c2e7cbaa-9410-466f-b912-b9666f5a02b7, ticket stapled
Apple notarization — DMG Accepted, submission ff776146-c989-45d6-99de-c93bb217d8c1, ticket stapled
Gatekeeper accepted, source Notarized Developer ID, for the app, the DMG and the app extracted from the ZIP
ORIXClean-1.0.0.dmg 5,133,225 bytes · SHA-256 c7a8b13f3db143e80d2902882d8fe069d68b6e89a1de212411dffdb7334bcd34
ORIXClean-1.0.0.zip 4,032,161 bytes · SHA-256 126a980dd974136fc958ff7ab9a9b735d68aa2cc162d2729fa4f7d55a03da9fd
SPB Git release Published: https://git.spboucher.ai/orix-clean/releases — tag v1.0.0; DMG, ZIP and SHA256SUMS.txt under https://git.spboucher.ai/releases/orix-clean/v1.0.0/
Download verification The three files were downloaded back from the forge: sizes and SHA-256 match, the DMG's stapled ticket validates and Gatekeeper answers accepted
README rendering on SPB Git Checked from the served page: 12 Mermaid blocks present, all 20 local images served (HTTP 200). Not looked at in a browser.

Full detail: RELEASE-REPORT.md.

# 33. Known Limitations

  • Tested on one configuration only: macOS 27.0 on Apple Silicon. Built for macOS 15.0+ and as a universal binary, but macOS 15, macOS 26 and the Intel slice have never been run.
  • Memory grows with the number of objects (about 125–160 bytes each). A whole-disk scan of a Mac with 15 million files would need roughly 2 GB; not measured.
  • Scan results are not persisted: quitting the app discards the tree.
  • Recoverable space is an estimate. Clones, snapshots and purgeable accounting are outside the app's control; private sizes of a folder are a lower bound.
  • Leftover and name-based associations are inferences, labelled as such.
  • The "unused" notion does not exist in the app: only modification dates and Spotlight's last-used date, when recorded.
  • Root-owned items (some App Store or system-wide installs, /Library caches) cannot be removed, because the app never elevates privileges; the attempt is reported as failed.
  • History is recorded only while the app runs.
  • Similar images: thresholds were calibrated on generated pictures; behaviour on large real photo collections has not been evaluated. Photo libraries are not opened.
  • Similar-image and treemap panning with a mouse wheel are not implemented (pinch and drag only).
  • Increase Contrast is not handled by the custom colours and the map; the density setting applies to the Large Files list only.
  • No automatic updates.
  • Sandbox: the app is not sandboxed (a disk analyzer needs broad read access), so it cannot be distributed on the Mac App Store as is.

# 34. ORIX Ecosystem

App Platform Purpose
ORIX Clean macOS Storage intelligence and cleanup
ORIX Scan iOS / iPadOS Document scanner with on-device OCR
ORIX Calc iOS / iPadOS Universal calculator
ORIX Audio iOS / iPadOS Recorder with on-device transcription

Shared principles: no account, everything on the device, every advertised feature works, every limitation is written down.

# 35. License

All rights reserved © 2026 Simon-Pierre Boucher. See LICENSE.


ORIX Clean — Reclaim Space. Keep Control.
Author: Simon-Pierre Boucher · contact@spboucher.ai
Files main
Menu

Theme