Apex Duplicate Analyzer
Deterministic. Intelligent. Offline-First.
Apex Duplicate Analyzer is more than a simple file hash checker; it is a high-performance visual analysis engine built to identify exact duplicates, compression variants, and burst-shot captures across complex directory trees. Combining perceptual hashing with locally hosted Vision LLMs (via Ollama), our solution delivers zero-latency visual audits without sending private photos over external cloud networks.
Featuring non-destructive Staged Quarantine Isolation, Automatic Keeper Resolution Heuristics, and Natural-Language Visual Retrieval, Apex prioritizes user data integrity above all else. All scanning, Hamming distance matching, and local AI reasoning execute strictly on your device.
216e6eb74649ade994fb0fe47d446cbe9155c18cc1c0331c8c3b4e401aa0cd61
Perceptual Hashing (pHash)
Multi-threaded OpenCV discrete cosine transform engine calculates visual fingerprints, detecting resaves, crops, and downscaled images in milliseconds.
Local Ollama Vision Integration
Search archives using conversational queries like "screenshots with terminal code" via locally hosted VLMs (such as LLaVA or Qwen2.5-VL).
Synchronized Visual Audit
Zinc-themed side-by-side comparison workspace with live metadata inspection, dimension matching, and granular single-duplicate actions.
Staged Quarantine Isolation
Non-destructive cleanup stages duplicates into an isolated, indexed directory backed by SQLite for instant selective or full-batch restoration.
Auto-Keeper Heuristics
Automate decision making across hundreds of clusters with selectable rules prioritizing highest resolution, least compression, or file age.
Air-Gapped Privacy
Zero telemetry, zero external API keys, and zero cloud uploads. Your personal and proprietary media never leaves your physical local storage.
See Apex in Action
An intuitive side-by-side audit environment built for high-throughput media libraries. Inspect micro-differences, execute granular per-item actions, and interact with your local Ollama vision models in real time.
01. Automated Sieve Engine
Ingests thousands of assets, applies pHash fingerprinting, and assigns auto-keeper rules by resolution.
02. Symmetrical Comparison
Simultaneously inspects protected source keepers against duplicate candidates with live metadata telemetry.
03. Granular Action Bar
Swap keepers, isolate individual candidates into recoverable staging, or route directly to the OS Recycle Bin.
04. Local Vision AI Dock
Integrated Ollama agent queries multimodal models locally to explain visual variances and perform natural-language search.
Auto-Keeper Decision Rules
Eliminate the friction of triaging hundreds of duplicate sets. Apex evaluates each cluster against user-selectable priority rules, automatically designating the optimal master file while preserving manual override capabilities.
Highest Resolution
Calculates total pixel density (Width × Height). Retains full-fidelity uncompressed masters and isolates web-downscaled versions.
Largest File Size
Identifies the candidate consuming the most raw disk bytes. Protects assets saved with the least aggressive JPEG/WebP compression ratios.
Newest Timestamp
Evaluates the latest filesystem modification date. Ideal for preserving newly retouched, filtered, or color-graded iterations of source photography.
Oldest Timestamp
Identifies the earliest filesystem creation/edit record. Guarantees protection for the untouched original camera ingest over downstream duplicates.
Tuning Perceptual Similarity
Hamming distance measures the bit variance between perceptual hashes. The real-time slider lets you toggle between strict clone verification and broad similarity clustering without re-reading files from disk.
Captures exact bit re-saves, slight format conversions (PNG to JPEG), and metadata additions. Near zero risk of false groupings.
Flags resized copies, social media compression artifacts, and minor edge cropping while ignoring unrelated compositions.
Clusters rapid camera burst shots taken fractions of a second apart, heavy crops, color adjustments, and subject micro-movements.
Built-In Quarantine Manager: Safe Audit & 1-Click Restore
Unlike cleaners that immediately wipe files, Apex isolates duplicates into an indexed .quarantine folder while mirroring your original subfolder structure. Each record is stored in duplicates.db with its original filesystem path and quarantine timestamp.
├── photos/IMG_001.JPG [KEEPER]
└── .quarantine/
└── photos/IMG_001_copy.JPG
original_path → photos/...
status: Ready to restore
Who Uses Apex Duplicate Analyzer?
Photographers & Creators
Sweep multi-gigabyte SD card dumps after photo shoots. Cluster continuous-shot burst sequences and instantly isolate identical exports without wading through thousands of thumbnails manually.
Archive Managers & Studios
Consolidate external hard drives, NAS volumes, and legacy photo backups. The local SQLite cache prevents re-hashing unchanged folders across weekly maintenance runs.
Privacy-Conscious Users
Audit personal, legal, or sensitive photo libraries with complete peace of mind. No telemetry, no background analytics, and no photos uploaded to cloud servers.
How It Works
Clean and organize mass media collections in three deterministic steps designed to eliminate accidental data destruction.
Index & Hash
Select a single directory or multi-select individual files. The background engine checks the SQLite cache for unchanged assets and executes SHA-256 and pHash computation without blocking the interface.
Inspect & Tune
Review clusters side-by-side with live Hamming distance readouts. Adjust the similarity slider in real time (0–15) to calibrate detection strictness or swap keepers with a single click.
Quarantine or Purge
Stage candidates into a reversible .quarantine hierarchy to audit disk space savings safely. Restore any file to its original location at will, or route redundant copies straight to the OS Recycle Bin.
Engineered Beyond Conventional Cleaners
Why perceptual discrete cosine transforms and local LLM evaluation provide a superior audit tier compared to legacy cleaners.
| Architecture & Features | Traditional File Cleaners | Cloud AI Applications | Apex Duplicate Analyzer |
|---|---|---|---|
| Near-Duplicate & Crop Detection | Fails (Byte-only SHA/MD5) | Supported | ✓ Full DCT Perceptual Hashing (pHash) |
| Natural-Language Image Search | Not Supported | Paid Cloud API | ✓ 100% Free Local Ollama Agent |
| Data Privacy & Telemetry | Uncertain Analytics | Uploads Private Photos | ✓ 100% Air-Gapped Local Processing |
| Data Safety Architecture | Hard Immediate Delete | Proprietary Cloud Trash | ✓ Reversible Staged Quarantine + Recycle Bin |
| Supported Image Types | Strictly JPEG / Fixed | Varies | ✓ .JPG, .JPEG, .PNG, .WEBP |
| Persistent Indexing Cache | None (Re-scans from scratch) | Remote Server | ✓ Embedded SQLite (duplicates.db) |
Technical Specifications
http://localhost:11434
Frequently Asked Questions
127.0.0.1:11434) without transmitting data outside your machine.
.quarantine subfolder while saving their exact original relative paths in the embedded SQLite store. You can open the Quarantine Reviewer at any time to inspect full-resolution previews and restore files to their exact original location with a single click.
