Unified AI-powered file organization platform — combining tag-based library management, 7-level classification, LLM intelligence, cleanup tools, duplicate detection, and media metadata into a single premium dark-themed desktop app.

UniFile merges the best ideas from five file organization projects into one cohesive tool:

git clone https://github.com/SysAdminDoc/UniFile.git
cd UniFile
python run.py  # Auto-installs all dependencies + Ollama on first run

On first launch, UniFile will:

1. Install PyQt6, SQLAlchemy, and other dependencies if missing
2. Download and install Ollama if not found
3. Start the Ollama server and pull the qwen2.5:7b model
4. Open the GUI with LLM mode enabled and ready

No manual setup required.

Full tag-based file management adapted from TagStudio's SQLAlchemy models:

The tag library stores data in .unifile/unifile_tags.sqlite within your library directory — non-destructive, no files are modified.

Movie and TV metadata lookup powered by TMDb, OMDb, and TVMaze APIs (adapted from mnamer):

Alternative local AI backend using Nexa SDK (adapted from Local-File-Organizer):

Enable in Settings > Ollama LLM > Alternative Backend: Nexa SDK. Requires pip install nexaai.

• Progressive hash-based detection: Size > Prefix hash > Suffix hash > Full SHA-256
• Perceptual image hashing for near-duplicate photos
• Side-by-side comparison dialog
• Configurable similarity tolerance

• EXIF metadata extraction (date, camera, GPS)
• Photo map view with geotagged markers (Leaflet)
• AI event grouping — cluster photos by vision descriptions
• Face detection and person-based organization (optional)
• Thumbnail grid view with flow layout

• Monitor folders and auto-organize new files
• System tray integration with minimize-to-tray
• Watch history log with timestamps

• Preview before apply — full destination tree preview before any files move
• Protected paths — system folders guarded at scan, apply, and delete layers
• Safe merge-move — merging into existing folders preserves all files
• Progressive hash dedup — SHA-256 + perceptual hash prevents overwrites
• Full undo log — every operation recorded with one-click rollback
• CSV audit trail — every classification logged with timestamp, method, confidence
• Crash handler — unhandled exceptions saved to crash log with MessageBox notification

unifile/
├── __init__.py          # Package version
├── __main__.py          # Entry point with crash handler
├── bootstrap.py         # Auto-dependency installer
├── config.py            # Settings, themes, protected paths
├── categories.py        # 384+ category definitions
├── classifier.py        # 7-level classification engine
├── engine.py            # Rule engine, scheduler, templates
├── naming.py            # Smart rename logic
├── metadata.py          # File metadata extraction
├── ollama.py            # Ollama LLM integration
├── photos.py            # Photo/EXIF/face processing
├── files.py             # PC file organizer logic
├── cache.py             # Classification cache, undo log
├── models.py            # Data models (ScanItem, etc.)
├── workers.py           # QThread workers for scanning/applying
├── plugins.py           # Plugin system, profiles, presets
├── profiles.py          # Scan profile management
├── cleanup.py           # Cleanup scanners (6 types)
├── duplicates.py        # Duplicate detection engine
├── widgets.py           # Custom Qt widgets (charts, map, preview)
├── main_window.py       # Main application window (UniFile class)
├── nexa_backend.py      # Nexa SDK AI backend (LLaVA + Llama 3.2)
├── scan_mixin.py        # Scan pipeline + auto-tag integration
├── apply_mixin.py       # Apply/move operations
├── tagging/
│   ├── db.py            # SQLAlchemy engine, Base, PathType
│   ├── models.py        # Tag, Entry, Folder, ValueType ORM models
│   └── library.py       # TagLibrary CRUD API
├── media/
│   └── providers.py     # TMDb, OMDb, TVMaze API + guessit parser
└── dialogs/
    ├── tag_library.py   # Tag Library browser panel
    ├── media_lookup.py  # Media Lookup panel (movie/TV search)
    ├── cleanup.py       # Cleanup results dialog
    ├── duplicates.py    # Duplicate comparison dialog
    ├── editors.py       # Rules/category editors
    ├── settings.py      # Settings dialog
    ├── theme.py         # Theme preview dialog
    └── tools.py         # Tool dialogs

Click Settings > Ollama LLM to configure:

Recommended models:

6 dark themes with live preview: Steam Dark (default), Catppuccin Mocha, OLED Black, GitHub Dark, Nord, Dracula.

python run.py                                    # Launch GUI
python run.py --source "C:/Users/You/Downloads"  # Auto-scan a folder
python run.py --profile MyProfile --auto-apply   # Automated profile scan
python run.py --dry-run --profile MyProfile      # Simulate without moving
python -m unifile                                 # Alternative launch

• Python 3.10+
• 8 GB RAM minimum (for Ollama LLM models)
• ~5 GB disk space for the default qwen2.5:7b model
• Internet connection for first launch only (Ollama install + model download)
• Works without Ollama — falls back to rule-based engine automatically

All dependencies auto-installed: PyQt6, SQLAlchemy, rapidfuzz, psd-tools, Pillow, and more.

UniFile is built directly on FileOrganizer's foundation. If you only need folder sorting and cleanup without tag-based library management or media metadata, FileOrganizer is the lighter option.

• Media Lookup — TMDb/TVDb/TVMaze metadata panel (from mnamer's provider system)
• Nexa SDK Backend — Alternative AI backend with Llama 3.2 + LLaVA vision (from Local-File-Organizer)
• Category Presets — Per-directory config, import/export, extension-based presets (from classifier)
• Search Query Language — Advanced tag search with boolean operators (tag:, ext:, field:, AND/OR/NOT)
• Preview Panel — Rich file preview with tag overlay, image thumbnails, metadata, and field display

MIT License — see LICENSE for details.