FileTidy
Sorts your Downloads the way you already file
A desktop app that tidies messy folders (Downloads, Desktop…) into the folders you have already organized, by learning how you file. Everything runs offline, and nothing moves without your approval.
- Platform
- Windows, macOS, Linux (Python 3.11+, Qt)
- Interface
- English, French, Spanish
- Network
- None: no telemetry, no downloads
- Status
- In development
Learns from your own folders
FileTidy indexes the folders you have already sorted. For each file to tidy, the 8 most similar filed files vote for a destination, and a move is pre-checked only when that vote proved right at least 98% of the time on your own files; the confidence shown is the cautious lower bound of that measurement.
No model required
The built-in vectorizer is plain numpy, about 0.14 ms per file. On the project’s test corpus (154 files in nine folders with overlapping vocabulary) it files every document correctly from its text. Optional local models help with long or multilingual files.
You decide, and can undo
Every proposal shows its confidence and its reason, and doubtful rows start unchecked. Each batch of moves is journaled before it runs and can be undone from the History tab.
Never destroys a file
Files are never deleted or overwritten: a name collision becomes “name (1).ext”.