No upload step. None.
Scanning, text and image extraction, and classification all run on your Mac through a bundled local runtime on your Apple Silicon GPU. The Gemma 4 model weights download once from public open-weights mirrors after onboarding, then Silo Tidy works offline for the lifetime of the app.
Describe the buckets in plain language.
Silo Tidy ships a sensible default taxonomy, but the real power is editing it: define 'Invoices from vendors', 'Kid's school paperwork', or 'Screenshots of bugs' in natural language and the local model classifies against your labels. The file system you end up with is yours.
Set it once. It stays tidy.
Keep Silo Tidy in your menu bar and it watches the folders you choose — Downloads, Desktop, a project dump — classifying new files as they arrive and organizing them by preview, copy, move, or an Apple Smart Folder that groups without moving a thing.
Documents, images, code, archives, and audio/video metadata are all in scope. Text files parse locally; images are read by Gemma 4 vision; scanned PDFs fall back to on-device Apple Vision OCR.
Edit, add, or remove categories in plain English. The model classifies against your list, not a fixed taxonomy, with a confidence score and a one-line reason for every file.
Dry-run the plan before a single write. Copy or move with conflict-safe renaming. Or generate an Apple Smart Folder (.savedSearch) that groups files by category without moving them — the macOS-native, non-destructive option.
Pick folders to watch and Silo Tidy classifies new arrivals as they land, while it's open. Lives in your menu bar so it's always one click away.