SlothSort: A Local AI That Tidies Your Downloads Folder and Never Moves a File Without Asking

Description and use

SlothSort uses a local Gemma 3 4B model via Ollama to scan a downloads folder, suggest structured filenames and categories, and, after user approval, safely move files while logging actions and supporting undo.

Problem

Sorting files in a downloads folder is tedious and error‑prone, especially when files are named generically and users must manually rename and move them.

Outcome

In a 29‑file synthetic dataset, SlothSort achieved 92.6% category accuracy on the first run and 88.9% on the second, with 100% of approved moves completed and all unsafe‑operation probes blocked. The system prevented accidental overwrites, ensured file integrity, and allowed undo of moves.

Workflow

• Point SlothSort at a target folder
• Scan supported files read‑only
• Extract text from PDFs, images, and text files
• Send extracted text to local Gemma 3 4B via Ollama
• Receive structured JSON with category, suggested filename, explanation, and confidence
• Validate the JSON against a schema and safety rules
• Present the full plan to the user for review and editing
• Await explicit user approval
• Execute safe file moves, logging actions in SQLite
• Allow undo of approved moves if original location remains available

Evidence

Point it at a folder, and SlothSort scans supported files, extracts available text, and asks an open-weight AI model running on your own machine to suggest meaningful filenames and categories.

Source

Nothing moves until you approve it. Existing files are never intentionally overwritten or deleted, and successful moves can be undone when the original location remains available and the integrity checks pass.

Source

I created a 29-file synthetic demo dataset containing invoices, receipts, assignments, notes, screenshots, photos, two pairs of exact duplicates, two unsupported files, and a prompt-injection example.

Source

Gemma 3 4B category accuracy 25/27 (92.6%)

Source