Sloth Tidy:let your AI clean your files
Description and use
Sloth Tidy organizes messy files automatically: it reads the first few words of each file, a local AI model (Gemma via Ollama) proposes a folder and clean name, shows a plan before moving anything, and supports undo. Three of four demo files sorted correctly; one needed a new Food category. Full workflow completed in seconds versus minutes of manual work.
Problem
Files are named like IMG_2231 and final_FINAL2, and the user never cleans them up, requiring manual sorting of every file into folders and renaming.
Outcome
For 4 demo files, three were sorted correctly on approval; one required adding a Food category. The full workflow completed in seconds versus minutes of manual effort. Undo restored all files to original locations.
Workflow
- Read first few words of each file
- AI model (Gemma) suggests a target folder and clean name
- Show user a plan before moving anything
- User approves or rejects the plan
- Move files to chosen folders
- If needed, add missing categories (e.g., Food)
- Run undo to restore original locations
Evidence
My files are named things like IMG_2231 and final_FINAL2, and I never clean them up. So I built Sloth Tidy. It reads the first few words of each file, and a local AI model picks a folder and a clean name. It shows me a plan first, moves nothing until I say yes, and has an undo command.
I no longer open every file, decide where it goes, and rename it. I just check the plan. For 4 demo files, sorting by hand took me about [X minutes], and Sloth Tidy made the plan in about [Y seconds].
Three of the four went to the right folder straight away. The recipe went to 'Other' because I had no food folder, so I added a Food category. Then I ran undo, and all 4 files went back to where they started.
It is free, with no API key and no bill. I can swap the model by changing one line. On my own laptop it can run offline, so private files never leave my computer.
It reads only text files, and uses just the file name for PDFs and images.
- Record date: 2026-10-24T18:26:09.032Z