SlothSort

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

SlothSort is a local-first AI file organizer. Choose a folder and it scans up to 100 top-level files, then suggests a category and a more descriptive filename. Review the suggestions and select which files to move into category folders.

It’s for anyone who wants a tidier downloads or documents folder without sorting everything by hand. The AI proposes changes; you stay in control.

Demo

SlothSort runs locally. Start it with Docker Compose, then open http://localhost:5173 in a recent version of Chrome or Edge. Choose a folder to see it in action.

Code

The project source is in this repository. To run it, follow the setup instructions in the README.

Repository link: https://github.com/aLok-1105/slothsort

How I Built It

SlothSort combines a React and Vite frontend with a Node.js and Express backend, packaged with Docker Compose. It uses the browser’s folder-access API to let you select a folder and apply approved moves.

For classification, the backend sends each file’s name and supported text content to Gemma 3 270M running locally through Ollama. It supports text extraction from TXT, Markdown, CSV, JSON, LOG, and text-based PDF files; images are categorized by extension, without OCR. After you download the model, classification runs locally without a cloud AI API. The initial model download requires internet access.

Why Does Open Innovation Matter?

Using an open-weight model with Ollama makes it possible to run classification on my own machine, rather than depend on a closed, cloud-hosted AI API. That gives users more control over where their file content goes and makes the project easier to inspect and adapt. SlothSort sends files to its local backend for classification, and only moves files after the user approves the suggestions.

I left out the optional agent-session and lazy-payoff sections because the repository doesn’t provide a session link or measured time savings.