This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
What I Built
SlothSort helps tidy a folder without the tedious drag-and-drop routine. It looks at filenames, asks a local AI model to suggest categories and clearer names, and presents the proposed moves for review. Files are only moved after the user confirms the plan.
Demo
The project currently runs locally. Install the dependencies and follow the Ollama setup in the README. There is no public demo yet.
Code
View the source code on GitHub.
How I Built It
The interface uses Streamlit, and the app asks a local Ollama server to run the open-weight qwen2.5:3b model. The model sees filenames only and returns a category plus a suggested name. The Python app validates that response, sanitizes names, checks for conflicts, and requires explicit approval before moving anything.
Why Does Open Innovation Matter?
Filenames can contain personal or work-related details. Local inference keeps them on the user's machine rather than sending them to a cloud API. After the one-time model download, there is no per-file inference fee. The prompt, category list, and model choice are all open to inspection and change.
My Agent Session
I do not have a shareable DevRelay session link for this project.
The Lazy Payoff
SlothSort turns a pile of filenames into a proposed folder structure in one step, while keeping the final decision with the user. Add a real time-saved estimate here after trying it on a folder and timing the manual alternative.
Top comments (0)