Recently, one of our customers told us about how they have been using some of the common data classification tools and how they are using classical machine learning. It fairly painful to operate, especially on-prem.

I pitched an idea to create a quick proof of concept for a classification service that uses GPT. It is incredibly easy to host and maintain. You can swap the model and the provider underneath based on your usage as well as environment.

I thought it might be a good idea to open-source it and put it out there.

Goal is that, you could just run:

docker run -p 8083:8083 -e OPENAI_API_KEY=your_openai_key ghcr.io/adaptive-scale/superclass:latest

And the classification service is available right away.

Implementation supports text files, pdf, docx, pptx, images like jpeg, png etc. Also, containerized service means, all dependencies like tessaract, documentation processing libraries are just baked into the containers. Just run the classification as a service.

I just verified some usecases, worked like charm for me. May be it is pretty useful for others.

Feel free to give me some feedback.