Hey everyone,

I wanted to share with you SimpleAI, a self-hosted alternative to OpenAI API.

The aim of this project is to replicate the (main) endpoints of OpenAI API, and to let you easily and quickly plug in any new model. It basically allows you to deploy your custom model wherever you want and easily, while minimizing the amount of changes both on server and client sides.

It's compatible with the OpenAI client so you don't have to change much in your existing code (or can use it to easily query your API).

Wether you like or not the AI-as-a-service approach of OpenAI, I think that project could be of interest to many. Even if you are fully satisfied with a paid API, you might be interested in this if:

• You need a model fine tuned on some specific language and don't see any good alternative, or your company data is too sensitive to send it to an external service

• You’ve developped your own awesome model, and want a drop-in replacement to switch to yours, to be able to A/B test the two approaches.

• You're deploying your services in an infrastructure with an unreliable internet connection, so you would rather have your service locally

• You're just another AI enthusiast with a lot of spare time and free GPU

I've personally really enjoyed how open the ML(Ops) community has been in the past years, and seeing how the industry seems to be moving towards paid API and black box systems can be a bit worrying. This project might be useful to expose great, community-based alternatives.

If that sounds interesting, please have a look at the examples. I also have a blogpost explaining a few more things.

Thank you!