Beaver is...
✔ The whole package • it's a framework to develop, deploy, and maintain machine learning models. And that includes feature engineering.
✔ No fuss • there's an SDK to do stuff, and a UI to see stuff.
✔ Online-first • it is designed for online machine learning models, while also supporting batch models.
✔ Opinionated • it encourages you to process data with SQL and define models in Python.
✔ Interfaces all the way down • you can plug in your existing message broker, stream processor, model store, etc. This is called a hexagonal architecture in software design.
✔ Batteries included • default infrastructure and monitoring are provided if needed.
There is a pre-built image you can pull and run 🐳
docker pull ghcr.io/online-ml/beaver:latest
docker run ghcr.io/online-ml/beaver
You can also build the provided docker-compose.yaml
yourself:
git clone git@github.com:online-ml/beaver.git
cd beaver
docker-compose up
Go to http://localhost:8501 to access the user interface.
This is all you need to run Beaver. Check out the examples to see how to use it.
Beaver is free and open-source software licensed under the 3-clause BSD license.