README.md

Zenbase Python SDK

Installation

Zenbase requires Python ≥3.10. You can install it using your favorite package manager:

pip install zenbase
poetry add zenbase
rye add zenbase

Usage

Zenbase is designed to require minimal changes to your existing codebase and integrate seamlessly with your existing eval/observability platforms. It works with any AI SDK (OpenAI, Anthropic, Cohere, Langchain, etc.).

Labeled Few-Shot Learning Cookbooks:

LabeledFewShot will be useful for tasks that are just one layer of prompts.

Cookbook	Run in Colab
without_integration.ipynb
langsmith.ipynb
arize.ipynb
langfuse.ipynb
parea.ipynb
lunary.ipynb

Bootstrap Few-Shot Learning Cookbooks:

BootstrapFewShot will be useful for tasks that are multiple layers of prompts.

Cookbook	Run in Colab
langsmith.ipynb
arize.ipynb
langfuse.ipynb
parea.ipynb
lunary.ipynb

Predefined Cookbooks:

Cookbook	Description	Run in Colab
Single Class Classifier	Basic implementation of a single class classifier
Synthetic Data Generation	Generate synthetic data for single class classification
Classifier with Synthetic Data	Train and test a single class classifier using synthetic data

Development setup

This repo uses Python 3.10 and rye to manage dependencies. Once you've gotten rye installed, you can install dependencies by running:

rye sync

And activate the virtualenv with:

. .venv/bin/activate

You can run tests with:

rye test # pytest -sv to see prints and verbose output
rye test -- -m helpers # integration tests with helpers