scikit-ai
scikit-ai provides task-oriented components for text, images, audio and video that follow the scikit-learn estimator API.
Components are named by what goes in and what comes out, like ImageClassifier, AudioSegmenter or TextToVideoGenerator, so the name tells you what each one does. Since they are estimators, they fit into Pipeline, GridSearchCV and the rest of scikit-learn.
Here is a text classifier that learns from a few labelled examples:
from skai.classification import TextClassifier
X = ['I loved this movie.', 'The product broke after a day.', 'Great service!', 'Never again.']
y = ['positive', 'negative', 'positive', 'negative']
classifier = TextClassifier(k_shot=2, random_state=0, backend='dummy').fit(X, y)
print(classifier.predict(['What a wonderful day.']))['negative']The backend parameter decides what actually runs the task. It can be local open-source models through Hugging Face, the OpenAI API, other providers through LiteLLM, or models hosted on Replicate. The code stays the same.
I used the dummy backend above, which is why it calls a wonderful day negative. It comes with the base install and needs no models or credentials, but its outputs are placeholders, so swap it for a real one to get real predictions.
Note that fit doesn’t train a model unless you ask for fine-tuning. A classifier, for example, keeps your samples as examples for the model to work from.
Install it with pip install scikit-ai, and add an extra such as scikit-ai[hf] or scikit-ai[openai] for the backend you want.