/hulse-py

The Python client for the Hulse platform

Primary LanguagePythonMIT LicenseMIT

hulse-py

Welcome to Hulse's Python Client! Hulse is currently in beta.

With your team's untapped computing power, Hulse makes self-hosting state-of-the-art open-source AI models easier. Start reading below to learn how to use the Hulse API, and set up the Hulse desktop app.

If you have questions or want to talk about anything related to Hulse, you are welcome to join the discussion on Github!

Installation

To install hulse-py:

pip install hulse

or from source, using this repository:

git clone git@github.com:hulsedev/hulse-py.git
pip install -e .

Setting Up

To run the example below, you'll need an active Hulse API key, to be part of a Hulse cluster, and at least one running host in your cluster. If you're not sure whether you have these three things set up, follow these steps:

hulse login
hulse create-cluster --name=<your-cluster-name> --description=<your-cluster-description>
hulse host

An alternative to running the host from the CLI is to download the macOS app (currently only available for Intel CPUs) from the dashboard. You may also manage your clusters and API key there.

Getting Started

At this stage, make sure you've retrieved your API key either using the CLI by running hulse get-api-key or from the dashboard.

Here is a simple example of how to run queries using Hulse, and the Hugging Face Transformers' pipeline:

import hulse

API_KEY = "<your-api-key>"
task = "text-classification"
# tweet https://twitter.com/GretaThunberg/status/1460159146720997377
data = "A reminder: the people in power don’t need conferences, treaties or agreements to start taking real climate action. They can start today. When enough people come together then change will come and we can achieve almost anything. So instead of looking for hope - start creating it."
client = hulse.Hulse(api_key=API_KEY)
client.query(task=task, data=data)

Here, we run a query using a text-classification model, which gives a prediction of the text's sentiment. The provided data comes from this tweet from Greta Thunberg.

Learn more