Python Packages

Three Python packages are published for working with Koios from your own code. All three install from PyPI and are independent of each other — take only the one that matches what you are doing.

pip install koios-client
pip install koios-component-builder
pip install koios-model-utils

Which one you need

Automating or integrating. If you want another system to read live values, create devices in bulk, pull trend data on a schedule, or drive backups, use the API client. It needs an API client ID and secret, created under System — see API Clients.

Adding logic to the platform. If you want to run your own calculation, control algorithm, or protocol inside Koios at a scan rate, that is a component. Build it with the component builder and upload the result. See Components.

Deploying a model. If you have a trained ONNX model, the metadata library lets you describe its inputs, outputs, and training cadence inside the file itself, so uploading it configures the bindings rather than leaving you to enter them. See Models.

Versions

Only the API client is tied to your platform version, because it is generated from the server's API schema. Install the line that matches the Koios you run:

pip install "koios-client~=1.2.0"

The component builder and the metadata library version independently. Newer releases stay readable by older platforms: a component library or model annotated by a newer release may carry fields your version does not use, and those are ignored rather than rejected.

Support

All three are published under the Apache License 2.0 and carry their full reference documentation on their package pages. The pages here cover the same ground for readers already inside Koios.