Skip to navigation
Kumo Studio Guide (Coming Soon)Preliminary DocumentationKumo Studio Code Guide

Installation

View as Markdown

This page describes the Kumo Studio SDK. For local Python execution in an Azure Databricks notebook, follow Azure Databricks Notebook Setup; its package and identity procedure is different.

The Kumo Studio SDK supports Python 3.10 and later with a minimal set of dependencies. Installation steps are identical for the SaaS, Databricks, and Snowflake editions of Kumo Studio.

Before installing, ensure pip is available in your environment.

Install the SDK:

pip install kumo-studio

Verify the installation using its kumoai_studio Python import:

>>> import kumoai_studio as kumo
>>> kumo.__version__
'2.22.0'

The version number shown may differ from the example above.

Export SDK Code from Studio

Use SDK Code to view and export Python code for an existing table, graph, training job, or prediction job.

  1. Open the object’s overview page in your Studio project.
  2. Open its action menu and select SDK Code.
  3. Review the generated Python in View SDK Code, then use the copy or download control to save it. Downloads are named sdk-code-<id>.py.

Before running the script, set API_URL to your Studio endpoint and API_KEY to your API key. Confirm the project in kumo.init and review the operations the script will execute. Warehouse access uses the existing project’s connectors and their assigned identities.

SDK export does not guarantee that every job can be replayed. DiskGraph forecasting batch prediction is not supported by SDK replay; generated code for that case stops before training with an explanation.