kfp

What it is and what it does

Kubeflow Pipelines is a Python SDK for building machine learning workflows as directed acyclic graphs of containerized tasks. You define pipeline components using Python decorators, compose them into workflows, and submit them to a Kubeflow backend running on Kubernetes. The SDK handles task orchestration, artifact management, and provides integration with Google Cloud services through its dependencies on google-cloud-storage and google-auth.

The package is designed for teams building reproducible, scalable ML systems. It lets you parameterize pipelines, schedule recurring runs, organize executions into experiments, and visualize results through the Kubeflow Dashboard. Dependencies include Kubernetes client libraries, protocol buffers for serialization, and CLI tools via click for command-line interaction.

Use it for:

• Define multi-step ML training workflows with data preprocessing, model training, and evaluation as separate containerized tasks.
• Schedule recurring pipeline runs on a Kubernetes cluster with different parameters or datasets.
• Organize and track multiple pipeline executions as experiments with artifact storage and metadata logging.
• Build reusable pipeline components that can be composed into different workflows across teams.
• Deploy production ML systems where task dependencies, retries, and resource allocation are managed by Kubernetes.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Kubeflow Pipelines is a Python SDK for defining, deploying, and managing machine learning workflows as containerized task graphs on Kubernetes clusters.

Yes, if you have a Kubeflow backend deployed on Kubernetes and need to orchestrate multi-step ML workflows. The package is actively maintained, permissively licensed, and has low install friction. No known vulnerabilities. Install only if you already have or plan to set up a Kubeflow infrastructure; it is not a standalone ML framework.

Install

kfp on PyPI

pip install kfp

uv add kfp

poetry add kfp

Installing kfp

Before you install

Low friction installation with a pure-Python wheel. The package is actively maintained with a recent release and has 4183 repository stars. Supports Python 3.9 through 3.13.

License in practice

Licensed under Apache Software License (permissive), which allows commercial and private use with minimal restrictions.

Quickstart

pip install kfp

from kfp import dsl
import kfp

@dsl.component
def add(a: float, b: float) -> float:
    return a + b

@dsl.pipeline(name='Addition pipeline')
def add_pipeline(a: float = 1.0, b: float = 7.0):
    task = add(a=a, b=b)

client = kfp.Client(host='<my-host-url>')
client.create_run_from_pipeline_func(add_pipeline, arguments={'a': 7.0, 'b': 8.0})

Requires a running Kubeflow Pipelines backend or Kubernetes cluster with Kubeflow deployed; a valid host URL is needed to connect the client.

Verify before relying

• Whether the package requires specific Kubernetes version compatibility or cluster configuration.
• Performance characteristics when scaling to large numbers of pipeline tasks or concurrent runs.
• Detailed compatibility matrix between kfp versions and Kubeflow backend versions.

Package facts

Evidence: kfp-2.17.0-py3-none-any.whl

Tags

More Software Development packages