codeflare-sdk

What it is and what it does

CodeFlare SDK is a Python client library that abstracts Kubernetes cluster management and batch job submission, letting developers request compute resources and submit work without writing raw Kubernetes manifests. It handles authentication through multiple methods (auto-detection, OIDC, OpenShift OAuth, token-based, kubeconfig) via its kube-authkit dependency, and provides a high-level Cluster and ClusterConfiguration API for resource provisioning.

The package is designed for developers who need to access high-performance compute on Kubernetes—either cloud-hosted or on-premises—without deep Kubernetes expertise. It depends on kubernetes, ray, pydantic, and rich for its core functionality, and includes demo notebooks and a migration path from deprecated authentication classes. Maintenance is active, with a recent release and no known security vulnerabilities.

Use it for:

• Submit distributed machine learning training jobs to a Kubernetes cluster without writing YAML manifests.
• Provision compute clusters on-demand for batch processing workloads in a Python notebook or script.
• Authenticate to OpenShift or cloud Kubernetes clusters using OIDC or token-based methods.
• Monitor and manage job lifecycle (submission, observation, resource cleanup) from Python code.
• Migrate legacy Kubernetes authentication code to modern kube-authkit patterns.

Worth the install?

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

CodeFlare SDK provides a Python interface for submitting and managing batch jobs and resource requests on Kubernetes clusters, supporting multiple authentication methods and simplifying access to high-performance compute resources.

Yes, if you need to submit batch jobs or provision compute resources on Kubernetes from Python. The low install friction, active maintenance, permissive license, and lack of known vulnerabilities make it a solid choice. The substantial dependency footprint (9 runtime packages) is justified by the abstraction it provides. Not necessary if you already have a working Kubernetes workflow or do not need programmatic cluster provisioning.

Install

codeflare-sdk on PyPI

pip install codeflare-sdk

uv add codeflare-sdk

poetry add codeflare-sdk

Installing codeflare-sdk

Before you install

Low friction installation as a pure Python wheel. Active maintenance with a release 24 days ago. Depends on 9 runtime packages including kubernetes, ray, and kube-authkit; all are established libraries, though the dependency footprint is substantial.

License in practice

Apache-2.0 permissive license allows use in commercial and proprietary projects without viral obligations, though derivative works must retain license attribution.

Quickstart

pip install codeflare-sdk

from codeflare_sdk import Cluster, ClusterConfiguration

cluster = Cluster(ClusterConfiguration(
    name='my-cluster',
    num_workers=2,
))
cluster.apply()

Requires Python 3.11 or 3.12; requires a Kubernetes cluster endpoint and valid authentication credentials (auto-detected from kubeconfig or in-cluster service account by default).

Verify before relying

• Whether the package works with Kubernetes versions beyond what the fact sheet specifies.
• Performance characteristics and scalability limits for large job submissions.
• Whether Ray integration requires Ray cluster pre-deployment or if CodeFlare manages that.

Package facts

Evidence: codeflare_sdk-0.38.2-py3-none-any.whl

Keywords: codeflare, python, sdk, client, batch, scale

Tags

More Distributed Computing packages