flyteorg

flyte: Orchestrate Python ML Pipelines, Models, and Agents

Flyte 2 is a Python-first runtime for reliably orchestrating machine-learning pipelines, models, and agents at scale. It supports async tasks and model serving, with Kubernetes-native backend infrastructure planned for its open-source release.

Stars
7.8k
Forks
918
Stars, last 30 days
+112
Last commit
2 days ago
Contributors
100+

Our take

Activeregular commits and releases

Health score

75/100

How it is scored
Activity
86
Community
94
Issues
55
Pull requests
63

+112 stars in the last 30 days

  • Merges more pull requests than 91% of the projects we track
  • Releases more often than 89% of the projects we track
  • More contributors than 78% of the projects we track

Flyte 2 is an actively maintained Python orchestration runtime with a clear pipeline, agent, and model-serving scope, but teams needing its open-source distributed backend today should account for its stated availability gap.

Good fit if

  • Your ML or AI workloads are Python-based and benefit from typed, composable task workflows.
  • You want to experiment locally with the SDK, CLI, or Devbox before choosing a deployment path.
  • You need an actively developed project: recent commits and releases are regular, and contributions are spread across many contributors rather than concentrated in one person.

Look elsewhere if

  • You require an open-source, Kubernetes-native Flyte 2 backend immediately, since the repository says it is coming soon.
  • Your workloads or team cannot use Python for task definitions and examples.
  • You need a ready-to-run backend from this repository rather than its SDK and local development experience.
All health signals
Last commit2026-10-31 (2 days ago)
Commits, last 90 days100+
Releases, last 12 months69 (latest v2.0.53, 2026-10-27)
Contributors100+ (top contributor: 8% of commits)
Issues closed, last 90 days5+ (typically closed in 4 days)
Pull requests merged, last 90 days87 (typically merged in 1 day)
Project age7 years

Checked on 2026-11-02 with the GitHub API.

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

7,839

Forks

916

Change since Nov 17, 2026: 0 stars·0 forks

Overview

Flyte 2 is a Python-first orchestration runtime for machine-learning pipelines, models, and agents, designed to run workloads reliably at scale. Its core is the Flyte Python SDK and CLI; the repository describes a Kubernetes-native backend for distributed execution, but says that open-source backend is still coming soon. Flyte 1 is maintained on the master branch.

The project targets Python teams that need to define typed tasks, compose asynchronous workflows, and run or serve them through Flyte. It also supports development from the command line and local Devbox environment. The repository is primarily Go and includes backend protocol definitions, while the user-facing examples are in Python.

Key Features

  • Define tasks with Python functions and typed inputs and outputs.
  • Compose asynchronous task execution using Python async and await.
  • Run workflows locally or through the Flyte CLI.
  • Build task environments with configured Python versions and packages.
  • Serve applications such as a FastAPI model endpoint through FastAPIAppEnvironment.
  • Use a local development experience with the optional TUI and Devbox.
  • Find backend protocol definitions and contribution guidance in the repository's backend documentation.

Use Cases

  • ML engineers can compose and run Python data and model pipelines as typed tasks.
  • AI teams can coordinate agent workflows alongside other Python workloads.
  • Model developers can expose a Python inference endpoint using the serving environment.
  • Developers evaluating Flyte can try workflows locally or in GitHub Codespaces before adopting a deployed backend.

What you need

Detected in the repository

  • Go 1.27.3 or newer (from go.mod)
  • A Dockerfile, so it can run in a container
  • A Helm chart for Kubernetes
  • Automated checks on GitHub Actions

License in plain words

Apache-2.0permissive

  • Commercial use: yes
  • Modify and redistribute: yes
  • You must keep: the license, the NOTICE file and a note of your changes
  • Share your changes: no
  • Includes an explicit patent grant from the contributors:

A summary, not legal advice: the LICENSE file is what applies.

Getting Started

Install the Python package:

uv pip install flyte

See the README for the workflow example, CLI usage, serving setup, and Devbox instructions. The full SDK and development tools are in flyte-sdk.

Alternatives

  • argo-workflows: Argo Workflows orchestrates Kubernetes container jobs, while Flyte provides a Python-first runtime geared toward ML pipelines, models, and agents.
ProjectLanguageLicenseStarsStatus
flyteGoApache-2.07.8kActive
argo-workflowsGoApache-2.017kActive

Considerations

  • The open-source backend for Flyte 2 is not yet available in this repository; the README points to Union.ai for a production-grade backend today.
  • The repository requires Go for development and lists Docker and Helm requirements. The README's user-facing setup is Python-based.
  • The project is actively maintained, with frequent commits and releases, broad contributor participation, and recent pull-request merges. These signals do not remove the need to assess the backend's readiness for your deployment.
  • The project uses the Apache-2.0 license. Review its terms when redistributing or integrating it into a product.

Found this useful?

Share it with someone who would like flyte.

Comparisons