Agents et multi-agents

Flyte 2: Dynamic, Resilient AI Orchestration

flyteorg/flyte

Flyte is an open-source orchestration framework for reliably building, scaling, and deploying machine learning pipelines, models, and agents using pure Python.

Ouvrir le dépôt officiel ↗ Page d'accueil du projet ↗
★ 7,2KÉtoiles
⑂ 861Forks
159Problèmes ouverts
GoLangue
Apache-2.0Licence
Q@project.QualityScoreScore éditorial

Captures d'écran du projet

Capture d’écran de Flyte 2: Dynamic, Resilient AI Orchestration Capture d’écran de Flyte 2: Dynamic, Resilient AI Orchestration Capture d’écran de Flyte 2: Dynamic, Resilient AI Orchestration Capture d’écran de Flyte 2: Dynamic, Resilient AI Orchestration Capture d’écran de Flyte 2: Dynamic, Resilient AI Orchestration Capture d’écran de Flyte 2: Dynamic, Resilient AI Orchestration

Vue d’ensemble

Flyte 2 is a graduated project of the LF AI & Data Foundation designed to coordinate data, models, and compute for AI workflows. It allows developers to define tasks and workflows natively in Python, supporting both local execution and distributed Kubernetes-native deployments. The framework emphasizes a resilient development experience, offering a terminal user interface (TUI) and browser-based demos to streamline the creation of ML pipelines and model serving endpoints.

Fonctionnalités clés

  • Pure Python workflow definition
  • Asynchronous task execution support
  • FastAPI integration for model serving
  • Terminal User Interface (TUI) for local development
  • Kubernetes-native backend architecture
  • CLI and SDK execution modes
  • LF AI & Data Foundation Graduated Project

Prérequis, installation et démarrage rapide

To install the base package: `uv pip install flyte`. For the full SDK and development tools, refer to the flyte-sdk repository. To install the local TUI experience: `uv pip install flyte[tui]`.

Utilisation

Define a task environment using `flyte.TaskEnvironment`, decorate functions with `@env.task`, initialize with `flyte.init()`, and run using `flyte.run()`. Workflows can be executed via standard Python execution (`python hello.py`) or via the Flyte CLI (`flyte run hello.py main --numbers '[1,2,3]'`). Models can be served using `flyte.serve()` with a FastAPI environment.

Compatibilité des modèles et cas d’usage

Not stated in the repository metadata

Licence et notes sur les risques

Licensed under Apache 2.0.

Editorial verification 2026-08-09: repository URL, owner, description, license and repository statistics were reviewed. License metadata: Apache-2.0. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.

Publication et maintenance

Flyte 2 is now generally available. The repository currently targets the Flyte 2 release, with Flyte 1 maintained on the master branch.

Firecrawl

firecrawl/firecrawl

★ 161,1KTypeScript

LangChain

langchain-ai/langchain

★ 143,6KPython

RAGFlow

infiniflow/ragflow

★ 86,7KGo