Agents et multi-agents

Agent Development Kit (ADK) 2.0 by Google

google/adk-python

An open-source, code-first Python framework for building, evaluating, and deploying sophisticated AI agents with flexibility and control.

Ouvrir le dépôt officiel ↗ Page d'accueil du projet ↗
★ 21KÉtoiles
⑂ 3,8KForks
554Problèmes ouverts
PythonLangue
Apache-2.0Licence
Q@project.QualityScoreScore éditorial

Captures d'écran du projet

Capture d’écran de Agent Development Kit (ADK) 2.0 by Google Capture d’écran de Agent Development Kit (ADK) 2.0 by Google Capture d’écran de Agent Development Kit (ADK) 2.0 by Google Capture d’écran de Agent Development Kit (ADK) 2.0 by Google Capture d’écran de Agent Development Kit (ADK) 2.0 by Google Capture d’écran de Agent Development Kit (ADK) 2.0 by Google

Vue d’ensemble

Google's Agent Development Kit (ADK) is a Python-based framework designed for creating agentic applications. Version 2.0 introduces a graph-based Workflow Runtime for deterministic execution flows and a Task API for structured agent-to-agent delegation. It allows developers to define individual AI agents and orchestrate them using workflows, supporting complex routing, loops, and human-in-the-loop interactions.

Fonctionnalités clés

  • Code-first Python toolkit
  • Graph-based Workflow Runtime
  • Task API for structured agent-to-agent delegation
  • Support for routing, fan-out/fan-in, loops, and nested workflows
  • Human-in-the-loop capabilities
  • Interactive CLI and Web UI for local running
  • Bi-weekly release cadence

Prérequis, installation et démarrage rapide

Install the core package using pip: `pip install google-adk`. For transitive dependency protection, it is recommended to download and use the constraints file matching your Python version (e.g., `curl -o constraints-3.10.txt https://raw.githubusercontent.com/google/adk-python/main/constraints-3.10.txt` and `pip install google-adk -c constraints-3.10.txt`). Optional integrations can be installed via `pip install "google-adk[extensions]"`.

Utilisation

Define an agent by instantiating the `Agent` class with a name, model, and instruction. Define a `Workflow` by specifying edges that connect agents in a graph. Run applications locally using the command line interface (`adk run path/to/my_agent`) or launch a web interface (`adk web path/to/agents_dir`).

Compatibilité des modèles et cas d’usage

The repository demonstrates usage with the 'gemini-2.5-flash' model. Specific supported model lists are not stated in the repository metadata.

Licence et notes sur les risques

Licensed under the Apache 2.0 License.

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

Version 2.0 introduces the Workflow Runtime (graph-based execution engine) and the Task API (structured agent-to-agent delegation). The project follows a roughly bi-weekly release cadence.

Firecrawl

firecrawl/firecrawl

★ 161,1KTypeScript

LangChain

langchain-ai/langchain

★ 143,6KPython

RAGFlow

infiniflow/ragflow

★ 86,7KGo