エージェントとマルチエージェント

LangChain

langchain-ai/langchain

An open-source Python framework for building LLM-powered applications and agents through interoperable, chainable components and third-party integrations.

★ 143.6Kスター
⑂ 23.9Kフォーク数
451未解決の問題
Python言語
MITライセンス
Q@project.QualityScore編集部スコア

プロジェクトのスクリーンショット

LangChainのスクリーンショット LangChainのスクリーンショット

概要

LangChain is a framework for building agents and LLM-powered applications. It helps developers chain together interoperable components and third-party integrations to simplify AI application development while aiming to future-proof decisions as the underlying technology evolves. The framework provides a standard interface for models, embeddings, vector stores, and more, allowing developers to work at different levels of abstraction. It is part of a broader ecosystem that includes Deep Agents, LangGraph, and LangSmith.

主な機能

  • Real-time data augmentation via integrations with model providers, tools, vector stores, and retrievers
  • Model interoperability allowing models to be swapped during experimentation
  • Modular, component-based architecture for rapid prototyping
  • Production-ready features with support for monitoring, evaluation, and debugging
  • Flexible abstraction layers from high-level chains to low-level components
  • Seamless integration with the LangChain ecosystem including Deep Agents, LangGraph, and LangSmith

要件、インストール、クイックスタート

The quickstart recommends using the `uv` package manager with the command: `uv add langchain`.

使用方法

A basic quickstart involves importing `init_chat_model` from `langchain.chat_models`, initializing a model (e.g., `init_chat_model("openai:gpt-5.5")`), and invoking it with a text prompt like `model.invoke("Hello, world!")`.

モデルの互換性とユースケース

The framework supports model interoperability and integrates with various model providers. Repository topics indicate relevance to Anthropic, ChatGPT, Gemini, and OpenAI.

ライセンスとリスクに関する注意事項

The repository is licensed under the MIT license.

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

リリースとメンテナンス

Not stated in the repository metadata.

Firecrawl

firecrawl/firecrawl

★ 161.1KTypeScript

RAGFlow

infiniflow/ragflow

★ 86.7KGo