Dive Into Langgraph

---
name: dive-into-langgraph
description: A comprehensive guide and reference for building agents using LangGraph 1.0, including ReAct agents, state graphs, and tool integrations.
---

# Dive Into LangGraph

LangGraph 是由 LangChain 团队开发的开源 Agent 框架。v1.0 是稳定版本，框架能力全面升级，支持中间件、状态图、多智能体等高级功能。本 skill 内容由《LangGraph 1.0 完全指南》提供。

**LangGraph 1.0 完全指南**：

- 在线文档：https://www.luochang.ink/dive-into-langgraph/
- GitHub：https://github.com/luochang212/dive-into-langgraph

## 安装依赖

基础依赖：

```bash
pip install \
  langgraph \
  "langchain[openai]" \
  langchain-community \
  langchain-mcp-adapters \
  python-dotenv \
  pydantic
```

## 环境变量

使用模型供应商的大模型需要设置环境变量，推荐使用阿里云百炼（DashScope）的模型：

```bash
# 阿里云百炼 (DashScope)
# 获取地址: https://bailian.console.aliyun.com/
DASHSCOPE_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
DASHSCOPE_API_KEY=your_api_key_here

# 火山方舟 (ARK)
# 获取地址: https://console.volcengine.com/ark/
ARK_BASE_URL=https://ark.cn-beijing.volces.com/api/v3
ARK_API_KEY=your_api_key_here

# 或者其他模型供应商...
```

请将环境变量添加到 `.env` 文件，并填入你的 API key。

## 章节概览

| 序号 | 章节 | 内容简介 | 在线阅读 | 离线阅读 |
|------|------|----------|----------|----------|
| 1 | **快速入门** | 创建你的第一个 ReAct Agent | [在线](https://www.luochang.ink/dive-into-langgraph/quickstart/) | [本地](references/1.quickstart.md) |
| 2 | **状态图** | 使用 StateGraph 创建工作流 | [在线](https://www.luochang.ink/dive-into-langgraph/stategraph/) | [本地](references/2.stategraph.md) |
| 3 | **中间件** | 预算控制、消息截断、敏感词过滤、PII 检测 | [在线](https://www.luochang.ink/dive-into-langgraph/middleware/) | [本地](references/3.middleware.md) |
| 4 | **人机交互** | 使用 HITL 中间件实现人机交互 | [在线](https://www.luochang.ink/dive-into-langgraph/human-in-the-loop/) | [本地](references/4.human_in_the_loop.md) |
| 5 | **记忆** | 短期记忆、长期记忆 | [在线](https://www.luochang.ink/dive-into-langgraph/memory/) | [本地](references/5.memory.md) |
| 6 | **上下文工程** | 使用 State、Store、Runtime 管理上下文 | [在线](https://www.luochang.ink/dive-into-langgraph/context/) | [本地](references/6.context.md) |
| 7 | **MCP Server** | 创建 MCP Server 并接入 LangGraph | [在线](https://www.luochang.ink/dive-into-langgraph/mcp-server/) | [本地](references/7.mcp_server.md) |
| 8 | **监督者模式** | 两种方法：tool-calling、langgraph-supervisor | [在线](https://www.luochang.ink/dive-into-langgraph/supervisor/) | [本地](references/8.supervisor.md) |
| 9 | **并行** | 节点并发、@task 装饰器、Map-reduce、Sub-graphs | [在线](https://www.luochang.ink/dive-into-langgraph/parallelization/) | [本地](references/9.parallelization.md) |
| 10 | **RAG** | 向量检索、关键词检索、混合检索 | [在线](https://www.luochang.ink/dive-into-langgraph/rag/) | [本地](references/10.rag.md) |
| 11 | **网络搜索** | DashScope、Tavily 和 DDGS | [在线](https://www.luochang.ink/dive-into-langgraph/web-search/) | [本地](references/11.web_search.md) |

## 官方资源

- [LangChain 官方文档](https://docs.langchain.com/oss/python/langchain/overview)
- [LangGraph 官方文档](https://docs.langchain.com/oss/python/langgraph/overview)
- [Deep Agents 官方文档](https://docs.langchain.com/oss/python/deepagents/overview)
- [LangMem 官方文档](https://langchain-ai.github.io/langmem/)
- [LangChain GitHub 仓库](https://github.com/langchain-ai/langchain)
- [LangGraph GitHub 仓库](https://github.com/langchain-ai/langgraph)
- [langchain-academy GitHub 仓库](https://github.com/langchain-ai/langchain-academy)