本指南向您展示如何在本地运行 LangGraph 应用程序。

1. 安装 LangGraph CLI & 创建一个 LangGraph 应用

从新的langgraph项目python模板创建新的应用程序。此模板演示了一个可以用自己的逻辑扩展的单节点应用程序。

1
2
3

# Python >= 3.11 is required.
pip install --upgrade "langgraph-cli[inmem]"
langgraph new /Users/king/workspace-test/llm/chatbot-server --template new-langgraph-project-python

如果遇见问题，也可以尝试下面的方式

# 找到 macOS 系统根证书export SSL_CERT_FILE=/etc/ssl/cert.pem

# 再执行
langgraph new /Users/king/workspace-test/llm/chatbot-server --template new-langgraph-project-python

如果您使用时langgraph new没有指定模板，您将看到一个交互式菜单，可让您从可用模板列表中进行选择。

将/src/agent/graph.py的内容：

"""LangGraph single-node graph template.

Returns a predefined response. Replace logic and configuration as needed.
"""

from __future__ import annotations

from dataclasses import dataclass
from typing import Any, Dict, TypedDict

from langgraph.graph import StateGraph
from langgraph.runtime import Runtime


class Context(TypedDict):
    """Context parameters for the agent.

    Set these when creating assistants OR when invoking the graph.
    See: https://langchain-ai.github.io/langgraph/cloud/how-tos/configuration_cloud/
    """

    my_configurable_param: str


@dataclass
class State:
    """Input state for the agent.

    Defines the initial structure of incoming data.
    See: https://langchain-ai.github.io/langgraph/concepts/low_level/#state
    """

    changeme: str = "example"


async def call_model(state: State, runtime: Runtime[Context]) -> Dict[str, Any]:
    param = (
        runtime.context.get("my_configurable_param")
        if runtime.context is not None
        else "default-value"
    )
    return {"changeme": f"output from call_model. Configured with {param}"}

# Define the graph
graph = (
    StateGraph(State, context_schema=Context)
    .add_node(call_model)
    .add_edge("__start__", "call_model")
    .compile(name="New Graph")
)

2. 安装依赖项

在新的 LangGraph 应用程序的根目录中，以edit模式安装依赖项，以便服务器使用您的本地更改：

1 2	cd /Users/king/workspace-test/llm/chatbot-server pip3 install -e .

3. 创建`.env`文件

您将在新LangGraph应用程序的根目录中找到一个.env.示例。在新的LangGraph应用程序的根目录中创建.env文件，并将.env.example文件的内容复制到其中，填写必要的API键：

1	LANGSMITH_API_KEY=lsv2...

这里按需去配置即可。

4. 启动 LangGraph 服务器

在本地启动 LangGraph API 服务器：

1	langgraph dev

➜  chatbot-server langgraph dev
INFO:langgraph_api.cli:

        Welcome to

╦  ┌─┐┌┐┌┌─┐╔═╗┬─┐┌─┐┌─┐┬ ┬
║  ├─┤││││ ┬║ ╦├┬┘├─┤├─┘├─┤
╩═╝┴ ┴┘└┘└─┘╚═╝┴└─┴ ┴┴  ┴ ┴

- 🚀 API: http://127.0.0.1:2024
- 🎨 Studio UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
- 📚 API Docs: http://127.0.0.1:2024/docs

This in-memory server is designed for development and testing.
For production use, please use LangGraph Platform.


2025-08-25T14:50:14.554870Z [info     ] Using langgraph_runtime_inmem  [langgraph_runtime] api_variant=local_dev langgraph_api_version=0.3.1 thread_name=MainThread
2025-08-25T14:50:14.561245Z [info     ] Using auth of type=noop        [langgraph_api.auth.middleware] api_variant=local_dev langgraph_api_version=0.3.1 thread_name=MainThread
2025-08-25T14:50:14.562535Z [info     ] Starting In-Memory runtime with langgraph-api=0.3.1 and in-memory runtime=0.8.1 [langgraph_runtime_inmem.lifespan] api_variant=local_dev langgraph_api_version=0.3.1 langgraph_runtime_inmem_version=0.8.1 thread_name=asyncio_0 version=0.3.1
2025-08-25T14:50:14.581597Z [info     ] Starting thread TTL sweeper with interval 5 minutes [langgraph_api.thread_ttl] api_variant=local_dev interval_minutes=5 langgraph_api_version=0.3.1 strategy=delete thread_name=asyncio_1
2025-08-25T14:50:14.583084Z [info     ] Registering graph with id 'agent' [langgraph_api.graph] api_variant=local_dev graph_id=agent langgraph_api_version=0.3.1 thread_name=asyncio_1
2025-08-25T14:50:14.647534Z [info     ] Starting 1 background workers  [langgraph_runtime_inmem.queue] api_variant=local_dev langgraph_api_version=0.3.1 thread_name=asyncio_0
2025-08-25T14:50:14.647917Z [info     ] Worker stats                   [langgraph_runtime_inmem.queue] active=0 api_variant=local_dev available=1 langgraph_api_version=0.3.1 max=1 thread_name=asyncio_1
2025-08-25T14:50:14.654076Z [info     ] 5 changes detected             [watchfiles.main] api_variant=local_dev langgraph_api_version=0.3.1 thread_name=MainThread
Server started in 1.04s
2025-08-25T14:50:14.861974Z [info     ] Server started in 1.04s        [browser_opener] api_variant=local_dev langgraph_api_version=0.3.1 message='Server started in 1.04s' thread_name='Thread-2 (_open_browser)'
🎨 Opening Studio in your browser...
2025-08-25T14:50:14.862062Z [info     ] 🎨 Opening Studio in your browser... [browser_opener] api_variant=local_dev langgraph_api_version=0.3.1 message='🎨 Opening Studio in your browser...' thread_name='Thread-2 (_open_browser)'
URL: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
2025-08-25T14:50:14.862123Z [info     ] URL: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024 [browser_opener] api_variant=local_dev langgraph_api_version=0.3.1 message='URL: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024' thread_name='Thread-2 (_open_browser)'
2025-08-25T14:50:15.149840Z [info     ] Queue stats                    [langgraph_runtime_inmem.queue] api_variant=local_dev langgraph_api_version=0.3.1 max_age_secs=None med_age_secs=None n_pending=0 n_running=0 thread_name=asyncio_0
2025-08-25T14:50:17.391294Z [info     ] Getting auth instance: None    [langgraph_api.auth.custom] api_variant=local_dev langgraph_api_version=0.3.1 langgraph_auth=None method=POST path=/assistants/search thread_name=MainThread
2025-08-25T14:51:14.796166Z [info     ] Worker stats                   [langgraph_runtime_inmem.queue] active=0 api_variant=local_dev available=1 langgraph_api_version=0.3.1 max=1 thread_name=asyncio_1
2025-08-25T14:51:15.298560Z [info     ] Queue stats                    [langgraph_runtime_inmem.queue] api_variant=local_dev langgraph_api_version=0.3.1 max_age_secs=None med_age_secs=None n_pending=0 n_running=0 thread_name=asyncio_0

该langgraph dev命令以内存模式启动 LangGraph 服务器。此模式适用于开发和测试。对于生产用途，请部署 LangGraph 服务器并使其能够访问持久存储后端。更多信息，请参阅部署选项。

5. 在 LangGraph Studio 中测试你的应用程序

LangGraph Studio是一个专门的 UI，您可以连接到 LangGraph API 服务器，在本地可视化、交互和调试您的应用程序。通过访问命令输出中提供的 URL，在 LangGraph Studio 中测试您的图表langgraph dev：

1	> - LangGraph Studio Web UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024

对于在自定义主机/端口上运行的 LangGraph 服务器，请更新 baseURL 参数。

使用--tunnel带有标志的命令来创建安全隧道，因为 Safari 在连接到本地主机服务器时有限制：
1
langgraph dev --tunnel

6. 测试 API

安装 LangGraph Python SDK：

1	pip3 install langgraph-sdk

向助手发送消息（无线程运行）：

from langgraph_sdk import get_client
import asyncio

client = get_client(url="http://localhost:2024")

async def main():
    async for chunk in client.runs.stream(
        None,  # Threadless run
        "agent", # Name of assistant. Defined in langgraph.json.
        input={
        "messages": [{
            "role": "human",
            "content": "What is LangGraph?",
            }],
        },
    ):
        print(f"Receiving new event of type: {chunk.event}...")
        print(chunk.data)
        print("\n\n")

asyncio.run(main())

客户端打印日志：

Receiving new event of type: metadata...
{'run_id': '0198e1c6-e74e-73f0-b3de-a68f59564c32', 'attempt': 1}



Receiving new event of type: values...
{'changeme': 'output from call_model. Configured with default-value'}

server端日志：

2025-08-25T15:10:24.742293Z [info     ] Getting auth instance: None    [langgraph_api.auth.custom] api_variant=local_dev langgraph_api_version=0.3.1 langgraph_auth=None method=POST path=/runs/stream request_id=0a96deec-40bb-48a1-a5ca-5ea9854e8042 thread_name=MainThread
2025-08-25T15:10:24.742589Z [info     ] Creating thread                [langgraph_runtime_inmem.ops] api_variant=local_dev langgraph_api_version=0.3.1 method=POST path=/runs/stream request_id=0a96deec-40bb-48a1-a5ca-5ea9854e8042 thread_id=UUID('1d81ffc1-ea38-4aa8-89f1-f9bb43ff14d0') thread_name=asyncio_1
2025-08-25T15:10:24.742894Z [info     ] Created run                    [langgraph_api.models.run] after_seconds=0 api_variant=local_dev assistant_id=fe096781-5601-53d2-b2f6-0d3403f7e9ca checkpoint_id=None if_not_exists=reject langgraph_api_version=0.3.1 method=POST multitask_strategy=enqueue path=/runs/stream request_id=0a96deec-40bb-48a1-a5ca-5ea9854e8042 run_create_ms=1 run_id=0198e1c7-cde6-775d-9fc2-5fffa7841986 run_put_ms=0 stream_mode=['values'] stream_resumable=False temporary=True thread_id=None thread_name=MainThread
2025-08-25T15:10:25.334572Z [info     ] Starting background run        [langgraph_api.worker] api_variant=local_dev assistant_id=fe096781-5601-53d2-b2f6-0d3403f7e9ca graph_id=agent langgraph_api_version=0.3.1 request_id=0a96deec-40bb-48a1-a5ca-5ea9854e8042 resumable=False run_attempt=1 run_creation_ms=1 run_id=0198e1c7-cde6-775d-9fc2-5fffa7841986 run_queue_ms=591 run_started_at=2025-08-25T15:10:25.333924+00:00 run_stream_start_ms=0 temporary=True thread_id=1d81ffc1-ea38-4aa8-89f1-f9bb43ff14d0 thread_name=asyncio_0
2025-08-25T15:10:25.352181Z [warning  ] /Library/Frameworks/Python.framework/Versions/3.13/lib/python3.13/site-packages/langgraph/pregel/main.py:2813: UserWarning: `durability` has no effect when no checkpointer is present.
  warnings.warn(
 [py.warnings] api_variant=local_dev assistant_id=fe096781-5601-53d2-b2f6-0d3403f7e9ca graph_id=agent langgraph_api_version=0.3.1 request_id=0a96deec-40bb-48a1-a5ca-5ea9854e8042 run_attempt=1 run_id=0198e1c7-cde6-775d-9fc2-5fffa7841986 thread_id=1d81ffc1-ea38-4aa8-89f1-f9bb43ff14d0 thread_name=MainThread
2025-08-25T15:10:25.357438Z [info     ] Background run succeeded       [langgraph_api.worker] api_variant=local_dev assistant_id=fe096781-5601-53d2-b2f6-0d3403f7e9ca graph_id=agent langgraph_api_version=0.3.1 request_id=0a96deec-40bb-48a1-a5ca-5ea9854e8042 run_attempt=1 run_completed_in_ms=616 run_created_at=2025-08-25T15:10:24.742858+00:00 run_ended_at=2025-08-25T15:10:25.357298+00:00 run_exec_ms=23 run_id=0198e1c7-cde6-775d-9fc2-5fffa7841986 run_started_at=2025-08-25T15:10:25.333924+00:00 thread_id=1d81ffc1-ea38-4aa8-89f1-f9bb43ff14d0 thread_name=asyncio_0
2025-08-25T15:10:26.837936Z [info     ] Worker stats                   [langgraph_runtime_inmem.queue] active=0 api_variant=local_dev available=1 langgraph_api_version=0.3.1 max=1 thread_name=asyncio_1
2025-08-25T15:10:27.340621Z [info     ] Queue stats                    [langgraph_runtime_inmem.queue] api_variant=local_dev langgraph_api_version=0.3.1 max_age_secs=None med_age_secs=None n_pending=0 n_running=0 thread_name=asyncio_0

也可以在UI直接进行：

花日の博客

LangGraph 案例：本地运行服务

1. 安装 LangGraph CLI & 创建一个 LangGraph 应用

2. 安装依赖项

3. 创建`.env`文件

4. 启动 LangGraph 服务器

5. 在 LangGraph Studio 中测试你的应用程序

6. 测试 API

下一篇：LangGraph 案例：时间旅行

花日の博客

LangGraph 案例：本地运行服务

1. 安装 LangGraph CLI & 创建一个 LangGraph 应用

2. 安装依赖项

3. 创建.env文件

4. 启动 LangGraph 服务器

5. 在 LangGraph Studio 中测试你的应用程序

6. 测试 API

下一篇：LangGraph 案例：时间旅行

3. 创建`.env`文件