Skip to content

如何使用 RemoteGraph 与部署进行交互

RemoteGraph 是一个接口,它允许你与 LangGraph 平台部署进行交互,就好像它是一个常规的、本地定义的 LangGraph 图(例如 CompiledGraph)一样。本指南将向你展示如何初始化 RemoteGraph 并与之交互。

初始化图

初始化 RemoteGraph 时,你必须始终指定: - name:你想要与之交互的图的名称。这与你在 langgraph.json 配置文件中用于部署的图名称相同。 - api_key:有效的 LangSmith API 密钥。可以将其设置为环境变量(LANGSMITH_API_KEY),也可以通过 api_key 参数直接传递。如果 LangGraphClient / SyncLangGraphClient 是使用 api_key 参数初始化的,那么 API 密钥也可以通过 client / sync_client 参数提供。

此外,你还必须提供以下其中一项: - url:你想要与之交互的部署的 URL。如果你传递了 url 参数,同步和异步客户端都将使用提供的 URL、头部信息(如果提供)和默认配置值(例如超时时间等)来创建。 - client:一个 LangGraphClient 实例,用于异步与部署进行交互(例如使用 .astream().ainvoke().aget_state().aupdate_state() 等)。 - sync_client:一个 SyncLangGraphClient 实例,用于同步与部署进行交互(例如使用 .stream().invoke().get_state().update_state() 等)。

注意

如果你同时传递了 clientsync_client 以及 url 参数,它们将优先于 url 参数。如果没有提供 client / sync_client / url 中的任何一个参数,RemoteGraph 将在运行时抛出 ValueError 异常。

使用 URL

from langgraph.pregel.remote import RemoteGraph

url = <DEPLOYMENT_URL>
graph_name = "agent"
remote_graph = RemoteGraph(graph_name, url=url)
import { RemoteGraph } from "@langchain/langgraph/remote";

const url = `<DEPLOYMENT_URL>`;
const graphName = "agent";
const remoteGraph = new RemoteGraph({ graphId: graphName, url });

使用客户端

from langgraph_sdk import get_client, get_sync_client
from langgraph.pregel.remote import RemoteGraph

url = <DEPLOYMENT_URL>
graph_name = "agent"
client = get_client(url=url)
sync_client = get_sync_client(url=url)
remote_graph = RemoteGraph(graph_name, client=client, sync_client=sync_client)
import { Client } from "@langchain/langgraph-sdk";
import { RemoteGraph } from "@langchain/langgraph/remote";

const client = new Client({ apiUrl: `<DEPLOYMENT_URL>` });
const graphName = "agent";
const remoteGraph = new RemoteGraph({ graphId: graphName, client });

调用图

由于 RemoteGraph 是一个 Runnable,它实现了与 CompiledGraph 相同的方法,因此你可以像通常与已编译图进行交互那样与它进行交互,即通过调用 .invoke().stream().get_state().update_state() 等方法(以及它们的异步对应方法)。

异步调用

注意

若要异步使用该图,你必须在初始化 RemoteGraph 时提供 urlclient

# 调用图
result = await remote_graph.ainvoke({
    "messages": [{"role": "user", "content": "what's the weather in sf"}]
})

# 从图中流式输出
async for chunk in remote_graph.astream({
    "messages": [{"role": "user", "content": "what's the weather in la"}]
}):
    print(chunk)
// 调用图
const result = await remoteGraph.invoke({
    messages: [{role: "user", content: "what's the weather in sf"}]
})

// 从图中流式输出
for await (const chunk of await remoteGraph.stream({
    messages: [{role: "user", content: "what's the weather in la"}]
})):
    console.log(chunk)

同步调用

注意

若要同步使用该图,你必须在初始化 RemoteGraph 时提供 urlsync_client

# 调用图
result = remote_graph.invoke({
    "messages": [{"role": "user", "content": "what's the weather in sf"}]
})

# 从图中流式输出
for chunk in remote_graph.stream({
    "messages": [{"role": "user", "content": "what's the weather in la"}]
}):
    print(chunk)

线程级持久化

默认情况下,图运行(即 .invoke().stream() 调用)是无状态的 —— 图的检查点和最终状态不会被持久化。如果你希望持久化图运行的输出(例如,启用人工介入功能),你可以创建一个线程,并通过 config 参数提供线程 ID,就像你对常规编译图所做的那样:

from langgraph_sdk import get_sync_client
url = <DEPLOYMENT_URL>
graph_name = "agent"
sync_client = get_sync_client(url=url)
remote_graph = RemoteGraph(graph_name, url=url)

# 创建一个线程(或者使用现有的线程)
thread = sync_client.threads.create()

# 使用线程配置调用图
config = {"configurable": {"thread_id": thread["thread_id"]}}
result = remote_graph.invoke({
    "messages": [{"role": "user", "content": "what's the weather in sf"}]
}, config=config)

# 验证状态是否已持久化到线程
thread_state = remote_graph.get_state(config)
print(thread_state)
import { Client } from "@langchain/langgraph-sdk";
import { RemoteGraph } from "@langchain/langgraph/remote";

const url = `<DEPLOYMENT_URL>`;
const graphName = "agent";
const client = new Client({ apiUrl: url });
const remoteGraph = new RemoteGraph({ graphId: graphName, url });

// 创建一个线程(或者使用现有的线程)
const thread = await client.threads.create();

// 使用线程配置调用图
const config = { configurable: { thread_id: thread.thread_id }};
const result = await remoteGraph.invoke({
  messages: [{ role: "user", content: "what's the weather in sf" }],
}, config);

// 验证状态是否已持久化到线程
const threadState = await remoteGraph.getState(config);
console.log(threadState);

作为子图使用

注意

如果你需要在包含 RemoteGraph 子图节点的图中使用 checkpointer,请确保使用 UUID 作为线程 ID。

由于 RemoteGraph 的行为与常规的 CompiledGraph 相同,因此它也可以在另一个图中用作子图。例如:

from langgraph_sdk import get_sync_client
from langgraph.graph import StateGraph, MessagesState, START
from typing import TypedDict

url = <DEPLOYMENT_URL>
graph_name = "agent"
remote_graph = RemoteGraph(graph_name, url=url)

# 定义父图
builder = StateGraph(MessagesState)
# 直接将远程图作为节点添加
builder.add_node("child", remote_graph)
builder.add_edge(START, "child")
graph = builder.compile()

# 调用父图
result = graph.invoke({
    "messages": [{"role": "user", "content": "what's the weather in sf"}]
})
print(result)

# 流式输出父图和子图的结果
for chunk in graph.stream({
    "messages": [{"role": "user", "content": "what's the weather in sf"}]
}, subgraphs=True):
    print(chunk)
import { MessagesAnnotation, StateGraph, START } from "@langchain/langgraph";
import { RemoteGraph } from "@langchain/langgraph/remote";

const url = `<DEPLOYMENT_URL>`;
const graphName = "agent";
const remoteGraph = new RemoteGraph({ graphId: graphName, url });

// 定义父图并直接将远程图作为节点添加
const graph = new StateGraph(MessagesAnnotation)
  .addNode("child", remoteGraph)
  .addEdge(START, "child")
  .compile()

// 调用父图
const result = await graph.invoke({
  messages: [{ role: "user", content: "what's the weather in sf" }]
});
console.log(result);

// 流式输出父图和子图的结果
for await (const chunk of await graph.stream({
  messages: [{ role: "user", content: "what's the weather in la" }]
}, { subgraphs: true })) {
  console.log(chunk);
}

Comments