Skip to main content

Documentation Index

Fetch the complete documentation index at: https://langchain-zh.cn/llms.txt

Use this file to discover all available pages before exploring further.

本笔记本提供了快速入门 DuckDuckGoSearch 工具 的概述。有关 DuckDuckGoSearch 所有功能和配置的详细文档,请参阅 API 参考 DuckDuckGoSearch 提供了一个专注于隐私的搜索 API,专为 LLM 智能体设计。它能够无缝集成多种数据源,同时优先考虑用户隐私和相关的搜索结果。

概述

集成详情

设置

该集成位于 @langchain/community 包中,同时需要 duck-duck-scrape 依赖:
npm install @langchain/community @langchain/core duck-duck-scrape

凭证

设置 LangSmith 以获得最佳的观测能力也很有帮助(但不是必需的):
process.env.LANGSMITH_TRACING="true"
process.env.LANGSMITH_API_KEY="your-api-key"

实例化

你可以像这样实例化一个 DuckDuckGoSearch 工具:
import { DuckDuckGoSearch } from "@langchain/community/tools/duckduckgo_search"

const tool = new DuckDuckGoSearch({ maxResults: 1 })

调用

直接使用参数调用

await tool.invoke("what is the current weather in sf?")
[{"title":"San Francisco, CA Current Weather | AccuWeather","link":"https://www.accuweather.com/en/us/san-francisco/94103/current-weather/347629","snippet":"<b>Current</b> <b>weather</b> <b>in</b> San Francisco, CA. Check <b>current</b> conditions in San Francisco, CA with radar, hourly, and more."}]

使用 ToolCall 调用

我们也可以使用模型生成的 ToolCall 来调用工具,此时将返回一个 ToolMessage
// 这通常由模型生成,但为了演示目的,我们直接创建一个工具调用。
const modelGeneratedToolCall = {
  args: {
    input: "what is the current weather in sf?"
  },
  id: "tool_call_id",
  name: tool.name,
  type: "tool_call",
}
await tool.invoke(modelGeneratedToolCall)
ToolMessage {
  "content": "[{\"title\":\"San Francisco, CA Weather Conditions | Weather Underground\",\"link\":\"https://www.wunderground.com/weather/us/ca/san-francisco\",\"snippet\":\"San Francisco <b>Weather</b> Forecasts. <b>Weather</b> Underground provides local & long-range <b>weather</b> forecasts, weatherreports, maps & tropical <b>weather</b> conditions for the San Francisco area.\"}]",
  "name": "duckduckgo-search",
  "additional_kwargs": {},
  "response_metadata": {},
  "tool_call_id": "tool_call_id"
}

链式调用

我们可以通过先将工具绑定到一个 工具调用模型,然后在链中使用它:
// @lc-docs-hide-cell

import { ChatOpenAI } from "@langchain/openai"

const llm = new ChatOpenAI({
  model: "gpt-4.1-mini",
})
import { HumanMessage } from "@langchain/core/messages";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { RunnableLambda } from "@langchain/core/runnables";

const prompt = ChatPromptTemplate.fromMessages(
  [
    ["system", "You are a helpful assistant."],
    ["placeholder", "{messages}"],
  ]
)

const llmWithTools = llm.bindTools([tool]);

const chain = prompt.pipe(llmWithTools);

const toolChain = RunnableLambda.from(
  async (userInput: string, config) => {
    const humanMessage = new HumanMessage(userInput,);
    const aiMsg = await chain.invoke({
      messages: [new HumanMessage(userInput)],
    }, config);
    const toolMsgs = await tool.batch(aiMsg.tool_calls, config);
    return chain.invoke({
      messages: [humanMessage, aiMsg, ...toolMsgs],
    }, config);
  }
);

const toolChainResult = await toolChain.invoke("how many people have climbed mount everest?");
const { tool_calls, content } = toolChainResult;

console.log("AIMessage", JSON.stringify({
  tool_calls,
  content,
}, null, 2));
AIMessage {
  "tool_calls": [],
  "content": "As of December 2023, a total of 6,664 different people have reached the summit of Mount Everest."
}

智能体

有关如何在智能体中使用 LangChain 工具的指南,请参阅 LangGraph.js 文档。

API 参考

有关 DuckDuckGoSearch 所有功能和配置的详细文档,请参阅 API 参考