跳到主要内容

Playwright

BrowserExecutionEngine 表示一个浏览器执行引擎,可用于响应用户操作在浏览器上执行动作。PlaywrightBrowserExecutionEngineBrowserExecutionEngine 的一个实现,使用 Playwright Java API 执行浏览器操作。BrowserUseTool 使用 BrowserExecutionEngine 执行浏览器操作,可用于浏览器使用(Browser-Use) Agent。你可以使用自然语言控制浏览器,例如:

  • open page 'https://docs.langchain4j.dev/', and summary the page text

Maven 依赖

<dependency>
<groupId>dev.langchain4j</groupId>
<artifactId>langchain4j-community-browser-execution-engine-playwright</artifactId>
<version>${latest version here}</version>
</dependency>
<dependency>
<groupId>dev.langchain4j</groupId>
<artifactId>langchain4j-community-tool-browser-use</artifactId>
<version>${latest version here}</version>
</dependency>

API 列表

  • BrowserExecutionEngine
  • PlaywrightBrowserExecutionEngine
  • BrowserUseTool

示例

Playwright playwright = Playwright.create();
BrowserType.LaunchOptions options = new BrowserType.LaunchOptions()
.setHeadless(false)
.setChannel("chrome")
.setChromiumSandbox(true)
.setSlowMo(500);
Browser browser = playwright.chromium().launch(options);

Assistant assistant = AiServices.builder(Assistant.class)
.chatModel(model)
.tools(BrowserUseTool.from(PlaywrightBrowserExecutionEngine.builder().browser(browser).build()))
.chatMemory(MessageWindowChatMemory.withMaxMessages(10))
.build();

String question = "open page 'https://docs.langchain4j.dev/', and summary the page text";
System.out.println(assistant.chat(question));