From 8e9e426d8e294c7d14e2d7c8c626b7a831a2f105 Mon Sep 17 00:00:00 2001 From: yuluo-yx Date: Mon, 19 Jan 2026 22:31:42 +0800 Subject: [PATCH] chore: update code block Signed-off-by: yuluo-yx --- .../agent-framework/advanced/agent-tool.md | 22 ++-- .../advanced/human-in-the-loop.md | 74 ++++++------ .../agent-framework/advanced/multi-agent.md | 110 +++++++++--------- .../tutorials/structured-output.md | 2 +- 4 files changed, 104 insertions(+), 104 deletions(-) diff --git a/docs/frameworks/agent-framework/advanced/agent-tool.md b/docs/frameworks/agent-framework/advanced/agent-tool.md index 7bf3676d..0ebb0553 100644 --- a/docs/frameworks/agent-framework/advanced/agent-tool.md +++ b/docs/frameworks/agent-framework/advanced/agent-tool.md @@ -85,7 +85,7 @@ ReactAgent blogAgent = ReactAgent.builder() .name("blog_agent") .model(chatModel) .instruction("根据用户给定的主题写一篇文章。使用写作工具来完成任务。") - .tools(AgentTool.getFunctionToolCallback(writerAgent)) // [!code highlight] + .tools(AgentTool.getFunctionToolCallback(writerAgent)) .build(); // 使用 @@ -149,7 +149,7 @@ ReactAgent writerAgent = ReactAgent.builder() .model(chatModel) .description("根据结构化输入写文章") .instruction("你是一个专业作家。请严格按照输入的主题、字数和风格要求创作文章。") - .inputSchema(writerInputSchema) // [!code highlight] + .inputSchema(writerInputSchema) .build(); ReactAgent coordinatorAgent = ReactAgent.builder() @@ -185,7 +185,7 @@ ReactAgent writerAgent = ReactAgent.builder() .model(chatModel) .description("根据类型化输入写文章") .instruction("你是一个专业作家。请严格按照输入的 topic（主题）、wordCount（字数）和 style（风格）要求创作文章。") - .inputType(ArticleRequest.class) // [!code highlight] + .inputType(ArticleRequest.class) .build(); ReactAgent coordinatorAgent = ReactAgent.builder() @@ -243,7 +243,7 @@ ReactAgent writerAgent = ReactAgent.builder() .model(chatModel) .description("写文章并返回结构化输出") .instruction("你是一个专业作家。请创作文章并严格按照指定的JSON格式返回结果。") - .outputSchema(format) // [!code highlight] + .outputSchema(format) .build(); ReactAgent coordinatorAgent = ReactAgent.builder() @@ -282,7 +282,7 @@ ReactAgent writerAgent = ReactAgent.builder() .model(chatModel) .description("写文章并返回类型化输出") .instruction("你是一个专业作家。请创作文章并返回包含 title、content 和 characterCount 的结构化结果。") - .outputType(ArticleOutput.class) // [!code highlight] + .outputType(ArticleOutput.class) .build(); ReactAgent coordinatorAgent = ReactAgent.builder() @@ -329,8 +329,8 @@ ReactAgent writerAgent = ReactAgent.builder() .model(chatModel) .description("完整类型化的写作工具") .instruction("根据结构化输入（topic、wordCount、style）创作文章，并返回结构化输出（title、content、characterCount）。") - .inputType(ArticleRequest.class) // [!code highlight] - .outputType(ArticleOutput.class) // [!code highlight] + .inputType(ArticleRequest.class) + .outputType(ArticleOutput.class) .build(); ReactAgent reviewerAgent = ReactAgent.builder() @@ -338,7 +338,7 @@ ReactAgent reviewerAgent = ReactAgent.builder() .model(chatModel) .description("完整类型化的评审工具") .instruction("对文章进行评审，返回评审意见（comment、approved、suggestions）。") - .outputType(ReviewOutput.class) // [!code highlight] + .outputType(ReviewOutput.class) .build(); ReactAgent orchestratorAgent = ReactAgent.builder() @@ -396,9 +396,9 @@ ReactAgent multiToolAgent = ReactAgent.builder() .instruction("你可以访问多个专业工具：写作、翻译和总结。" + "根据用户需求选择合适的工具来完成任务。") .tools( - AgentTool.getFunctionToolCallback(writerAgent), // [!code highlight] - AgentTool.getFunctionToolCallback(translatorAgent), // [!code highlight] - AgentTool.getFunctionToolCallback(summarizerAgent) // [!code highlight] + AgentTool.getFunctionToolCallback(writerAgent), + AgentTool.getFunctionToolCallback(translatorAgent), + AgentTool.getFunctionToolCallback(summarizerAgent) ) .build(); diff --git a/docs/frameworks/agent-framework/advanced/human-in-the-loop.md b/docs/frameworks/agent-framework/advanced/human-in-the-loop.md index 5b2a304c..45844a88 100644 --- a/docs/frameworks/agent-framework/advanced/human-in-the-loop.md +++ b/docs/frameworks/agent-framework/advanced/human-in-the-loop.md @@ -57,22 +57,22 @@ import com.alibaba.cloud.ai.graph.checkpoint.savers.MemorySaver; MemorySaver memorySaver = new MemorySaver(); // 创建人工介入Hook -HumanInTheLoopHook humanInTheLoopHook = HumanInTheLoopHook.builder() // [!code highlight] - .approvalOn("write_file", ToolConfig.builder() // [!code highlight] - .description("文件写入操作需要审批") // [!code highlight] - .build()) // [!code highlight] - .approvalOn("execute_sql", ToolConfig.builder() // [!code highlight] - .description("SQL执行操作需要审批") // [!code highlight] - .build()) // [!code highlight] - .build(); // [!code highlight] +HumanInTheLoopHook humanInTheLoopHook = HumanInTheLoopHook.builder() + .approvalOn("write_file", ToolConfig.builder() + .description("文件写入操作需要审批") + .build()) + .approvalOn("execute_sql", ToolConfig.builder() + .description("SQL执行操作需要审批") + .build()) + .build(); // 创建Agent ReactAgent agent = ReactAgent.builder() .name("approval_agent") .model(chatModel) .tools(writeFileTool, executeSqlTool, readDataTool) - .hooks(List.of(humanInTheLoopHook)) // [!code highlight] - .saver(memorySaver) // [!code highlight] + .hooks(List.of(humanInTheLoopHook)) + .saver(memorySaver) .build();`} @@ -98,24 +98,24 @@ import com.alibaba.cloud.ai.graph.action.InterruptionMetadata; // 人工介入利用检查点机制。 // 你必须提供线程ID以将执行与会话线程关联， // 以便可以暂停和恢复对话（人工审查所需）。 -String threadId = "user-session-123"; // [!code highlight] -RunnableConfig config = RunnableConfig.builder() // [!code highlight] - .threadId(threadId) // [!code highlight] - .build(); // [!code highlight] +String threadId = "user-session-123"; +RunnableConfig config = RunnableConfig.builder() + .threadId(threadId) + .build(); // 运行图直到触发中断 -Optional result = agent.invokeAndGetOutput( // [!code highlight] +Optional result = agent.invokeAndGetOutput( "删除数据库中的旧记录", config ); // 检查是否返回了中断 -if (result.isPresent() && result.get() instanceof InterruptionMetadata) { // [!code highlight] - InterruptionMetadata interruptionMetadata = (InterruptionMetadata) result.get(); // [!code highlight] +if (result.isPresent() && result.get() instanceof InterruptionMetadata) { + InterruptionMetadata interruptionMetadata = (InterruptionMetadata) result.get(); // 中断包含需要审查的工具反馈 - List toolFeedbacks = // [!code highlight] - interruptionMetadata.toolFeedbacks(); // [!code highlight] + List toolFeedbacks = + interruptionMetadata.toolFeedbacks(); for (InterruptionMetadata.ToolFeedback feedback : toolFeedbacks) { System.out.println("工具: " + feedback.getName()); @@ -150,9 +150,9 @@ if (result.isPresent() && result.get() instanceof InterruptionMetadata) { // [!c // 对每个工具调用设置批准决策 interruptionMetadata.toolFeedbacks().forEach(toolFeedback -> { InterruptionMetadata.ToolFeedback approvedFeedback = - InterruptionMetadata.ToolFeedback.builder(toolFeedback) // [!code highlight] - .result(InterruptionMetadata.ToolFeedback.FeedbackResult.APPROVED) // [!code highlight] - .build(); // [!code highlight] + InterruptionMetadata.ToolFeedback.builder(toolFeedback) + .result(InterruptionMetadata.ToolFeedback.FeedbackResult.APPROVED) + .build(); feedbackBuilder.addToolFeedback(approvedFeedback); }); @@ -161,7 +161,7 @@ if (result.isPresent() && result.get() instanceof InterruptionMetadata) { // [!c // 使用批准决策恢复执行 RunnableConfig resumeConfig = RunnableConfig.builder() .threadId(threadId) // 相同的线程ID以恢复暂停的对话 - .addMetadata(RunnableConfig.HUMAN_FEEDBACK_METADATA_KEY, approvalMetadata) // [!code highlight] + .addMetadata(RunnableConfig.HUMAN_FEEDBACK_METADATA_KEY, approvalMetadata) .build(); // 第二次调用以恢复执行 @@ -186,12 +186,12 @@ if (result.isPresent() && result.get() instanceof InterruptionMetadata) { // [!c interruptionMetadata.toolFeedbacks().forEach(toolFeedback -> { // 修改工具参数 String editedArguments = toolFeedback.getArguments() - .replace("DELETE FROM records", "DELETE FROM old_records"); // [!code highlight] + .replace("DELETE FROM records", "DELETE FROM old_records"); InterruptionMetadata.ToolFeedback editedFeedback = InterruptionMetadata.ToolFeedback.builder(toolFeedback) - .arguments(editedArguments) // [!code highlight] - .result(InterruptionMetadata.ToolFeedback.FeedbackResult.EDITED) // [!code highlight] + .arguments(editedArguments) + .result(InterruptionMetadata.ToolFeedback.FeedbackResult.EDITED) .build(); feedbackBuilder.addToolFeedback(editedFeedback); }); @@ -227,8 +227,8 @@ if (result.isPresent() && result.get() instanceof InterruptionMetadata) { // [!c interruptionMetadata.toolFeedbacks().forEach(toolFeedback -> { InterruptionMetadata.ToolFeedback rejectedFeedback = InterruptionMetadata.ToolFeedback.builder(toolFeedback) - .result(InterruptionMetadata.ToolFeedback.FeedbackResult.REJECTED) // [!code highlight] - .description("不允许删除操作，请使用归档功能代替。") // [!code highlight] + .result(InterruptionMetadata.ToolFeedback.FeedbackResult.REJECTED) + .description("不允许删除操作，请使用归档功能代替。") .build(); feedbackBuilder.addToolFeedback(rejectedFeedback); }); @@ -457,8 +457,8 @@ ReactAgent qaAgent = ReactAgent.builder() .model(chatModel) .instruction("你是一个问答专家，负责回答用户的问题。如果需要搜索信息，请使用search工具。\n用户问题：{cleaned_input}") .outputKey("qa_result") - .saver(saver) // [!code highlight] - .hooks(HumanInTheLoopHook.builder() // [!code highlight] + .saver(saver) + .hooks(HumanInTheLoopHook.builder() .approvalOn("search", ToolConfig.builder() .description("搜索操作需要人工审批，请确认是否执行搜索") .build()) @@ -507,7 +507,7 @@ workflow.addNode("preprocess", node_async(new PreprocessorNode())); workflow.addNode("validate", node_async(new ValidatorNode())); // 添加Agent Node（嵌套的ReactAgent） -workflow.addNode(qaAgent.name(), qaAgent.asNode( // [!code highlight] +workflow.addNode(qaAgent.name(), qaAgent.asNode( true, // includeContents: 传递父图的消息历史 false // includeReasoning: 不返回推理过程 )); @@ -531,9 +531,9 @@ workflow.addConditionalEdges( ); // 8. 编译工作流（必须在CompileConfig中注册检查点保存器） -CompiledGraph compiledGraph = workflow.compile( // [!code highlight] +CompiledGraph compiledGraph = workflow.compile( CompileConfig.builder() - .saverConfig(SaverConfig.builder().register(saver).build()) // [!code highlight] + .saverConfig(SaverConfig.builder().register(saver).build()) .build() ); @@ -542,14 +542,14 @@ String threadId = "workflow-hilt-001"; Map input = Map.of("input", "请解释量子计算的基本原理"); // 第一次调用 - 可能触发中断 -Optional nodeOutputOptional = compiledGraph.invokeAndGetOutput( // [!code highlight] +Optional nodeOutputOptional = compiledGraph.invokeAndGetOutput( input, RunnableConfig.builder().threadId(threadId).build() ); // 检查是否发生中断 if (nodeOutputOptional.isPresent() - && nodeOutputOptional.get() instanceof InterruptionMetadata interruptionMetadata) { // [!code highlight] + && nodeOutputOptional.get() instanceof InterruptionMetadata interruptionMetadata) { System.out.println("工作流被中断，等待人工审核。"); System.out.println("中断节点: " + interruptionMetadata.node()); @@ -581,11 +581,11 @@ if (nodeOutputOptional.isPresent() // 使用批准决策恢复执行 RunnableConfig resumableConfig = RunnableConfig.builder() .threadId(threadId) // 相同的线程ID - .addHumanFeedback(approvalMetadata) // [!code highlight] + .addHumanFeedback(approvalMetadata) .build(); // 恢复工作流执行（传入空Map，因为状态已保存在检查点中） - nodeOutputOptional = compiledGraph.invokeAndGetOutput(Map.of(), resumableConfig); // [!code highlight] + nodeOutputOptional = compiledGraph.invokeAndGetOutput(Map.of(), resumableConfig); }`} diff --git a/docs/frameworks/agent-framework/advanced/multi-agent.md b/docs/frameworks/agent-framework/advanced/multi-agent.md index 9623c40d..669e3beb 100644 --- a/docs/frameworks/agent-framework/advanced/multi-agent.md +++ b/docs/frameworks/agent-framework/advanced/multi-agent.md @@ -139,8 +139,8 @@ ReactAgent writerAgent = ReactAgent.builder() .name("writer_agent") .model(chatModel) .description("专业写作Agent") - .instruction("你是一个知名的作家，擅长写作和创作。请根据用户的提问进行回答：{input}。") // [!code highlight] - .outputKey("article") // [!code highlight] + .instruction("你是一个知名的作家，擅长写作和创作。请根据用户的提问进行回答：{input}。") + .outputKey("article") .build(); ReactAgent reviewerAgent = ReactAgent.builder() @@ -148,16 +148,16 @@ ReactAgent reviewerAgent = ReactAgent.builder() .model(chatModel) .description("专业评审Agent") .instruction("你是一个知名的评论家，擅长对文章进行评论和修改。" + - "对于散文类文章，请确保文章中必须包含对于西湖风景的描述。待评论文章：\n\n {article}" + // [!code highlight] + "对于散文类文章，请确保文章中必须包含对于西湖风景的描述。待评论文章：\n\n {article}" + "最终只返回修改后的文章，不要包含任何评论信息。") - .outputKey("reviewed_article") // [!code highlight] + .outputKey("reviewed_article") .build(); // 创建顺序Agent -SequentialAgent blogAgent = SequentialAgent.builder() // [!code highlight] +SequentialAgent blogAgent = SequentialAgent.builder() .name("blog_agent") .description("根据用户给定的主题写一篇文章，然后将文章交给评论员进行评论") - .subAgents(List.of(writerAgent, reviewerAgent)) // [!code highlight] + .subAgents(List.of(writerAgent, reviewerAgent)) .build(); // 使用 @@ -167,14 +167,14 @@ if (result.isPresent()) { OverAllState state = result.get(); // 访问第一个Agent的输出 - state.value("article").ifPresent(article -> { // [!code highlight] + state.value("article").ifPresent(article -> { if (article instanceof AssistantMessage) { System.out.println("原始文章: " + ((AssistantMessage) article).getText()); } }); // 访问第二个Agent的输出 - state.value("reviewed_article").ifPresent(reviewedArticle -> { // [!code highlight] + state.value("reviewed_article").ifPresent(reviewedArticle -> { if (reviewedArticle instanceof AssistantMessage) { System.out.println("评审后文章: " + ((AssistantMessage) reviewedArticle).getText()); } @@ -198,16 +198,16 @@ if (result.isPresent()) { {`ReactAgent writerAgent = ReactAgent.builder() .name("writer_agent") .model(chatModel) - .returnReasoningContents(true) // [!code highlight] + .returnReasoningContents(true) .outputKey("article") .build(); ReactAgent reviewerAgent = ReactAgent.builder() .name("reviewer_agent") .model(chatModel) - .instruction("请对文章进行评审修正：\n{article}，最终返回评审修正后的文章内容") // [!code highlight] - .includeContents(true) // 包含上一个Agent的推理内容 // [!code highlight] - .returnReasoningContents(true) // [!code highlight] + .instruction("请对文章进行评审修正：\n{article}，最终返回评审修正后的文章内容") + .includeContents(true) // 包含上一个Agent的推理内容 + .returnReasoningContents(true) .outputKey("reviewed_article") .build(); @@ -220,7 +220,7 @@ Optional result = blogAgent.invoke("帮我写一个100字左右的 if (result.isPresent()) { // 消息历史将包含所有工具调用和推理过程 - List messages = (List) result.get().value("messages").orElse(List.of()); // [!code highlight] + List messages = (List) result.get().value("messages").orElse(List.of()); System.out.println("消息数量: " + messages.size()); // 包含所有中间步骤 }`} @@ -251,8 +251,8 @@ ReactAgent proseWriterAgent = ReactAgent.builder() .model(chatModel) .description("专门写散文的AI助手") .instruction("你是一个知名的散文作家，擅长写优美的散文。" + - "用户会给你一个主题：{input}，你只需要创作一篇100字左右的散文。") // [!code highlight] - .outputKey("prose_result") // [!code highlight] + "用户会给你一个主题：{input}，你只需要创作一篇100字左右的散文。") + .outputKey("prose_result") .build(); ReactAgent poemWriterAgent = ReactAgent.builder() @@ -260,8 +260,8 @@ ReactAgent poemWriterAgent = ReactAgent.builder() .model(chatModel) .description("专门写现代诗的AI助手") .instruction("你是一个知名的现代诗人，擅长写现代诗。" + - "用户会给你的主题是：{input}，你只需要创作一首现代诗。") // [!code highlight] - .outputKey("poem_result") // [!code highlight] + "用户会给你的主题是：{input}，你只需要创作一首现代诗。") + .outputKey("poem_result") .build(); ReactAgent summaryAgent = ReactAgent.builder() @@ -269,17 +269,17 @@ ReactAgent summaryAgent = ReactAgent.builder() .model(chatModel) .description("专门做内容总结的AI助手") .instruction("你是一个专业的内容分析师，擅长对主题进行总结和提炼。" + - "用户会给你一个主题：{input}，你只需要对这个主题进行简要总结。") // [!code highlight] - .outputKey("summary_result") // [!code highlight] + "用户会给你一个主题：{input}，你只需要对这个主题进行简要总结。") + .outputKey("summary_result") .build(); // 创建并行Agent -ParallelAgent parallelAgent = ParallelAgent.builder() // [!code highlight] +ParallelAgent parallelAgent = ParallelAgent.builder() .name("parallel_creative_agent") .description("并行执行多个创作任务，包括写散文、写诗和做总结") - .mergeOutputKey("merged_results") // [!code highlight] - .subAgents(List.of(proseWriterAgent, poemWriterAgent, summaryAgent)) // [!code highlight] - .mergeStrategy(new ParallelAgent.DefaultMergeStrategy()) // [!code highlight] + .mergeOutputKey("merged_results") + .subAgents(List.of(proseWriterAgent, poemWriterAgent, summaryAgent)) + .mergeStrategy(new ParallelAgent.DefaultMergeStrategy()) .build(); // 使用 @@ -289,15 +289,15 @@ if (result.isPresent()) { OverAllState state = result.get(); // 访问各个Agent的输出 - state.value("prose_result").ifPresent(r -> // [!code highlight] + state.value("prose_result").ifPresent(r -> System.out.println("散文: " + r)); - state.value("poem_result").ifPresent(r -> // [!code highlight] + state.value("poem_result").ifPresent(r -> System.out.println("诗歌: " + r)); - state.value("summary_result").ifPresent(r -> // [!code highlight] + state.value("summary_result").ifPresent(r -> System.out.println("总结: " + r)); // 访问合并后的结果 - state.value("merged_results").ifPresent(r -> // [!code highlight] + state.value("merged_results").ifPresent(r -> System.out.println("合并结果: " + r)); }`} @@ -313,7 +313,7 @@ if (result.isPresent()) { {`public class CustomMergeStrategy implements ParallelAgent.MergeStrategy { @Override - public Map merge(Map mergedState, OverAllState state) { // [!code highlight] + public Map merge(Map mergedState, OverAllState state) { // 从每个Agent的状态中提取输出 state.data().forEach((key, value) -> { if (key.endsWith("_result")) { @@ -335,8 +335,8 @@ if (result.isPresent()) { ParallelAgent parallelAgent = ParallelAgent.builder() .name("parallel_agent") .subAgents(List.of(agent1, agent2, agent3)) - .mergeStrategy(new CustomMergeStrategy()) // [!code highlight] - .mergeOutputKey("final_merged_result") // [!code highlight] + .mergeStrategy(new CustomMergeStrategy()) + .mergeOutputKey("final_merged_result") .build();`} @@ -392,8 +392,8 @@ ReactAgent translatorAgent = ReactAgent.builder() LlmRoutingAgent routingAgent = LlmRoutingAgent.builder() .name("content_routing_agent") .description("根据用户需求智能路由到合适的专家Agent") - .model(chatModel) // [!code highlight] - .subAgents(List.of(writerAgent, reviewerAgent, translatorAgent)) // [!code highlight] + .model(chatModel) + .subAgents(List.of(writerAgent, reviewerAgent, translatorAgent)) .build(); // 使用 - LLM会自动选择最合适的Agent @@ -426,8 +426,8 @@ Optional result3 = routingAgent.invoke("请将以下内容翻译 ReactAgent codeAgent = ReactAgent.builder() .name("code_agent") .model(chatModel) - .description("专门处理编程相关问题，包括代码编写、调试、重构和优化。" + // [!code highlight] - "擅长Java、Python、JavaScript等主流编程语言。") // [!code highlight] + .description("专门处理编程相关问题，包括代码编写、调试、重构和优化。" + + "擅长Java、Python、JavaScript等主流编程语言。") .instruction("你是一个资深的软件工程师...") .build(); @@ -435,8 +435,8 @@ ReactAgent codeAgent = ReactAgent.builder() ReactAgent businessAgent = ReactAgent.builder() .name("business_agent") .model(chatModel) - .description("专门处理商业分析、市场研究和战略规划问题。" + // [!code highlight] - "不处理技术实现细节。") // [!code highlight] + .description("专门处理商业分析、市场研究和战略规划问题。" + + "不处理技术实现细节。") .instruction("你是一个资深的商业分析师...") .build(); @@ -505,7 +505,7 @@ LlmRoutingAgent routingAgent = LlmRoutingAgent.builder() .name("content_routing_agent") .description("根据用户需求智能路由到合适的专家Agent") .model(chatModel) - .systemPrompt(ROUTING_SYSTEM_PROMPT) // [!code highlight] + .systemPrompt(ROUTING_SYSTEM_PROMPT) .subAgents(List.of(writerAgent, reviewerAgent, translatorAgent)) .build();`} @@ -532,7 +532,7 @@ LlmRoutingAgent routingAgent = LlmRoutingAgent.builder() .name("content_routing_agent") .description("根据用户需求智能路由到合适的专家Agent") .model(chatModel) - .instruction(ROUTING_INSTRUCTION) // [!code highlight] + .instruction(ROUTING_INSTRUCTION) .subAgents(List.of(writerAgent, reviewerAgent, translatorAgent)) .build();`} @@ -577,8 +577,8 @@ LlmRoutingAgent routingAgent = LlmRoutingAgent.builder() .name("content_routing_agent") .description("根据用户需求智能路由到合适的专家Agent") .model(chatModel) - .systemPrompt(ROUTING_SYSTEM_PROMPT) // [!code highlight] - .instruction(ROUTING_INSTRUCTION) // [!code highlight] + .systemPrompt(ROUTING_SYSTEM_PROMPT) + .instruction(ROUTING_INSTRUCTION) .subAgents(List.of(writerAgent, reviewerAgent, translatorAgent)) .build();`} @@ -643,8 +643,8 @@ ReactAgent translatorAgent = ReactAgent.builder() SupervisorAgent supervisorAgent = SupervisorAgent.builder() .name("content_supervisor") .description("内容管理监督者，负责协调写作、翻译等任务") - .model(chatModel) // [!code highlight] - .subAgents(List.of(writerAgent, translatorAgent)) // [!code highlight] + .model(chatModel) + .subAgents(List.of(writerAgent, translatorAgent)) .build(); // 使用 - 监督者会根据任务自动路由并支持多步骤处理 @@ -705,7 +705,7 @@ SupervisorAgent supervisorAgent = SupervisorAgent.builder() .name("content_supervisor") .description("内容管理监督者") .model(chatModel) - .systemPrompt(SUPERVISOR_SYSTEM_PROMPT) // [!code highlight] + .systemPrompt(SUPERVISOR_SYSTEM_PROMPT) .subAgents(List.of(writerAgent, translatorAgent)) .build();`} @@ -724,7 +724,7 @@ ReactAgent articleWriterAgent = ReactAgent.builder() .model(chatModel) .description("专业写作Agent，负责创作文章") .instruction("你是一个知名的作家，擅长写作和创作。请根据用户的提问进行回答：{input}。") - .outputKey("article_content") // [!code highlight] + .outputKey("article_content") .build(); // 监督者的子Agent @@ -748,7 +748,7 @@ ReactAgent reviewerAgent = ReactAgent.builder() final String SUPERVISOR_INSTRUCTION = """ 你是一个智能的内容处理监督者，你可以看到前序Agent的聊天历史与任务处理记录。当前，你收到了以下文章内容： -{article_content} // [!code highlight] +{article_content} 请根据文章内容的特点，决定是进行翻译还是评审： - 如果文章是中文且需要翻译，选择 translator_agent @@ -779,7 +779,7 @@ SupervisorAgent supervisorAgent = SupervisorAgent.builder() .description("内容处理监督者，根据前序Agent的输出决定翻译或评审") .model(chatModel) .systemPrompt(SUPERVISOR_SYSTEM_PROMPT) - .instruction(SUPERVISOR_INSTRUCTION) // [!code highlight] + .instruction(SUPERVISOR_INSTRUCTION) .subAgents(List.of(translatorAgent, reviewerAgent)) .build(); @@ -787,7 +787,7 @@ SupervisorAgent supervisorAgent = SupervisorAgent.builder() SequentialAgent sequentialAgent = SequentialAgent.builder() .name("content_processing_workflow") .description("内容处理工作流：先写文章，然后根据文章内容决定翻译或评审") - .subAgents(List.of(articleWriterAgent, supervisorAgent)) // [!code highlight] + .subAgents(List.of(articleWriterAgent, supervisorAgent)) .build(); // 使用 @@ -983,9 +983,9 @@ Predicate> isUrgent = state -> { ConditionalAgent conditionalAgent = ConditionalAgent.builder() .name("priority_router") .description("根据紧急程度路由请求") - .condition(isUrgent) // [!code highlight] - .trueAgent(urgentAgent) // [!code highlight] - .falseAgent(normalAgent) // [!code highlight] + .condition(isUrgent) + .trueAgent(urgentAgent) + .falseAgent(normalAgent) .build(); // 使用 @@ -1071,7 +1071,7 @@ ReactAgent webResearchAgent = ReactAgent.builder() .name("web_research") .model(chatModel) .description("从互联网搜索信息") - .instruction("请搜索并收集关于以下主题的信息：{input}") // [!code highlight] + .instruction("请搜索并收集关于以下主题的信息：{input}") .outputKey("web_data") .build(); @@ -1079,7 +1079,7 @@ ReactAgent dbResearchAgent = ReactAgent.builder() .name("db_research") .model(chatModel) .description("从数据库查询信息") - .instruction("请从数据库中查询并收集关于以下主题的信息：{input}") // [!code highlight] + .instruction("请从数据库中查询并收集关于以下主题的信息：{input}") .outputKey("db_data") .build(); @@ -1095,7 +1095,7 @@ ReactAgent analysisAgent = ReactAgent.builder() .name("analysis_agent") .model(chatModel) .description("分析研究数据") - .instruction("请分析以下收集到的数据并提供见解：{research_data}") // [!code highlight] + .instruction("请分析以下收集到的数据并提供见解：{research_data}") .outputKey("analysis_result") .build(); @@ -1109,7 +1109,7 @@ ReactAgent pdfReportAgent = ReactAgent.builder() 研究结果：{research_data} 分析结果：{analysis_result} - """) // [!code highlight] + """) .outputKey("pdf_report") .build(); @@ -1122,7 +1122,7 @@ ReactAgent htmlReportAgent = ReactAgent.builder() 研究结果：{research_data} 分析结果：{analysis_result} - """) // [!code highlight] + """) .outputKey("html_report") .build(); diff --git a/docs/frameworks/agent-framework/tutorials/structured-output.md b/docs/frameworks/agent-framework/tutorials/structured-output.md index e26f9e81..16206bb3 100644 --- a/docs/frameworks/agent-framework/tutorials/structured-output.md +++ b/docs/frameworks/agent-framework/tutorials/structured-output.md @@ -251,7 +251,7 @@ public static class ContactInfo { ReactAgent agent = ReactAgent.builder() .name("contact_extractor") .model(chatModel) - .outputType(ContactInfo.class) // [!code highlight] + .outputType(ContactInfo.class) .saver(new MemorySaver()) .build();