AgentScope Harness 深度实战：让Java智能体从“Demo可用”走向“生产可用”

<!-- AgentScope 核心 Harness 能力包 -->
<dependency>
    <groupId>com.agentscope</groupId>
    <artifactId>agentscope-java-sdk</artifactId>
    <version>1.1.0</version>
</dependency>

<!-- OpenAI模型适配包（可替换通义千问/文心一言） -->
<dependency>
    <groupId>com.agentscope</groupId>
    <artifactId>agentscope-model-openai</artifactId>
    <version>1.1.0</version>
</dependency>

spring:
  application:
    name: agentscope-harness-travel
# AgentScope Harness 生产配置
agentscope:
  model:
    type: openai
    api-key: sk-xxxxxxxxxxxxxxxxxxxxxx
    model-name: gpt-3.5-turbo
    base-url: https://api.openai.com/v1
  harness:
    workspace-path: ./travel-agent-workspace
    memory-enable: true
    memory-compress-enable: true
    sandbox-mode: normal
    # 开启跨会话状态恢复
    session-recovery-enable: true

import com.agentscope.HarnessAgent;
import com.agentscope.model.openai.OpenAIModel;
import org.springframework.beans.factory.annotation.Value;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;

@Configuration
public class TravelHarnessConfig {

    @Value("${agentscope.model.api-key}")
    private String apiKey;

    @Value("${agentscope.model.model-name}")
    private String modelName;

    @Value("${agentscope.harness.workspace-path}")
    private String workspacePath;

    // 统一初始化大模型
    @Bean
    public OpenAIModel openAIModel() {
        return new OpenAIModel(apiKey, modelName);
    }

    // 主调度Agent：全局任务分发、会话维护、结果汇总
    @Bean
    public HarnessAgent travelMasterAgent(OpenAIModel openAIModel) {
        return HarnessAgent.builder()
                .name("TravelMasterAgent")
                .model(openAIModel)
                .workspace(workspacePath + "/master")
                .enableLongTermMemory(true)
                .enableContextCompress(true)
                .enableSandbox(true)
                .systemPrompt("你是企业差旅智能助手总调度，负责接收用户消息，判断场景，分发任务给对应子Agent，汇总结果并回复用户。")
                .build();
    }

    // 子Agent1：意图识别Agent
    @Bean
    public HarnessAgent intentAgent(OpenAIModel openAIModel) {
        return HarnessAgent.builder()
                .name("TravelIntentAgent")
                .model(openAIModel)
                .workspace(workspacePath + "/intent")
                .enableContextCompress(true)
                .systemPrompt("你是差旅意图识别专家，仅输出结果：行程申请/进度查询/报销咨询/其他。无需多余解释。")
                .build();
    }

    // 子Agent2：信息收集Agent
    @Bean
    public HarnessAgent collectAgent(OpenAIModel openAIModel) {
        return HarnessAgent.builder()
                .name("TravelCollectAgent")
                .model(openAIModel)
                .workspace(workspacePath + "/collect")
                .enableContextCompress(true)
                .systemPrompt("你负责收集用户差旅所需信息，缺失则礼貌询问：出发地、目的地、出行时间、出行人员。")
                .build();
    }

    // 子Agent3：规则校验Agent
    @Bean
    public HarnessAgent ruleAgent(OpenAIModel openAIModel) {
        return HarnessAgent.builder()
                .name("TravelRuleAgent")
                .model(openAIModel)
                .workspace(workspacePath + "/rule")
                .enableContextCompress(true)
                .systemPrompt("你是企业差旅规则校验助手，判断用户行程是否合规，不合规则说明原因。企业规则：工作日短途出差优先高铁，夜间出行可申请机票。")
                .build();
    }
}

import com.agentscope.HarnessAgent;
import com.agentscope.request.ChatRequest;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RequestParam;
import org.springframework.web.bind.annotation.RestController;

@RestController
@RequestMapping("/travel/agent")
public class TravelAgentController {

    @Autowired
    private HarnessAgent travelMasterAgent;

    @Autowired
    private HarnessAgent intentAgent;

    @Autowired
    private HarnessAgent collectAgent;

    @Autowired
    private HarnessAgent ruleAgent;

    /**
     * 企业差旅智能问答统一入口
     * 完整多Agent协同链路：意图识别 -> 信息收集 -> 规则校验 -> 结果汇总
     */
    @GetMapping("/chat")
    public String chat(@RequestParam String message) {
        // 1. 意图识别
        String intent = intentAgent.chat(ChatRequest.builder().prompt(message).build()).getContent();

        // 2. 根据意图分流执行不同子Agent任务
        StringBuilder result = new StringBuilder();
        if ("行程申请".equals(intent)) {
            // 信息补齐
            String collectInfo = collectAgent.chat(ChatRequest.builder().prompt(message).build()).getContent();
            // 规则校验
            String ruleCheck = ruleAgent.chat(ChatRequest.builder().prompt(message).build()).getContent();
            result.append("【信息收集】").append(collectInfo).append("\n");
            result.append("【规则校验】").append(ruleCheck);
        } else {
            // 通用咨询由主Agent直接回复
            String masterReply = travelMasterAgent.chat(ChatRequest.builder().prompt(message).build()).getContent();
            result.append(masterReply);
        }
        return result.toString();
    }
}