进行端到端测试

使用工具端到端运行智能体，验证自主 API 调用，并确认完整的智能体管道。

连接工具并更新智能体合同后，您就可以测试完整管道了。

步骤 5 - 端到端测试

To test the agent end-to-end, use the Debug button from the toolbar at the top of the agent builder. This opens up the Debug configuration dialog, then do the following:

1. Open the Entrypoint arguments tab
2. Enter a quest description from the code block below
3. Then select Save & Debug to run it

The villagers report a massive creature has been destroying farms on the edge of the forest at night.

After you click on Save & Debug, you can watch the Execution Trail at the bottom of the agent builder as it executes. The Execution Trail shows the agent's full decision process. For a tool-using agent, expect at least three events:

1. The agent's initial analysis of the quest
2. The tool call with the search term it chose
3. The tool response with the candidate list

The agent then reasons over those candidates before producing its final output.

您应该会看到：

1. 智能体使用根据任务选择的搜索词来调用Unattended Query 。
2. 该工具会返回候选对象列表。
3. 智能体会选择最佳匹配项并返回五个输出字段。

Select the HTTP Request - Open5e Monster Query span to see the request the tool actually sent: the v2/creatures/ URL, the search term the agent chose in name__icontains, and the document__key and fields values you configured in Step 1. The response below it carries the key identifier for each candidate.

Verify the output contains all five fields: monsterIndex, monsterName, monsterType, monsterCr, and monsterReasoning. monsterIndex is the API's key, so it carries the srd_ prefix - for example srd_giant-ape. The monsterReasoning field should explain why the agent chose this monster for the quest.

---

您构建的内容

已经构建了 UiPathfinder 的第一个组件：

• Built a live API connector that searches the 5e SRD: one input, one HTTP request, one output.
• 已将其作为工具连接到智能体，只需在智能体构建器中进行选择。
• 现在，智能体可以自主决定何时调用工具、搜索什么内容，以及哪个候选对象最适合任务。

两步推理链（搜索词选择和候选对象评估）正是会生成有意义的评估信号的轨迹；这就是下一个实验室的内容。

后续操作

• Open5e API 文档- 探索完整的此项搜索 API，了解哪些字段可用。
• UiPath Integration Service - 构建连接到 SaaS API 的工具，而无需编写自定义工作流。
• UiPath 社区- 论坛、操作方法和开发者讨论。