anthropics / anthropics/claude-agent-sdk-python

StructuredOutput gets in a loop

未關閉
#775 0 則留言 5 個 reaction 已指派 0 人 在 GitHub 檢視
主要語言
Python
星號
8.1k
分支
1.3k
平均合併
2 天 31 分鐘
30 天內合併 PR
1

描述

Sometimes I have that `StructuredOutput` gets in a loop where it keeps trying to correct itself. It basically goes like this:

```
AssistantMessage(content=[TextBlock(text='\nfalse\nanswertext\n[]\n')], model='claude-sonnet-4-6'....
```

```
UserMessage(content=[TextBlock(text='Stop hook feedback:\nYou MUST call the StructuredOutput tool to complete this request. Call this tool now.')], uuid='191538a1-0dd7-415c-a728-50165282a087', parent_tool_use_id=None, tool_use_result=None)
```

And then it just keeps repeating the same pattern, it never calls the StructuredOutput tool, it just keeps on trying this odd `` type of text output. It somehow thinks that using `` is the same thing as calling it?

It also gets stuck inside this loop for WAY too long (basically until it hit the max budget), this should be restricted somehow.

貢獻指南

這個儲存庫沒有索引到貢獻指南

評估

這個 Issue 還沒有評估資料。

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。