s02的normalize函数是不是有点问题

[meichiguozhuro]

下面是一个日志：
normalize前原始消息历史：[{'role': 'user', 'content': '你好'}, {'role': 'assistant', 'content': [TextBlock(citations=None, text='你好！很高兴见到你！有什么我可以帮助你的吗？', type='text')]}, {'role': 'user', 'content': '帮我在D盘创建一个txt文件'}]
normalize后消息历史：[{'role': 'user', 'content': '你好'}, {'role': 'assistant', 'content': []}, {'role': 'user', 'content': '帮我在D盘创建一个txt文件'}]

可见TextBlock被清洗掉了

是不是因为这部分代码 cleaned = []
for msg in messages:
clean = {"role": msg["role"]}
if isinstance(msg.get("content"), str):
clean["content"] = msg["content"]
elif isinstance(msg.get("content"), list):
clean["content"] = [
{k: v for k, v in block.items()
if not k.startswith("_")}
for block in msg["content"]
if isinstance(block, dict)
]
else:
clean["content"] = msg.get("content", "")
cleaned.append(clean)
这一行if isinstance(block, dict)无法兼容返回的TextBlock类

[Simon-in]

下面是一个日志： normalize前原始消息历史：[{'role': 'user', 'content': '你好'}, {'role': 'assistant', 'content': [TextBlock(citations=None, text='你好！很高兴见到你！有什么我可以帮助你的吗？', type='text')]}, {'role': 'user', 'content': '帮我在D盘创建一个txt文件'}] normalize后消息历史：[{'role': 'user', 'content': '你好'}, {'role': 'assistant', 'content': []}, {'role': 'user', 'content': '帮我在D盘创建一个txt文件'}]

可见TextBlock被清洗掉了

是不是因为这部分代码 cleaned = [] for msg in messages: clean = {"role": msg["role"]} if isinstance(msg.get("content"), str): clean["content"] = msg["content"] elif isinstance(msg.get("content"), list): clean["content"] = [ {k: v for k, v in block.items() if not k.startswith("_")} for block in msg["content"] if isinstance(block, dict) ] else: clean["content"] = msg.get("content", "") cleaned.append(clean) 这一行if isinstance(block, dict)无法兼容返回的TextBlock类

用ai修了下， 兼容后就能跑了
def normalize_messages(messages: list) -> list:
"""Clean up messages before sending to the API.

Three jobs:
1. Strip internal metadata fields the API doesn't understand
2. Ensure every tool_use has a matching tool_result (insert placeholder if missing)
3. Merge consecutive same-role messages (API requires strict alternation)
"""
cleaned = []
def block_to_dict(block):
    if isinstance(block, dict):
        return {k: v for k, v in block.items() if not k.startswith("_")}
    # 兼容模型返回的工具块对象（如 tool_use），尝试提取常见属性
    attrs = {}
    for attr in ("type", "id", "name", "input", "tool_use_id", "content", "text"):
        v = getattr(block, attr, None)
        if v is not None:
            attrs[attr] = v
    if attrs:
        return attrs
    # fallback: str 表示
    return {"text": str(block)}

for msg in messages:
    clean = {"role": msg["role"]}
    if isinstance(msg.get("content"), str):
        clean["content"] = msg["content"]
    elif isinstance(msg.get("content"), list):
        clean["content"] = [block_to_dict(block) for block in msg["content"]]
    else:
        clean["content"] = msg.get("content", "")
    cleaned.append(clean)

[meichiguozhuro]

下面是一个日志： normalize前原始消息历史：[{'role': 'user', 'content': '你好'}, {'role': 'assistant', 'content': [TextBlock(citations=None, text='你好！很高兴见到你！有什么我可以帮助你的吗？', type='text')]}, {'role': 'user', 'content': '帮我在D盘创建一个txt文件'}] normalize后消息历史：[{'role': 'user', 'content': '你好'}, {'role': 'assistant', 'content': []}, {'role': 'user', 'content': '帮我在D盘创建一个txt文件'}]
可见TextBlock被清洗掉了
是不是因为这部分代码 cleaned = [] for msg in messages: clean = {"role": msg["role"]} if isinstance(msg.get("content"), str): clean["content"] = msg["content"] elif isinstance(msg.get("content"), list): clean["content"] = [ {k: v for k, v in block.items() if not k.startswith("_")} for block in msg["content"] if isinstance(block, dict) ] else: clean["content"] = msg.get("content", "") cleaned.append(clean) 这一行if isinstance(block, dict)无法兼容返回的TextBlock类

用ai修了下， 兼容后就能跑了 def normalize_messages(messages: list) -> list: """Clean up messages before sending to the API.

Three jobs:
1. Strip internal metadata fields the API doesn't understand
2. Ensure every tool_use has a matching tool_result (insert placeholder if missing)
3. Merge consecutive same-role messages (API requires strict alternation)
"""
cleaned = []
def block_to_dict(block):
    if isinstance(block, dict):
        return {k: v for k, v in block.items() if not k.startswith("_")}
    # 兼容模型返回的工具块对象（如 tool_use），尝试提取常见属性
    attrs = {}
    for attr in ("type", "id", "name", "input", "tool_use_id", "content", "text"):
        v = getattr(block, attr, None)
        if v is not None:
            attrs[attr] = v
    if attrs:
        return attrs
    # fallback: str 表示
    return {"text": str(block)}

for msg in messages:
    clean = {"role": msg["role"]}
    if isinstance(msg.get("content"), str):
        clean["content"] = msg["content"]
    elif isinstance(msg.get("content"), list):
        clean["content"] = [block_to_dict(block) for block in msg["content"]]
    else:
        clean["content"] = msg.get("content", "")
    cleaned.append(clean)

确实，问题不大