Cruz Rowe

https://github.com/vnpy/vnpy_ctastrategy/blob/1125ae953b6721f45ecfd7c212f91abddcaaa6c6/vnpy_ctastrategy/backtesting.py#L389 ``` import math annual_return: float = (math.pow(end_balance / self.capital, 240 / total_days) - 1)*100 ```

highlevel 应该和 capital 再比较一次

1

https://github.com/vnpy/vnpy_ctastrategy/blob/1125ae953b6721f45ecfd7c212f91abddcaaa6c6/vnpy_ctastrategy/backtesting.py#L343C68-L343C74 ``` _balance = df["balance"].copy() if _balance.iloc[0] < self.capital: _balance.iloc[0] = self.capital df["highlevel"] = _balance.rolling(min_periods=1, window=len(df), center=False).max() ```

关于examples/csqa kag 对比lightrag 的结果较差的原因

1

以下是我的配置： #------------project configuration start----------------# openie_llm: &openie_llm api_key: base_url: https://dashscope.aliyuncs.com/compatible-mode/v1 model: qwen3-8b type: maas stream: true extra_body: {chat_template_kwargs: {enable_thinking: false}} chat_llm: &chat_llm api_key: base_url: https://dashscope.aliyuncs.com/compatible-mode/v1 model: qwen-plus-latest type: maas stream: true...