[Feature]

[fwt11]

MaxKB Version

all

Please describe your needs or suggestions for improvements

Please provide more flexible chunking schemes, such as parent-child documents

Please describe the solution you suggest

No response

Additional Information

No response

[baixin513]

感谢反馈，当前默认逻辑就是父子分段，系统会先分段，在针对分段中的句子进行子分块，向量化的是子分块，检索也是检索的子分块召回的父分段。

[shaohuzhang1]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Thanks for the feedback. The current default logic is parent-child segmentation. The system will segment first, and then perform sub-chunking on the sentences in the segment. The vectorized sub-chunks are also the parent segments recalled by the retrieved sub-chunks.