如何上传文件到知识库但跳过自动拆分?想保留原文件 + 使用自己拆分的数据 #6371
Replies: 2 comments
-
|
模板导入 |
Beta Was this translation helpful? Give feedback.
-
|
This kind of failure is often less about one broken component and more about a pipeline mismatch between chunking, embeddings, reranking, and the retrieval path that consumes them. The frustrating part is that each stage can look healthy on its own while the end-to-end flow still feels wrong. If I were debugging it, I would validate each stage independently with one sample document and one known query before scaling back up. That usually exposes whether the problem is compatibility, configuration drift, or simply an unexpected assumption between adjacent stages. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
大家好,想请教一个知识库使用的问题。
我的场景
我们有大量文档需要导入知识库,但 FastGPT 的自动拆分效果对我们的文档不太理想,所以我们通过其他方式预处理好了问答对。
现在想把这些数据导入知识库,遇到了一个两难的问题:
我想要的效果
想问下大家有没有遇到类似的需求?是怎么解决的?
或者有没有什么参数/方法可以实现"只上传文件不处理,然后手动推数据"?
感谢🙏
Beta Was this translation helpful? Give feedback.
All reactions