infiniflow
[Bug]: Keywords and questions should not be left out, especially the last dozen or so chunks.
Is there an existing issue for the same bug?
- [x] I have checked the existing issues.
RAGFlow workspace code commit ID
commit 5fdfb8d465826bf27bff61a6c5fb5f8340b0b5cc
RAGFlow image version
v0.16.0-177-g5fdfb8d4 slim
Other environment information
ubuntu 24
ollama qwen2.5:32b
bgm-m3
Actual behavior
I have been conducting tests for several days. When using academic monographs to create a knowledge base, I noticed that in the process of assigning keywords and questions to the chunks, the first 20 chunks are often omitted. That is, when I open the document and check the first and second pages of the chunks, there are no keywords or questions. The parsing settings are such that everything is enabled except for the knowledge graph.
Expected behavior
https://567.daoson.top:8443/#s/_U6PL34A ---> test document
Steps to reproduce
normal control
Additional information
No response
I did not reproduce that. Is that possiblly caused by unstable LLM invokation? How to reproduce that? What's the chunking method?
