infiniflow
[Bug]: When the keyword match exceeds a certain length, the prompt is replaced by the system.
Self Checks
- [x] I have searched for existing issues search for existing issues, including closed ones.
- [x] I confirm that I am using English to submit this report (Language Policy).
- [x] Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
- [x] Please do not modify this template :) and fill in all the required fields.
RAGFlow workspace code commit ID
8e96504
RAGFlow image version
v0.17.1
Other environment information
deepseek r1 14b
nginx
Actual behavior
When the knowledge base parses the Excel file in table mode, and the matched data exceeds a certain length, the chatbot prompt is replaced by the system. As a result, the large language model is not triggered and returns the data directly.
When the volume of data matched by the keyword is not large, the performance is normal, as shown in the figure:
Additionally, the conversion of Markdown is quite unstable. Sometimes it converts into an HTML table, and sometimes it doesn't.
Expected behavior
It can handle large Excel files with a large amount of data, and the performance is relatively normal.
Steps to reproduce
1. Create a knowledge base in table analysis mode.
2. Import an Excel file containing 145 data entries.
3. Write chatbot prompts and reference the knowledge base.
4. Test the questions.
Additional information
No response
For Table parsed KB, the result is not responded by LLM. So, no prompt is needed.
LLM just turns your question to ES SQL to select data out.
For parsed KB, the result is not responded by LLM. So, no prompt is needed. LLM just turns your question to ES SQL to select data out.
Table
Then why does the LLM generate additional content based on the prompt when the volume of matched data is small? Also, I don't want the query to return results directly. I'd like the LLM to analyze and summarize the knowledge base. What should I do?