Question Classifcation function in workflow maybe abnormal

[tigflanker]

Self Checks

• This is only for bug report, if you would like to ask a question, please head to Discussions.
• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
• Please do not modify this template :) and fill in all the required fields.

Dify version

v0.7.3

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

Hello experts, I conducted a comparison test in version v0.7.3.

On the left is the regular model invocation, and on the right is the issue classification within the workflow.

The green box represents the same prompt, the yellow box is the test question and the response, and the red box on the right indicates potential issues.

In this code(https://github.com/langgenius/dify/blob/main/api/core/workflow/nodes/question_classifier/template_prompts.py), it seems that some prompt words were not cleaned up and removed, which might be the reason for the inaccurate classification. Please check this issue.

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

[tigflanker]

In addition, I replaced the Question Classifcation module with a general large model + prompt, and it operates quite smoothly with a high accuracy rate. 👍

Another idea I have is to collect bad cases using a knowledge base approach to create a mapping from questions to classifications. Therefore, this issue is not too urgent.

[tu-jiajun]

请问找到解决方法了吗

[Copilotes]

What's your prompt for the llm, can you share it with me? I'm afraid I was in the same situation.

[Copilotes]

Perhaps it's time to enhance this classifier, as it currently seems very poorly functioning to me, with most of the time allocating randomly.