[Question]: categorize in agent works weird.

[RNGMARTIN]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (Language Policy).
• 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).
• Please do not modify this template :) and fill in all the required fields.

Describe your problem

you can see very clearly in the picture. The answer I sent to the AI is clearly complaining, but the "categorize" treated it as "casual". why??

[RNGMARTIN]

this one is more clearly. There are only two options, but can not choose the right one.

[KevinHuSh]

I recomand you switch to another LLM to give a try.
It highly depends on the capability of LLM.

[RNGMARTIN]

I recomand you switch to another LLM to give a try. It highly depends on the capability of LLM.

Thanks. I will take another try

[RNGMARTIN]

capability of LLM

I just tried the glm4 which is only 9B. It can choose the right way of the "categorize". However, the deepseek:32B and the qwq which is also 32B can not make the right decision. It's quite confused.

[hero8152]

兄弟，他这个分类节点几乎没法用，你如果跟dify对比一下就知道了，这个分类是纯电子垃圾，我都不知道他的运行逻辑是什么。

你要用的话，只能连接<回答问题>，然后定义关键词是：
你是一个XX专家，你判断对话内容，分析属于什么分类。你只能回答“生气”、“开心”，不能回复多余的内容，回答简洁。
生气：明显表达生气了。
开心：语气轻松。

类似于这种，可以实现分类的功能。但是需要获取两种结果来进行分类流程，就需要加入<判断>节点，if 的条件写<对话>的输出，包含<开心>，就走下一个流程，else就走另一个流程。

可以说是非常折磨。