chatglm2 couldn't remember chat history, cause the lack of ability of multi-turn conversations #725

Closed
opened 2026-02-21 17:28:11 -05:00 by yindo · 1 comment
Owner

Originally created by @lchustc on GitHub (Nov 21, 2023).

Originally assigned to: @takatost on GitHub.

Dify version

0.3.31

Cloud or Self Hosted

Self Hosted

Steps to reproduce

正常部署

✔️ Expected Behavior

If i chat with chatglm2, it seems that dify demo couldn't remember chat history, which means the "history" will be empty list.
It doesn't have the ability for multi-turn conversations because the history is always empty.

image image

Actual Behavior

No response

Originally created by @lchustc on GitHub (Nov 21, 2023). Originally assigned to: @takatost on GitHub. ### Dify version 0.3.31 ### Cloud or Self Hosted Self Hosted ### Steps to reproduce 正常部署 ### ✔️ Expected Behavior If i chat with chatglm2, it seems that dify demo couldn't remember chat history, which means the "history" will be empty list. It doesn't have the ability for multi-turn conversations because the history is always empty. <img width="779" alt="image" src="https://github.com/langgenius/dify/assets/27990344/1d8c0069-9be0-4d46-8058-3284c4dba249"> <img width="848" alt="image" src="https://github.com/langgenius/dify/assets/27990344/73d2d14b-13a6-4a46-bcb6-4da58fa5c05c"> ### ❌ Actual Behavior _No response_
yindo added the 🐞 bug label 2026-02-21 17:28:11 -05:00
yindo closed this issue 2026-02-21 17:28:11 -05:00
Author
Owner

@takatost commented on GitHub (Nov 22, 2023):

Thanks a lot for the feedback! 😊
I actually noticed this issue too, and we'll definitely fix it in the upcoming version. We'll also make sure to add support for ChatGLM3. In the meantime, you can use Xinference as an alternative solution to deploy the ChatGLM2 model.

@takatost commented on GitHub (Nov 22, 2023): Thanks a lot for the feedback! 😊 I actually noticed this issue too, and we'll definitely fix it in the upcoming version. We'll also make sure to add support for ChatGLM3. In the meantime, you can use `Xinference` as an alternative solution to deploy the ChatGLM2 model.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langgenius/dify#725