mirror of
https://github.com/langgenius/dify-official-plugins.git
synced 2026-07-22 01:55:27 -04:00
The deepseek-r1 model does not support streaming output. #224
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @Moiyu1998 on GitHub (Apr 24, 2025).
Self Checks
Dify version
1.3.0
Cloud or Self Hosted
Self Hosted (Docker)
Steps to reproduce
Why does the deepseek-r1 model no longer support streaming output? Versions prior to 1.0.0 allowed streaming, but now with r1, you have to wait until the entire output is generated before you can see the results. This is not very user-friendly. Could you please consider supporting streaming output again, as it was in previous versions?
✔️ Expected Behavior
No response
❌ Actual Behavior
No response
@Moiyu1998 commented on GitHub (Apr 24, 2025):
It seems that it’s not just the r1 model—other inference models that previously supported streaming output, such as o1-preview, can no longer stream output either. In earlier versions, streaming output was available.
@Moiyu1998 commented on GitHub (Apr 26, 2025):
https://github.com/user-attachments/assets/1466717f-f608-4eb0-b12b-af01d1ffaa8b
As shown in the video above, the current r1 model does not support streaming output. For some questions, the waiting time is very long, which is not very user-friendly. Please support streaming output for this type of inference model as in previous versions. Thank you.
@Moiyu1998 commented on GitHub (Apr 26, 2025):
https://github.com/user-attachments/assets/3e0d17f5-0810-4587-aae1-3f248d76085e
This is version 0.15.5, which supports streaming output. You can see the output process. I also noticed that previous versions provided a small notification about time consumption and token usage after the output was finished. This feature was very useful, but I’m not sure why it is no longer available. Could you please consider bringing this feature back as well?
@Moiyu1998 commented on GitHub (Apr 27, 2025):
https://github.com/user-attachments/assets/f051b44d-ce65-45a7-9f38-ae24121c0d66
In addition, I have found another issue. In version 1.3.0, not only is the small notification no longer available as it was in version 0.15.5, but also clicking the log icon does not trigger any response. Please kindly look into this issue and fix it as well.
@Moiyu1998 commented on GitHub (Apr 29, 2025):
@crazywoola When you have time, could you please take a look at these issues?
@crazywoola commented on GitHub (Apr 29, 2025):
Okay. :)
@Moiyu1998 commented on GitHub (Apr 30, 2025):
I just realized that it’s not only the inference models that no longer have streaming output—it seems that none of the models support streaming output anymore. Have we disabled streaming output altogether? I feel that this might negatively affect the user experience.
@Moiyu1998 commented on GitHub (May 6, 2025):
@crazywoola Am I the only one experiencing the issue of not having streaming output? Why hasn’t anyone else reported it? Are you encountering this problem as well?
@Moiyu1998 commented on GitHub (May 7, 2025):
@crazywoola Could you please take a look at this issue? I’m not sure if it’s just happening to me or if it’s indeed a bug. If it is a bug, could it be fixed? Thank you.
@Moiyu1998 commented on GitHub (May 8, 2025):
https://github.com/user-attachments/assets/f37dd0a4-8eb8-4bdf-97aa-22d99a501a32
@crazywoola I tried our Dify Cloud version, and it’s the same—there’s no streaming output, or the output comes in large chunks, as if something is being blocked. There must be an issue here. I’m really asking for your help—could you please take a look at this problem?