Replies: 3 comments
-
Unable to reproduce from our end, keep up updated. |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Oh, I may have configured that folder to use a knowledge collection. Maybe that did it. Other than that, IIRC I think I was using Devstral through ollama (hence the mention of a Modelfile). |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
In addition, I would like to be able to control whether the folder system prompt is appended to the model system prompt or whether it overrides the model system prompt. What do you think? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment

Uh oh!
There was an error while loading. Please reload this page.
-
Check Existing Issues
Installation Method
Docker
Open WebUI Version
v0.6.16
Ollama Version (if applicable)
No response
Operating System
Windows 11
Browser (if applicable)
No response
Confirmation
README.md.Expected Behavior
New Chat -> Click Folder -> Input -> Output Per Model Default System Prompt -> Regenerate -> Output Per Model Default System Prompt
Actual Behavior
New Chat -> Click Folder -> Input -> Output -> Regenerate -> Output Per EMPTY System Prompt
Steps to Reproduce
Use ollama create, etc, to install a model with a Modelfile having a system prompt that clearly differs from the AI LLM model training (eg. specifying to output flow charts in properly block quoted mermaid instead of text).
New Chat -> Click Folder -> Input -> Output -> Regenerate -> Output Per EMPTY System Prompt
Logs & Screenshots
Nothing relevant. Things seem to work normally without obvious errors, except that the system prompt is ignored (which is of course the worst possible situation since the lack of anything other than unexpected AI behavior confuses users).
Additional Information
This is DEFINITELY VERY BAD . Wherever an explicit system prompt is not configured by the user, the default system prompt shipped with the model absolutely should be used, as is the case with every other system prompt configuration text input configuration anywhere in OpenWebUI . Best case this can lead to confusion why model outputs are bad, worst case is very unexpected, not aligned, model outputs that the users unwittingly see and use, possibly forward to third-parties, as-is.
Beta Was this translation helpful? Give feedback.
All reactions