[GH-ISSUE #8798] When Autocomplete fails to find local configured model, it should fail instead of sending everything online #30792

New Issue

GiteaMirror · 2026-04-25T05:00:19-05:00

GiteaMirror commented

2026-04-25 05:00:19 -05:00

Originally created by @TFWol on GitHub (Jan 23, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/8798

Feature Request

Is your feature request related to a problem? Please describe.
If I configure a local model to perform Autocomplete (which I find useful) while I use an online model for the actual chat, Autocomplete will send everything to the online model if it can't find the local one. I learned about this after seeing my bill double in price.

You can reproduce this by setting up a local model for both Local and External Model dropdowns under Set Task Model, killing the ollama instance, and then start typing in a chat with an online model. You'll see web-traffic start leaking out.

Describe the solution you'd like
Instead of using the current model as a silent fallback, have an option to let it fail or indicate there's an issue with the configured model. For all affected tasks (chat title creation, autocomplete, etc.), it could show placeholder text like 'failed to contact model'

Describe alternatives you've considered
Turning off Autocomplete all together, but I really want to use it - offline though.

Originally created by @TFWol on GitHub (Jan 23, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/8798 # Feature Request **Is your feature request related to a problem? Please describe.** If I configure a local model to perform Autocomplete (which I find useful) while I use an online model for the actual chat, Autocomplete will send everything to the online model if it can't find the local one. I learned about this after seeing my bill double in price. You can reproduce this by setting up a local model for both **Local** and **External Model** dropdowns under **Set Task Model**, killing the ollama instance, and then start typing in a chat with an online model. You'll see web-traffic start leaking out. **Describe the solution you'd like** Instead of using the current model as a silent fallback, have an option to let it fail or indicate there's an issue with the configured model. For all affected tasks (chat title creation, autocomplete, etc.), it could show placeholder text like _'failed to contact model'_ **Describe alternatives you've considered** Turning off Autocomplete all together, but I really want to use it - offline though.

GiteaMirror closed this issue

2026-04-25 05:00:19 -05:00

GiteaMirror commented

2026-04-25 05:00:20 -05:00

@flefevre commented on GitHub (Jan 23, 2025):

This seems to be really important if true because it is a security breach with potentially data leak through prompt leak.

@flefevre commented on GitHub (Jan 23, 2025): This seems to be really important if true because it is a security breach with potentially data leak through prompt leak.

GiteaMirror commented

2026-04-25 05:00:20 -05:00

@tjbck commented on GitHub (Jan 23, 2025):

It uses a task model you defined in the admin settings.

@tjbck commented on GitHub (Jan 23, 2025): It uses a task model you defined in the admin settings.

Sign in to join this conversation.

Branches Tags

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/open-webui#30792