issue: Custom Local Model Uses External Task Model for Auto-Prompt/Title/Follow-Up #6334

New Issue

2025-11-11T16:51:48-06:00

GiteaMirror commented

2025-11-11 16:51:48 -06:00

Originally created by @imbible on GitHub (Sep 7, 2025).

Check Existing Issues

I have searched the existing issues and discussions.
I am using the latest version of Open WebUI.

Installation Method

Docker

Open WebUI Version

v0.6.26

Ollama Version (if applicable)

No response

Operating System

macOS Tahoe

Browser (if applicable)

Chrome

Confirmation

I have read and followed all instructions in README.md.
I am using the latest version of both Open WebUI and Ollama.
I have included the browser console logs.
I have included the Docker container logs.
I have provided every relevant configuration, setting, and environment variable used in my setup.
I have clearly listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
I have documented step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation. My steps:
Start with the initial platform/version/OS and dependencies used,
Specify exact install/launch/configure commands,
List URLs visited, user input (incl. example values/emails/passwords if needed),
Describe all options and toggles enabled or changed,
Include any files or environmental changes,
Identify the expected and actual result at each stage,
Ensure any reasonably skilled user can follow and hit the same issue.

Expected Behavior

When a custom model is created in Workspace based on a local model, the chat responses should use the local model and all other features—including auto-prompt suggestions, title generation, and follow-up queries—should exclusively use the Local Task Model specified in Admin Panel > Settings > Interface. No data related to user prompts or system instructions should be sent to the External Task Model.

Actual Behavior

While chat responses correctly use the selected local base model configured in Workplace, auto-prompt suggestions, title generation, and follow-up queries are instead handled by the External Task Model configured in Admin Panel > Settings > Interface. This results in user data being sent to external models, contrary to user expectations and privacy needs.

Steps to Reproduce

Navigate to Workspace in WebUI.
Create a custom model, ensuring to select a local model as the base.
Configure a custom system prompt and save the model.
Start a chat with this custom model.
Observe that while direct chat responses come from the local model, features such as auto-prompt, title generation, and follow-up use the external model specified in Admin Panel > Settings > Interface.

Logs & Screenshots

I used gpt-5-mini as the external task model. When I disconnect the wifi, the auto complete shows this error in the prompt input box, which means it attempted to use the external task model.

After sending the prompt, the local model generated the response, but no AI-generated title or follow-ups. So I'm sure it uses external task model to generate title and auto-complete for local custom model.

Additional Information

Originally created by @imbible on GitHub (Sep 7, 2025). ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of Open WebUI. ### Installation Method Docker ### Open WebUI Version v0.6.26 ### Ollama Version (if applicable) _No response_ ### Operating System macOS Tahoe ### Browser (if applicable) Chrome ### Confirmation - [x] I have read and followed all instructions in `README.md`. - [x] I am using the latest version of **both** Open WebUI and Ollama. - [x] I have included the browser console logs. - [x] I have included the Docker container logs. - [x] I have **provided every relevant configuration, setting, and environment variable used in my setup.** - [x] I have clearly **listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup** (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc). - [x] I have documented **step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation**. My steps: - Start with the initial platform/version/OS and dependencies used, - Specify exact install/launch/configure commands, - List URLs visited, user input (incl. example values/emails/passwords if needed), - Describe all options and toggles enabled or changed, - Include any files or environmental changes, - Identify the expected and actual result at each stage, - Ensure any reasonably skilled user can follow and hit the same issue. ### Expected Behavior When a custom model is created in Workspace based on a local model, the chat responses should use the local model and all other features—including auto-prompt suggestions, title generation, and follow-up queries—should exclusively use the **Local Task Model** specified in Admin Panel > Settings > Interface. No data related to user prompts or system instructions should be sent to the **External Task Model**. ### Actual Behavior While chat responses correctly use the selected local base model configured in Workplace, auto-prompt suggestions, title generation, and follow-up queries are instead handled by the External Task Model configured in Admin Panel > Settings > Interface. This results in user data being sent to external models, contrary to user expectations and privacy needs. ### Steps to Reproduce 1. Navigate to Workspace in WebUI. 2. Create a custom model, ensuring to select a local model as the base. 3. Configure a custom system prompt and save the model. 4. Start a chat with this custom model. 5. Observe that while direct chat responses come from the local model, features such as auto-prompt, title generation, and follow-up use the external model specified in Admin Panel > Settings > Interface. ### Logs & Screenshots I used gpt-5-mini as the external task model. When I disconnect the wifi, the auto complete shows this error in the prompt input box, which means it attempted to use the external task model. <img width="784" height="139" alt="Image" src="https://github.com/user-attachments/assets/288610df-0d7f-4745-a1c9-cdb2d7cca367" /> After sending the prompt, the local model generated the response, but no AI-generated title or follow-ups. So I'm sure it uses **external** task model to generate title and auto-complete for **local** custom model. ### Additional Information

GiteaMirror added the bug label 2025-11-11 16:51:48 -06:00

GiteaMirror closed this issue

2025-11-11 16:51:48 -06:00

GiteaMirror referenced this issue

2026-04-19 20:44:02 -05:00

[GH-ISSUE #6334] too painful to install #14326

GiteaMirror referenced this issue

2026-04-25 04:15:19 -05:00

[GH-ISSUE #6334] too painful to install #29854

GiteaMirror referenced this issue

2026-05-05 14:13:32 -05:00

[GH-ISSUE #6334] too painful to install #52992