github-starred/open-webui
2
0
Fork 0
mirror of https://github.com/open-webui/open-webui.git synced 2026-05-06 10:58:17 -05:00
Code Issues 1.3k Packages Projects Releases 126 Wiki Activity

[GH-ISSUE #16973] issue: Post-tool “thinking” text leaks outside reasoning tags #56784

New Issue
Closed
opened 2026-05-05 20:06:34 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-05-05 20:06:34 -05:00
Owner
Copy Link

Originally created by @FabioPolito24 on GitHub (Aug 27, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16973

Check Existing Issues

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

Installation Method

Other

Open WebUI Version

v0.6.25

Ollama Version (if applicable)

No response

Operating System

Linux

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

Any internal reasoning produced after the tool result should remain hidden within the reasoning wrapper/tags.
No orphan closing tag should appear without a corresponding opening tag

Actual Behavior

The model “thinks,” then issues the MCP tool call.
After the tool result is shown, additional “thinking” tokens are displayed to the user (not wrapped).
Only the closing reasoning tag appears afterward (no opening tag).
Internal reasoning becomes visible in the transcript

Steps to Reproduce

Configure an MCP tool (In my case a Tavily web search tool) in Open WebUI.
Run Qwen/Qwen3-32B-AWQ in native function-calling mode with streaming enabled.
Prompt the model with with a question that force him to use the tool and to keep thinking after using it
Observe the streamed transcript in the chat UI.

Logs & Screenshots

Image

Additional Information

Reproduced consistently with Qwen/Qwen3-32B-AWQ. Similar models that emit structured reasoning may also trigger it.

Originally created by @FabioPolito24 on GitHub (Aug 27, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/16973 ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of Open WebUI. ### Installation Method Other ### Open WebUI Version v0.6.25 ### Ollama Version (if applicable) _No response_ ### Operating System Linux ### 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 Any internal reasoning produced after the tool result should remain hidden within the reasoning wrapper/tags. No orphan closing tag should appear without a corresponding opening tag ### Actual Behavior The model “thinks,” then issues the MCP tool call. After the tool result is shown, additional “thinking” tokens are displayed to the user (not wrapped). Only the closing reasoning tag appears afterward (no opening tag). Internal reasoning becomes visible in the transcript ### Steps to Reproduce Configure an MCP tool (In my case a Tavily web search tool) in Open WebUI. Run Qwen/Qwen3-32B-AWQ in native function-calling mode with streaming enabled. Prompt the model with with a question that force him to use the tool and to keep thinking after using it Observe the streamed transcript in the chat UI. ### Logs & Screenshots <img width="1390" height="1313" alt="Image" src="https://github.com/user-attachments/assets/33d860e6-99f5-455c-9de5-a488e2d0b778" /> ### Additional Information Reproduced consistently with Qwen/Qwen3-32B-AWQ. Similar models that emit structured reasoning may also trigger it.
GiteaMirror added the bug label 2026-05-05 20:06:34 -05:00
GiteaMirror commented 2026-05-05 20:06:37 -05:00
Author
Owner
Copy Link

@tjbck commented on GitHub (Aug 27, 2025):

This behaviour depends on the model/inference engine you're using.

<!-- gh-comment-id:3228817379 --> @tjbck commented on GitHub (Aug 27, 2025): This behaviour depends on the model/inference engine you're using.
GiteaMirror commented 2026-05-05 20:06:37 -05:00
Author
Owner
Copy Link

@gkachru commented on GitHub (Nov 3, 2025):

I am also facing this with Qwen3 reasoning models running via VLLM. Anyone has any way to fix this?

<!-- gh-comment-id:3481799999 --> @gkachru commented on GitHub (Nov 3, 2025): I am also facing this with Qwen3 reasoning models running via VLLM. Anyone has any way to fix this?
GiteaMirror commented 2026-05-05 20:06:38 -05:00
Author
Owner
Copy Link

@Classic298 commented on GitHub (Nov 3, 2025):

@gkachru can you test the /dev branch? A fix related to thinking tags was deployed there.

<!-- gh-comment-id:3482100706 --> @Classic298 commented on GitHub (Nov 3, 2025): @gkachru can you test the /dev branch? A fix related to thinking tags was deployed there.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
bug
confirmed
confirmed issue
core
documentation
enhancement
good first issue
help wanted
non-core
pull-request
Mirrored from GitHub Pull Request
 python
question
testing wanted
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#56784
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?