I have searched the existing issues and discussions.
I am using the latest version of Open WebUI.
Installation Method
Docker
Open WebUI Version
0.6.10
Ollama Version (if applicable)
0.7.0
Operating System
debian bookworm
Browser (if applicable)
No response
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 listed steps to reproduce the bug in detail.
Expected Behavior
Streaming should work with native tool calling.
Actual Behavior
With native tool calling and at least 1 active tool, streaming seems to break (waits until the full response is done, then instantly appears, like streaming is turned off). Doesn't matter if the model uses a tool or not.
Steps to Reproduce
To reproduce:
Ensure Streaming is On
Use a model that supports native tool calling (e.g. Qwen3, Mistral Small ...)
Assign at least one active tool
Ask for anything, instead of streaming tokens while generating, it now waits until the response is done then appears.
I noticed this since my last update and was trying to search around/try different combos of things, but couldn't dial it in more on what's breaking it. Can't see any trace in logs anywhere really. Searching around I only found a reddit post referencing it here, without any solution.
Originally created by @rvkwi on GitHub (May 23, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/14221
### 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
0.6.10
### Ollama Version (if applicable)
0.7.0
### Operating System
debian bookworm
### Browser (if applicable)
_No response_
### 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 listed steps to reproduce the bug in detail.
### Expected Behavior
Streaming should work with native tool calling.
### Actual Behavior
With native tool calling and at least 1 active tool, streaming seems to break (waits until the full response is done, then instantly appears, like streaming is turned off). Doesn't matter if the model uses a tool or not.
### Steps to Reproduce
To reproduce:
- Ensure Streaming is On
- Use a model that supports native tool calling (e.g. Qwen3, Mistral Small ...)
- Assign at least one active tool
- Ask for anything, instead of streaming tokens while generating, it now waits until the response is done then appears.
### Logs & Screenshots
[logs.txt](https://github.com/user-attachments/files/20406940/logs.txt)
### Additional Information
I noticed this since my last update and was trying to search around/try different combos of things, but couldn't dial it in more on what's breaking it. Can't see any trace in logs anywhere really. Searching around I only found a reddit post referencing it [here](https://www.reddit.com/r/OpenWebUI/comments/1kmdjro/streaming_dont_work_when_using_native_calling/), without any solution.
GiteaMirror
added the bug label 2026-04-19 22:54:49 -05:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @rvkwi on GitHub (May 23, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/14221
Check Existing Issues
Installation Method
Docker
Open WebUI Version
0.6.10
Ollama Version (if applicable)
0.7.0
Operating System
debian bookworm
Browser (if applicable)
No response
Confirmation
README.md.Expected Behavior
Streaming should work with native tool calling.
Actual Behavior
With native tool calling and at least 1 active tool, streaming seems to break (waits until the full response is done, then instantly appears, like streaming is turned off). Doesn't matter if the model uses a tool or not.
Steps to Reproduce
To reproduce:
Logs & Screenshots
logs.txt
Additional Information
I noticed this since my last update and was trying to search around/try different combos of things, but couldn't dial it in more on what's breaking it. Can't see any trace in logs anywhere really. Searching around I only found a reddit post referencing it here, without any solution.
@tjbck commented on GitHub (May 23, 2025):
PLEASE check for existing issues/discussions first. This is Ollama issue.