I have searched the existing issues and discussions.
Problem Description
I run local LLMs and serve them through Open WebUI. Because some of my LLMs are slower CPU inference systems, I turn off title generation because I don't want it to waste time having to generate that, I'd prefer it be quicker to respond to my second query immediately.
Desired Solution you'd like
It would be very helpful to have an option like:
Settings > Interface > Title generation style
with two options (maybe a drop-down?):
Use task model
Use start of prompt
The start of the prompt would be good enough if many cases to jog my memory about what a conversation was (far better than the "New Chat" that they all have right now with title generation off), without incurring the non-trivial cost of summarization for slower local models.
Alternatives Considered
For now I just turn off title generation, which isn't optimal because it's hard to then find a specific conversation in the history.
Additional Context
No response
Originally created by @VanceVagell on GitHub (Jun 11, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/14895
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
### Problem Description
I run local LLMs and serve them through Open WebUI. Because some of my LLMs are slower CPU inference systems, I turn off title generation because I don't want it to waste time having to generate that, I'd prefer it be quicker to respond to my second query immediately.
### Desired Solution you'd like
It would be very helpful to have an option like:
Settings > Interface > Title generation style
with two options (maybe a drop-down?):
1) Use task model
2) Use start of prompt
The start of the prompt would be good enough if many cases to jog my memory about what a conversation was (far better than the "New Chat" that they all have right now with title generation off), without incurring the non-trivial cost of summarization for slower local models.
### Alternatives Considered
For now I just turn off title generation, which isn't optimal because it's hard to then find a specific conversation in the history.
### Additional Context
_No response_
While the development team works on implementing the feature, you can meanwhile use this title generation prompt as a workaround. Even a lighter model like Ollama (LLaMA 3.2:1B) running on a CPU can achieve this relatively quickly.
Task:
Extract the first 4-6 words generated by the assistant
Guidelines:
The title should be the first 4-6 words of the response from assistant as it is
Extract only the first 4-6 words, depending on their length and clarity.
If the response starts with many words like "The", "is", "a", you can extract the first 6 words
End the title with an ellipsis (...)
Output:
JSON format: { "title": "your extracted title here ..." }
Examples:
{ "title": "Stock Market Trends ..." }
{ "title": "Perfect Chocolate Chip Recipe ..." }
{ "title": "Evolution of Music Streaming ..." }
Chat History:
<chat_history>
{{MESSAGES:END:2}}
</chat_history>
<!-- gh-comment-id:2975543681 -->
@sreesdas commented on GitHub (Jun 16, 2025):
While the development team works on implementing the feature, you can meanwhile use this title generation prompt as a workaround. Even a lighter model like Ollama (LLaMA 3.2:1B) running on a CPU can achieve this relatively quickly.
### Task:
Extract the first 4-6 words generated by the assistant
### Guidelines:
- The title should be the first 4-6 words of the response from assistant as it is
- Extract only the first 4-6 words, depending on their length and clarity.
- If the response starts with many words like "The", "is", "a", you can extract the first 6 words
- End the title with an ellipsis (...)
### Output:
JSON format: { "title": "your extracted title here ..." }
### Examples:
- { "title": "Stock Market Trends ..." }
- { "title": "Perfect Chocolate Chip Recipe ..." }
- { "title": "Evolution of Music Streaming ..." }
### Chat History:
<chat_history>
{{MESSAGES:END:2}}
</chat_history>
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 @VanceVagell on GitHub (Jun 11, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/14895
Check Existing Issues
Problem Description
I run local LLMs and serve them through Open WebUI. Because some of my LLMs are slower CPU inference systems, I turn off title generation because I don't want it to waste time having to generate that, I'd prefer it be quicker to respond to my second query immediately.
Desired Solution you'd like
It would be very helpful to have an option like:
Settings > Interface > Title generation style
with two options (maybe a drop-down?):
The start of the prompt would be good enough if many cases to jog my memory about what a conversation was (far better than the "New Chat" that they all have right now with title generation off), without incurring the non-trivial cost of summarization for slower local models.
Alternatives Considered
For now I just turn off title generation, which isn't optimal because it's hard to then find a specific conversation in the history.
Additional Context
No response
@sreesdas commented on GitHub (Jun 16, 2025):
While the development team works on implementing the feature, you can meanwhile use this title generation prompt as a workaround. Even a lighter model like Ollama (LLaMA 3.2:1B) running on a CPU can achieve this relatively quickly.
Task:
Extract the first 4-6 words generated by the assistant
Guidelines:
Output:
JSON format: { "title": "your extracted title here ..." }
Examples:
Chat History:
<chat_history>
{{MESSAGES:END:2}}
</chat_history>
@tjbck commented on GitHub (Jun 16, 2025):
Addressed with
ea578af45f