I have searched the existing issues and discussions.
Problem Description
I've recently been using RAG quite heavily and it's fantastic.
Unfortunately... with only 16GB of VRAM, using ~14B parameter models + the RAG model staking up around 8GB after a while, things get tight pretty quickly.
Desired Solution you'd like
I'd like to ask for it to be possible to unload models after a period of time (or immediately once they've done their job), similar to the KEEP_ALIVE variable in Ollama?
Alternatives Considered
Manually restarting the container, various "action functions" which don't really solve the issue.
Additional Context
No response
Originally created by @dojoca on GitHub (Aug 19, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16724
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
### Problem Description
I've recently been using RAG quite heavily and it's fantastic.
Unfortunately... with only 16GB of VRAM, using ~14B parameter models + the RAG model staking up around 8GB after a while, things get tight pretty quickly.
### Desired Solution you'd like
I'd like to ask for it to be possible to unload models after a period of time (or immediately once they've done their job), similar to the KEEP_ALIVE variable in Ollama?
### Alternatives Considered
Manually restarting the container, various "action functions" which don't really solve the issue.
### Additional Context
_No response_
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 @dojoca on GitHub (Aug 19, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16724
Check Existing Issues
Problem Description
I've recently been using RAG quite heavily and it's fantastic.
Unfortunately... with only 16GB of VRAM, using ~14B parameter models + the RAG model staking up around 8GB after a while, things get tight pretty quickly.
Desired Solution you'd like
I'd like to ask for it to be possible to unload models after a period of time (or immediately once they've done their job), similar to the KEEP_ALIVE variable in Ollama?
Alternatives Considered
Manually restarting the container, various "action functions" which don't really solve the issue.
Additional Context
No response