feat: Add an option to set Full Context as the default for file uploads

[Zijie-wang1125]

Check Existing Issues

• I have searched the existing issues and discussions.

Problem Description

Many users, like myself, frequently use both RAG and Full Context Mode. However, in OpenWebUI, Full Context Mode (also known as Focus Retrieval Mode) is unreliable. Unless embedding and the bypass feature are completely disabled, all uploaded documents will still be embedded, which can caused token waste and significantly increase the upload time. Also, if i upload multiple files at the same time and do not want to use Rag, it can be painful to toggle Focus Reetrieval for every files one-by-one manually.

Desired Solution you'd like

It would be a huge improvement for experience if we can have an option to set the default behavior of attaching knowledge to be regular Rag. And the default behavior of of uploading file to be full context.

Alternatives Considered

Or just have a more flexible and controllable interface. For example, in the toolbar, we can choose to turn on rag or not. This can be problematic, as the current logic of OpenWebUI's toolbar only applies tool or process after a user sends a message, where the user's request is the trigger. Our goal is to completely bypass the embedding process for user-uploaded files before user even send a message.

Additional Context

This feature can be really helpful to people working on text writting or reading tasks. because we always want LLM to be able to access the full context of our document as a whole. Currently the only solution is to turn on 'Bypass Embedding and Retrival' in settings, which will cause Rag to be disabled completely. However, Rag is also a great functionality that we don't want to give up, we just don't want it to take effect in these cases.

[FraYoshi]

It's super painfull that this has not already been implemented... Each time I upload documents have to go one-by-one and enable the full context for each and every one of them. Then, wait for the vectorialization to finish, to just discard it because going full context :')
It's a waste of time, resources, and patience ^^"
So, super arrow up from me!