[GH-ISSUE #18431] feat: Option to define default mode for file uploads #18597

Open
opened 2026-04-20 00:48:55 -05:00 by GiteaMirror · 8 comments
Owner

Originally created by @Master-Pr0grammer on GitHub (Oct 19, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/18431

Check Existing Issues

  • I have searched all existing open AND closed issues and discussions for similar requests. I have found none that is comparable to my request.

Problem Description

When uploading files in a chat, all files are first required to be split up and processed for rag before you can do anything, even if you only intend to use them as a full file without using rag at all.

This process can take a while especially for local models because the current LLM needs to be unloaded (deleting prompt cache), then load in the embedding model, process the document, then reload the LLM, and reprocess the entire chat history because the prompt cache has been reset, which can take very long for longer chats.

It is also quite annoying to have to remember to switch it to use the whole file.

Currently, there is no way to have it default to "Using Entire Document" Vs "Using Focused Retrieval" for a regular file upload in the chat UI, without completely disabling RAG for the entire service as a whole by turning on "bypass RAG" in the admin panel. RAG is a very useful feature, but isn't always needed.

Desired Solution you'd like

An option in settings to set the default mode for chat file uploads where you can select it to default to "Using Entire Document" Vs "Using Focused Retrieval".

Alternatives Considered

No response

Additional Context

No response

Originally created by @Master-Pr0grammer on GitHub (Oct 19, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/18431 ### Check Existing Issues - [x] I have searched all existing open AND closed issues and discussions for similar requests. I have found none that is comparable to my request. ### Problem Description When uploading files in a chat, all files are first required to be split up and processed for rag before you can do anything, even if you only intend to use them as a full file without using rag at all. This process can take a while especially for local models because the current LLM needs to be unloaded (deleting prompt cache), then load in the embedding model, process the document, then reload the LLM, and reprocess the entire chat history because the prompt cache has been reset, which can take very long for longer chats. It is also quite annoying to have to remember to switch it to use the whole file. Currently, there is no way to have it default to "Using Entire Document" Vs "Using Focused Retrieval" for a regular file upload in the chat UI, without completely disabling RAG for the entire service as a whole by turning on "bypass RAG" in the admin panel. RAG is a very useful feature, but isn't always needed. ### Desired Solution you'd like An option in settings to set the default mode for chat file uploads where you can select it to default to "Using Entire Document" Vs "Using Focused Retrieval". ### Alternatives Considered _No response_ ### Additional Context _No response_
Author
Owner

@silentoplayz commented on GitHub (Oct 20, 2025):

Closing this issue; further discussion can continue on the related request here: https://github.com/open-webui/open-webui/issues/12228

Here's how:

The two posts are about the same underlying pain-point (Open-WebUI always insists on locally parsing / chunking / embedding every file that is dropped into a chat) but they describe two different use-cases and therefore ask for two different knobs.

  1. First post (yours)
    User persona: “I’m a normal chat user, I just want to send the file as-is to the model 90 % of the time and only occasionally use RAG.”
    Requested knob: A per-user default that says “when I drop a file, start in Entire Document mode instead of Focused Retrieval.”
    Scope: UI-only, no server-level change; RAG stays enabled globally.

  2. Second post (https://github.com/open-webui/open-webui/issues/12228)
    User persona: “I’m running a custom pipeline that needs the raw bytes of the uploaded file; I never want Open-WebUI to open, OCR, chunk or embed anything.”
    Requested knob: A server-level switch that completely bypasses the content-extraction pipeline and exposes an URL from which the pipeline can fetch the original file.
    Scope: Admin / back-end; RAG is disabled for the whole instance.

So the root cause (“Open-WebUI always extracts text”) is identical, but the remedies are orthogonal:

  • A default-UI-toggle (post 1) does not help the pipeline author.
  • A full bypass (post 2) would also solve post 1’s complaint, but is over-kill for ordinary chat users who still want RAG to be available on demand.

In short: same river, two bridges.

Also related - https://github.com/open-webui/open-webui/issues/15964

<!-- gh-comment-id:3420191848 --> @silentoplayz commented on GitHub (Oct 20, 2025): Closing this issue; further discussion can continue on the related request here: https://github.com/open-webui/open-webui/issues/12228 Here's how: The two posts are **about the same underlying pain-point** (Open-WebUI always insists on locally parsing / chunking / embedding every file that is dropped into a chat) but they describe **two different use-cases** and therefore ask for **two different knobs**. 1. First post (yours) – **User persona**: “I’m a normal chat user, I just want to send the file **as-is** to the model 90 % of the time and only occasionally use RAG.” – **Requested knob**: A **per-user default** that says “when I drop a file, start in *Entire Document* mode instead of *Focused Retrieval*.” – **Scope**: UI-only, no server-level change; RAG stays enabled globally. 2. Second post (https://github.com/open-webui/open-webui/issues/12228) – **User persona**: “I’m running a **custom pipeline** that needs the **raw bytes** of the uploaded file; I never want Open-WebUI to open, OCR, chunk or embed anything.” – **Requested knob**: A **server-level switch** that completely bypasses the content-extraction pipeline and exposes an URL from which the pipeline can fetch the original file. – **Scope**: Admin / back-end; RAG is disabled for the whole instance. So the **root cause** (“Open-WebUI always extracts text”) is identical, but the **remedies** are orthogonal: - A default-UI-toggle (post 1) does **not** help the pipeline author. - A full bypass (post 2) would **also** solve post 1’s complaint, but is over-kill for ordinary chat users who still want RAG to be available on demand. In short: same river, two bridges. Also related - https://github.com/open-webui/open-webui/issues/15964
Author
Owner

@sanchitbhavsar commented on GitHub (Oct 24, 2025):

@silentoplayz what is the update here, as when i upload file i need to process them in full context as default. Right now it is only possible when user click on the uploaded file and enable the toggle option for full context but it should achievable through some system settings or configurations.

<!-- gh-comment-id:3442035971 --> @sanchitbhavsar commented on GitHub (Oct 24, 2025): @silentoplayz what is the update here, as when i upload file i need to process them in full context as default. Right now it is only possible when user click on the uploaded file and enable the toggle option for full context but it should achievable through some system settings or configurations.
Author
Owner

@silentoplayz commented on GitHub (Oct 24, 2025):

Right now it is only possible when user click on the uploaded file and enable the toggle option for full context but it should achievable through some system settings or configurations.

You can also click on an attached knowledgebase collection or a note to a model and toggle on Using Entire Document for selected attached files/documents.
Image

A global toggle already exists in the admin settings to toggle Full Context Mode too.
Image

<!-- gh-comment-id:3442608430 --> @silentoplayz commented on GitHub (Oct 24, 2025): > Right now it is only possible when user click on the uploaded file and enable the toggle option for full context but it should achievable through some system settings or configurations. You can also click on an attached knowledgebase collection or a note to a model and toggle on `Using Entire Document` for selected attached files/documents. <img width="2313" height="747" alt="Image" src="https://github.com/user-attachments/assets/7d4ed6d4-1141-47d4-8fa2-dce31696a3a3" /> A global toggle already exists in the admin settings to toggle `Full Context Mode` too. <img width="2282" height="538" alt="Image" src="https://github.com/user-attachments/assets/c443d31d-deda-424d-924f-d564a6a49b8f" />
Author
Owner

@sanchitbhavsar commented on GitHub (Oct 24, 2025):

Right now it is only possible when user click on the uploaded file and enable the toggle option for full context but it should achievable through some system settings or configurations.

You can also click on an attached knowledgebase collection or a note to a model and toggle on Using Entire Document for selected attached files/documents. Image

A global toggle already exists in the admin settings to toggle Full Context Mode too. Image

@silentoplayz Thanks for sharing. I am aware of these settings but we do not want to have full content enabled for all. you want your RAG to works as default and do the full context on file uploads. The KB flag is for the files attached to the model but to the files that are dropped in the chat.

<!-- gh-comment-id:3443022281 --> @sanchitbhavsar commented on GitHub (Oct 24, 2025): > > Right now it is only possible when user click on the uploaded file and enable the toggle option for full context but it should achievable through some system settings or configurations. > > You can also click on an attached knowledgebase collection or a note to a model and toggle on `Using Entire Document` for selected attached files/documents. <img alt="Image" width="2000" height="747" src="https://private-user-images.githubusercontent.com/50341825/505239541-7d4ed6d4-1141-47d4-8fa2-dce31696a3a3.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.nbBpric9InfiYwDxxVqce8ujnkA-4_VzNbLnsH9n9aM"> > > A global toggle already exists in the admin settings to toggle `Full Context Mode` too. <img alt="Image" width="2000" height="538" src="https://private-user-images.githubusercontent.com/50341825/505240648-c443d31d-deda-424d-924f-d564a6a49b8f.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.AytZCzJgeaoNaSOs_OUxkCYLjppG_bR2NnwgHs_M5sU"> @silentoplayz Thanks for sharing. I am aware of these settings but we do not want to have full content enabled for all. you want your RAG to works as default and do the full context on file uploads. The KB flag is for the files attached to the model but to the files that are dropped in the chat.
Author
Owner

@adhusch commented on GitHub (Oct 29, 2025):

@silentoplayz i am quite surprised that this was closed after your nice and detailled analysis. Its pretty clear both bridges are needed and brige one should be the default highway. Why should anyone in 2025 (with the context sizes we have now) want RAG as default on simple document attachement upload? RAG is for cases that overshoot context, but thats not the default anymore, but a tool to deal with very large docs or knowlege bases. A full bypass (2) does btw not fully solve this basic requirement, one wants the document to be converted to markdown and injected to the context, not a binary stream injection.

<!-- gh-comment-id:3460826148 --> @adhusch commented on GitHub (Oct 29, 2025): @silentoplayz i am quite surprised that this was closed after your nice and detailled analysis. Its pretty clear both bridges are needed and brige one should be the default highway. Why should anyone in 2025 (with the context sizes we have now) want RAG as default on simple document attachement upload? RAG is for cases that overshoot context, but thats not the default anymore, but a tool to deal with very large docs or knowlege bases. A full bypass (2) does btw *not* fully solve this basic requirement, one wants the document to be converted to markdown and injected to the context, not a binary stream injection.
Author
Owner

@adhusch commented on GitHub (Nov 26, 2025):

Hello,

i have a question to the community: How is anyone using openwebui with RAG enabled in production without this feature? For me the only option is to disable RAG globally, as users will absolutly need full-context mode for normal single-file uploads as default. It will otherwise complety break user experience?

Cheers

<!-- gh-comment-id:3580671344 --> @adhusch commented on GitHub (Nov 26, 2025): Hello, i have a question to the community: How is anyone using openwebui with RAG enabled in production without this feature? For me the only option is to disable RAG globally, as users will absolutly need full-context mode for normal single-file uploads as default. It will otherwise complety break user experience? Cheers
Author
Owner

@wsomm commented on GitHub (Apr 19, 2026):

Continued from discussion #18632.

@Classic298

but i wanna add setting it to default full context might be dangerous. Our users upload PDFs with 20000 pages. thats simply too much

So we have two situations which lead to problems:
"focused retrieval": Nothing found on small PDF files. User confusion. You tell them to choose "full context".
"full context": Context overflow on a PDF file with 20000 pages. You tell them to choose "focused retrievel".

I my opinion, the users that use small ones, are the less educated.
It is far more easy to tell a user who throws in a bunch of PDFs or large ones, to change the option,
than the occasional user, who just wants translation of a document.

And shouldn't it be the responsibility of the admin, to know its user base and choose what is the best fit?

I would be more than thankful for a global option for this. Patching is an option, but really only the the second best.

<!-- gh-comment-id:4276137100 --> @wsomm commented on GitHub (Apr 19, 2026): Continued from discussion #18632. @Classic298 > but i wanna add setting it to default full context might be dangerous. Our users upload PDFs with 20000 pages. thats simply too much So we have two situations which lead to problems: "focused retrieval": Nothing found on small PDF files. User confusion. You tell them to choose "full context". "full context": Context overflow on a PDF file with 20000 pages. You tell them to choose "focused retrievel". I my opinion, the users that use small ones, are the less educated. It is far more easy to tell a user who throws in a bunch of PDFs or large ones, to change the option, than the occasional user, who just wants translation of a document. And shouldn't it be the responsibility of the admin, to know its user base and choose what is the best fit? I would be more than thankful for a global option for this. Patching is an option, but really only the the second best.
Author
Owner

@d1ffuser commented on GitHub (Apr 19, 2026):

Hello,

i have a question to the community: How is anyone using openwebui with RAG enabled in production without this feature? For me the only option is to disable RAG globally, as users will absolutly need full-context mode for normal single-file uploads as default. It will otherwise complety break user experience?

Cheers

Completely agree. Users shouldn't have to disable RAG just to have the convenience of uploading docs for queries without clicking through submenus every single time.

Tbh, the analysis is sound. In the long term, the handling of documents in owui has to change. But short term, a user setting to choose a default upload mode is the way to go.

<!-- gh-comment-id:4276951375 --> @d1ffuser commented on GitHub (Apr 19, 2026): > Hello, > > i have a question to the community: How is anyone using openwebui with RAG enabled in production without this feature? For me the only option is to disable RAG globally, as users will absolutly need full-context mode for normal single-file uploads as default. It will otherwise complety break user experience? > > Cheers Completely agree. Users shouldn't have to disable RAG just to have the convenience of uploading docs for queries without clicking through submenus every single time. Tbh, the analysis is sound. In the long term, the handling of documents in owui has to change. But short term, a user setting to choose a default upload mode is the way to go.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#18597