[GH-ISSUE #19281] issue: RAG Template applied with "Bypass Embedding and Retrieval" enabled #34357

New Issue

GiteaMirror · 2026-04-25T08:17:59-05:00

GiteaMirror commented

2026-04-25 08:17:59 -05:00

Originally created by @lucyknada on GitHub (Nov 19, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/19281

Check Existing Issues

I have searched for any existing and/or related issues.
I have searched for any existing and/or related discussions.
I have also searched in the CLOSED issues AND CLOSED discussions and found no related items (your issue might already be addressed on the development branch!).
I am using the latest version of Open WebUI.

Installation Method

Git Clone

Open WebUI Version

v0.6.36

Ollama Version (if applicable)

No response

Operating System

debian 13

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 provided every relevant configuration, setting, and environment variable used in my setup.
I have clearly listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
I have documented step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation. My steps:
Start with the initial platform/version/OS and dependencies used,
Specify exact install/launch/configure commands,
List URLs visited, user input (incl. example values/emails/passwords if needed),
Describe all options and toggles enabled or changed,
Include any files or environmental changes,
Identify the expected and actual result at each stage,
Ensure any reasonably skilled user can follow and hit the same issue.

Expected Behavior

When "Bypass Embedding and Retrieval" is enabled in settings, the system should bypass both:

Vector database embedding/retrieval
The verbose RAG template prompt that includes citation and formatting instructions

Actual Behavior

When "Bypass Embedding and Retrieval" is enabled:

The verbose RAG template is still applied, injecting unwanted citation instructions (incl. with citations off in the model settings)
"N/A" citations appear in the context when bypass is enabled (due to source IDs being set to "N/A" when bypassing)

Steps to Reproduce

Install OpenWebUI (tested on latest version)
Go to Settings > Documents
Enable "Bypass Embedding and Retrieval" toggle
Upload a document or attach a file to a chat
Send a query (e.g., "summary?")
Check the actual request sent to the model (via inference logs)
Observe that the verbose RAG template with citation instructions is still being applied

Logs & Screenshots

Additional Information

The code applies the RAG template whenever context_string != "" without checking if BYPASS_EMBEDDING_AND_RETRIEVAL is enabled:

if context_string != "":
    form_data["messages"] = add_or_update_user_message(
        rag_template(
            request.app.state.config.RAG_TEMPLATE,
            context_string,
            prompt,
        ),
        form_data["messages"],
        append=False,
    )

Originally created by @lucyknada on GitHub (Nov 19, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/19281 ### Check Existing Issues - [x] I have searched for any existing and/or related issues. - [x] I have searched for any existing and/or related discussions. - [x] I have also searched in the CLOSED issues AND CLOSED discussions and found no related items (your issue might already be addressed on the development branch!). - [x] I am using the latest version of Open WebUI. ### Installation Method Git Clone ### Open WebUI Version v0.6.36 ### Ollama Version (if applicable) _No response_ ### Operating System debian 13 ### 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 **provided every relevant configuration, setting, and environment variable used in my setup.** - [x] I have clearly **listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup** (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc). - [x] I have documented **step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation**. My steps: - Start with the initial platform/version/OS and dependencies used, - Specify exact install/launch/configure commands, - List URLs visited, user input (incl. example values/emails/passwords if needed), - Describe all options and toggles enabled or changed, - Include any files or environmental changes, - Identify the expected and actual result at each stage, - Ensure any reasonably skilled user can follow and hit the same issue. ### Expected Behavior When "Bypass Embedding and Retrieval" is enabled in settings, the system should bypass both: 1. Vector database embedding/retrieval 2. The verbose RAG template prompt that includes citation and formatting instructions ### Actual Behavior When "Bypass Embedding and Retrieval" is enabled: - The verbose RAG template is still applied, injecting unwanted citation instructions (incl. with citations off in the model settings) - "N/A" citations appear in the context when bypass is enabled (due to source IDs being set to "N/A" when bypassing) ### Steps to Reproduce 1. Install OpenWebUI (tested on latest version) 2. Go to Settings > Documents 3. Enable "Bypass Embedding and Retrieval" toggle 4. Upload a document or attach a file to a chat 5. Send a query (e.g., "summary?") 6. Check the actual request sent to the model (via inference logs) 7. Observe that the verbose RAG template with citation instructions is still being applied ### Logs & Screenshots - ### Additional Information The code applies the RAG template whenever `context_string != ""` without checking if `BYPASS_EMBEDDING_AND_RETRIEVAL` is enabled: ```python if context_string != "": form_data["messages"] = add_or_update_user_message( rag_template( request.app.state.config.RAG_TEMPLATE, context_string, prompt, ), form_data["messages"], append=False, ) ```

GiteaMirror added the bug label 2026-04-25 08:18:00 -05:00

GiteaMirror closed this issue

2026-04-25 08:18:01 -05:00

GiteaMirror commented

2026-04-25 08:18:01 -05:00

@tjbck commented on GitHub (Nov 19, 2025):

BYPASS_EMBEDDING_AND_RETRIEVAL is separate from RAG_TEMPLATE, did you have any specific use cases in mind?

@tjbck commented on GitHub (Nov 19, 2025): `BYPASS_EMBEDDING_AND_RETRIEVAL` is separate from `RAG_TEMPLATE`, did you have any specific use cases in mind?

Sign in to join this conversation.

Branches Tags

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/open-webui#34357