[GH-ISSUE #16474] issue: Permission issue - User cannot use full context on RAG #33449

Closed
opened 2026-04-25 07:22:18 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @frenzybiscuit on GitHub (Aug 11, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16474

Check Existing Issues

  • I have searched the existing issues and discussions.
  • I am using the latest version of Open WebUI.

Installation Method

Pip Install

Open WebUI Version

v0.6.21

Ollama Version (if applicable)

N/A (OpenAI)

Operating System

Debian 12

Browser (if applicable)

Firefox

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

End-users should be able to use full context mode on the knowledgebase, but cannot.

Actual Behavior

Doesn't work

Steps to Reproduce

I have tried both docker and pip.

On a PIP install, configure the following:

Image Image

CREATE A NEW GROUP

Image

ADD USER TO GROUP

As the user:

Image Image

CLICK USE ENTIRE DOCUMENT

Image

Left: User | Right: Admin

Image

GO:

Image

Note there is no citation on the left (user).

If it's relevant (I dont think it is, as I've tried different models):

Image

Startup command for PIP:

WEBUI_SECRET_KEY=whoop VECTOR_DB=pgvector PGVECTOR_DB_URL=postgresql://openwebui:whoop@127.0.0.1:5432/openwebui \
DATABASE_URL=postgresql://openwebui:whoop@127.0.0.1:5432/openwebui \
WEBUI_URL=https://chat.ready.art ENABLE_OAUTH_SIGNUP=false THREAD_POOL_SIZE=80 DATABASE_POOL_SIZE=0 \
AIOHTTP_CLIENT_TIMEOUT=120 AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST=15 AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST=15 \
ENABLE_USER_INFO_HEADERS=true CORS_ALLOW_ORIGIN=https://chat.ready.art OPENWEBUI_USE_COLORS=true ENABLE_RAG_WEB_SEARCH=true \
ENABLE_WEB_SEARCH=true ENABLE_SEARCH_QUERY_GENERATION=true RAG_WEB_SEARCH_RESULT_COUNT=3 RAG_WEB_SEARCH_CONCURRENT_REQUESTS=3 \
RAG_WEB_SEARCH_ENGINE=google_pse RAG_WEB_LOADER_ENGINE=safe_web ENABLE_WEBSOCKET_SUPPORT=true ENABLE_WEBSOCKET_SUPPORT=true \
WEBSOCKET_REDIS_RUL=redis://127.0.0.1:6379/0 REDIS_URL=redis://127.0.0.1:6379/0 CUSTOM_NAME=Ready.Art WEBUI_NAME=Ready.Art \
USER_AGENT="Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36" \
open-webui serve```


### Logs & Screenshots

.

### Additional Information

_No response_
Originally created by @frenzybiscuit on GitHub (Aug 11, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/16474 ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of Open WebUI. ### Installation Method Pip Install ### Open WebUI Version v0.6.21 ### Ollama Version (if applicable) N/A (OpenAI) ### Operating System Debian 12 ### Browser (if applicable) Firefox ### 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 End-users should be able to use full context mode on the knowledgebase, but cannot. ### Actual Behavior Doesn't work ### Steps to Reproduce I have tried both docker and pip. On a PIP install, configure the following: <img width="4183" height="1017" alt="Image" src="https://github.com/user-attachments/assets/6969a3a9-3c75-4ae9-8795-4af3393e6537" /> <img width="1455" height="1025" alt="Image" src="https://github.com/user-attachments/assets/88633fdd-542e-43bd-81c3-734374db02ba" /> CREATE A NEW GROUP <img width="1585" height="1059" alt="Image" src="https://github.com/user-attachments/assets/3108b099-4bf6-481d-965b-10a4cad1ef12" /> ADD USER TO GROUP As the user: <img width="1608" height="1207" alt="Image" src="https://github.com/user-attachments/assets/0acd1879-4735-45d7-8ee6-9286f46ff917" /> <img width="1984" height="471" alt="Image" src="https://github.com/user-attachments/assets/466790f1-946f-4907-b33d-b52d34b0bb5e" /> CLICK USE ENTIRE DOCUMENT <img width="2081" height="772" alt="Image" src="https://github.com/user-attachments/assets/7588e197-f3de-4456-b636-5721b0b550a9" /> Left: User | Right: Admin <img width="5063" height="1728" alt="Image" src="https://github.com/user-attachments/assets/03faceb9-f9c7-4e2f-99b4-bbb6ef126e76" /> GO: <img width="5087" height="1302" alt="Image" src="https://github.com/user-attachments/assets/68acf662-b56a-4336-80af-b62d417d59b6" /> Note there is no citation on the left (user). If it's relevant (I dont think it is, as I've tried different models): <img width="1649" height="2444" alt="Image" src="https://github.com/user-attachments/assets/84618c59-1590-4662-a3a4-c8d9a150a5f3" /> Startup command for PIP: ```DATA_DIR=/home/openwebui/openwebui/data/open-webui MODEL_DOWNLOAD_DIR=/home/openwebui/openwebui/data/models LOG_LEVEL=info \ WEBUI_SECRET_KEY=whoop VECTOR_DB=pgvector PGVECTOR_DB_URL=postgresql://openwebui:whoop@127.0.0.1:5432/openwebui \ DATABASE_URL=postgresql://openwebui:whoop@127.0.0.1:5432/openwebui \ WEBUI_URL=https://chat.ready.art ENABLE_OAUTH_SIGNUP=false THREAD_POOL_SIZE=80 DATABASE_POOL_SIZE=0 \ AIOHTTP_CLIENT_TIMEOUT=120 AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST=15 AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST=15 \ ENABLE_USER_INFO_HEADERS=true CORS_ALLOW_ORIGIN=https://chat.ready.art OPENWEBUI_USE_COLORS=true ENABLE_RAG_WEB_SEARCH=true \ ENABLE_WEB_SEARCH=true ENABLE_SEARCH_QUERY_GENERATION=true RAG_WEB_SEARCH_RESULT_COUNT=3 RAG_WEB_SEARCH_CONCURRENT_REQUESTS=3 \ RAG_WEB_SEARCH_ENGINE=google_pse RAG_WEB_LOADER_ENGINE=safe_web ENABLE_WEBSOCKET_SUPPORT=true ENABLE_WEBSOCKET_SUPPORT=true \ WEBSOCKET_REDIS_RUL=redis://127.0.0.1:6379/0 REDIS_URL=redis://127.0.0.1:6379/0 CUSTOM_NAME=Ready.Art WEBUI_NAME=Ready.Art \ USER_AGENT="Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36" \ open-webui serve``` ### Logs & Screenshots . ### Additional Information _No response_
GiteaMirror added the bug label 2026-04-25 07:22:18 -05:00
Author
Owner

@frenzybiscuit commented on GitHub (Aug 13, 2025):

It looks like your issue matches Problem #5 – "Semantic ≠ Embedding" from our AI failure pattern map. This happens when a cosine similarity match doesn’t equal the true semantic meaning, so the retrieved chunk isn’t contextually correct even if the embedding vector is mathematically close.

We’ve helped over 80 developers resolve this by refining embedding strategies and implementing semantic alignment checks. If you’d like, I can share our tested approach and code snippet that improves retrieval accuracy for your RAG setup — just let me know.

You didnt read what I wrote at all did you?

<!-- gh-comment-id:3182015662 --> @frenzybiscuit commented on GitHub (Aug 13, 2025): > It looks like your issue matches Problem [#5](https://github.com/open-webui/open-webui/pull/5) – "Semantic ≠ Embedding" from our AI failure pattern map. This happens when a cosine similarity match doesn’t equal the true semantic meaning, so the retrieved chunk isn’t contextually correct even if the embedding vector is mathematically close. > > We’ve helped over 80 developers resolve this by refining embedding strategies and implementing semantic alignment checks. If you’d like, I can share our tested approach and code snippet that improves retrieval accuracy for your RAG setup — just let me know. You didnt read what I wrote at all did you?
Author
Owner

@Classic298 commented on GitHub (Dec 21, 2025):

End-users should be able to use full context mode on the knowledgebase, but cannot.

Cannot reproduce, works

<!-- gh-comment-id:3678752830 --> @Classic298 commented on GitHub (Dec 21, 2025): > End-users should be able to use full context mode on the knowledgebase, but cannot. Cannot reproduce, works
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#33449