I have searched the existing issues and discussions.
I am using the latest version of Open WebUI.
Installation Method
Docker
Open WebUI Version
v0.6.18
Ollama Version (if applicable)
0.9.6
Operating System
macOS Sequoia
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
Attaching a document using Models -> Knowledge -> 'Upload Files' should result in chats that can reference the attached files as if I had made a Knowledge base first and attached that collection to the Model instead (i.e. using Select Knowledge in the Model creation).
Actual Behavior
Model knowledge added using the direct 'Upload Files' system are completely inaccessible by any chats using that Model.
Using 'Upload Files':
Model setup:
Chat:
Using Knowledge base collection:
Model setup:
Chat:
Steps to Reproduce
Create a new Model and upload a document using 'Upload Files'.
Create chat using that model and ask a question about the file or information in it.
Observe chat says "Searching Knowledge for 'prompt'"
Observe that final response has used none of the attached knowledge.
Logs & Screenshots
Refer above for screenshots.
I can see in logs that when prompting using attached Knowledge base I can see: open-webui | 2025-07-24 18:29:19.303 | INFO | open_webui.retrieval.utils:query_collection_with_hybrid_search:352 - Starting hybrid search for 3 queries in 1 collections... - {}
When prompting using 'Upload Files' I see nothing in the logs indicating that a RAG search is actually taking place, despite the UI saying "Searching Knowledge for "can you tell me about the documents you have access to" there seems to be nothing actually happening.
Additional Information
No response
Originally created by @apotenza92 on GitHub (Jul 24, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/15977
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
- [x] I am using the latest version of Open WebUI.
### Installation Method
Docker
### Open WebUI Version
v0.6.18
### Ollama Version (if applicable)
0.9.6
### Operating System
macOS Sequoia
### 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
Attaching a document using Models -> Knowledge -> 'Upload Files' should result in chats that can reference the attached files as if I had made a Knowledge base first and attached that collection to the Model instead (i.e. using Select Knowledge in the Model creation).
### Actual Behavior
Model knowledge added using the direct 'Upload Files' system are completely inaccessible by any chats using that Model.
## Using 'Upload Files':
Model setup:
<img width="1139" height="1046" alt="Image" src="https://github.com/user-attachments/assets/a3047fc7-1949-4d7c-8fc3-4e203dc55b75" />
Chat:
<img width="1138" height="887" alt="Image" src="https://github.com/user-attachments/assets/6c14c4f4-a605-4d4c-9183-cb5c415f6e83" />
## Using Knowledge base collection:
Model setup:
<img width="1137" height="1043" alt="Image" src="https://github.com/user-attachments/assets/ea835e7f-b9c6-45a3-85a2-43e187832d84" />
Chat:
<img width="1147" height="723" alt="Image" src="https://github.com/user-attachments/assets/85fd510b-73d0-40c7-a071-5d498b1e406d" />
### Steps to Reproduce
1. Create a new Model and upload a document using 'Upload Files'.
2. Create chat using that model and ask a question about the file or information in it.
3. Observe chat says "Searching Knowledge for 'prompt'"
4. Observe that final response has used none of the attached knowledge.
### Logs & Screenshots
Refer above for screenshots.
I can see in logs that when prompting using attached Knowledge base I can see:
`open-webui | 2025-07-24 18:29:19.303 | INFO | open_webui.retrieval.utils:query_collection_with_hybrid_search:352 - Starting hybrid search for 3 queries in 1 collections... - {}`
When prompting using 'Upload Files' I see nothing in the logs indicating that a RAG search is actually taking place, despite the UI saying "Searching Knowledge for "can you tell me about the documents you have access to" there seems to be nothing actually happening.
### Additional Information
_No response_
GiteaMirror
added the bug label 2026-04-25 07:11:20 -05:00
I have tested it again using a completely clean setup with reset databases for all documents and vectors.
When I click "Upload Files" I see activity in the logs that the file is being processed successfully:
open-webui | 2025-07-24 18:39:18.424 | INFO | open_webui.routers.files:upload_file:95 - file.content_type: application/pdf - {}
tika | INFO [qtp402207819-135] 08:39:18,522 org.apache.tika.server.core.resource.TikaResource /tika (application/pdf)
open-webui | 2025-07-24 18:39:22.594 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:1137 - save_docs_to_vector_db: document 1170.2-2011(+A5).pdf file-499c5552-1d51-4cb4-9fd5-e89da9d9d5eb - {}
open-webui | 2025-07-24 18:39:22.599 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:1163 - Using token text splitter: cl100k_base - {}
open-webui | 2025-07-24 18:39:22.654 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:1266 - adding to collection file-499c5552-1d51-4cb4-9fd5-e89da9d9d5eb - {}
Batches: 100%|██████████| 3/3 [00:02<00:00, 1.19it/s]
However, when creating a chat and asking it anything using this Model I still see the same result as my original post - there's nothing in the logs indicating any RAG retrieval taking place and the chat knows nothing about the attached file.
Essentially the "Upload Files" functionality is broken at the moment.
<!-- gh-comment-id:3112588711 -->
@apotenza92 commented on GitHub (Jul 24, 2025):
I have tested it again using a completely clean setup with reset databases for all documents and vectors.
When I click "Upload Files" I see activity in the logs that the file is being processed successfully:
```
open-webui | 2025-07-24 18:39:18.424 | INFO | open_webui.routers.files:upload_file:95 - file.content_type: application/pdf - {}
tika | INFO [qtp402207819-135] 08:39:18,522 org.apache.tika.server.core.resource.TikaResource /tika (application/pdf)
open-webui | 2025-07-24 18:39:22.594 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:1137 - save_docs_to_vector_db: document 1170.2-2011(+A5).pdf file-499c5552-1d51-4cb4-9fd5-e89da9d9d5eb - {}
open-webui | 2025-07-24 18:39:22.599 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:1163 - Using token text splitter: cl100k_base - {}
open-webui | 2025-07-24 18:39:22.654 | INFO | open_webui.routers.retrieval:save_docs_to_vector_db:1266 - adding to collection file-499c5552-1d51-4cb4-9fd5-e89da9d9d5eb - {}
Batches: 100%|██████████| 3/3 [00:02<00:00, 1.19it/s]
```
However, when creating a chat and asking it anything using this Model I still see the same result as my original post - there's nothing in the logs indicating any RAG retrieval taking place and the chat knows nothing about the attached file.
Essentially the "Upload Files" functionality is broken at the moment.
It work properly, it's the way you ask:
As you ask, model understand documents as a file added or accesible to it but knowledge is added as part of the context. You have to reference it as context, this way you are indicating to the model that it have to search in context.
Feel free to close the issue.
.
<!-- gh-comment-id:3112650234 -->
@rgaricano commented on GitHub (Jul 24, 2025):
It work properly, it's the way you ask:
As you ask, model understand documents as a file added or accesible to it but knowledge is added as part of the context. You have to reference it as context, this way you are indicating to the model that it have to search in context.
Feel free to close the issue.
.
Maybe you misunderstood. I can see that your knowledge is attached to the Model as a 'collection' - this works properly. Instead try to attach knowledge using the 'Upload Files' button.
<!-- gh-comment-id:3112770870 -->
@apotenza92 commented on GitHub (Jul 24, 2025):
Maybe you misunderstood. I can see that your knowledge is attached to the Model as a 'collection' - this works properly. Instead try to attach knowledge using the 'Upload Files' button.
Apologies but this is not closed. You may have misunderstood the issue I've submitted. I have no issue using # to attach files in a regular chat window - the issue is specifically that when uploading files in a templated 'Model' that context is not used by any chats you make using that Model.
Instead of selecting "Select Knowledge", select "Upload Files" and add a file to a Model then try to speak to that model in a new chat without manually uploading any files or collections into that chat.
<!-- gh-comment-id:3113698628 -->
@apotenza92 commented on GitHub (Jul 24, 2025):
Apologies but this is not closed. You may have misunderstood the issue I've submitted. I have no issue using # to attach files in a regular chat window - the issue is specifically that when uploading files in a templated 'Model' that context is not used by any chats you make using that Model.
Instead of selecting "Select Knowledge", select "Upload Files" and add a file to a Model then try to speak to that model in a new chat without manually uploading any files or collections into that chat.
@spammenotinoz commented on GitHub (Jul 27, 2025):
Agree this is a current bug, you can attached all the files from the knowledge individually and it works fine, so that would discount a token limit issue, or document processing issue. However when referencing the "Collection", regardless of prompt, the LLM appears not to have any knowledge.
Last known working version was v0.6.15,
v0.6.16 and v0.6.18 both have this issue.
Re-produceable when model is created by admin, but the collection and chat is by an end-user.
Note: I was unable to re-produce as an Admin.
<!-- gh-comment-id:3124733660 -->
@spammenotinoz commented on GitHub (Jul 27, 2025):
Agree this is a current bug, you can attached all the files from the knowledge individually and it works fine, so that would discount a token limit issue, or document processing issue. However when referencing the "Collection", regardless of prompt, the LLM appears not to have any knowledge.
Last known working version was v0.6.15,
v0.6.16 and v0.6.18 both have this issue.
Re-produceable when model is created by admin, but the collection and chat is by an end-user.
Note: I was unable to re-produce as an Admin.
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 @apotenza92 on GitHub (Jul 24, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/15977
Check Existing Issues
Installation Method
Docker
Open WebUI Version
v0.6.18
Ollama Version (if applicable)
0.9.6
Operating System
macOS Sequoia
Browser (if applicable)
No response
Confirmation
README.md.Expected Behavior
Attaching a document using Models -> Knowledge -> 'Upload Files' should result in chats that can reference the attached files as if I had made a Knowledge base first and attached that collection to the Model instead (i.e. using Select Knowledge in the Model creation).
Actual Behavior
Model knowledge added using the direct 'Upload Files' system are completely inaccessible by any chats using that Model.
Using 'Upload Files':
Model setup:
Chat:
Using Knowledge base collection:
Model setup:
Chat:
Steps to Reproduce
Logs & Screenshots
Refer above for screenshots.
I can see in logs that when prompting using attached Knowledge base I can see:
open-webui | 2025-07-24 18:29:19.303 | INFO | open_webui.retrieval.utils:query_collection_with_hybrid_search:352 - Starting hybrid search for 3 queries in 1 collections... - {}When prompting using 'Upload Files' I see nothing in the logs indicating that a RAG search is actually taking place, despite the UI saying "Searching Knowledge for "can you tell me about the documents you have access to" there seems to be nothing actually happening.
Additional Information
No response
@apotenza92 commented on GitHub (Jul 24, 2025):
I have tested it again using a completely clean setup with reset databases for all documents and vectors.
When I click "Upload Files" I see activity in the logs that the file is being processed successfully:
However, when creating a chat and asking it anything using this Model I still see the same result as my original post - there's nothing in the logs indicating any RAG retrieval taking place and the chat knows nothing about the attached file.
Essentially the "Upload Files" functionality is broken at the moment.
@rgaricano commented on GitHub (Jul 24, 2025):
It work properly, it's the way you ask:
As you ask, model understand documents as a file added or accesible to it but knowledge is added as part of the context. You have to reference it as context, this way you are indicating to the model that it have to search in context.
Feel free to close the issue.
.
@apotenza92 commented on GitHub (Jul 24, 2025):
Maybe you misunderstood. I can see that your knowledge is attached to the Model as a 'collection' - this works properly. Instead try to attach knowledge using the 'Upload Files' button.
@rgaricano commented on GitHub (Jul 24, 2025):
attach knowledgfe with # or uploading files with +?
File added with + Upload file:
Collection added with #
@tjbck commented on GitHub (Jul 24, 2025):
https://docs.openwebui.com/troubleshooting/rag
@apotenza92 commented on GitHub (Jul 24, 2025):
Apologies but this is not closed. You may have misunderstood the issue I've submitted. I have no issue using # to attach files in a regular chat window - the issue is specifically that when uploading files in a templated 'Model' that context is not used by any chats you make using that Model.
Instead of selecting "Select Knowledge", select "Upload Files" and add a file to a Model then try to speak to that model in a new chat without manually uploading any files or collections into that chat.
@spammenotinoz commented on GitHub (Jul 27, 2025):
Agree this is a current bug, you can attached all the files from the knowledge individually and it works fine, so that would discount a token limit issue, or document processing issue. However when referencing the "Collection", regardless of prompt, the LLM appears not to have any knowledge.
Last known working version was v0.6.15,
v0.6.16 and v0.6.18 both have this issue.
Re-produceable when model is created by admin, but the collection and chat is by an end-user.
Note: I was unable to re-produce as an Admin.
@apotenza92 commented on GitHub (Jul 28, 2025):
Thanks, appreciate your agreement. Are you able to reopen the issue?
@spammenotinoz commented on GitHub (Aug 5, 2025):
@tjbck can this issue please be re-opened?
@tjbck commented on GitHub (Aug 6, 2025):
https://github.com/open-webui/open-webui/issues/16318#issuecomment-3158257257
@spammenotinoz commented on GitHub (Aug 7, 2025):
Confirmed as resolved for myself in the latest DEV