I have searched the existing issues and discussions.
Problem Description
Currently, enabling document Bypass Embedding and Retrieval will work well for most use cased where you need to work with files, but it seems like this also applies to knowledge resources.
For knowledge, the use case is usually different, and using Embedding and Retrieval is actually more useful, especially for interacting with documentation.
It would be very useful to have an option to disable/enable this for specific knowledge resources, to enable full control and better experience.
Desired Solution you'd like
When creating/editing knowledge resource, this feature can be enabled/disabled (per knowledge, not file), just like in the admin documents settings. The default values should be the one that is set in the admin settings.
When enabled, the files in the knowledge will be used just like any other files when this setting is globally enabled, and if disabled, it will use the vector DB and embedding model.
I assume that embedding the knowledge files would need to be triggered when this is disabled. When enabled, the chat should not use the embedded files but just attach the files.
Alternatives Considered
No response
Additional Context
Appreciate your great work on this project 🤝
Originally created by @roeizavida on GitHub (Jul 23, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/15964
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
### Problem Description
Currently, enabling document `Bypass Embedding and Retrieval` will work well for most use cased where you need to work with files, but it seems like this also applies to knowledge resources.
For knowledge, the use case is usually different, and using Embedding and Retrieval is actually more useful, especially for interacting with documentation.
It would be very useful to have an option to disable/enable this for specific knowledge resources, to enable full control and better experience.
### Desired Solution you'd like
When creating/editing knowledge resource, this feature can be enabled/disabled (per knowledge, not file), just like in the admin documents settings. The default values should be the one that is set in the admin settings.
When enabled, the files in the knowledge will be used just like any other files when this setting is globally enabled, and if disabled, it will use the vector DB and embedding model.
I assume that embedding the knowledge files would need to be triggered when this is disabled. When enabled, the chat should not use the embedded files but just attach the files.
### Alternatives Considered
_No response_
### Additional Context
Appreciate your great work on this project 🤝
@spammenotinoz commented on GitHub (Jul 27, 2025):
Indeed this would be a useful feature, or perhaps building on the "Bypass Embeddings" feature based on TOKEN\Document size.
eg: Document < n mb "Bypass Embeddings" otherwise use RAG.
<!-- gh-comment-id:3124728109 -->
@spammenotinoz commented on GitHub (Jul 27, 2025):
Indeed this would be a useful feature, or perhaps building on the "Bypass Embeddings" feature based on TOKEN\Document size.
eg: Document < n mb "Bypass Embeddings" otherwise use RAG.
Hi I am facing similar issue, I want to Bypass embedding for csv/xlsx files , but keep for all other types
i kept Allowed File Extensions as pdf,docx etc , but now when i try to upload a csv , it would say error uploading file
<!-- gh-comment-id:3409585517 -->
@parthraut45 commented on GitHub (Oct 16, 2025):
Hi I am facing similar issue, I want to Bypass embedding for csv/xlsx files , but keep for all other types
i kept Allowed File Extensions as pdf,docx etc , but now when i try to upload a csv , it would say error uploading file
When creating/editing knowledge resource, this feature can be enabled/disabled (per knowledge, not file), just like in the admin documents settings. The default values should be the one that is set in the admin settings.
When enabled, the files in the knowledge will be used just like any other files when this setting is globally enabled, and if disabled, it will use the vector DB and embedding model.
I assume that embedding the knowledge files would need to be triggered when this is disabled. When enabled, the chat should not use the embedded files but just attach the files.
If I understood your request correctly, then this is already in place.
Attach a knowledge collection to a chat, then click on the collection and enable full context mode.
Same behaviour when attaching a knowledge base to custom models in the workspace.
Attach knowledge base, click on it, enable full context mode.
<!-- gh-comment-id:3409685495 -->
@Classic298 commented on GitHub (Oct 16, 2025):
> When creating/editing knowledge resource, this feature can be enabled/disabled (per knowledge, not file), just like in the admin documents settings. The default values should be the one that is set in the admin settings.
> When enabled, the files in the knowledge will be used just like any other files when this setting is globally enabled, and if disabled, it will use the vector DB and embedding model.
> I assume that embedding the knowledge files would need to be triggered when this is disabled. When enabled, the chat should not use the embedded files but just attach the files.
If I understood your request correctly, then this is already in place.
Attach a knowledge collection to a chat, then click on the collection and enable full context mode.
Same behaviour when attaching a knowledge base to custom models in the workspace.
Attach knowledge base, click on it, enable full context mode.
@silentoplayz i am confused, i think this was solved more than a month ago already. Unless i am misunderstanding the original request. Can this be closed?
<!-- gh-comment-id:3589260086 -->
@Classic298 commented on GitHub (Nov 28, 2025):
@silentoplayz i am confused, i think this was solved more than a month ago already. Unless i am misunderstanding the original request. Can this be closed?
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 @roeizavida on GitHub (Jul 23, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/15964
Check Existing Issues
Problem Description
Currently, enabling document
Bypass Embedding and Retrievalwill work well for most use cased where you need to work with files, but it seems like this also applies to knowledge resources.For knowledge, the use case is usually different, and using Embedding and Retrieval is actually more useful, especially for interacting with documentation.
It would be very useful to have an option to disable/enable this for specific knowledge resources, to enable full control and better experience.
Desired Solution you'd like
When creating/editing knowledge resource, this feature can be enabled/disabled (per knowledge, not file), just like in the admin documents settings. The default values should be the one that is set in the admin settings.
When enabled, the files in the knowledge will be used just like any other files when this setting is globally enabled, and if disabled, it will use the vector DB and embedding model.
I assume that embedding the knowledge files would need to be triggered when this is disabled. When enabled, the chat should not use the embedded files but just attach the files.
Alternatives Considered
No response
Additional Context
Appreciate your great work on this project 🤝
@spammenotinoz commented on GitHub (Jul 27, 2025):
Indeed this would be a useful feature, or perhaps building on the "Bypass Embeddings" feature based on TOKEN\Document size.
eg: Document < n mb "Bypass Embeddings" otherwise use RAG.
@parthraut45 commented on GitHub (Oct 16, 2025):
Hi I am facing similar issue, I want to Bypass embedding for csv/xlsx files , but keep for all other types
i kept Allowed File Extensions as pdf,docx etc , but now when i try to upload a csv , it would say error uploading file
@Classic298 commented on GitHub (Oct 16, 2025):
If I understood your request correctly, then this is already in place.
Attach a knowledge collection to a chat, then click on the collection and enable full context mode.
Same behaviour when attaching a knowledge base to custom models in the workspace.
Attach knowledge base, click on it, enable full context mode.
@Classic298 commented on GitHub (Nov 28, 2025):
@silentoplayz i am confused, i think this was solved more than a month ago already. Unless i am misunderstanding the original request. Can this be closed?