feat: vectordb explorer #6052

Open
opened 2025-11-11 16:43:23 -06:00 by GiteaMirror · 5 comments
Owner

Originally created by @MattBash17 on GitHub (Aug 12, 2025).

Originally assigned to: @tjbck on GitHub.

Check Existing Issues

  • I have searched the existing issues and discussions.

Problem Description

Not a problem, just a suggested feature.

Desired Solution you'd like

Would be nice to have a feature where OpenWebUI is capable of reading existing collections from the selected vector database (configured from environment variables).
This is particularly useful for cases in which the vector db gets populated through another tool/workflow, but you later want to chat with its content through owui.

Alternatives Considered

No response

Additional Context

No response

Originally created by @MattBash17 on GitHub (Aug 12, 2025). Originally assigned to: @tjbck on GitHub. ### Check Existing Issues - [x] I have searched the existing issues and discussions. ### Problem Description Not a problem, just a suggested feature. ### Desired Solution you'd like Would be nice to have a feature where OpenWebUI is capable of reading existing collections from the selected vector database (configured from environment variables). This is particularly useful for cases in which the vector db gets populated through another tool/workflow, but you later want to chat with its content through owui. ### Alternatives Considered _No response_ ### Additional Context _No response_
Author
Owner

@tjbck commented on GitHub (Aug 12, 2025):

Great idea, will investigate!

@tjbck commented on GitHub (Aug 12, 2025): Great idea, will investigate!
Author
Owner

@jsweetzer-ea commented on GitHub (Aug 13, 2025):

Yeah, this is a must have for business use IMO.

We have existing vector stores (Pinecone, Elastic) that I'd prefer to integrate with vs re-create solely for OpenWebUI.

It adds a lot of overhead to have to do so, as then your pipeline to maintain the vector db has to hit multiple targets.

For Pinecone/Elastic, it would be helpful to connect to the instance and then provide a query which dictates how vectors are interacted with by OpenWebUI (filtering by namespace/index, etc).

@jsweetzer-ea commented on GitHub (Aug 13, 2025): Yeah, this is a must have for business use IMO. We have existing vector stores (Pinecone, Elastic) that I'd prefer to integrate with vs re-create solely for OpenWebUI. It adds a lot of overhead to have to do so, as then your pipeline to maintain the vector db has to hit multiple targets. For Pinecone/Elastic, it would be helpful to connect to the instance and then provide a query which dictates how vectors are interacted with by OpenWebUI (filtering by namespace/index, etc).
Author
Owner

@flefevre commented on GitHub (Aug 26, 2025):

Does your proposition matched with or is it different?

Feature Request: Support for External Read-Only Knowledge Bases (Vector DBs) #15282

@flefevre commented on GitHub (Aug 26, 2025): Does your proposition matched with or is it different? Feature Request: Support for External Read-Only Knowledge Bases (Vector DBs) #15282
Author
Owner

@tan-yong-sheng commented on GitHub (Aug 31, 2025):

Just a suggestion: it would be great if OpenWebUI could support contextual RAG in the future. On the UI, one possible feature is to display the summarized context (as one of the outputs of contextual RAG), similar to what we see in NotebookLM:

NotebookLM example:
Image

As a comparison, OpenWebUI's current implementation as below:

Image

More info about contextual rag: https://www.anthropic.com/news/contextual-retrieval


Although I believe contextual RAG could improve retrieval accuracy, it may also be more difficult to maintain and could slow down OpenWebUI, since it requires heavy operations and multiple LLM requests to generate and maintain summaries for each context.

@tan-yong-sheng commented on GitHub (Aug 31, 2025): Just a suggestion: it would be great if OpenWebUI could support contextual RAG in the future. On the UI, one possible feature is to display the summarized context (as one of the outputs of contextual RAG), similar to what we see in NotebookLM: NotebookLM example: ![Image](https://github.com/user-attachments/assets/b2e43e61-fd1e-4649-9f54-c68f6f62317d) As a comparison, OpenWebUI's current implementation as below: ![Image](https://github.com/user-attachments/assets/e530cc8a-c281-4e69-8fc7-bf82c7b8b142) More info about contextual rag: https://www.anthropic.com/news/contextual-retrieval ----- Although I believe contextual RAG could improve retrieval accuracy, it may also be more difficult to maintain and could slow down OpenWebUI, since it requires heavy operations and multiple LLM requests to generate and maintain summaries for each context.
Author
Owner

@Busboy3129 commented on GitHub (Nov 4, 2025):

ragflow seesm to have a very good vectordb ui

they also use https://github.com/infiniflow/ragflow/tree/main/deepdoc to do very good document parsing of pdf and doc.

@Busboy3129 commented on GitHub (Nov 4, 2025): ragflow seesm to have a very good vectordb ui <img src="https://raw.githubusercontent.com/infiniflow/ragflow-docs/refs/heads/image/image/chunking.gif"/> they also use https://github.com/infiniflow/ragflow/tree/main/deepdoc to do very good document parsing of pdf and doc.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#6052