[GH-ISSUE #4967] [FEAT]: Allow to manage vector databases #3117

Closed
opened 2026-02-22 18:32:41 -05:00 by yindo · 1 comment
Owner

Originally created by @fkdp06 on GitHub (Feb 5, 2026).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4967

What would you like to see?

AnythingLLM is very user friendly and easy to use. And I hope it stays like that in future.

A saw multiple propositions and ideas about customizing RAG functionalities or editing vector DBs. There are many existing RAG tools which can produce high quality vector databases. I don't think AnythingLLM should expand to such point, instead I wish it could allow user more flexibly of switching or managing existing vector databases.

The only piece I miss is to attach already existing vector databases globally or against particular workspace.

The way I can see that is (for example):

  1. With LanceDB allow to upload an existing .lance files.
  2. With PGVector is to point to the existing db name/table.

So it allows to prepare vector databases by any other existing tools and import/attach them to the AnythingLLM globally or per Workspace.

Please note, that I am not talking about having multiple vector db providers configured against AnythingLLM or its Workspaces, nor about importing eg. .lance files into PGVector. But rather about, once you configured your VectorDB globally, to allow to attach/manage existing databases globally or against workspaces.

For example. My AnythingLLM instance uses LanceDb. I have .lance file generated by some other RAG tool. How do I attach/include that file to the AnythingLLM so I can use it in my Workspaces?

Originally created by @fkdp06 on GitHub (Feb 5, 2026). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4967 ### What would you like to see? AnythingLLM is very user friendly and easy to use. And I hope it stays like that in future. A saw multiple propositions and ideas about customizing RAG functionalities or editing vector DBs. There are many existing RAG tools which can produce high quality vector databases. I don't think AnythingLLM should expand to such point, instead I wish it could allow user more flexibly of switching or managing existing vector databases. The only piece I miss is to attach already existing vector databases globally or against particular workspace. The way I can see that is (for example): 1. With LanceDB allow to upload an existing `.lance` files. 2. With PGVector is to point to the existing db name/table. So it allows to prepare vector databases by any other existing tools and import/attach them to the AnythingLLM globally or per Workspace. Please note, that I am not talking about having multiple vector db providers configured against AnythingLLM or its Workspaces, nor about importing eg. `.lance` files into PGVector. But rather about, once you configured your VectorDB globally, to allow to attach/manage existing databases globally or against workspaces. For example. My AnythingLLM instance uses LanceDb. I have `.lance` file generated by some other RAG tool. How do I attach/include that file to the AnythingLLM so I can use it in my Workspaces?
yindo added the enhancementfeature request labels 2026-02-22 18:32:41 -05:00
yindo closed this issue 2026-02-22 18:32:41 -05:00
Author
Owner

@timothycarambat commented on GitHub (Feb 5, 2026):

The main reason we dont allow foreign or externally created vector tables of any provider is because we rely on a specific schema format to show citations and inject data.

For example, we absolutely need a text field in metadata so that we can inject it to the context window correctly, if it is missing it will all blow up or just you wont get context.

In fact, there is a way to do this where you create a workspace, which produces a unique slug (found in Workspace Settings > Vector Database) and then have the foreign DB collection name the same slug. Then if the collection is named correctly and in the right spot (for lancedb this is in the storage/lancedb folder) it will actually work

@timothycarambat commented on GitHub (Feb 5, 2026): The main reason we dont allow foreign or externally created vector tables of any provider is because we rely on a specific schema format to show citations and inject data. For example, we **absolutely** need a `text` field in metadata so that we can inject it to the context window correctly, if it is missing it will all blow up or just you wont get context. In fact, there is a way to do this where you create a workspace, which produces a unique slug (found in Workspace Settings > Vector Database) and then have the foreign DB collection name the same slug. Then if the collection is named correctly and in the right spot (for lancedb this is in the storage/lancedb folder) it will actually work
yindo changed title from [FEAT]: Allow to manage vector databases to [GH-ISSUE #4967] [FEAT]: Allow to manage vector databases 2026-06-05 14:50:22 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#3117