Best Practices and Support for Multi-Tenancy / Data Isolation

[harshitraj2409]

I'm exploring the possibility of using lightRAG to power a multi-tenant application (e.g., a SaaS platform) where different users or organizations would index their own private documents. A critical requirement for this use case is strict data isolation—ensuring that one tenant's data (indexed documents, extracted knowledge graph entities/relationships, and vector embeddings) is never accessible or returned in queries made by another tenant.

Any guidance, insights, code pointers, or discussion on the best way to architect this within the lightRAG ecosystem would be greatly appreciated!

[ArindamRoy23]

If separation is paramount, different workspaces for clients is the only way at the moment.

Actually , you can achieve everything that you are looking for in the current soln apart from graph item seperation. If you inspect the nodes and edges data, you will find different items add to the node, separated by . The IDs col has the relevant ID that added this. If you can figure out logic for this, it would be your required soln in one workspace,

[eddyjm09]

Hi guys, have you observe any news on this topic? I'm also exploring the option of using LightRAG as multitenant application where data isolation is a must.
@harshitraj2409 @ArindamRoy23

[BireleyX]

this is the same topic covered at #1016

[BireleyX]

there's already an ongoing discussion there

[eddyjm09]

thanks!