Managing user permissions with self-hosted chat UI

[bruno-oliveira]

Enforcing user-level policies via Chat UI

Currently, we are using a self-hosted version of chat UI, deployed in a cluster, using the official docker image.

We've set up OIDC which works amazingly well and, we also have set up some custom OpenAI-compliant endpoints to connect to some internal services.

My question is basically the title.

Suppose that we'd like to enforce something along the lines of: if or then else .

Essentially, what I am asking is: Could we, using exclusively tweaks to the .env file, make this type of mechanism work, where we'd make our data store for RAG be "user-aware"? Think kind of like a multitenancy style scenario or some sort of DB row-level filtering based on permissions from an authenticated user.

I know most of this work is custom one that needs to happen from our side, but I assume the OIDC setup would need to expose/allow Chat UI "requests" to a backend endpoint to use some login info? Is this possible?