[Obs AI Assistant] Knowledge base should create user scoped IDs

[miltonhultgren]

Summary

An entry in the knowledge base should be a piece of content that is bound to a user and a namespace meaning that two users (A and B) should be able to create an entry into the knowledge base in the same space and use the same ID without creating a conflict because the "primary key" for the entry should be username-space-id (to allow for easier wildcard queries).

Today we only use the id, meaning that if User A has created the entry "my-entry", when user B tries to create "my-entry" the system reports a conflict which has two issues:

1. We leak across the access control boundary that some entry with this ID already exists
2. User B has to modify their ID to avoid a conflict that shouldn't be visible to the user at all

If User B follows up their create request with a "get all entries" request to check what they have, they will get back a list that doesn't include "my-entry" since it doesn't belong to them.

Possible solutions

One approach that could work is to append the user name and space name to each ID behind the scenes, so that what is stored has the "full" ID schema, but what we return from the API to the user only has the base ID part.
This means that all APIs and the Knowledge base service have to consistently do this packing/unpacking.

Related

Beta Give feedback

1. [Obs AI Assistant] Knowledge base entries with the same name overwrites each other without warning #184069
   8.15 candidate Team:Obs AI Assistant bug +3
   

[elasticmachine]

Pinging @elastic/obs-knowledge-team (Team:obs-knowledge)

[miltonhultgren]

Duplicate of #184069