AI
Retrieval is filtered by tenant, course, and role
Retrieval is filtered by tenant, course, and role before ranking. Student requests see only published allowed chunks; instructor-only chunks require an instructor context.
Retrieval receives tenant, user, role, and permitted course ids from authenticated server context.
Students can retrieve only published chunks explicitly allowed for the student role.
Instructor-only and draft chunks stay hidden from student retrieval, even when the text matches the query.
Cross-tenant chunks are excluded before scoring and never become citations.
Restricted source records are rejected during indexing and are not available for tutor retrieval.
A no-match query returns not_covered so the future tutor can avoid inventing answers.