openai/gpt-chat-latest → A (3:1)

Side A fixes a concrete UI correctness bug by changing rank gradient calculations from global ranking offsets to per-group indexing, so highlighting behaves correctly across multiple ranking groups, and adds targeted tests to lock in that behavior. Side B mixes several unrelated changes (route rename, removing a button, delegate plumbing, and fallback behavior), with much of the diff consisting of mechanical URL updates rather than a single substantive improvement.