openai/gpt-chat-latest → B (5:1)

Side B introduces a substantial architectural improvement by moving entity fetching to an SSE-based workflow, adding a dedicated fetch module, asynchronous streaming, completion/error signaling via oneshot channels, client-side SSE handling, and integration tests for the new behavior. Side A is a focused UI correctness fix that changes unranked siblings from a single group into one group per item and adds a regression test, but its scope and long-term impact are much smaller than the new fetch infrastructure.