openai/gpt-chat-latest → B (4:3)

Side B fixes the ranking model to compare individual commits rather than short-circuiting on contributors, introduces per-commit ranking with contributor rollups, updates prompts/evidence, and adds tests covering same-contributor multi-commit cases. Side A adds a useful SSE-based entity fetch flow with streaming UI updates and worker completion signaling, but it is partly a refactor plus feature split, while Side B corrects a core ownership-allocation algorithm in a way that more directly affects project correctness.