openai/gpt-chat-latest → B (3:2)

Side B fixes a correctness issue by enforcing meaningful vote ratio bounds (both sides must be at least 1 and at most 100) consistently in the DSL parser, UI handler, and reducer, preventing zero-weight or extreme votes from creating invalid graph behavior. It also updates affected tests and adds parser, integration, and UI regression tests, whereas Side A primarily adds a useful developer-facing offline tool without changing core runtime correctness.