v1.0.0
中文：DSH 升级后重打「模型调优」补丁的幂等脚本：DeepSeek 推理强度 7 档拓宽、workflow/ralph 子 agent 默认模型指向 deepseek-v4-flash。
English: An idempotent script to re-apply DSH model-tuning patches after an upgrade: widens DeepSeek reasoning to 7 levels and points workflow/ralph sub-agent default models at deepseek-v4-flash.