dsh-think-ultra
官方 DeepSeek Harness 的推理层：所有请求保持原生最大努力模式，并在其之上构建隔离的 Flash/Vision/Pro 深度控制、稳定性轴和推理工具箱。仅提供构建版本；允许商业用途，禁止逆向/修改/提取。
Reasoning layer for the official DeepSeek Harness: every request stays on native max effort, with isolated Flash/Vision/Pro depth controls, stability axes and reasoning toolboxes built above it. Delivered build only; commercial use open, reverse/modify/extract closed.
怎么安装
dsh plugin add github:YUEYUEXYS/dsh-think-ultra 简介
dsh-think-ultra A reasoning layer for DeepSeek Harness (DSH). Current build is 0.10.0-experimental-preview. It runs only inside an official, unmodified DSH, and this repository ships the delivered build — the readable source is not published, and the licence does not grant reverse analysis, modification or extraction (see LICENSE). What "Ultra" actually is DeepSeek's API exposes four reasoning efforts: off, low, high, max. There is no fifth one. Send reasoning_effort="ultra" and the provider rejects the request outright (... does not support reasoning effort "ultra"). This plugin never makes t…
推荐参考
信息
- 协议
- NOASSERTION
- 语言
- –
- GitHub 星标
- 0
- 月下载
- –
- 最近更新
- 2026-08-31
- 创建于
- 2026-08-30
基础安全检查
- 检查结果
- curated 收录但无 npm 包/安装命令
- 收录来源
- curated:0xsline/awesome-deepseek-harness
- 主题标签
- ai, deepseek, deepseek-harness, dsh, dsh-plugin, harness-plugin, llm, max, plugin, reasoning, ultra
同类推荐
dsh-commandcode-provider
Mars-Sea/dsh-commandcode-provider
非官方 Command Code 模型接入插件：注册 `commandcode` 路由，带实时模型目录与推理强度支持。
claude-code-router
musistudio/claude-code-router
每个AI代理的本地控制平面：跨模型路由、融合新能力、编排工具，并始终保持完全控制。