Yuer6327/NoLetMe 预览 preview

Yuer6327/NoLetMe

Deepseek v4 神鬼二象性观测面板

Project Overview项目介绍

NoLetMe is a native web plugin built exclusively for DeepSeek Harness (DSH). It adds a floating real-time reasoning keyword statistics panel that mounts to the right edge of the DSH conversation page. The plugin only counts feature keywords that appear within the model’s reasoning blocks, and categorizes the current reasoning style into three buckets based on empirically validated lexical rules: efficient direct action, hesitant probing, and neutral task recap. All classification rules are sourced from public research on DeepSeek V4 Pro model training overfitting, so every category has concrete empirical support from real trajectory data.

When a model streams output during a DSH conversation, NoLetMe updates the keyword counts incrementally with each new reasoning chunk, so you see updated style stats in real time without re-scanning the entire conversation. The plugin is designed for developers and power users who use DSH to debug LLM behavior or work on development tasks. It helps you quickly spot shifts in reasoning style that may indicate problems with the model’s output or approach, and all processing happens locally in your browser. To use it after installation, you just open any DSH conversation and the panel will automatically load the cached stats for that session.

NoLetMe is distributed under the open source MIT license. To build from source, you need to have pnpm installed, run pnpm install to get all dev dependencies, then pnpm build to generate the output bundles compatible with multiple DSH versions ranging from 0.1.0-rc.7 to recent 0.1.x releases. All per-session statistics are persisted to your browser’s localStorage, so stats are retained when you close and re-open DSH, and no data is sent to external servers. The plugin does not modify any core DSH UI or code, and integrates via DSH’s official overlay slot system, so it works reliably across DSH updates without modifications.

NoLetMe 是专为 DeepSeek Harness(DSH)网页端开发的原生插件,会在会话页面右侧边缘挂载一块实时推理关键词统计面板。它只统计模型推理块内的特征关键词,基于公开研究得到的词频特征将推理风格划分为高效直接、犹豫试探、中性复述三类,帮助开发者快速诊断当前会话的推理风格。

该插件基于DeepSeek V4 Pro训练过拟合事件的公开调研数据构建分类规则,所有词法风格划分都有实测轨迹数据支撑,不添加无依据的主观分类逻辑。它面向使用DSH进行开发调试的用户,帮助用户观测模型推理风格的变化,辅助排查开发任务中遇到的推理异常问题。

插件通过DSH官方的shell.overlay插槽挂载,完全复用DSH的设计系统,不会修改或补丁任何原有UI。所有统计都在浏览器本地完成,统计数据存储在本地localStorage,不会向外传输。它遵循MIT许可证开源,开发构建使用pnpm,产物兼容多个DSH预发布和正式版本。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意
  • Only 8 stars - very few users, little community feedback星标只有 8,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile demo add dsh-noletme

把 Yuer6327/NoLetMe 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

NoLetMe · dsh 推理轨迹面板

npm version dsh-std Community v0.15 Awesome dsh-plugin

NoLetMe 是 DeepSeek Harness(dsh)网页端插件,在会话页右侧边缘挂载一块实时推理关键词统计面板。

模型流式输出时,NoLetMe 只统计其**推理块(reasoning blocks)**中出现的特征词,据此反映当前推理风格:

分类 关键词 依据
🟢 高效 · 直接行动 We need… Let's… We should… We can… We will…;首行 Good./Great./Excellent. minimal 类高分轨迹
🟠 犹豫 · 第一人称试探 Let me… I think… I'm not sure… I wonder… I guess… maybe perhaps standard 类低分轨迹
⚪ 中性 · 复述任务 The user wants… The user asked… this task… the request… Standard 目录开场框架

灰测是另一层、不改 0813 词表。面板只报命中态与数字;完整判定规则见 灰测如何判定。

面板是原生融入的悬浮层:复用 harness 设计系统(--dsw-alias-* 语义令牌、DetailsPanel 头/体结构、CSS Modules、明暗与 reduced-motion),通过官方 shell.overlay 插槽挂载,不改动、不补丁任何既有 UI。

证据链

本插件的分类体系并非杜撰,每一条都追溯至公开的 xiaobright/modeltest 仓库,及其对 DeepSeek V4 Pro GA 0813 后训练过拟合事件的调研:在 DeepSeek Harness「Minimal」预设(RL 训练所用的双工具脚手架)上训练出的 checkpoint,换到更宽的 Standard 工具目录后能力崩塌。

测试集:Project2 V4.1b —— 一个真实损坏的 ESP-IDF 嵌入式工程任务,已正式冻结(PROJECT_FROZEN.md,2026-07-23 冻结;评分规则与隐藏测试于 2026-07-19 做 SHA-256 固定)。

实测数据:evaluator/trajectory_evidence/derived/trajectory_stats.json(每次运行做 SHA-256 固定;只统计已完成的助手推理块,排除流式分块):

运行(模型 / 配置) 得分 we let me let's I 可见回复
V4 Pro / Minimal WSL 99 272 0 101 17 1
V4 Pro / Minimal WSL 96 231 0 117 18 1
V4 Pro / anchored-standard Win 98 179 1 88 17 1
V4 Pro / anchored-standard Win 99 165 0 98 18 1
V4 Pro / Standard WSL 91 11 208 2 137 55
V4 Pro / PTC WSL 92 16 194 0 237 33

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev dsh-codex-pet 下一个 Next dsh-self-evolve →