npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

dsh-plugin-deep-scan

v0.1.0

Published

dsh plugin: iterative GitHub deep search — multi-query × 4-channel search (phrase / loose / README full-text / topic sweep), mine new query terms from hit topics, iterate rounds with a diminishing-returns stop rule, and rank by channel precision + stars.

Readme

dsh-plugin-deep-scan

🧩 dsh 插件家族(22 件):总目录 dsh-plugin-family | 明星插件:ide-hub 跨 IDE 统一管理 · task-forge 跨窗口无损交接 · quota 实时用量仪表

EN · Iterative GitHub deep search: four channels (exact phrase, broad match, README body, topic expansion), new keywords mined out of the previous round, and a diminishing-returns stop rule instead of a fixed page count. /deep-scan returns a ranked pool annotating which channel found each hit and which words matched. · Limits: the search API caps at 1,000 results per query; generic cross-industry words sink to the noisy channels.

DeepSeek Harness (dsh) 插件:GitHub 赛道深度扫描器。一条链路跑完"多词搜索 → 挖词 → 迭代 → 止损",自动把一个口语化的赛道需求变成几百个候选仓库的精准榜单。

适合回答:"这个赛道有哪些开源项目/竞品?我下一个产品该卡哪个位?哪个是漏网大鱼?"

同系列:price-aware · cost-ledger · relay-quota · transcript · transcript-search · obsidian-push · session-insights · eco-scan。

链路

词1|词2|词3
  └─ 第1轮:每词 × 四通道并行搜索
       短语(phrase)   "词" 精确短语匹配 —— 精准度最高
       泛搜(loose)    词 名字/简介/topic —— 按 star 排
       正文(readme)   词 in:readme —— 挖简介没写但 README 详述的
       赛道(topic)    topic:x —— 从精准命中挖出的标签横扫
  └─ 挖词:统计精准命中池的 topics 高频标签(滤掉 ai/llm 等噪声标签,
          排除已搜过的词,出现≥2次才入选)
  └─ 第2..N轮:用挖出的 topic:xxx 当新一轮搜索词,重复上述链路
  └─ 止损:某轮新发现占比 < newKeepRatio(默认25%),或挖不出新词 → 收
  └─ 排序:通道精准度优先(短语>泛搜>正文>赛道),同通道按 star
  └─ 深挖:Top N 仓库补 README 摘要,尾部附完整候选池(标注命中通道+词)

设计依据(AI 小说赛道实测):泛词必被巨库淹没(搜"AI 小说"出来的是 LangChain);精确短语一击命中真竞品;in:readme 捞回简介漏写的大鱼(AI_NovelGenerator 6.1k★ 靠"伏笔"命中);第 3 轮起边际收益明显下降——所以默认 mineRounds: 2(最多 3 轮)。

用法

  • /deep-scan 写小说|网文|AI novel:多词用 | 分隔(含全角),上限 6 个
  • deep_scan 工具(给 agent 用):{ queries: "写小说|网文|AI novel", language?, minStars? }

搜索词要具体:给 2-4 个同义词/中英文变体。词内可带 GitHub 限定符(pushed:>2025-06-01)。

配置

| 字段 | 默认 | 说明 | |---|---|---| | tokenEnv | [GITHUB_TOKEN, GH_TOKEN] | GitHub token(无 token 搜索限流 10 次/分,建议设置) | | perChannel | 30 | 每通道每词拉取条数 | | mineRounds | 2 | 首轮之后自动挖词迭代几轮 | | deep | 5 | 深挖(README 摘要)的仓库数 | | newKeepRatio | 0.25 | 一轮新发现占比低于此值 → 止损 |

已知边界

  • 挖词只走 topics 标签(确定性可测),简介里的功能词(如"去AI味/拆文")由 agent 看完报告后自行追加搜索——这是分工:插件负责可复现的链路,agent 负责语义跳转
  • 跨行业通用词(审稿/续写/工作流)会漏噪声进池,但噪声沉在泛搜/正文通道底部,不动摇深挖区
  • GitHub 搜索 API 上限 1000 条/查询,超大赛道建议加 minStars 收窄

安装

三步,实测于 @deepseek-ai/[email protected](需 pnpm 在 PATH 上):

# ① 装进 profile:dsh plugin 把参数原样转发给 pnpm,git 包会自动跑 prepare 构建 lib/
dsh plugin --profile web add github:121212165/dsh-plugin-deep-scan

② 把本仓库根目录 cordis.patch.yml 的内容并进 $DSH_HOME/profiles/web/cordis.patch.yml。 该文件默认是 [],所以要么整份替换,要么把 insert 条目并进同一个数组;不要直接追加—— 追加会形成两个 YAML 文档,启动即报 failed to parse overlay ... end of the stream or a document separator is expected(本机实测踩过)。

③ 重启 dsh。配置层与 client 半都要重启才生效(客户端按 boot 时算出的内容 rev 下发,硬刷新浏览器没用)。

自检挂载:dsh --profile web --dump-config | grep dsh-plugin-deep-scan,应看到该条目。