# The platform open-sourced it, and on day two someone filled the gap it left: how dsh-context got over four thousand stars among dozens of similar plugins · GitHub open-source projects

> Source: https://stringers.io/en/github/gh_bowenliang123__dsh-context.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> The official side only offered a context-occupancy percentage. The day after the platform opened source, dsh-context filled in the detail with that name, and it's more thorough than any alternative. After DSH lead Cui Tianyi recommended it, the author put the endorsement in the README, Weibo influencers and Zhidx followed with coverage, and it gained 684 stars in one day in October 22.

[GitHub open-source project](https://stringers.io/en/github/index.md) · written 11-6

bowenliang123/dsh-context · bowenliang123

# The platform open-sourced it, and on day two someone filled the gap it left: how dsh-context got over four thousand stars among dozens of similar plugins

**What it is** A free open-source plugin for DeepSeek Harness that breaks down the agent's context window: what's in it, how it grows after each request, when it gets compressed, and what each agent is doing.

**Why it took off** The official release only showed a context-usage percentage. On the second day after the platform went open source, dsh-context filled in the detail, and did it more thoroughly than any comparable tool. After DSH lead Cui Tianyi recommended it, the author put the recommendation in the README, and big Weibo accounts and Zhidx covered it. Stars rose by 684 in one day in October 22.

**B** **How you could follow this** Don't build another context plugin for dsh; that spot is already taken. Instead, look for other features in dsh where the official release gives only pointers and no detail, or enter the first week of any new agent platform that opens a plugin marketplace. Use names users actually search for, read the official data sources, ship releases frequently for several months, and aim to get named by the platform itself.

Grade B: the play is clear and copyable without any personal following (the author has only 298 followers), but this slot is already taken, and the first-week window of the dsh plugin ecosystem has passed. Following it means finding a new empty slot yourself or waiting for the next platform, and being ready to ship nearly every day for months. It's spreading in small circles

Data (updated daily) · Repo created 2026-08-14 · 4,546 stars · 7 days +610

GitHub · stars

Data 2026-09-30 → 2026-11-08 (40 days, 32 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

The day after DeepSeek Harness (dsh) was open-sourced, a developer from Guangzhou published the package `dsh-context` to npm, doing the half of the job the official project left undone: letting you see exactly what's packed into the agent's context window. Since then it has shipped a new version almost every day for nearly three months. At the end of September, DSH lead Cui Tianyi named it on X as a recommendation, and at the end of October it gained 684 stars in a single day. Dozens of plugins do the same thing, and the largest of them has only 640 stars.

### A tweet with just one sentence

On the evening of Sep 30, 2026 (Beijing time), DeepSeek product lead and DSH team head Cui Tianyi posted on X a single line: "Recommending a feature-rich DSH context management plugin today: dsh-context," followed by a GitHub link. The tweet later reached 64.6k views and 339 likes. The same day, another account reposted it as "Cui's personal endorsement of plugins, round four," meaning this was already the fourth third-party plugin he had recommended, making the recommendation a recurring segment.

Five days earlier, Cui Tianyi had also posted that, according to official API statistics, about 60% of dsh users had installed at least one third-party plugin, and the team would "keep supporting the flourishing of the third-party plugin ecosystem."

The most-liked reply under the recommendation tweet read: "Why not build an app store? How many people can see your posts on X?" It points to a fact: the tweet itself brought few stars (9/30 to 10/9 rose by only about 280). Its real value came afterward. The author put it into the README's second line: "Recommended and starred by DeepSeek Harness team members @tianyicui and @turtle1999." Every outlet that covered it later repeated that line.

### What it is

dsh is an agent framework open-sourced by DeepSeek in Aug 13, with the slogan "everything is a plugin." It passed 70k stars in under a day and has 274k now. dsh includes a small ring showing "context usage percentage," so users can see the window is nearly full but not what is filling it.

dsh-context breaks that ring apart:

- **Context tab** : the current window stacked in colors across six content types (system prompt, tool definitions, tool results, messages, etc.), showing how context grows after each request, with compaction and pruning events marked directly on the chart. You can also open any request to see the exact raw content sent to the model at that moment.

- **Context insights page** : token use and cost over 7 days across sessions, cache hit rate, a 8-week activity heatmap, and usage stats for skills and tools.

- **Fleet tab** : when multiple agents run together, which is the parent, which are children, and what each is doing.

- A right-sidebar panel and a `/context` slash command (the naming is clearly borrowed from Claude Code).

Installation is done by searching `dsh-context` in dsh's "Add plugin" wizard and clicking install, with no build step and no restart. It reads dsh's official metering data, so its numbers match the official ring, which matters later. The README and UI come in 8 languages: Chinese, English, Japanese, Spanish, French, German, Russian, and Portuguese. It is open source under Apache-2.0 and free.

The author, Liang Bowen (GitHub bowenliang123), is a PMC member of Apache Kyuubi and a contributor to Dify. He previously wrote several Dify plugins and a Grafana dashboard with 32 stars. He understands plugin ecosystems and observability but has no personal audience: 298 followers on GitHub.

### How the stars stacked up, layer by layer

| Date | Cumulative stars / event |
|---|---|
| 8-13 | dsh open-sourced |
| 8-14 | Repo created, npm package published the same day |
| 8-15 | Added to the deepseek.club plugin directory |
| 8-21 | Yu Pi's "16 hot DeepSeek Harness plugins" includes it |
| 9-05 | An overseas individual account posts a recommendation thread, 5,225 views |
| 9-30 | Cui Tianyi recommends it; about 1,644 stars that day |
| 10-09 | 1,925, then gaining 20–50 per day |
| 10-15/16 | Zhihu "There's a recently viral DSH plugin, dsh-context" and WeChat official account "AI Coding Guide" |
| 10-21 | At 2 pm, 46 Weibo account AiKeKe-LifeLove recommends it; at 5 pm, 42 Zhidx publishes "12 must-have DeepSeek Harness tools, including Cui Tianyi's picks," with it ranked #2 |
| 10-22 | 3,085 (+684, single-day high) |
| 10-23 | 3,453 (+368), #12 on the TypeScript daily trending chart |
| 11-02 | The author posts a 4-minute walkthrough on Bilibili, 53.9k views |
| 11-06 | 4,245 (+102 that day), 180 forks |

(GitHub daily star records; events from original posts on each platform, to be verified by 11-06)

The jump in Oct 22 stands out: in the same week, the dsh main repo and several other popular plugins grew steadily, and only this one spiked, so it was its own exposure, not a rising ecosystem. How that day's volume splits between Weibo and Zhidx can't be separated. After the Bilibili video went up, daily gains went from thirty-some to sixty-some or around a hundred. Whether the video caused that is an inference.

On usage, npm downloads over the last 30 days (10/7–11/5) totaled 380k, with 2–28k per day in the last week of October. The number is inflated by frequent releases and auto-updates (84 days, 159 versions), so it's only useful for comparison with peers: over the same period, the cost-tracking plugin dsh-cost-meter had 135k.

### Same idea, dozens of people building it

"See how many tokens dsh uses, how much it costs, and what's in the context" was one of the most popular plugin directions in dsh's first week:

| Add-ons | Launched | stars (approx., 11-06) | How it differs from dsh-context |
|---|---|---|---|
| dsh's built-in ring and usage popup | 8-13 | All users | Only a percentage and a few rough categories |
| dsh-cost-meter | 8-13 | 640 | Tracks money: cost, budget, balance, without breaking down context |
| dsh-usage-stats and dozens of other usage plugins | From ¥8-14 | Largest at 188, most single digits | A heatmap plus balance, all looking alike |
| dsh-context-doctor | 8-12 | 67 | Counts only how many tokens "static injections" take up: AGENTS.md, skills, tool definitions |
| dsh-context-control / -mode, etc. | 8–October | All under 30 | Actively compresses and prunes context; dsh-context is about "seeing" |
| Later entrants like dsh-context-explorer and dsh-command-context | October | All under 15 | Same direction, covering only one facet |

A pile of forks even copy the descriptions, including the author of dsh-cost-meter, who forked one too.

The most notable is dsh-context-doctor: it launched two days earlier and was also careful work, with 67 stars. A two-day gap and a feature gap, but no one recommended it, and the result was sixty times smaller.

One more note: dsh-context is not dsh's biggest plugin. In the Zhidx 10/21 article, the sidebar enhancement plugin DSH Better Sidebar has about 4,400 stars and was also recommended by Cui Tianyi. dsh-context won the "context" slot.

### Why it made the charts

**The need is real, and the official project did only half the job.** Running long tasks or multiple agents in dsh consumes a lot of context, and users hit "why did it suddenly get dumber" and "why did tokens spike." The official ring tells you it's nearly full but not why. dsh-context sits exactly in that gap, and because it uses the official metering, its numbers match the ring. Users trust it, and the official side is willing to promote it.

**The name is the search term, and it arrived the next day.** A user searching "context" in the Add plugin wizard finds the most literal package name is theirs. How the wizard ranks results or whether it has featured slots is not visible from outside, but this is probably its largest sustained source of traffic (inference).

**Exposure in layers, each citing the one before.** Early plugin roundups → official lead's recommendation → recommendation written into the README → media copying it into headlines → the author makes a video. No paid acquisition or sock-puppet starring was found.

**Built deep, and kept building.** Nearly one release a day, growing from a single `/context` command into cross-session insights, a multi-agent view, and 8 languages, with UI using dsh's native highlighting, sidebar, and preview bar so it looks like an official feature. As dsh itself iterated from 0.2.0 to 0.2.2 and its API changed, the author also built a GitHub Action to check plugin compatibility so the plugin wouldn't break. Most similar plugins cover one facet and stop there.

### If it were you

**One judgment:** "In the first days of a major platform opening its plugin ecosystem, use the most literal name to fill a native feature the official project only half built, then keep building deeper and aim for an official shout-out" is a real play and can be copied. But dsh's "context" slot is already taken, so building another dsh context plugin isn't worth it. What is worth it is finding the next empty slot in dsh, or moving on the first week of the next plugin-open platform.

**Who it fits:** people who can ship almost daily for several months; people who know the internals of an agent framework well enough to read its metering and event data; people who have already built things in plugin ecosystems like Dify, VS Code, or Obsidian. No personal audience is needed. Liang Bowen has only 298 followers.

**Who it doesn't fit:** people who want a weekend project they can ship and walk away from. dsh-context-doctor launched two days earlier and got only 67 stars, and dozens of usage plugins mostly stalled in single digits. It also doesn't suit anyone planning a "better context panel" next to dsh-context: it has official backing, dozens of articles, and 4,000 stars, and late entrants since October have all stayed under 15.

**How to do it:**

- Look for places in dsh where the official project "gave a pointer but not the details." dsh-context fills the context gap; the sidebar is the other one Cui recommended. Check dsh's issue tracker and the replies under Cui's tweets, where people ask for features directly (for example, someone under his tweet asked whether there's a plugin like VS Code remote workspace).

- Read the official data sources instead of estimating yourself. Matching the native numbers is the prerequisite for official endorsement.

- Mirror the host's look so it feels like it should have been an official feature from the start.

- Port the same play to another platform: Codex, Claude Code, or any new open plugin marketplace for agent frameworks. If the official side only offers a rough context indicator, the "context observatory" slot may still be empty. Get in within a week of the platform launch.

**How to get users** : name the package after the word users actually search for; in week one, find the authors of "must-install N plugins" lists; put any likes or stars from official figures in the second line of the README; the README in 8 languages gives Japanese and Russian aggregator sites material to repost; once it's solid, make a few-minute walkthrough video on Bilibili yourself.

**When to move** : the first week the platform opens plugins. dsh is already three months old, so entering dsh now means finding a niche with no leader yet, not trying to beat existing ones.

**What would mean you read this wrong** : ① dsh's official team builds context breakdown into the native UI. Then every plugin that "fills the gap the official side leaves" gets taken back, and the play has to shift to things the official side won't build. ② In the first month of the next open plugin ecosystem, the top plugins are all from official teams or big companies, and no generic-name plugin from an individual developer passes 1,000 stars. That would mean this window only existed once, for dsh. ③ dsh launches an official plugin store that sorts by downloads or editor picks. Then the value of "grabbing the name" drops and the value of "getting picked for featuring" rises.

### Data and sources

GitHub repo page, README, and daily star history (as of 11-06); npm download API (10/7–11/5); X: @tianyi 9-25, 9-30 posts and replies, @NFT_Chen, @ItsmeAjayKV, @yupi996; Zhidx article "DeepSeek Harness: 12 must-have tools, plus Cui Tianyi's picks" (10-21); Weibo account 爱可可-爱生活 (10-21); Zhihu column and the AI Coding 指北 WeChat account (10-15/16); Bilibili BV1bn4K6TEEV (published 11-02, checked 11-06); deepseek.club plugin directory. Star counts for similar plugins are from GitHub as of around 11-06; figures marked "about" are rounded.

[GitHub repo ↗](https://github.com/bowenliang123/dsh-context)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| dsh's built-in context ring / per-turn usage popup | Big-company features | 2026-08-13 | all dsh users (2.21M monthly npm downloads) | The official side only shows an occupancy percentage and rough categories. dsh-context reads the same official metering and extends it into per-request trends, raw-content browsing, events, and multi-agent views. |
| Han-1413141/dsh-cost-meter | Open source | 2026-08-13 | 640 stars, 135k npm downloads per month | Covers cost, budget, and balance, but doesn't break down context composition |
| Ychris12138/dsh-usage-stats and dozens of other usage/token stats plugins | Open source | From ¥2026-08-14 | biggest has 188 stars, most have single digits | Only usage heatmaps and balances. Thin features, all alike. |
| Zhenyu98/dsh-context-doctor | Open source | 2026-08-12 | 67 stars | Audits only the token cost of static injection, not how context evolves at runtime |
| dsh-context-control / dsh-context-mode / compression-selector, etc. | Open source | 2026-08~10 | none above 30 stars | They act on context: compressing and pruning it. dsh-context does the "seeing". The two complement each other. |
| Claude Code /context, ccusage | Other | 2025 | Big | Belong to a different ecosystem. dsh-context borrowed the /context name. |

Where users come from

- **Verified** **Platform event + early name claim** : DSH 8/13 went open source, the repo and npm package dsh-context were published 8/14, and 8/21 it was included in Yupi's "16 hot DSH plugins". You can install it by searching the package name in the Add plugin wizard.

- **Verified** **Official figure endorsement (X)** : Cui Tianyi recommended it 9/30 as "a feature-rich DSH context management plugin", with 64.6k views. @NFT_Chen posted "Cui's personal endorsement, fourth round" the same day. The README's second line lists the official team member's recommendation and star.

- **Verified** **Chinese media / Weibo KOLs** : 10/21 14:46 爱可可-爱生活 posted on Weibo, 17:42 Zhidx's "12 must-have tools" ranked it #2. From 10/22 to 23 it gained +1,052 stars, while daily gains for other repos in the ecosystem stayed flat over the same period.

- **Verified** **Zhihu / WeChat Official Accounts / tech community roundups** : Zhihu's "the viral DSH plugin dsh-context" (10/15), AI Coding 指北 (10/16), Alibaba Cloud developer community, 什么值得买, Tencent Cloud, and other plugin roundups

- **Verified** **Author-made Bilibili video** : 11/2 posted a 4-minute walkthrough, 53k views, 4,727 favorites. Afterward daily gains went from 30–40 back to 60–102 (attributing the gains to the video is an inference).

- **Verified** **overseas X / multilingual SEO / plugin directory sites** : @ItsmeAjayKV 9/5 post (5,225 views); README in 8 languages; listed on a dozen-plus directory sites including dshfind, deepseekplugins, and dsh-plugin.org; reposted on Japanese and Russian sites

- Inferred **Search / recommendation slots in dsh's built-in plugin wizard** : the README pushes searching for dsh-context in the wizard to install it. Ranking and placement weren't verified.

## Why dsh-context (bowenliang123/dsh-context) took off

> Channel: GitHub | Research date: 2026-11-06 | Data card: repo created 2026-08-14,84 days ago, 4245 stars / 180 forks (GitHub API live check that day: 4374 stars)

### Bottom line

Within a day of DeepSeek Harness (dsh, DeepSeek's official open-source agent framework, released 2026-08-13, slogan "everything is a plugin") launching, its author claimed the "context visualization" slot and got the most straightforward package name, `dsh-context`. Over the following nearly three months it shipped almost daily (159 versions on npm), steadily deepening the features and expanding dsh's built-in "context occupancy ring" into a full observatory. Its exposure stacked in layers: early plugin roundups → DSH lead Cui Tianyi 9/30 publicly recommending it on X → 10/21 爱可可's Weibo post plus a Zhidx write-up (this drove +684 stars on 10/22) → 11/2 the author's own Bilibili video (53k views). Similar plugins number in the dozens, and the largest has only 640 stars. It is the kind of winner that combines an early-ecosystem claim, official endorsement, and a clearly deeper product.

### 1. What it is

A third-party plugin for dsh that opens up the agent's context window for inspection, with five entry points:

- A cross-session "Context Insights" page: 7-day tokens / cost, cache hit rate, 8-week activity heatmap, and call stats for skills / tools / commands;

- A per-session Context tab: the current window stacked by color across six content types, a per-request trend (Total and Delta views, with compression and pruning events marked on the chart), a Context Browser (expand to the raw content of any request, including system prompts, tool schemas, and tool results), a context events log, file activity, and command execution;

- A Fleet tab: a family tree of parent and child agents, cross-agent swimlanes, and tasks;

- A right-sidebar panel, plus the `/context` slash command;

- Preferences, including an optional system prompt prefix / suffix.

Install: search for `dsh-context` in the Add plugin wizard on the dsh desktop or web app and click install. You can also run `dsh plugin --profile web add dsh-context`. No build step, no restart. The README is in 8 languages: Chinese, English, Japanese, Spanish, French, German, Russian, and Portuguese. The UI uses the same 8 languages (source: GitHub README).

The author, Liang Bowen (GitHub bowenliang123, Guangzhou), is a PMC member of Apache Kyuubi and a Dify contributor. He previously wrote dify-grafana-dashboard (32 stars) and several Dify plugins. He knows the plugin ecosystem and observability well but has no personal audience (298 GitHub followers).

### 2. Growth data and timeline

| Timeline | Stars / event | Source |
|---|---|---|
| 2026-08-13 | DeepSeek Harness open-sourced, passed 70k stars in under a day, now at 274k | Tencent Cloud developer article; GitHub API |
| 2026-08-14 | dsh-context repo created, npm package published the same day | GitHub;npm registry(created 2026-08-14T09:26) |
| 2026-08-15 | Listed in the deepseek.club plugin directory (v1.0.1) | deepseek.club/plugins/dsh-context |
| 2026-08-21 | Included in Yupi's "16 viral DeepSeek Harness plugins" (X long post, reposted to Tencent Cloud 8/31) | x.com/yupi996/article/2090649288535921094 |
| 2026-09-05 | Overseas individual @ItsmeAjayKV posted a recommendation thread, 5,225 views | x.com/ItsmeAjayKV/status/2096173850136559639 |
| 2026-09-29~10-01 | dsh v0.2 desktop released with a built-in plugin install wizard; DeepSeek's official account promoted it on X | Sina / IT Home; x.com/deepseek_ai/status/2105915715241062644 |
| 2026-09-25 | Cui Tianyi: about 60% of dsh users have installed at least one third-party plugin | x.com/tianyi/status/2103534313463783831 |
| 2026-09-30 | Cui Tianyi (DeepSeek product team, DSH team lead) recommended on X: "Today I recommend a feature-rich DSH context management plugin, dsh-context", 64.6k views, 339 likes. The same day @NFT_Chen reposted it as "Cui's personal endorsement of plugin no. 4". | x.com/tianyi/status/2105256532657267119;x.com/NFT_Chen/status/2105258670850494961 |
| 2026-09-30 | 1,644 stars | Our GitHub repo (Wayback Machine) |
| 2026-10-09 | 1,925 stars, then gained 20–50 per day | github repo_day |
| 2026-10-15/16 | Zhihu: "Recently the viral DSH plugin dsh-context"; WeChat Official Account "AI Coding 指北" also wrote about it | zhuanlan.zhihu.com/p/2094125855421248833;ai-coding-guide.com/articles/dsh-context |
| 2026-10-21 14:46 | 爱可可-爱生活 (Weibo uid 1402400261) posted a recommendation | sina.cn/news/detail/5356469010042330.html |
| 2026-10-21 17:42 | Zhidx "12 must-have DeepSeek Harness tools, plus Cui Tianyi's picks": dsh-context ranked #2 (about 2,700 stars at the time) | zhidx.com/p/602940.html |
| 2026-10-22 / 23 | +684 / +368 stars (single-day peaks); 10/23 hit #12 on GitHub's TypeScript daily trending (+296 that day) | GitHub repo_day, trending |
| 2026-10-21~22 | Japanese review sites, the Russian DevTrends, AITOP and other aggregators reposted in succession | tameshita.net, devtrends.ru, aitop.news |
| 2026-11-02 | The author posted a 4-minute video walkthrough on Bilibili (account "博文博文"): 53k views, 4,727 favorites, 381 comments | Bilibili API BV1bn4K6TEEV |
| 2026-11-04~06 | Daily gains recovered to 60–102 | GitHub repo |

Control over the same period: across 10/19–10/24, the daily gains for the dsh main repo, awesome-dsh-plugin, dsh-market, DSH-better-sidebar, and dsh-cost-meter stayed flat. Only dsh-context jumped on 10/22. So this spike came from its own exposure, not a rising ecosystem tide.

Usage cross-check: on npm, dsh-context got 30 days of downloads totaling 380,611 (10/7–11/5), with the last week of October at 2–28k per day. Over the same period dsh-better-sidebar got 282k, dsh-cost-meter 135k, and the dsh core package `@deepseek-ai/dsh` 2.22M (source: npm API). Note that download counts get inflated by frequent releases and auto-updates (159 versions), so treat them as relative only.

### 3. Competitors

Plenty of plugins do "look at dsh's context / tokens / cost", but apart from this one they're all tiny (star counts from GitHub search / our github database, 2026-11-06):

| Name | Where | Launched | Scale | How it differs from dsh-context |
|---|---|---|---|---|
| dsh's built-in context ring + per-turn usage popup (contextPressure / contextBreakdown) | Big-company features | 2026-08, shipped with dsh | all dsh users | Shows one occupancy percentage and a coarse breakdown. dsh-context reads the same official metering and extends it with per-request trends, raw text browsing, events, and multi-agent views. Its numbers match the official ring. |
| Han-1413141/dsh-cost-meter | Open source | 2026-08-13 | 640 stars, 135k npm downloads per month | Tracks money: cost, budget, balance, quota. Doesn't break down the context composition. |
| Ychris12138/dsh-usage-stats and dozens of dsh-usage / dsh-token-usage plugins | Open source | From ¥2026-08-14 | Biggest is 188 stars; most are single digits | Heatmaps and balance widgets. Thin features, and all look alike. |
| Zhenyu98/dsh-context-doctor | Open source | 2026-08-12 | 67 stars | Audits only the token cost of static injections (AGENTS.md, skills, tool schemas). Doesn't track how things evolve at runtime. |
| dsh-context-control / dsh-context-mode / dsh-context-compression-selector, etc. | Open source | 2026-08~10 | none above 30 stars | These actively manage context (compression, pruning). dsh-context watches it. The two complement each other. |
| dsh-xray, dsh-llm-profile, Lovenndme/dsh-command-context (10/31, also a /context command), epool86/dsh-monitor | Open source | 2026-08~10 | 1–12 stars | Same direction, but each covers only one slice of it. |
| Claude Code's built-in `/context`, ccusage-style tools | Other ecosystems | 2025 | Big | Not part of the dsh ecosystem. The author clearly borrowed the `/context` name. |

Verdict: direct competitors exist, but none match this one's completeness, and none has official backing. The closest, dsh-cost-meter, takes a different path (managing money).

### 4. Growth tactics (where users come from)

1. **Platform timing + first-mover naming** : the repo was created the day after dsh launched. It arrived in the first week, when plugin roundups ("must-install N plugins") were everywhere, and early roundups like Fishskin's picked it up. The package name `dsh-context` is exactly what users type into the Add plugin wizard. (Confirmed: repo / package creation dates and roundup articles. Ranking in the wizard search not verified.)

2. **Official endorsement from a person** : Cui Tianyi 9/30 recommended it and starred it, and another team member @turtle1999 also starred it. The author put this in the second line of the README, and it became the tag every media article repeats (the Zhidx headline literally reads "also recommended by Cui Tianyi"). (Confirmed.)

3. **Chinese tech media and KOL reposts** : plugin roundups on Zhihu, a WeChat Official Account (AI Coding 指北), Weibo (爱可可), Zhidx, Sina, Smzdm and the Alibaba Cloud developer community. 10/21 Weibo and Zhidx landed on the same day, matching the +1,052 stars over 10/22–23. (Confirmed: the timing lines up; the causation is an inference)

4. **The author makes their own content** : 11/2's Bilibili video got 53k views, after which daily star growth dropped from 30–40 back to 60–102. (Video data confirmed; attributing the daily gains to it is inference.)

5. **Overseas and long-tail SEO** : a recommendation thread on X by @ItsmeAjayKV; dozens of plugin directory sites (dshfind, deepseekplugins, dsh-plugin.org, dshbase, etc.) list it; the README in 8 languages brought reposts in Japanese and Russian. (Confirmed.)

6. **No paid promotion or sock-puppet network found** . Our database shows no suspicious flags either, and the fork-to-star ratio is 4.2%.

### 5. Why it charted

- **The need is real, and the official side only did half the job** : long tasks and multi-agent runs in dsh eat context, and users do hit "why did it suddenly get dumb / why did tokens spike." The official tool only gives a percentage ring. Complaints on Zhihu and Reddit describe exactly this pain point.

- **Timing** : it entered the day after dsh launched, riding the surge of "plugins as an ecosystem" (the dsh main repo has 270k stars, and thousands of repos carry the dsh-plugin topic), and official data says about 60% of users install third-party plugins.

- **The official endorsement became a long-term tag** : Cui Tianyi's single tweet alone only drove a moderate bump (9/30–10/9 added about 280 stars). But it went into the README and got quoted by media again and again, becoming a hook for every later round of promotion.

- **Product depth set it apart** : 84 days, 159 releases, nearly daily commits. It grew from a /context command into cross-session insights, a Fleet multi-agent view, and 8 languages. It also stays strictly aligned with official metering and dsh's native UI style (using the host's shiki highlighting, preview pane, and sidebar). Most similar plugins cover just one facet.

- **The author understands the plugin ecosystem** : experience building Dify plugins and a background as an Apache PMC member. They also built dsh-plugin-checker (a GitHub Action that checks plugins) and compatibility docs, so the plugin doesn't break through dsh's fast iteration (0.2.0 → 0.2.2).

### 6. Differentiation

The differentiator is depth and completeness of "seeing": others give one number, this gives the whole chain: composition, per-request evolution, raw text, events, files and commands, multiple agents, all matching official metering. That's why the official team singled it out ("feature-rich"), and why media picked it. So differentiation really is part of why it broke out. But going solo also took first-mover naming and official backing. Among similar plugins, dsh-context-doctor is also solid, and it only got 67 stars.

### 7. Visibility and copycats

- **Visibility** : in the Chinese dsh community it's a recognized top plugin (covered by Zhidx, Zhihu, Weibo influencers, and multiple plugin roundups, and recommended by a DSH lead). Outside that circle, almost nobody knows it. The English world has only scattered tweets and reposts on aggregator sites. It's open source and free, with no revenue.

- **Copycats** :
    - Straight copies: many forks copied the description word for word, including xiaweiliang060035, Michaelwucoc, localchart, Foam123, ccch1mneyyy, and even the dsh-cost-meter author Han-1413141 forked it. There's also an install-package mirror run by DSH-Extensions-list, 0.76.0.
    - Similar new plugins: dsh-context-explorer (10/21), Lovenndme/dsh-command-context (10/31, the /context command), dsh-llm-profile, and mitmplay/dsh-insight (multi-agent observation).
    - None of these has reached 15 stars, and none poses a real threat yet.

### 8. What others can learn

- In the first few days after a major platform opens its plugin ecosystem, grab a plain, generic name (`<platform>-context`) and fill in the native feature the official side only half-built.

- Read the official data source (here, the official token metering) so the numbers match the native UI. That earns trust and the chance of official recommendation.

- Put the official figure's endorsement in the README's second line as reusable social proof. Media will copy it.

- Ship releases often and keep multilingual READMEs: a steady supply of "fresh material" for roundup articles and aggregator sites.

- It can be copied to other agent ecosystems, e.g. building a "context observatory" for Codex, Claude Code, or other open plugin harnesses. The prerequisite is a plugin marketplace on the platform, with only coarse official context indicators.

### 9. Still unclear

- How dsh's Add plugin wizard search ranks results, and whether there's an official recommendation slot: not verified. This could be the biggest source of ongoing traffic.

- How the 10/22 spike splits between the Ackeri Kuang Weibo post and Zhidx can't be separated. There may also be WeChat reposts from the same period that my search didn't catch.

- The GitHub stargazers API was rate-limited the day of the check, so I couldn't pull daily star counts or the accounts that starred, to run a quality check.

- npm download numbers are inflated by auto-updates and frequent releases, so they can't be converted into a real install base.

### Sources

- [github.com/bowenliang123/dsh-context](https://github.com/bowenliang123/dsh-context) (README, commit history)

- [x.com/tianyi/status/2105256532657267119](https://x.com/tianyi/status/2105256532657267119) ;[x.com/tianyi/status/2103534313463783831](https://x.com/tianyi/status/2103534313463783831)

- [x.com/NFT_Chen/status/2105258670850494961](https://x.com/NFT_Chen/status/2105258670850494961) ;[x.com/ItsmeAjayKV/status/2096173850136559639](https://x.com/ItsmeAjayKV/status/2096173850136559639) ;[x.com/yupi996/article/2090649288535921094](https://x.com/yupi996/article/2090649288535921094)

- [www.zhidx.com/p/602940.html](https://www.zhidx.com/p/602940.html) ;[www.sina.cn/news/detail/5356469010042330.html](https://www.sina.cn/news/detail/5356469010042330.html)

- [zhuanlan.zhihu.com/p/2094125855421248833](https://zhuanlan.zhihu.com/p/2094125855421248833) ;[ai-coding-guide.com/articles/dsh-context](https://ai-coding-guide.com/articles/dsh-context)

- [www.bilibili.com/video/BV1bn4K6TEEV/](https://www.bilibili.com/video/BV1bn4K6TEEV/)(Bilibili API,2026-11-06)

- npm API: api.npmjs.org/downloads (dsh-context, dsh-cost-meter, dsh-better-sidebar, @deepseek-ai/dsh)

- Our github database: repo_day, repo_profile, trending (2026-11-06)

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.