dsh-armory 1.9.3 → 1.9.4
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.en.md +37 -10
- package/README.md +49 -11
- package/lib/client.js +827 -561
- package/lib/expert-market.js +172 -4
- package/lib/index.js +46 -8
- package/lib/mcp-config.js +4 -0
- package/lib/tool-ledger.js +111 -14
- package/lib/whats-new.js +21 -0
- package/lib/workspace-runtime.js +3 -1
- package/package.json +10 -2
package/README.en.md
CHANGED
|
@@ -43,8 +43,8 @@ A **unified board** for DeepSeek Harness (Web and desktop): five top-level tabs
|
|
|
43
43
|
|---|---|
|
|
44
44
|
| **MCP** | MCP servers: connection status, tool list, enable/disable, import, market install |
|
|
45
45
|
| **Skill** | Skills: enable/disable of `SKILL.md` folders, source directories, session audit scan |
|
|
46
|
-
| **Expert** | WorkBuddy expert directory (429 experts / 15 industries): browse, search, and "summon" one into a new session as an independent agent preset |
|
|
47
|
-
| **Tool Tailor** | Per-tool call counts, error rates, latency and context-token cost from local session logs;
|
|
46
|
+
| **Expert** | WorkBuddy expert directory (429 experts / 15 industries, including 53 expert **teams**): browse, search, and "summon" one into a new session as an independent agent preset. Team experts summon with their full roster — member personas are materialised locally and the lead is handed a runtime bridge that maps WorkBuddy's `TeamCreate` / `Agent(name=…)` onto DSH's `subagent` |
|
|
47
|
+
| **Tool Tailor** | Per-tool call counts, error rates, latency and context-token cost from local session logs; dead tools shown inline and prunable. Shows **all tools with their real call counts by default** — toggle "zero-call only (N)" to narrow. A persistent "showing N/M tools" read-out makes the filter state unmistakable |
|
|
48
48
|
| **Market** | Three sub-tabs: WorkBuddy connectors (316), curated built-ins (47), SkillHub skills (190k+) — browse, search, one-click install |
|
|
49
49
|
|
|
50
50
|
> The Skill panel reads DSH's custom skill directory (default `~/.agents/skills`): each skill is a folder with a `SKILL.md`,
|
|
@@ -53,19 +53,45 @@ A **unified board** for DeepSeek Harness (Web and desktop): five top-level tabs
|
|
|
53
53
|
|
|
54
54
|
## Screenshots
|
|
55
55
|
|
|
56
|
-
|
|
56
|
+
One shot per board tab (captured on the Windows desktop app, light theme):
|
|
57
57
|
|
|
58
|
-
|
|
58
|
+
### MCP — global board
|
|
59
59
|
|
|
60
|
-
|
|
60
|
+
Global / per-project scope tabs, inline enable/disable toggles; global and project servers may share a name while staying independent:
|
|
61
61
|
|
|
62
|
-
|
|
63
63
|
|
|
64
|
-
###
|
|
64
|
+
### MCP — connection details
|
|
65
65
|
|
|
66
|
-
|
|
66
|
+
Transport, connection parameters, call timeout and config source at a glance; below is the server's registered tool list (170 here — expand any row for its full input JSON Schema):
|
|
67
67
|
|
|
68
|
-
|
|
69
|
+
|
|
70
|
+
### Tool Tailor — tool-usage ledger
|
|
71
|
+
|
|
72
|
+
Four metric cards for the 30-day window (total calls / tools seen / zero-call MCP tools / overall error rate), one row per tool with calls, errors, error rate, Schema context tax and last use, and a persistent tailor suggestion at the bottom. Shows all tools with real call counts by default; "zero-call only (N)" narrows the view:
|
|
73
|
+
|
|
74
|
+
|
|
75
|
+
|
|
76
|
+
### Experts — summoned list
|
|
77
|
+
|
|
78
|
+
One card per summoned expert / expert team: bundled skills, details, dismiss; team experts expose a "roster" block in the detail dialog (each member's duty and persona file path):
|
|
79
|
+
|
|
80
|
+
|
|
81
|
+
|
|
82
|
+
### Market — three direct sources
|
|
83
|
+
|
|
84
|
+
Connectors (official WorkBuddy distribution, 316 entries; installable ones are labeled "install", CLI-based / token-dependent ones honestly show "view only" with the reason):
|
|
85
|
+
|
|
86
|
+
|
|
87
|
+
|
|
88
|
+
Skills (public SkillHub API, 190k+ skills; search + sort + pagination, installed ones show an "installed" badge with one-click uninstall):
|
|
89
|
+
|
|
90
|
+
|
|
91
|
+
|
|
92
|
+
Experts (WorkBuddy expert center, 429 experts / 15 industries; category filter + keyword search, "summon" straight from the card):
|
|
93
|
+
|
|
94
|
+
|
|
69
95
|
|
|
70
96
|
## Features (MCP board)
|
|
71
97
|
|
|
@@ -83,7 +109,8 @@ A **unified board** for DeepSeek Harness (Web and desktop): five top-level tabs
|
|
|
83
109
|
- **Directory**: reads the official WorkBuddy expert center distribution (`expert_center.json`, 429 experts / 15 industries, ETag-cached under `~/.dsh/wb-market/experts/`) with category filter, keyword search, and a detail dialog (persona, bundled skills, sample prompts) before you summon
|
|
84
110
|
- **Summoning** = appending one independent **agent preset** row to the profile patch (a `@deepseek-ai/dsh-agent-preset` entry that mirrors the built-in `standard` preset, with only the persona prefix swapped for the expert's full persona). New sessions then offer **WorkBuddy Expert · …** and that expert takes over the session, while the standard tools / planning / compaction / delegation stay intact. Multiple experts coexist; writes go through the existing patch pipeline (file lock + version check + atomic write), preserving other rows, comments and `!!js` expressions
|
|
85
111
|
- **Bundled skills**: tick to install them too — a full local WorkBuddy cache (`~/.workbuddy/plugins/cache/experts/`) is copied wholesale into `~/.agents/skills/` (prefixed `wbexpert__`, discovered by the skill registry immediately); without a cache it falls back to the CDN's single `SKILL.md` and labels the source honestly
|
|
86
|
-
- **
|
|
112
|
+
- **Team experts (summonable since 1.9.4)**: WorkBuddy's 53 "expert teams" assume `TeamCreate` and `Agent(name=…)`, neither of which DSH has. Summoning now also writes one file per member persona under `~/.dsh/wb-market/experts/teams/<expert>/`, and appends a **runtime adaptation** block to the lead's persona: creating the team becomes a no-op (the roster is fixed at summon time); dispatching a member becomes `subagent` (after `read`-ing that member's persona into the prompt), with sequential SOPs passing `run_in_background: false` so the lead waits for the result; follow-ups use `send_message`. The original rules — never ghost-write a member's output, never connect members directly, everything relayed through the lead — are preserved verbatim. Member personas stay **out of the system prompt** (otherwise every turn would pay for all five), and are read only when dispatching
|
|
113
|
+
- **Dismissing** removes that preset, its skill directory and the team member personas (hot-reloaded, no restart)
|
|
87
114
|
- **Summon menu (in-session quick access, enhanced in 1.9.3)** — the small board icon at the right of the composer toolbar opens three categories (**Skills / MCP / Experts**); hovering a category expands its submenu, and clicking an entry writes a reference into the composer (skills = `/skill-name`, MCP = names the `mcp__server__*` tools, experts = attaches to the current session):
|
|
88
115
|
- **Six rows per screen, scrollable**: a submenu shows six entries at a time and scrolls for the rest (thin themed scrollbar), while the **Manage skills / MCP / experts…** button stays pinned to the bottom instead of scrolling away
|
|
89
116
|
- **Most-recently-used first**: entries you just used float to the top with a "recent" badge, the rest keep their original order; the recency pool holds 40 entries
|
package/README.md
CHANGED
|
@@ -38,19 +38,45 @@ dsh plugin --profile web add github:2026heshao/dsh-armory
|
|
|
38
38
|
|
|
39
39
|
## 界面预览
|
|
40
40
|
|
|
41
|
-
|
|
41
|
+
看板五个分类页各一张（Windows 桌面端实拍，浅色主题）：
|
|
42
42
|
|
|
43
|
-
|
|
43
|
+
### MCP — 全局管理面板
|
|
44
44
|
|
|
45
|
-
|
|
45
|
+
全局 / 项目双作用域切换，行内启停；全局与各项目之间可以同名共存、各自独立连接：
|
|
46
46
|
|
|
47
|
-
|
|
48
48
|
|
|
49
|
-
###
|
|
49
|
+
### MCP — 连接详情
|
|
50
50
|
|
|
51
|
-
|
|
51
|
+
传输方式、连接参数、调用超时、配置来源一目了然，下方是该服务器注册的工具清单（本例 170 个，展开可看完整输入 JSON Schema）：
|
|
52
52
|
|
|
53
|
-
|
|
54
|
+
|
|
55
|
+
### 工具裁缝 — 工具使用账本
|
|
56
|
+
|
|
57
|
+
30 天窗口四张指标卡（调用总次数 / 出现过的工具 / 零调用 MCP 工具 / 整体错误率），逐工具列出调用、错误、错误率、Schema 上下文税与最后使用时间，底部常驻裁缝建议。默认显示全部工具与真实调用次数，需要时点「只看零调用（N）」收窄：
|
|
58
|
+
|
|
59
|
+
|
|
60
|
+
|
|
61
|
+
### 专家 — 已召唤清单
|
|
62
|
+
|
|
63
|
+
每个召唤的专家 / 专家团一张卡片：绑定技能、详情、送还；团队型专家在详情弹层里有「团队编制」区块（每位成员的职责与人设路径）：
|
|
64
|
+
|
|
65
|
+
|
|
66
|
+
|
|
67
|
+
### 市场 — 三处直连
|
|
68
|
+
|
|
69
|
+
连接器（WorkBuddy 官方分发 316 个；可安装的标「可安装」，CLI 型 / 需 token 注入的如实标「仅展示」并给出原因）：
|
|
70
|
+
|
|
71
|
+
|
|
72
|
+
|
|
73
|
+
技能（SkillHub 公开 API，19 万+ 技能；搜索 + 排序 + 分页，已装的标「已安装」可一键卸载）：
|
|
74
|
+
|
|
75
|
+
|
|
76
|
+
|
|
77
|
+
专家（WorkBuddy 专家中心 429 位 / 15 个行业；分类筛选 + 关键词搜索，卡片直接「召唤」）：
|
|
78
|
+
|
|
79
|
+
|
|
54
80
|
|
|
55
81
|
## 功能
|
|
56
82
|
|
|
@@ -64,7 +90,7 @@ dsh plugin --profile web add github:2026heshao/dsh-armory
|
|
|
64
90
|
- 跟随 DSH 深色/浅色主题，适配窄屏与移动宽度
|
|
65
91
|
- 更新提示与**一键升级**：有新版时在面板顶部显示可关闭的提示条（每天最多查一次，可用 `DSH_ARMORY_DISABLE_UPDATE_CHECK` 关闭）。点了「升级」才动，**不会自动升级**；升级由 DSH 官方的 pluginManager 执行，装完提示重启宿主生效。用 `link:` / `file:` / `github:` 装的本地开发副本不给按钮（那会把链接换成 npm 包）
|
|
66
92
|
- **升级后主动说一次变了什么**：版本变了后第一次打开面板弹一张小卡片，列出本次更新内容，并指路新功能在哪（例如「入口」开关在面板右上角）；确认一次后不再提示，不联网、不发任何数据
|
|
67
|
-
- **工具裁缝**（1.7.0 新增）：扫描本机会话日志（只解析 `tool/call` / `tool/result` 两类事件，不读消息正文），统计每个工具近 7/30/90 天的调用次数、错误率与最后使用时间；同时按当前注册表估算每个工具 Schema 每轮请求占用的上下文 token。窗口内零调用的 MCP
|
|
93
|
+
- **工具裁缝**（1.7.0 新增）：扫描本机会话日志（只解析 `tool/call` / `tool/result` 两类事件，不读消息正文），统计每个工具近 7/30/90 天的调用次数、错误率与最后使用时间；同时按当前注册表估算每个工具 Schema 每轮请求占用的上下文 token。窗口内零调用的 MCP 工具自动进「死工具清单」，行内可一键禁用/启用对应服务器（复用官方写链，随时可恢复）。**默认显示全部工具与真实调用次数**，零调用的行以底色标出；想只看死工具时点「只看零调用（N）」切换。顶部说明栏常驻回显「显示 N/M 个工具」，避免筛选态被误读成「所有计数都是 0」。全程只读，多帧 zstd 日志逐帧解析、个别帧损坏跳过不中断
|
|
68
94
|
- **WB 市场（直连安装，1.8.0 新增）**：看板第四个分类，两个市场一处浏览——
|
|
69
95
|
- **连接器（MCP）**：直读 WorkBuddy 官方分发 ZIP（`static.workbuddy.cn`，与 WorkBuddy 桌面端同一数据源），316 个连接器按需解析；可安装的 242 个（streamable-http 为主），CLI 型 / 需 WorkBuddy token 注入的 74 个如实标注「仅展示」并给出原因。安装 = 把该连接器的 `mcpServers` 合并写入 MCP 配置，与「导入 JSON」同一链路（准入、同名跳过、entry id 预留一致），全局与项目作用域都可选；ZIP 按 ETag 缓存在 `~/.dsh/wb-market/`，解压用 Node 内置 zlib，**零新增运行时依赖**
|
|
70
96
|
- **技能（SkillHub）**：直读 SkillHub 公开 API（`api.skillhub.cn`，19 万+ 技能），支持搜索（`keyword`）、排序（下载/安装/星标/评分）、分页；详情含发布方认证与安全扫描状态。安装 = 逐文件下载并按清单 sha256 校验后写入 `~/.agents/skills/`（与官方 skill 子系统同一发现范围），装完即刻生效；已装的显示「已安装」，可一键卸载、重装即更新
|
|
@@ -73,7 +99,8 @@ dsh plugin --profile web add github:2026heshao/dsh-armory
|
|
|
73
99
|
- **目录**：直读 WorkBuddy 专家中心官方分发（`expert_center.json`，429 位专家 / 15 个行业分类，ETag 缓存在 `~/.dsh/wb-market/experts/`），分类筛选 + 关键词搜索 + 详情弹层（召唤前看人设正文、捆绑技能、试试这样问）
|
|
74
100
|
- **召唤** = 往 profile 补丁写入一条**独立 Agent 预设**（`@deepseek-ai/dsh-agent-preset` 行，plugins 复刻内置 standard 预设、仅 persona 的 prefix 换成专家人设全文）——新建会话时选择「WorkBuddy 专家 · …」即专家接管该会话，标准工具/规划/压缩/委派原样保留；多个专家可共存，写入走既有 patch 写链（文件锁 + 版本校验 + 原子写），其余条目、注释与 `!!js` 表达式原样保留
|
|
75
101
|
- **捆绑技能**：召唤时可勾选一并安装——本机 WorkBuddy 缓存（`~/.workbuddy/plugins/cache/experts/`）有完整技能树就整目录复制到 `~/.agents/skills/`（`wbexpert__` 前缀，即刻被技能注册表发现）；无缓存时按 CDN 的 SKILL.md 单文件兜底并如实标注来源
|
|
76
|
-
-
|
|
102
|
+
- **团队型专家（1.9.4 可召唤）**：WorkBuddy 的 53 个「专家团」人设原本依赖 `TeamCreate` / `Agent(name=…)` 两个 DSH 没有的工具。召唤时除了写主理人预设，还会把每位成员的人设各自落成一份文件（`~/.dsh/wb-market/experts/teams/<专家>/`），并在主理人人设末尾追加一段**运行时适配**：建团队 → 无需调用（编制已在召唤时固定）；派活 → `subagent`（先把该成员人设 `read` 出来拼进 prompt），顺序 SOP 传 `run_in_background: false` 当场等结果；续聊用 `send_message`。原人设的铁律（不得代写成员产出、成员之间不直连、一切经主理人中转）原样保留。成员人设**不进系统提示**（否则每轮都要为 5 份人设白付上下文费），只在调度时按需读取
|
|
103
|
+
- **送还** = 摘除该召唤预设、对应技能目录与团队成员人设（热加载，无需重启）
|
|
77
104
|
- **召唤菜单（会话内快速调用，1.9.3 增强）**：会话输入框工具行右侧的看板小图标 → 「技能 / MCP / 专家」三分类，悬停展开子菜单，点一条即把引用写进输入框（技能 = `/技能名`、MCP = 点名 `mcp__服务__*`、专家 = 挂到当前会话）——
|
|
78
105
|
- **每屏 6 行 + 滚动**：子菜单一次显示 6 条，超出的往下滚（细滚动条跟随主题）；「管理技能 / MCP / 专家…」固定在底部，不被列表滚走
|
|
79
106
|
- **最近使用置顶**：刚用过的条目自动排到最前并带「最近」标记，其余保持原有顺序；最近使用记录池 40 条
|
|
@@ -196,13 +223,24 @@ dsh plugin --profile web remove dsh-armory
|
|
|
196
223
|
|
|
197
224
|
```sh
|
|
198
225
|
npm install --no-save yaml@2 # 测试唯一的裸依赖（node:test 内置）
|
|
199
|
-
npm test #
|
|
226
|
+
npm test # 全 6 套件（与 CI / 发版门槛同一条）
|
|
200
227
|
npm run test:e2e # 可选：用宿主补丁算法验证专家预设进入组合树
|
|
201
228
|
dsh --profile web --dump-config # 看 profile 的有效配置
|
|
202
229
|
dsh web # 起 Web 做真实操作验证
|
|
203
230
|
```
|
|
204
231
|
|
|
205
|
-
`npm test` 跑的是 `test/
|
|
232
|
+
`npm test` 跑的是 `test/run-all.mjs`，依次执行 6 个独立套件（各自是独立 node 进程，任一非零退出即整体失败）：
|
|
233
|
+
|
|
234
|
+
| 套件 | 覆盖 |
|
|
235
|
+
| --- | --- |
|
|
236
|
+
| `run-expert-manual.mjs` | 专家市场写链：真网拉 WorkBuddy 专家目录（ETag 命中后走磁盘缓存）与 senior-developer 人设，对 `lib/expert-market.js` 的 patch 写链做 33 条断言——含 1.9.1 修复的回归用例（召唤预设必须写成顶层 `- insert:` 指令、旧版顶层裸条目要能被迁移） |
|
|
237
|
+
| `upgrade-plan.mjs` | 一键升级安全判据：本地安装 / 来源未知一律不写盘 |
|
|
238
|
+
| `session-persona.mjs` | 会话级专家随时挂 / 换 / 摘（替身宿主验证 systemPrompt 作用域与 ctx 重绑两个假设） |
|
|
239
|
+
| `team-bridge.mjs` | 团队专家桥接：成员人设落盘、桥接段内容、preset 不含成员正文、送还清理、取不到人设时降级 |
|
|
240
|
+
| `verify-install-info.mjs` | 真 profile 数据下的安装来源判定（profile 不存在时自动 SKIP） |
|
|
241
|
+
| `verify-host-installinfo.mjs` | 宿主读取路径下的安装来源判定 |
|
|
242
|
+
|
|
243
|
+
本机存在 `~/.dsh/profiles/desktop/cordis.patch.yml` 时，`run-expert-manual.mjs` 会追加一组真实 profile 往返断言，没有则记 `SKIP`。
|
|
206
244
|
|
|
207
245
|
CI（push 到 main / PR，见 `.github/workflows/ci.yml`）与发版门槛（推 tag，见 `.github/workflows/release.yml`）跑的是**逐字相同**的命令（装 `yaml@2` → `node ./test/run-expert-manual.mjs`），两处口径不许分叉；发版流程见 [安装与升级](docs/installation.md#发布流程维护者)。
|
|
208
246
|
|