dsh-workbuddy-connect 0.7.6 → 0.7.8

package/README.en.md

@@ -13,7 +13,7 @@ Brings the models available in your signed-in WorkBuddy / WorkBuddy AI desktop a
 ```text
 Please install and verify the DSH (DeepSeek Harness) plugin dsh-workbuddy-connect for me (project: https://github.com/corrinehu/dsh-workbuddy-connect).
 Start by reading the project README and confirming which interface I actually use (Web / Desktop / TUI) and its DSH core version — for the Desktop app, the bundled core, never the standalone CLI's version.
-Then install from the official npm package per the version mapping (0.2.0-rc.2 -> 0.7.6, rc.1 -> 0.7.0, 0.1.5-0.1.7 -> 0.6.5), preferring a locally supported non-market channel; on some Desktop distributions (the official Electron app in particular) the profile is app-managed and rules vary by distribution — check mine first, then explore its compliant install route, but never bypass its validation.
+Then install from the official npm package per the version mapping (0.2.1-rc.2 -> 0.7.8, 0.2.1-rc.1 -> 0.7.7, 0.2.0-rc.2 -> 0.7.6, rc.1 -> 0.7.0, 0.1.5-0.1.7 -> 0.6.5), preferring a locally supported non-market channel; on some Desktop distributions (the official Electron app in particular) the profile is app-managed and rules vary by distribution — check mine first, then explore its compliant install route, but never bypass its validation.
 Ask for my consent and make backups before changing any local files or configuration.
 After installing, verify the WorkBuddy group appears in the model picker and walk me through one short conversation.
 If the Desktop install still fails generation validation (e.g. the @types/retry false positive), report it honestly, reference upstream dataelement/dsh-desktop#688, and suggest switching to the standalone Web / TUI — do not modify dependency packages or bypass the check.
@@ -23,7 +23,9 @@ Core-to-plugin quick reference (full table in "Compatibility and release history
 
 | DSH core | Recommended plugin |
 |---|---|
-| 0.2.0-rc.2 | 0.7.6 (npm latest) |
+| 0.2.1-rc.2 | 0.7.8 (npm latest) |
+| 0.2.1-rc.1 | 0.7.7 |
+| 0.2.0-rc.2 | 0.7.6 |
 | 0.2.0-rc.1 | 0.7.0 |
 | 0.1.5 – 0.1.7 | 0.6.5 |
 
@@ -31,8 +33,8 @@ Core-to-plugin quick reference (full table in "Compatibility and release history
 <summary>Manual install (command line, Web / TUI only)</summary>
 
 ```sh
-# Web (pin per your actual core — example: 0.2.0-rc.2 -> 0.7.6; or install from source via github:corrinehu/dsh-workbuddy-connect)
-dsh plugin --profile web add dsh-workbuddy-connect@0.7.6
+# Web (pin per your actual core — example: 0.2.1-rc.2 -> 0.7.8; or install from source via github:corrinehu/dsh-workbuddy-connect)
+dsh plugin --profile web add dsh-workbuddy-connect@0.7.8
 dsh web
 
 # TUI (example for a 0.1.x core; that profile needs pnpm 11, e.g. npx pnpm@11)
@@ -58,11 +60,11 @@ dsh --profile dsh-tui
 
 ![WorkBuddy AI models in the DSH model picker](assets/5.png)
 
-- **Image input**: most models accept images pasted or dropped straight into the conversation (GLM-5.3-Flash, GLM-5.2, the DeepSeek-V4 series, and more); the few text-only models (e.g. GLM-5.1) clearly say so.
+- **Image input**: most models accept images pasted or dropped straight into the conversation (GLM-5.3-Flash, the DeepSeek-V4 series, and more); the few text-only models (e.g. GLM-5.1 / GLM-5.2 on the China endpoint) clearly say so.
 
 - **Reasoning levels**: levels explicitly declared by WorkBuddy appear directly — for example, GLM-5.3 and GLM-5.3-Flash offer low / high / max. For models without declared levels, Web and Desktop offer a manual check in the model picker (it sends a few requests and may consume credit); models without a check result or selectable levels keep using WorkBuddy's default.
 
-- **Balance at a glance**: with a WorkBuddy / WorkBuddy AI model selected, the composer shows the account's remaining credit next to the model selector (e.g. `WB.3,401 credits`, `WB AI.3,401 credits` on the international app). Credit is not polled automatically (every status read is a real billing query): one read on mount, then refresh on demand with the ↻ beside it. During session switching, the controls may briefly hide until the model selection is ready, then return automatically. The refresh button remains available even when the model has no reasoning-level detection entry or the balance cannot be read.
+- **Balance at a glance**: with a WorkBuddy / WorkBuddy AI model selected, the composer shows the account's remaining credit next to the model selector (e.g. `WB.3,401 credits`, `WB AI.3,401 credits` on the international app). Credit is not polled automatically (every status read is a real billing query): one read on mount, then refresh on demand with the ↻ before the balance. During session switching, the controls may briefly hide until the model selection is ready, then return automatically. The refresh button remains available even when the model has no reasoning-level detection entry or the balance cannot be read.
 
 - **Status and detection**: Settings → Plugins → the matching card shows the account, token validity, remaining credit, and model offers; it also refreshes the model list manually, shows whether the current list came from the upstream or the built-in fallback, and runs manual reasoning-level detection for eligible models. On DSH 0.1.5 the entry lives under Settings → Plugins; on 0.1.6+ it moved to the left-sidebar Plugins panel (workbuddy-connect → View) — and the Settings → Models page shows no WorkBuddy entries on either generation, by design.
 
@@ -126,7 +128,9 @@ The plugin version must match the DSH core — a mismatched combination fails to
 
 | DSH core | Recommended plugin | Notes |
 |---|---|---|
-| `0.2.0-rc.2` | `0.7.6` (npm latest) | Current line; fixes enterprise catalog gaps and international chat requests blocked by the upstream channel check. Avoid 0.7.2 with its registry regression. |
+| `0.2.1-rc.2` | `0.7.8` | Targets this core only; `0.2.0-rc.2` and `0.2.1-rc.2` differ. Check the core bundled in your Desktop app. |
+| `0.2.1-rc.1` | `0.7.7` | The core moved to `@earendil-works/pi-ai` 1.3.0, so the plugin bundles 1.3.0 too — a leftover 0.87.1 overrides entry must be removed (see the pi-ai note below). |
+| `0.2.0-rc.2` | `0.7.6` | Previous core line; fixes enterprise catalog gaps and international chat requests blocked by the upstream channel check. Avoid 0.7.2 with its registry regression. |
 | `0.2.0-rc.1` | `0.7.0` | 0.7.1 drops rc.1 support; stay on this version. |
 | `0.1.5` – `0.1.7` | `0.6.5` | Final release for the old cores; 0.7.0 targets 0.2.0 only. |
 | `0.1.2-rc.1` | `0.3.1` | Older line, historical maintenance only. |
@@ -134,6 +138,8 @@ The plugin version must match the DSH core — a mismatched combination fails to
 
 Release history (full records at [releases](https://github.com/corrinehu/dsh-workbuddy-connect/releases)):
 
+- **0.7.8**: adapts to DSH 0.2.1-rc.2; bundles the official atomic-write helper instead of relying on host module exposure; refreshes the China catalog snapshot and distinguishes Hy3 / Hy3-X; improves probe result labels and fixes the credit-refresh placement and tooltip. Thanks to @chenpiyanghappy for PR [#156](https://github.com/corrinehu/dsh-workbuddy-connect/pull/156).
+- **0.7.7**: adapts to the DSH 0.2.1-rc.1 core: peers narrow to 0.2.1-rc.1 and the plugin-bundled `@earendil-works/pi-ai` moves to 1.3.0 with it (a profile-level leftover 0.87.1 overrides entry must be removed, or the plugin fails to load).
 - **0.7.6**: fixes enterprise accounts missing `glm-5.3-flashx` from the model catalog ([#123](https://github.com/corrinehu/dsh-workbuddy-connect/issues/123)), and fixes international chat requests being blocked by the upstream channel check ([#139](https://github.com/corrinehu/dsh-workbuddy-connect/issues/139) / [#142](https://github.com/corrinehu/dsh-workbuddy-connect/issues/142)).
 - **0.7.5**: restores composer controls after session switching ([#125](https://github.com/corrinehu/dsh-workbuddy-connect/issues/125) / [#132](https://github.com/corrinehu/dsh-workbuddy-connect/issues/132)), fixes WSL helper environment forwarding ([#133](https://github.com/corrinehu/dsh-workbuddy-connect/issues/133)), and adds community projects to both READMEs.
 - **0.7.4**: Windows diagnostics and stability improvements: accurate classification of oversized registry-subprocess output, cross-platform test-assumption fixes ([#62](https://github.com/corrinehu/dsh-workbuddy-connect/issues/62)), and a reworked install section.
@@ -143,9 +149,9 @@ Release history (full records at [releases](https://github.com/corrinehu/dsh-wor
 - **0.7.0**: adapts to DSH 0.2.0's settings-service rework and drops 0.1.x support.
 - **0.6.5**: final release of the 0.1.x line.
 
-Pinned-version install example: `dsh plugin --profile web add dsh-workbuddy-connect@0.7.6`.
+Pinned-version install example: `dsh plugin --profile web add dsh-workbuddy-connect@0.7.8`.
 
-**Do not upgrade the profile-level `@earendil-works/pi-ai` to 1.x**: this plugin and the `0.2.0-rc.2` generation of the DSH core both require `^0.87.1` (bundled by the plugin since 0.7.2), and third-party plugin managers' "updatable" prompts do not check peer constraints — do not act on them ([#86](https://github.com/corrinehu/dsh-workbuddy-connect/issues/86)); likewise, do not blindly remove a leftover `overrides: {'@earendil-works/pi-ai': '0.87.1'}` until you have confirmed nothing else depends on it.
+**pi-ai versions are per core line and must not be mixed**: the 0.2.1-rc.1 / rc.2 core lines (plugin 0.7.7 / 0.7.8 respectively) moved to `@earendil-works/pi-ai` 1.3.0 (bundled by the plugin), so a profile-level leftover `overrides: {'@earendil-works/pi-ai': '0.87.1'}` **must be removed** or the plugin fails to load; the 0.2.0-rc.2 core line (plugin 0.7.6) still requires `^0.87.1` — on that line, **do not** upgrade the profile-level pi-ai to 1.x, and do not act on third-party plugin managers' "updatable" prompts, which do not check peer constraints ([#86](https://github.com/corrinehu/dsh-workbuddy-connect/issues/86)).
 
 ## Disclaimer
 

package/README.md

@@ -13,7 +13,7 @@
 ```text
 请帮我安装并验证 DSH（DeepSeek Harness）插件 dsh-workbuddy-connect（项目地址：https://github.com/corrinehu/dsh-workbuddy-connect）：
 先阅读项目 README，确认我实际使用的界面（Web / Desktop / TUI）与 DSH 内核版本，桌面版以 App 内置核心为准，不要用独立 CLI 的版本代替；
-再按版本对应关系从 npm 正式包安装（0.2.0-rc.2→0.7.6、rc.1→0.7.0、0.1.5–0.1.7→0.6.5），优先走本机受支持的非插件市场渠道；部分 Desktop 发行版（尤其官方 Electron 版）的 profile 由 App 管理，各发行版规则未必相同，先确认我所用的发行版，再探查其合规的安装方式，但不得绕过其校验；
+再按版本对应关系从 npm 正式包安装（0.2.1-rc.2→0.7.8、0.2.1-rc.1→0.7.7、0.2.0-rc.2→0.7.6、rc.1→0.7.0、0.1.5–0.1.7→0.6.5），优先走本机受支持的非插件市场渠道；部分 Desktop 发行版（尤其官方 Electron 版）的 profile 由 App 管理，各发行版规则未必相同，先确认我所用的发行版，再探查其合规的安装方式，但不得绕过其校验；
 任何本地文件或配置改动，先征得我同意并做好备份；
 装完后验证模型选择器里出现 WorkBuddy 分组，并陪我完成一次简短对话；
 若 Desktop 安装仍因 generation 校验失败（如 @types/retry 误报），请如实报告、参照上游 dataelement/dsh-desktop#688，并建议我改用独立的 Web / TUI，不要修改依赖包或绕过校验。
@@ -23,7 +23,9 @@
 
 | DSH 内核 | 推荐插件版 |
 |---|---|
-| 0.2.0-rc.2 | 0.7.6（npm latest） |
+| 0.2.1-rc.2 | 0.7.8（npm latest） |
+| 0.2.1-rc.1 | 0.7.7 |
+| 0.2.0-rc.2 | 0.7.6 |
 | 0.2.0-rc.1 | 0.7.0 |
 | 0.1.5 – 0.1.7 | 0.6.5 |
 
@@ -31,8 +33,8 @@
 <summary>手动安装（命令行，仅 Web / TUI）</summary>
 
 ```sh
-# Web（按实际内核固定版本，示例为 0.2.0-rc.2 → 0.7.6；也可用 github:corrinehu/dsh-workbuddy-connect 从源码安装）
-dsh plugin --profile web add dsh-workbuddy-connect@0.7.6
+# Web（按实际内核固定版本，示例为 0.2.1-rc.2 → 0.7.8；也可用 github:corrinehu/dsh-workbuddy-connect 从源码安装）
+dsh plugin --profile web add dsh-workbuddy-connect@0.7.8
 dsh web
 
 # TUI（示例为 0.1.x 内核；该 profile 需用 pnpm 11 安装，如 npx pnpm@11）
@@ -58,11 +60,11 @@ dsh --profile dsh-tui
 
 ![WorkBuddy AI 模型出现在 DSH 模型选择器中](assets/5.png)
 
-- **图片输入**：大部分模型支持在对话里直接粘贴或拖入图片（GLM-5.3-Flash、GLM-5.2、DeepSeek-V4 系列等）；少数纯文字模型（如 GLM-5.1）会明确提示不支持。
+- **图片输入**：大部分模型支持在对话里直接粘贴或拖入图片（GLM-5.3-Flash、DeepSeek-V4 系列等）；少数纯文字模型（如 GLM-5.1、GLM-5.2）会明确提示不支持。
 
 - **推理档位**：WorkBuddy 明确声明的档位会直接显示，例如 GLM-5.3 和 GLM-5.3-Flash 可选 low / high / max。未声明档位的模型可在 Web 和 Desktop 的模型选择器中手动检测（发送少量请求，可能消耗积分）；未检测或没有可用档位的模型使用 WorkBuddy 默认档位。
 
-- **余额速览**：选中 WorkBuddy / WorkBuddy AI 模型时，输入框底部在模型选择器旁显示当前账号的剩余积分（如 `WB.3,401（积分）`、国际版 `WB AI.3,401（积分）`）。积分不自动轮询（每次读取都是真实计费查询），挂载时读取一次后可点旁边的 ↻ 手动刷新。切换会话时组件可能短暂隐藏，模型选择就绪后自动恢复；即使模型没有推理档位检测入口或余额读取失败，刷新按钮仍保留。
+- **余额速览**：选中 WorkBuddy / WorkBuddy AI 模型时，输入框底部在模型选择器旁显示当前账号的剩余积分（如 `WB.3,401（积分）`、国际版 `WB AI.3,401（积分）`）。积分不自动轮询（每次读取都是真实计费查询），挂载时读取一次后可点积分文字前方的 ↻ 手动刷新。切换会话时组件可能短暂隐藏，模型选择就绪后自动恢复；即使模型没有推理档位检测入口或余额读取失败，刷新按钮仍保留。
 
 - **信息查看与检测**：设置 → 插件 → 对应卡片可查看账号、令牌有效期、剩余积分和模型优惠，也可手动刷新模型列表并查看当前列表来自上游还是内置兜底，或对可检测模型手动检测推理档位。DSH 0.1.5 的入口在「设置 → 插件」，0.1.6+ 移到主界面左侧栏「插件」面板（workbuddy-connect → 查看）；「设置 → 模型」页两代都不显示 WorkBuddy 条目，属有意设计。
 
@@ -127,7 +129,9 @@ dsh plugin --profile web exec dsh-workbuddy-connect doctor --provider workbuddy-
 
 | DSH 内核 | 推荐插件版 | 注意事项 |
 |---|---|---|
-| `0.2.0-rc.2` | `0.7.6`（npm latest） | 当前最新线；修复企业账号模型目录缺项与国际版对话被渠道校验拦截，勿停留在有注册表回归的 0.7.2。 |
+| `0.2.1-rc.2` | `0.7.8` | 仅适配这一内核；`0.2.0-rc.2` 与 `0.2.1-rc.2` 不同，请核对 App 内置版本。 |
+| `0.2.1-rc.1` | `0.7.7` | 内核升级到 `@earendil-works/pi-ai` 1.3.0，插件同步自带 1.3.0，历史遗留的 0.87.1 overrides 必须移除（见下方 pi-ai 说明）。 |
+| `0.2.0-rc.2` | `0.7.6` | 上一代内核线；修复企业账号模型目录缺项与国际版对话被渠道校验拦截，勿停留在有注册表回归的 0.7.2。 |
 | `0.2.0-rc.1` | `0.7.0` | 0.7.1 起不再支持 rc.1，请停留本版。 |
 | `0.1.5` – `0.1.7` | `0.6.5` | 旧内核最终版；0.7.0 起仅面向 0.2.0。 |
 | `0.1.2-rc.1` | `0.3.1` | 更早旧线，仅历史维护。 |
@@ -135,6 +139,8 @@ dsh plugin --profile web exec dsh-workbuddy-connect doctor --provider workbuddy-
 
 历次更新（完整记录见 [releases](https://github.com/corrinehu/dsh-workbuddy-connect/releases)）：
 
+- **0.7.8**：适配 DSH 0.2.1-rc.2；将官方原子写入工具打包到插件，避免依赖宿主暴露该模块；更新国内模型目录快照，区分 Hy3 / Hy3-X；优化推理检测结果展示，调整积分刷新按钮位置并修正悬停提示。感谢 @chenpiyanghappy 的 PR [#156](https://github.com/corrinehu/dsh-workbuddy-connect/pull/156)。
+- **0.7.7**：适配 DSH 0.2.1-rc.1 内核：peer 收窄到 0.2.1-rc.1，插件自带的 `@earendil-works/pi-ai` 随内核升级到 1.3.0（profile 级历史遗留的 0.87.1 overrides 必须移除，否则插件无法加载）；修复国际版与国内版推理档位检测停留「检测未完成」——上游把档位拒绝码改拼为 `invalid_parameter_value`（[#143](https://github.com/corrinehu/dsh-workbuddy-connect/issues/143) / [#148](https://github.com/corrinehu/dsh-workbuddy-connect/issues/148)，PR [#145](https://github.com/corrinehu/dsh-workbuddy-connect/pull/145)）；修复中文 Windows 控制台下注册表缺键诊断因 GBK 编码被误判（zh-CN 实机验证）。
 - **0.7.6**：修复企业账号模型目录缺少 `glm-5.3-flashx`（[#123](https://github.com/corrinehu/dsh-workbuddy-connect/issues/123)），修复国际版对话请求被上游渠道校验拦截（[#139](https://github.com/corrinehu/dsh-workbuddy-connect/issues/139) / [#142](https://github.com/corrinehu/dsh-workbuddy-connect/issues/142)）。
 - **0.7.5**：修复切换会话后输入框余额与刷新按钮消失（[#125](https://github.com/corrinehu/dsh-workbuddy-connect/issues/125) / [#132](https://github.com/corrinehu/dsh-workbuddy-connect/issues/132)），修复 WSL 调用 Windows 凭据解密程序的环境转发（[#133](https://github.com/corrinehu/dsh-workbuddy-connect/issues/133)），README 收录社区项目（[#130](https://github.com/corrinehu/dsh-workbuddy-connect/issues/130)）。
 - **0.7.4**：Windows 诊断与稳定性改进：注册表子进程输出超限准确分类、跨平台测试假设修正（[#62](https://github.com/corrinehu/dsh-workbuddy-connect/issues/62)），安装说明重排。
@@ -144,9 +150,9 @@ dsh plugin --profile web exec dsh-workbuddy-connect doctor --provider workbuddy-
 - **0.7.0**：适配 DSH 0.2.0 的设置服务改造，不再支持 0.1.x。
 - **0.6.5**：0.1.x 线最终版本。
 
-固定版本安装示例：`dsh plugin --profile web add dsh-workbuddy-connect@0.7.6`。
+固定版本安装示例：`dsh plugin --profile web add dsh-workbuddy-connect@0.7.8`。
 
-**不要把 profile 级的 `@earendil-works/pi-ai` 升到 1.x**：本插件与 `0.2.0-rc.2` 这一代内核都要求 `^0.87.1`（0.7.2 起由插件自带），第三方插件管理器的「可更新」提示不校验 peer 约束、不能照做（[#86](https://github.com/corrinehu/dsh-workbuddy-connect/issues/86)）；为此历史遗留的 `overrides: {'@earendil-works/pi-ai': '0.87.1'}` 在确认其他插件不依赖前也不要盲目移除。
+**pi-ai 版本按内核线区分，不能混**：0.2.1-rc.1 / rc.2 内核线（分别对应插件 0.7.7 / 0.7.8）已升级到 `@earendil-works/pi-ai` 1.3.0（由插件自带），profile 级历史遗留的 `overrides: {'@earendil-works/pi-ai': '0.87.1'}` **必须移除**，否则插件会因加载到旧一代 pi-ai 而启动失败；0.2.0-rc.2 内核线（插件 0.7.6）仍要求 `^0.87.1`，在这条线上**不要**把 profile 级的 pi-ai 升到 1.x，第三方插件管理器的「可更新」提示不校验 peer 约束、不能照做（[#86](https://github.com/corrinehu/dsh-workbuddy-connect/issues/86)）。
 
 ## 免责声明
 

package/THIRD_PARTY_NOTICES.txt

@@ -0,0 +1,24 @@
+@deepseek-ai/dsh-atomic-write 0.2.1-rc.2
+Bundled from https://github.com/deepseek-ai/deepseek-harness/tree/main/packages/util/atomic-write
+
+MIT License
+
+Copyright (c) 2026 DeepSeek
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/lib/bin.js

@@ -1,5 +1,5 @@
 #!/usr/bin/env node
-import { C as WORKBUDDY_VARIANTS, O as WorkBuddyUpstreamClient, Q as resolveAppVersion, S as CN_VARIANT, a as readHostHeartbeat, b as atRestKeyProviderFor, c as WORKBUDDY_CONNECT_VERSION, l as FALLBACK_WORKBUDDY_AI_MODELS, m as WorkBuddyCredentialStore, o as workbuddyHostHeartbeatPath, r as isHeartbeatProcessAlive, u as FALLBACK_WORKBUDDY_MODELS, w as variantFor } from "./host-heartbeat-vU86u9aU.js";
+import { C as WORKBUDDY_VARIANTS, O as WorkBuddyUpstreamClient, Q as resolveAppVersion, S as CN_VARIANT, a as readHostHeartbeat, b as atRestKeyProviderFor, c as WORKBUDDY_CONNECT_VERSION, l as FALLBACK_WORKBUDDY_AI_MODELS, m as WorkBuddyCredentialStore, o as workbuddyHostHeartbeatPath, r as isHeartbeatProcessAlive, u as FALLBACK_WORKBUDDY_MODELS, w as variantFor } from "./host-heartbeat-sjotAEdb.js";
 import { realpathSync } from "node:fs";
 import { fileURLToPath } from "node:url";
 //#region src/bin.ts

package/lib/client.js

@@ -456,6 +456,37 @@ window.__ModuleLoader__.load({
 			color: "var(--dsw-alias-state-success-primary, #22a06b)"
 		};
 		/**
+		* Chip for a probe that came back `non-validating`.
+		*
+		* Amber rather than the success green: the upstream accepted a sentinel effort,
+		* which tells us it does not check the parameter — not that any level works.
+		* The state is an observation, and it used to share the green chip with the
+		* verified levels, which read as "these are the supported levels".
+		*/
+		const probeChipNotValidatingStyle = {
+			padding: "1px 8px",
+			borderRadius: 999,
+			fontSize: 11,
+			lineHeight: "18px",
+			background: "var(--dsw-alias-state-warning-subtle, rgba(214, 142, 23, 0.14))",
+			color: "var(--dsw-alias-state-warning-primary, #b26a00)"
+		};
+		/** Chip for a probe that never produced an answer (cancelled or failed). */
+		const probeChipUnknownStyle = {
+			padding: "1px 8px",
+			borderRadius: 999,
+			fontSize: 11,
+			lineHeight: "18px",
+			background: "var(--dsw-alias-fill-tertiary, rgba(127, 127, 127, 0.12))",
+			color: "var(--dsw-alias-label-tertiary, #8a8a8a)"
+		};
+		/** Pick the chip style that matches what the observation actually says. */
+		function probeChipStyleFor(result) {
+			if (result.validation === "validating" && result.efforts.length > 0) return modelBadgeChipStyle;
+			if (result.validation === "non-validating") return probeChipNotValidatingStyle;
+			return probeChipUnknownStyle;
+		}
+		/**
 		* Localize an upstream promotional badge label, with an unknown-badge fallback.
 		*
 		* The CN catalog spells badges in Chinese (`限时免费`, `夜间折扣`); the
@@ -467,9 +498,32 @@ window.__ModuleLoader__.load({
 		function modelBadgeLabel(badge, t) {
 			if (badge === "限时免费") return t("badgeLimitedFree");
 			if (badge === "夜间折扣") return t("badgeNightDiscount");
+			if (badge === "夜间免费") return t("badgeNightFree");
+			if (badge === "独家限免") return t("badgeExclusiveFree");
 			if (badge === "Free now") return t("badgeFreeNow");
 			return badge;
 		}
+		/**
+		* Upstream badge labels that already say "you pay nothing", in either language.
+		*
+		* The catalog reports both `free: true` and a promotional badge for the same
+		* row (`hy3` carries `限时免费`, `step-5-preview` carries `独家限免`). Rendering
+		* the generic free chip on top of those printed the same fact twice and, worse,
+		* pushed the second chip past the row's horizontal budget — it collapsed into a
+		* sliver instead of wrapping. One chip per row is enough: a badge that already
+		* states the offer is preferred over the bare "Free".
+		*/
+		const FREE_STATEMENT_BADGES = /* @__PURE__ */ new Set([
+			"限时免费",
+			"夜间免费",
+			"独家限免",
+			"Free now"
+		]);
+		/** Whether a row still needs the generic free chip next to its badges. */
+		function needsFreeChip(free, badges) {
+			if (free !== true) return false;
+			return !(badges ?? []).some((badge) => FREE_STATEMENT_BADGES.has(badge));
+		}
 		const progressTrackStyle = {
 			height: 8,
 			overflow: "hidden",
@@ -718,7 +772,7 @@ window.__ModuleLoader__.load({
 						children: [model.badges?.map((badge) => /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
 							style: modelBadgeChipStyle,
 							children: modelBadgeLabel(badge, t)
-						}, badge)), model.free === true ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+						}, badge)), needsFreeChip(model.free, model.badges) ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
 							style: modelBadgeChipStyle,
 							children: t("freeModel")
 						}) : null]
@@ -822,7 +876,7 @@ window.__ModuleLoader__.load({
 													style: modelBadgeChipStyle,
 													children: modelBadgeLabel(badge, t)
 												}, badge)),
-												model.free === true ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+												needsFreeChip(model.free, model.badges) ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
 													style: modelBadgeChipStyle,
 													children: t("freeModel")
 												}) : null
@@ -923,7 +977,9 @@ window.__ModuleLoader__.load({
 						style: quotaGroupStyle,
 						children: probe.candidates.map((id) => {
 							const result = probe.results.find((entry) => entry.id === id);
-							const name = models?.find((model) => model.id === id)?.name ?? result?.name ?? id;
+							const row = models?.find((model) => model.id === id);
+							const name = row?.name ?? result?.name ?? id;
+							const declared = (row?.declaredEfforts?.length ?? 0) > 0 ? row.declaredEfforts.join(" / ") : row?.defaultEffort;
 							return /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 								style: modelOfferStyle,
 								children: [
@@ -932,8 +988,8 @@ window.__ModuleLoader__.load({
 										children: [/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", { children: name }), /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
 											style: probeRowEndStyle,
 											children: [result === void 0 ? null : /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-												style: modelBadgeChipStyle,
-												children: result.validation === "validating" && result.efforts.length > 0 ? result.efforts.join(" / ") : t(result.validation === "non-validating" ? "probeResultNotValidating" : "probeResultUnknown")
+												style: probeChipStyleFor(result),
+												children: result.validation === "validating" && result.efforts.length > 0 ? result.efforts.join(" / ") : result.validation === "non-validating" ? t("probeResultNotValidating") + (declared === void 0 ? "" : t("probeDeclaredSuffix", { levels: declared })) : t("probeResultUnknown")
 											}), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
 												type: "button",
 												style: buttonStyle$2,
@@ -1560,7 +1616,7 @@ window.__ModuleLoader__.load({
 			lineHeight: "16px",
 			color: "var(--dsw-alias-label-tertiary)",
 			whiteSpace: "nowrap",
-			padding: "0 4px"
+			padding: "0 4px 0 0"
 		};
 		/**
 		* The badge's ↻ (#99): the only re-read after the mount fetch. An icon-only
@@ -1890,17 +1946,7 @@ window.__ModuleLoader__.load({
 			const showTooltip = tooltipVisible && !confirming && note === void 0;
 			return /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
 				style: wrapperStyle,
-				onMouseEnter: () => {
-					setTooltipVisible(true);
-				},
-				onMouseLeave: () => {
-					setTooltipVisible(false);
-				},
 				children: [
-					credit === void 0 ? null : /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-						style: creditStyle,
-						children: credit
-					}),
 					/* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
 						type: "button",
 						style: {
@@ -1929,6 +1975,10 @@ window.__ModuleLoader__.load({
 							children: [/* @__PURE__ */ (0, react_jsx_runtime.jsx)("path", { d: "M20 12a8 8 0 1 1-2.34-5.66" }), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("path", { d: "M20 4v4h-4" })]
 						})
 					}),
+					credit === void 0 ? null : /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+						style: creditStyle,
+						children: credit
+					}),
 					visible && /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("button", {
 						type: "button",
 						"aria-label": text,
@@ -1939,6 +1989,12 @@ window.__ModuleLoader__.load({
 						onClick: () => {
 							setConfirming(true);
 						},
+						onMouseEnter: () => {
+							setTooltipVisible(true);
+						},
+						onMouseLeave: () => {
+							setTooltipVisible(false);
+						},
 						onFocus: () => {
 							setTooltipVisible(true);
 						},
@@ -2772,7 +2828,7 @@ window.__ModuleLoader__.load({
 		};
 		//#endregion
 		//#region src/version.ts
-		const WORKBUDDY_CONNECT_VERSION = "0.7.6";
+		const WORKBUDDY_CONNECT_VERSION = "0.7.8";
 		//#endregion
 		//#region src/client/WorkBuddyConfigPage.tsx
 		/**
@@ -2815,15 +2871,15 @@ window.__ModuleLoader__.load({
 		//#region src/client/locales.ts
 		/** Plugin-card copy registered under the settings.workbuddy locale namespace. */
 		const en = {
-			title: "DSH WorkBuddy Connect",
-			intro: "Use the models in the WorkBuddy desktop app directly in DSH — zero configuration, ready out of the box.",
-			titleAI: "DSH WorkBuddy AI Connect",
+			title: "DSH WorkBuddy Connect (China)",
+			intro: "Use the models in the WorkBuddy desktop app (China edition) directly in DSH — zero configuration, ready out of the box.",
+			titleAI: "DSH WorkBuddy AI Connect (International)",
 			introAI: "Use the models in the WorkBuddy AI international desktop app directly in DSH — zero configuration, ready out of the box.",
 			expand: "Expand",
 			collapse: "Collapse",
 			loading: "Loading account…",
 			signedOut: "Not signed in",
-			signedOutHint: "Sign in once in the WorkBuddy desktop app; this plugin follows that sign-in automatically.",
+			signedOutHint: "Sign in once in the WorkBuddy desktop app (China edition); this plugin follows that sign-in automatically.",
 			signedOutHintAI: "Sign in once in the WorkBuddy AI desktop app; this plugin follows that sign-in automatically.",
 			signedInAs: "Signed in as {nickname}",
 			accessTokenExpires: "Access token expires {time} (refresh is automatic)",
@@ -2873,17 +2929,19 @@ window.__ModuleLoader__.load({
 			badgeLimitedFree: "Limited-time free",
 			badgeNightDiscount: "Night discount",
 			badgeFreeNow: "Free now",
+			badgeExclusiveFree: "Exclusive free",
+			badgeNightFree: "Free at night",
 			rate: "{rate} credits per message",
 			rateUnknown: "Price unavailable — refresh to update",
 			probeLabel: "Reasoning levels",
 			probeTooltipIdle: "Detect the reasoning levels {model} accepts",
 			probeTooltipVerified: "Accepted levels: {levels} · click to detect again",
-			probeTooltipNotValidating: "This model does not check the effort parameter",
+			probeTooltipNotValidating: "This model accepts any effort value, so the probe cannot tell the levels apart · click to re-probe",
 			probeTooltipRetry: "Detection did not complete · click to retry",
 			probeBubbleBody: "Send test requests to confirm the available reasoning levels. May consume a small amount of credit.",
 			probeConfirmAction: "Confirm",
 			probeNoteVerified: "Detected: {levels}",
-			probeNoteNotValidating: "This model does not check the effort parameter",
+			probeNoteNotValidating: "Could not probe the levels: the upstream does not validate the effort parameter — that is not the same as no level being supported",
 			probeNoteUnknown: "Detection did not complete",
 			probeNoteDismiss: "Got it",
 			probeHeading: "Reasoning effort detection",
@@ -2899,7 +2957,8 @@ window.__ModuleLoader__.load({
 			probeConfirmBody: "Send test requests to {model} to confirm its available reasoning levels. May consume a small amount of credit.",
 			cancel: "Cancel",
 			probeResultVerified: "Verified levels: {levels}",
-			probeResultNotValidating: "This model does not check the effort parameter",
+			probeResultNotValidating: "Upstream ignores the effort parameter",
+			probeDeclaredSuffix: " (catalogue declares {levels})",
 			probeResultUnknown: "Detection did not complete",
 			probeResultAt: "Detected {time}",
 			probeResultEmpty: "No detectable models right now.",
@@ -2939,15 +2998,15 @@ window.__ModuleLoader__.load({
 			agentUpgradePrompt: "Please open {repository}, check its latest version, and install or update the plugin \"dsh-workbuddy-connect\" in my current DSH profile following the project README (the install command differs by profile). After upgrading, fully quit and restart DSH."
 		};
 		const zh = {
-			title: "DSH WorkBuddy Connect",
-			intro: "在 DSH 中直接使用 WorkBuddy 桌面 App 包含的模型，开箱即用，无需额外配置。",
-			titleAI: "DSH WorkBuddy AI Connect",
+			title: "DSH WorkBuddy Connect（国内版）",
+			intro: "在 DSH 中直接使用 WorkBuddy 国内版桌面 App 包含的模型，开箱即用，无需额外配置。",
+			titleAI: "DSH WorkBuddy AI Connect（国际版）",
 			introAI: "在 DSH 中直接使用 WorkBuddy AI 国际版桌面 App 包含的模型，开箱即用，无需额外配置。",
 			expand: "展开",
 			collapse: "收起",
 			loading: "正在读取账号…",
 			signedOut: "未登录",
-			signedOutHint: "在 WorkBuddy 桌面 App 里登录一次即可，插件会自动跟随当前登录的账号。",
+			signedOutHint: "在 WorkBuddy 国内版桌面 App 里登录一次即可，插件会自动跟随当前登录的账号。",
 			signedOutHintAI: "在 WorkBuddy AI 国际版桌面 App 里登录一次即可，插件会自动跟随当前登录的账号。",
 			signedInAs: "已登录：{nickname}",
 			accessTokenExpires: "访问令牌 {time} 过期（自动续期）",
@@ -2997,17 +3056,19 @@ window.__ModuleLoader__.load({
 			badgeLimitedFree: "限时免费",
 			badgeNightDiscount: "夜间折扣",
 			badgeFreeNow: "限时免费",
+			badgeExclusiveFree: "独家限免",
+			badgeNightFree: "夜间免费",
 			rate: "{rate} 积分/次",
 			rateUnknown: "价格未知 — 刷新后更新",
 			probeLabel: "推理等级",
 			probeTooltipIdle: "检测 {model} 可用的推理档位",
 			probeTooltipVerified: "已接受：{levels} · 点击可重新检测",
-			probeTooltipNotValidating: "该模型不校验该参数",
+			probeTooltipNotValidating: "该模型接受任意档位参数，探测无法区分可用档位 · 点击可重新检测",
 			probeTooltipRetry: "检测未完成 · 点击重试",
 			probeBubbleBody: "发送探测请求以确认可用推理档位。可能消耗少量积分。",
 			probeConfirmAction: "确认检测",
 			probeNoteVerified: "已检测：{levels}",
-			probeNoteNotValidating: "该模型不校验该参数",
+			probeNoteNotValidating: "无法探测档位：上游不校验该参数 —— 这不等于该档位不被支持",
 			probeNoteUnknown: "检测未完成",
 			probeNoteDismiss: "知道了",
 			probeHeading: "推理档位检测",
@@ -3023,7 +3084,8 @@ window.__ModuleLoader__.load({
 			probeConfirmBody: "向 {model} 发送探测请求，以确认可用推理档位。可能消耗少量积分。",
 			cancel: "取消",
 			probeResultVerified: "已验证接受的档位：{levels}",
-			probeResultNotValidating: "该模型不校验该参数",
+			probeResultNotValidating: "上游不校验档位参数",
+			probeDeclaredSuffix: "（目录声明 {levels}）",
 			probeResultUnknown: "检测未完成",
 			probeResultAt: "检测于 {time}",
 			probeResultEmpty: "当前没有可检测的模型。",

package/lib/{host-heartbeat-vU86u9aU.js → host-heartbeat-sjotAEdb.js}

@@ -1,11 +1,226 @@
-import { readFile, rm, writeFile } from "node:fs/promises";
+import { lstat, mkdir, readFile, rename, rm, writeFile } from "node:fs/promises";
 import { homedir, release } from "node:os";
 import { basename, dirname, join } from "node:path";
-import { withFileLock, writeFileAtomic } from "@deepseek-ai/dsh-atomic-write";
-import { resolveDshHome } from "@deepseek-ai/dsh-home-paths";
 import { createDecipheriv, createHash, randomBytes, randomUUID } from "node:crypto";
+import { setTimeout as setTimeout$1 } from "node:timers/promises";
+import { resolveDshHome } from "@deepseek-ai/dsh-home-paths";
 import { execFile, execFileSync } from "node:child_process";
 import { accessSync, constants, readFileSync, readdirSync, realpathSync, statSync } from "node:fs";
+//#region node_modules/.pnpm/@deepseek-ai+dsh-atomic-write@0.2.1-rc.2_@deepseek-ai+cordis@4.0.5/node_modules/@deepseek-ai/dsh-atomic-write/lib/index.js
+/**
+* Zero-dependency atomic file replacement and writer coordination.
+* `writeFileAtomic` writes a random-suffix sibling with exclusive create and
+* the caller's permission bits, then renames it over the target, so readers
+* observe either the old or the new complete content and a replaced file ends
+* up with exactly the stated mode. `withFileLock` serializes cross-process
+* writers of one file through a `wx`-created `<file>.lock` sibling, so a
+* read-modify-write cycle can never resurrect a state another writer just
+* replaced; readers stay lock-free because the rename commit is atomic. A lock
+* whose recorded holder process no longer exists is taken over.
+* @module @deepseek-ai/dsh-atomic-write
+*/
+const WINDOWS_TRANSIENT_RENAME_ERRORS = /* @__PURE__ */ new Set([
+	"EACCES",
+	"EBUSY",
+	"EPERM"
+]);
+const WINDOWS_RENAME_RETRY_INITIAL_MS = 20;
+const WINDOWS_RENAME_RETRY_MAX_MS = 200;
+const WINDOWS_RENAME_RETRY_LIMIT = 8;
+/** Whether Windows reported temporary interference with an atomic replacement. */
+function isTransientWindowsRenameError(error) {
+	if (process.platform !== "win32") return false;
+	return WINDOWS_TRANSIENT_RENAME_ERRORS.has(error?.code ?? "");
+}
+/** Replace the target after bounded retries for transient Windows interference. */
+async function renameAtomicTemp(temp, filename) {
+	let delay = WINDOWS_RENAME_RETRY_INITIAL_MS;
+	for (let retries = 0;; retries += 1) {
+		try {
+			await rename(temp, filename);
+			return;
+		} catch (error) {
+			if (!isTransientWindowsRenameError(error)) throw error;
+			if (retries >= WINDOWS_RENAME_RETRY_LIMIT) throw error;
+		}
+		await new Promise((resolve) => setTimeout(resolve, delay));
+		delay = Math.min(delay * 2, WINDOWS_RENAME_RETRY_MAX_MS);
+	}
+}
+/**
+* Replace `filename` with `content` in one atomic step, creating parent
+* directories. The content is first written to a random-suffix sibling opened
+* with exclusive create (`wx`): the open refuses to follow a symlink planted
+* at the temp path, and the fresh inode carries `options.mode` through the
+* rename, so replacing a wider-permission file narrows it without a chmod
+* race. The rename also replaces a symlinked target itself instead of writing
+* through to its referent, and the same-directory sibling keeps the rename on
+* one filesystem. Windows replacement retries transient `EACCES`, `EBUSY`,
+* and `EPERM` failures for a bounded interval while the complete temp file
+* remains the rename source. On any remaining failure the temp file is
+* removed and the failure rethrown. Crash durability (fsync) is out of scope.
+* @param filename - final path receiving the content.
+* @param content - complete next file content.
+* @param options - permission bits for the replacement inode.
+*/
+async function writeFileAtomic(filename, content, options) {
+	await mkdir(dirname(filename), {
+		recursive: true,
+		...options.dirMode === void 0 ? {} : { mode: options.dirMode }
+	});
+	const temp = `${filename}.${randomBytes(6).toString("hex")}.tmp`;
+	try {
+		await writeFile(temp, content, {
+			mode: options.mode,
+			flag: "wx"
+		});
+		await renameAtomicTemp(temp, filename);
+	} catch (error) {
+		await rm(temp, { force: true });
+		throw error;
+	}
+}
+/** Whether an exclusive create found an existing lock. */
+async function isLockContention(error, lockPath) {
+	const code = error?.code;
+	if (code === "EEXIST") return true;
+	if (code !== "EPERM") return false;
+	try {
+		await lstat(lockPath);
+		return true;
+	} catch {
+		return false;
+	}
+}
+/** Whether the holder a `<pid>\n` record names is proven gone: a signal probe finds no such process. */
+function holderExited(record) {
+	if (!/^\d+\n$/.test(record)) return false;
+	const pid = Number(record.trim());
+	if (pid === 0 || pid > 2147483647) return false;
+	if (pid === process.pid) return false;
+	try {
+		process.kill(pid, 0);
+		return false;
+	} catch (error) {
+		return error.code === "ESRCH";
+	}
+}
+/** The lock file's content, or undefined when it cannot be read. */
+async function readLockRecord(lockPath) {
+	try {
+		return await readFile(lockPath, "utf8");
+	} catch (error) {
+		return;
+	}
+}
+/**
+* Remove the lock when its recorded holder exited. Contenders that read the
+* same record serialize on a claim file named after it. Under the claim, the
+* claimant re-reads the lock and probes its PID again, and removes it only
+* when it still holds that record and that PID is still gone: the record's
+* holder can no longer release it, and no other contender can remove it
+* without the claim, so a removal never deletes a lock another contender
+* acquired after the dead holder's, including one whose holder reused the PID.
+* @returns Whether this call removed the dead holder's lock.
+*/
+async function takeOverExitedLock(lockPath) {
+	const record = await readLockRecord(lockPath);
+	if (record === void 0 || !holderExited(record)) return false;
+	const claim = `${lockPath}.takeover-${createHash("sha256").update(record).digest("hex").slice(0, 16)}`;
+	try {
+		await writeFile(claim, `${process.pid}\n`, {
+			mode: 384,
+			flag: "wx"
+		});
+	} catch (error) {
+		const code = error.code;
+		if (code === "EEXIST" || code === "EPERM") return false;
+		throw error;
+	}
+	try {
+		if (await readLockRecord(lockPath) !== record || !holderExited(record)) return false;
+		try {
+			await rm(lockPath, { force: true });
+		} catch (error) {
+			return false;
+		}
+		return true;
+	} finally {
+		await rm(claim, { force: true }).catch((error) => {});
+	}
+}
+/**
+* Retry cadence for a contended lock. These stay robustness invariants of the
+* cross-process write protocol rather than deployment tunables: they govern how
+* often a contender asks, which no caller has a reason to vary.
+*/
+const LOCK_RETRY_INITIAL_MS = 20;
+const LOCK_RETRY_MAX_MS = 200;
+/**
+* How long a contender waits when the caller states no limit — sized for the
+* render-and-rename cycle every call site had when this package was written.
+* Expiry fails the contender rather than guessing whether the existing lock
+* still has an owner. How long is *worth* waiting is a property of the
+* operation the lock holder runs, which is why {@link FileLockOptions.waitMs}
+* exists; the value here is the floor for an operation that does file work
+* alone.
+*/
+const DEFAULT_LOCK_WAIT_MS = 2e3;
+/**
+* Hold the cross-process writer lock for `filename` around one operation. The
+* lock is a `wx`-created sibling (`<filename>.lock`); paired with the
+* rename-based commit of {@link writeFileAtomic}, readers stay lock-free and
+* only writers contend. `EEXIST` is contention directly; an `EPERM` is
+* contention only when a fresh `lstat` confirms the lock path exists, covering
+* Windows exclusive-create behavior. Windows retries one unconfirmed EPERM
+* because the holder can release before the probe; a repeated unconfirmed
+* permission error is rethrown. The lock records its holder's PID. A contender
+* removes the lock and retries at once when no process with that PID exists
+* (`ESRCH`); any other lock, including one whose holder exists under another
+* user (`EPERM`) or whose record is incomplete, is waited for. Contention backs
+* off exponentially and times out after the deadline. A holder whose PID a
+* live process reused keeps its lock until an operator removes it. Takeover
+* proves only that the recorded process exited: an operation that starts other
+* writers must stop them with it or leave its successor a way to find them.
+* PIDs are compared on the contender's host, so writers on other hosts or in
+* other PID namespaces sharing the file are unsupported and could both hold
+* the lock. The parent directory must exist.
+* @param filename - the file whose writers this lock serializes.
+* @param operation - the read-render-commit cycle to run while holding the lock.
+* @param options - acquisition options; omitted waits {@link DEFAULT_LOCK_WAIT_MS}.
+* @returns the operation's result; the lock releases on both outcomes.
+*/
+async function withFileLock(filename, operation, options) {
+	const lockPath = `${filename}.lock`;
+	const deadline = Date.now() + (options?.waitMs ?? DEFAULT_LOCK_WAIT_MS);
+	let delay = LOCK_RETRY_INITIAL_MS;
+	let retriedUnconfirmedPermissionError = false;
+	for (;;) {
+		options?.signal?.throwIfAborted();
+		try {
+			await writeFile(lockPath, `${process.pid}\n`, {
+				mode: 384,
+				flag: "wx"
+			});
+			break;
+		} catch (error) {
+			if (!await isLockContention(error, lockPath)) {
+				if (process.platform !== "win32" || error?.code !== "EPERM" || retriedUnconfirmedPermissionError) throw error;
+				retriedUnconfirmedPermissionError = true;
+			} else if (await takeOverExitedLock(lockPath)) continue;
+		}
+		if (Date.now() >= deadline) throw new Error(`atomic-write: timed out waiting for the writer lock at ${lockPath}`);
+		await setTimeout$1(delay, void 0, { signal: options?.signal });
+		delay = Math.min(delay * 2, LOCK_RETRY_MAX_MS);
+	}
+	try {
+		options?.signal?.throwIfAborted();
+		return await operation();
+	} finally {
+		await rm(lockPath, { force: true });
+	}
+}
+//#endregion
 //#region src/app-version.ts
 /**
 * The international desktop app's version, used as the `/v3/config` UA.
@@ -396,22 +611,37 @@ function randomSentinel() {
 * - `model_param_invalid` — the global endpoint, measured 2026-10-01. A
 *   non-canonical value is answered `400` / `11133` with this code, which names
 *   no field (`extError.param` is empty), instead of one that names the effort.
-*   It is generic enough to be readable here only because the baseline step has
-*   already proved the *same* request without `reasoning_effort` succeeds,
-*   leaving the sentinel as the only difference between the two attempts. A
-*   sibling code in the same `11133` envelope that names another parameter
-*   (`integer_below_min_value`, the `max_tokens` floor) and a body carrying no
-*   `extError` at all (`11102`, unknown model) are therefore not mistaken for
-*   it, and neither is a level-sweep answer. It is *not* added to `cn`: that
-*   endpoint was measured answering the specific code, and reading a generic
-*   code there would widen attribution beyond what was observed.
+* - `invalid_parameter_value` — both endpoints. Measured on the global
+*   endpoint 2026-10-16 (issue #143); observed on the China endpoint the same
+*   day (`deepseek-v4-pro`, PR #145 discussion) and reproduced there
+*   2026-10-17 on WorkBuddy CN 5.8.2. The *same* `400` / `11133` envelope as
+*   `model_param_invalid`, with `extError.code` respelled and the `msg`
+*   unchanged; a non-canonical value is again what triggers it and the
+*   baseline still answers 200. Kept alongside the old spellings rather than
+*   replacing them: the rename was observed on the gateways, not promised for
+*   every model behind them.
+*
+* The generic codes name no effort field. They are readable here only
+* because the baseline step has already proved the *same* request without
+* `reasoning_effort` succeeds, leaving the sentinel as the only difference
+* between the two attempts — an isolation that holds per endpoint, which is
+* why both regions read the respelled spelling once each was observed on it.
+* A sibling code in the same `11133` envelope that names another parameter
+* (`integer_below_min_value`, the `max_tokens` floor) and a body carrying no
+* `extError` at all (`11102`, unknown model) are therefore not mistaken for
+* them, and neither is a level-sweep answer.
 *
-* Neither is treated as a permanent protocol promise: anything unrecognized
-* still degrades to `unknown` rather than to a capability conclusion.
+* None of these is treated as a permanent protocol promise: anything
+* unrecognized still degrades to `unknown` rather than to a capability
+* conclusion.
 */
 const INVALID_EFFORT_CODES = {
-	cn: /* @__PURE__ */ new Set(["invalid_reasoning_effort"]),
-	global: /* @__PURE__ */ new Set(["invalid_reasoning_effort", "model_param_invalid"])
+	cn: /* @__PURE__ */ new Set(["invalid_reasoning_effort", "invalid_parameter_value"]),
+	global: /* @__PURE__ */ new Set([
+		"invalid_reasoning_effort",
+		"model_param_invalid",
+		"invalid_parameter_value"
+	])
 };
 /** Whether an attempt is an attributable rejection of the effort value. */
 function isEffortRejection(attempt, region) {
@@ -1504,7 +1734,41 @@ function parseModelCatalog(data, international = false, promoBadges) {
 	}
 	const models = cliIds.map((id) => byId.get(id)).filter((model) => model !== void 0);
 	if (models.length === 0) throw new Error("workbuddy model catalog resolved to an empty list");
-	return models;
+	return disambiguateDisplayNames(models);
+}
+/**
+* Make display names unique within one catalog.
+*
+* The CN document names both `hy3` and `hy3-x` "Hy3" — two neighbouring rows
+* whose rates differ (`x0.00` free versus `x0.05`), so the picker and the probe
+* card render two identical rows and the user cannot tell which one they are
+* paying for. The static fallback spells the second one "Hy3-X", but the live
+* fetch overwrote that correction: as soon as the roster gained the id, the
+* collision came back through the parse path.
+*
+* The suffix comes from the id itself — whatever follows the shared name stem
+* — so `hy3-x` reads "Hy3-X" again. Ids are untouched: they are the wire
+* contract the request path and the shim route on. Only exact (case-
+* insensitive) collisions are rewritten; a distinct upstream name such as
+* `MiniMax-M3` for `minimax-m3-pay` keeps the upstream's own wording.
+*/
+function disambiguateDisplayNames(models) {
+	const seen = /* @__PURE__ */ new Set();
+	return models.map((model) => {
+		const key = model.name.trim().toLowerCase();
+		if (!seen.has(key)) {
+			seen.add(key);
+			return model;
+		}
+		const tail = model.id.toLowerCase().startsWith(key) ? model.id.slice(key.length) : "";
+		if (tail === "") return model;
+		const name = `${model.name}${tail.toUpperCase()}`;
+		seen.add(name.toLowerCase());
+		return {
+			...model,
+			name
+		};
+	});
 }
 /** Extract the promotions covering `model` from the `modelPromotions` array. */
 function parsePromotions(value, model) {
@@ -2410,7 +2674,7 @@ function workBuddyWindowsDiscoveryTools() {
 			if (settled) return;
 			settled = true;
 			if (error !== null && error !== void 0) {
-				if (error.killed !== true && (error.code === 1 || error.code === "1") && windowsRegistryKeyMissing(stderr.toString("utf8"))) {
+				if (error.killed !== true && (error.code === 1 || error.code === "1") && windowsRegistryKeyMissingFromBytes(stderr)) {
 					resolve(Buffer.alloc(0));
 					return;
 				}
@@ -2829,6 +3093,24 @@ function windowsRegistryKeyMissing(stderr) {
 	return /^ERROR:\s*The system was unable to find the specified registry key or value\.?$/iu.test(detail) || /^错误[:：]\s*系统找不到指定的注册表项或值[。.]?$/u.test(detail);
 }
 /**
+* Reg.exe speaks the console OEM code page, not UTF-8: on a zh-CN install the
+* missing-key diagnostic arrives as cp936 bytes, which UTF-8 decoding turns
+* into mojibake no diagnostic regex can match — the query then reads as
+* "incomplete" on every absent root. Extension of #88's byte-level principle:
+* the bytes decide, so decode as UTF-8 first and then as GBK. ASCII — every
+* English diagnostic — is identical under both, so a match is unambiguous and
+* recognition only ever widens; a genuine failure (access denied, unknown
+* diagnostics) still stays incomplete in either decode.
+*/
+function windowsRegistryKeyMissingFromBytes(stderr) {
+	if (windowsRegistryKeyMissing(stderr.toString("utf8"))) return true;
+	try {
+		return windowsRegistryKeyMissing(new TextDecoder("gbk").decode(stderr));
+	} catch {
+		return false;
+	}
+}
+/**
 * Parse the value columns emitted by `reg query ... /s`.
 *
 * Entries are filtered by the product's DisplayName pattern *before* the
@@ -3403,12 +3685,12 @@ var WorkBuddyCredentialStore = class {
 */
 /**
 * Static CLI models observed on the CN endpoint (re-verified against the live
-* `/v3/config` document 2026-09-23, including the thinking-effort and billing
+* `/v3/config` document 2026-10-17, including the thinking-effort and billing
 * metadata). The upstream refresh replaces this list at startup; it exists so
 * the provider registers with a usable catalog even while the first fetch is
 * in flight or offline.
 *
-* The list tracks the `cli` agent's model roster exactly — the 16 models it
+* The list tracks the `cli` agent's model roster exactly — the 19 models it
 * offered that day. The roster churns quickly (`auto`, `kimi-k3-1`,
 * `minimax-m3` each appeared or vanished within days, and a competing patch's
 * 2026-09-22 snapshot named three ids that were gone a day later), so this
@@ -3419,6 +3701,11 @@ var WorkBuddyCredentialStore = class {
 * document, and the `free` flag follows the normalized `x0.00` credits
 * marker.
 *
+* Image capability is also verbatim, and it moved since the previous capture:
+* `glm-5.2` and `glm-5.1` now report `supportsImages: false`. Claiming it for
+* them admitted an attachment the upstream then rejected after the message was
+* durable, so this table follows the document rather than the old snapshot.
+*
 * Deliberately NOT baked in: promotional badges. `限时免费` and friends are
 * dynamic console-side promotions with no reliable validity window, so a
 * static table would keep them alive long after the offers end. The live
@@ -3430,7 +3717,7 @@ const FALLBACK_WORKBUDDY_MODELS = [
 	{
 		id: "hy4-preview",
 		name: "Hy4 preview",
-		contextWindow: 1e6,
+		contextWindow: 96e4,
 		maxTokens: 64e3,
 		supportsImages: true,
 		reasoning: {
@@ -3479,6 +3766,52 @@ const FALLBACK_WORKBUDDY_MODELS = [
 			free: false
 		}
 	},
+	{
+		id: "step-5-preview",
+		name: "Step-5-Preview",
+		contextWindow: 1e6,
+		maxTokens: 64e3,
+		supportsImages: true,
+		reasoning: {
+			supports: true,
+			onlyReasoning: true,
+			supportedEfforts: [
+				"low",
+				"medium",
+				"high"
+			],
+			defaultEffort: "high",
+			canDisableThinking: false
+		},
+		billing: {
+			credits: "x0.00 credits",
+			free: true
+		}
+	},
+	{
+		id: "space-bunny",
+		name: "Space-Bunny",
+		contextWindow: 1e6,
+		maxTokens: 128e3,
+		supportsImages: true,
+		reasoning: {
+			supports: true,
+			onlyReasoning: true,
+			supportedEfforts: [
+				"low",
+				"medium",
+				"high",
+				"xhigh",
+				"max"
+			],
+			defaultEffort: "max",
+			canDisableThinking: false
+		},
+		billing: {
+			credits: "x0.08 credits",
+			free: false
+		}
+	},
 	{
 		id: "deepseek-v4.1-flash",
 		name: "Deepseek-V4.1-Flash",
@@ -3488,11 +3821,16 @@ const FALLBACK_WORKBUDDY_MODELS = [
 		reasoning: {
 			supports: true,
 			onlyReasoning: true,
+			supportedEfforts: [
+				"low",
+				"high",
+				"max"
+			],
 			defaultEffort: "high",
-			canDisableThinking: false
+			canDisableThinking: true
 		},
 		billing: {
-			credits: "x0.03 credits",
+			credits: "x0.11 credits",
 			free: false
 		}
 	},
@@ -3505,7 +3843,12 @@ const FALLBACK_WORKBUDDY_MODELS = [
 		reasoning: {
 			supports: true,
 			onlyReasoning: true,
-			defaultEffort: "medium",
+			supportedEfforts: [
+				"low",
+				"high",
+				"max"
+			],
+			defaultEffort: "max",
 			canDisableThinking: false
 		},
 		billing: {
@@ -3528,19 +3871,41 @@ const FALLBACK_WORKBUDDY_MODELS = [
 				"max"
 			],
 			defaultEffort: "high",
-			canDisableThinking: true
+			canDisableThinking: false
 		},
 		billing: {
 			credits: "x0.06 credits",
 			free: false
 		}
 	},
+	{
+		id: "glm-5.3-flashx",
+		name: "GLM-5.3-FlashX",
+		contextWindow: 1e6,
+		maxTokens: 131072,
+		supportsImages: true,
+		reasoning: {
+			supports: true,
+			onlyReasoning: true,
+			supportedEfforts: [
+				"low",
+				"high",
+				"max"
+			],
+			defaultEffort: "high",
+			canDisableThinking: false
+		},
+		billing: {
+			credits: "x0.14 credits",
+			free: false
+		}
+	},
 	{
 		id: "glm-5.2",
 		name: "GLM-5.2",
 		contextWindow: 1e6,
 		maxTokens: 64e3,
-		supportsImages: true,
+		supportsImages: false,
 		reasoning: {
 			supports: true,
 			onlyReasoning: true,
@@ -3557,7 +3922,7 @@ const FALLBACK_WORKBUDDY_MODELS = [
 		name: "GLM-5.1",
 		contextWindow: 2e5,
 		maxTokens: 48e3,
-		supportsImages: true,
+		supportsImages: false,
 		reasoning: {
 			supports: true,
 			onlyReasoning: true,
@@ -3587,10 +3952,10 @@ const FALLBACK_WORKBUDDY_MODELS = [
 		}
 	},
 	{
-		id: "minimax-m3",
+		id: "minimax-m3-pay",
 		name: "MiniMax-M3",
 		contextWindow: 512e3,
-		maxTokens: 64e3,
+		maxTokens: 128e3,
 		supportsImages: true,
 		reasoning: {
 			supports: true,
@@ -3616,12 +3981,12 @@ const FALLBACK_WORKBUDDY_MODELS = [
 			canDisableThinking: false
 		},
 		billing: {
-			credits: "x0.19 credits",
+			credits: "x0.26 credits",
 			free: false
 		}
 	},
 	{
-		id: "kimi-k3-1",
+		id: "kimi-k3-2",
 		name: "Kimi-K3",
 		contextWindow: 1e6,
 		maxTokens: 32e3,
@@ -3633,7 +3998,7 @@ const FALLBACK_WORKBUDDY_MODELS = [
 			canDisableThinking: false
 		},
 		billing: {
-			credits: "x1.62 credits",
+			credits: "x1.62",
 			free: false
 		}
 	},
@@ -4177,7 +4542,7 @@ var WorkBuddyCatalog = class {
 };
 //#endregion
 //#region src/version.ts
-const WORKBUDDY_CONNECT_VERSION = "0.7.6";
+const WORKBUDDY_CONNECT_VERSION = "0.7.8";
 //#endregion
 //#region src/host-heartbeat.ts
 /**

package/lib/index.d.ts

@@ -1129,12 +1129,12 @@ declare class WorkBuddyCredentialStore {
 type WorkBuddyModelInfo = WorkBuddyUpstreamModel;
 /**
  * Static CLI models observed on the CN endpoint (re-verified against the live
- * `/v3/config` document 2026-09-23, including the thinking-effort and billing
+ * `/v3/config` document 2026-10-17, including the thinking-effort and billing
  * metadata). The upstream refresh replaces this list at startup; it exists so
  * the provider registers with a usable catalog even while the first fetch is
  * in flight or offline.
  *
- * The list tracks the `cli` agent's model roster exactly — the 16 models it
+ * The list tracks the `cli` agent's model roster exactly — the 19 models it
  * offered that day. The roster churns quickly (`auto`, `kimi-k3-1`,
  * `minimax-m3` each appeared or vanished within days, and a competing patch's
  * 2026-09-22 snapshot named three ids that were gone a day later), so this
@@ -1145,6 +1145,11 @@ type WorkBuddyModelInfo = WorkBuddyUpstreamModel;
  * document, and the `free` flag follows the normalized `x0.00` credits
  * marker.
  *
+ * Image capability is also verbatim, and it moved since the previous capture:
+ * `glm-5.2` and `glm-5.1` now report `supportsImages: false`. Claiming it for
+ * them admitted an attachment the upstream then rejected after the message was
+ * durable, so this table follows the document rather than the old snapshot.
+ *
  * Deliberately NOT baked in: promotional badges. `限时免费` and friends are
  * dynamic console-side promotions with no reliable validity window, so a
  * static table would keep them alive long after the offers end. The live

package/lib/index.js

@@ -1,9 +1,9 @@
-import { $ as validAppVersion, A as extractDisplayErrorMessage, B as randomSentinel, C as WORKBUDDY_VARIANTS, D as WORKBUDDY_UPDATE_PATH, F as prepareCnChatBody, G as readCliVersion, H as FALLBACK_CN_APP_VERSION, I as prepareInternationalChatBody, J as WORKBUDDY_APP_VERSION_FILENAME, K as resolveChatIdentity, L as regionOf, M as normalizeCredits, N as parseModelCatalog, O as WorkBuddyUpstreamClient, P as prepareChatBody, Q as resolveAppVersion, R as PROBE_EFFORT_CANDIDATES, S as CN_VARIANT, U as chatUserAgent, V as CN_APP_VERSION_FILENAME, W as fallbackChatIdentity, X as installedAppVersion, Y as appUserAgent, Z as readBundleVersion, _ as desktopAuthCandidatesFor, a as readHostHeartbeat, b as atRestKeyProviderFor, c as WORKBUDDY_CONNECT_VERSION, d as WorkBuddyCatalog, f as WORKBUDDY_AUTH_FILENAME, g as defaultDesktopAuthPath, h as defaultDesktopAuthCandidates, i as processStartTimeMs, j as modelWithCurrentPromotion, k as classifyUpstreamError, l as FALLBACK_WORKBUDDY_AI_MODELS, m as WorkBuddyCredentialStore, n as clearHostHeartbeat, o as workbuddyHostHeartbeatPath, p as WORKBUDDY_AUTH_FILE_ENV, q as validCliVersion, r as isHeartbeatProcessAlive, s as writeHostHeartbeat, t as WORKBUDDY_HOST_HEARTBEAT_FILENAME, u as FALLBACK_WORKBUDDY_MODELS, v as parseWorkBuddyAuth, w as variantFor, x as AI_VARIANT, y as workbuddyOwnAuthPath, z as probeModel } from "./host-heartbeat-vU86u9aU.js";
+import { $ as validAppVersion, A as extractDisplayErrorMessage, B as randomSentinel, C as WORKBUDDY_VARIANTS, D as WORKBUDDY_UPDATE_PATH, F as prepareCnChatBody, G as readCliVersion, H as FALLBACK_CN_APP_VERSION, I as prepareInternationalChatBody, J as WORKBUDDY_APP_VERSION_FILENAME, K as resolveChatIdentity, L as regionOf, M as normalizeCredits, N as parseModelCatalog, O as WorkBuddyUpstreamClient, P as prepareChatBody, Q as resolveAppVersion, R as PROBE_EFFORT_CANDIDATES, S as CN_VARIANT, U as chatUserAgent, V as CN_APP_VERSION_FILENAME, W as fallbackChatIdentity, X as installedAppVersion, Y as appUserAgent, Z as readBundleVersion, _ as desktopAuthCandidatesFor, a as readHostHeartbeat, b as atRestKeyProviderFor, c as WORKBUDDY_CONNECT_VERSION, d as WorkBuddyCatalog, f as WORKBUDDY_AUTH_FILENAME, g as defaultDesktopAuthPath, h as defaultDesktopAuthCandidates, i as processStartTimeMs, j as modelWithCurrentPromotion, k as classifyUpstreamError, l as FALLBACK_WORKBUDDY_AI_MODELS, m as WorkBuddyCredentialStore, n as clearHostHeartbeat, o as workbuddyHostHeartbeatPath, p as WORKBUDDY_AUTH_FILE_ENV, q as validCliVersion, r as isHeartbeatProcessAlive, s as writeHostHeartbeat, t as WORKBUDDY_HOST_HEARTBEAT_FILENAME, u as FALLBACK_WORKBUDDY_MODELS, v as parseWorkBuddyAuth, w as variantFor, x as AI_VARIANT, y as workbuddyOwnAuthPath, z as probeModel } from "./host-heartbeat-sjotAEdb.js";
 import z from "@deepseek-ai/schemastery";
 import { resolveImageAttachmentAccess, resolveRetryPolicy } from "@deepseek-ai/dsh-llm";
 import { dirname, join, resolve } from "node:path";
-import { resolveDshHome } from "@deepseek-ai/dsh-home-paths";
 import { createHash, randomBytes, timingSafeEqual } from "node:crypto";
+import { resolveDshHome } from "@deepseek-ai/dsh-home-paths";
 import { existsSync, mkdirSync, readFileSync, renameSync, writeFileSync } from "node:fs";
 import { createProvider } from "@earendil-works/pi-ai";
 import { openAICompletionsApi } from "@earendil-works/pi-ai/api/openai-completions.lazy";
@@ -1159,7 +1159,7 @@ var WorkBuddyProbeService = class {
 //#region src/pi-ai-guard.ts
 /**
 * Confirm the pi-ai copy this plugin actually loaded is the generation the DSH
-* 0.2.0-rc.2 host was built against.
+* 0.2.1-rc.1 host was built against.
 *
 * #92's failure mode: a pnpm profile with `nodeLinker: hoisted` and
 * `autoInstallPeers: false` installs no peer for us, so a co-installed
@@ -1178,7 +1178,7 @@ var WorkBuddyProbeService = class {
 * @module dsh-workbuddy-connect/pi-ai-guard
 */
 /** The pi-ai generation this plugin is built and tested against. */
-const REQUIRED_PI_AI_VERSION = "0.87.1";
+const REQUIRED_PI_AI_VERSION = "1.3.0";
 /**
 * Read the version of the pi-ai copy Node resolves for *this* module.
 *
@@ -1225,7 +1225,7 @@ function resolvedPiAiVersion() {
 */
 function piAiVersionWarning(resolved) {
 	const version = arguments.length === 0 ? resolvedPiAiVersion() : resolved;
-	if (version === void 0 || version === "0.87.1") return void 0;
+	if (version === void 0 || version === "1.3.0") return void 0;
 	return `loaded @earendil-works/pi-ai ${version}, but this plugin is built for ${REQUIRED_PI_AI_VERSION}. The wrong copy is being resolved — check the profile's pi-ai install (a pnpm profile with nodeLinker "hoisted" and autoInstallPeers false can expose another package's older copy). Reinstalling the plugin so it installs its own ${REQUIRED_PI_AI_VERSION} copy resolves it`;
 }
 //#endregion
@@ -1282,6 +1282,8 @@ async function workBuddyWebStatus(deps) {
 			...model.billing?.badges !== void 0 && model.billing.badges.length > 0 ? { badges: model.billing.badges } : {},
 			...rate === void 0 ? {} : { credits: rate },
 			...model.billing?.rateUnknown === true ? { rateUnknown: true } : {},
+			...(model.reasoning?.supportedEfforts?.length ?? 0) > 0 ? { declaredEfforts: model.reasoning.supportedEfforts } : {},
+			...model.reasoning?.defaultEffort !== void 0 ? { defaultEffort: model.reasoning.defaultEffort } : {},
 			...typeof model.contextWindow === "number" && model.contextWindow > 0 ? { contextWindow: model.contextWindow } : {},
 			...typeof defaultContextWindow === "number" && defaultContextWindow > 0 && defaultContextWindow < model.contextWindow ? { defaultContextWindow } : {},
 			...maxContextWindow === void 0 || maxContextWindow <= defaultContextWindow ? {} : { maxContextWindow },

package/package.json

@@ -2,7 +2,7 @@
   "name": "dsh-workbuddy-connect",
   "displayName": "DSH WorkBuddy Connect",
   "description": "将 WorkBuddy 桌面 App 包含的模型自动接入 DeepSeek Harness — bring WorkBuddy desktop app models into DeepSeek Harness with zero configuration.",
-  "version": "0.7.6",
+  "version": "0.7.8",
   "author": "corrinehu",
   "keywords": [
     "dsh-plugin",
@@ -41,7 +41,8 @@
     "cordis.patch.yml",
     "README.md",
     "README.en.md",
-    "LICENSE"
+    "LICENSE",
+    "THIRD_PARTY_NOTICES.txt"
   ],
   "license": "MIT",
   "repository": {
@@ -67,42 +68,41 @@
     }
   },
   "dependencies": {
-    "@earendil-works/pi-ai": "0.87.1"
+    "@earendil-works/pi-ai": "1.3.0"
   },
   "peerDependencies": {
-    "@deepseek-ai/cordis": "^4.0.2",
-    "@deepseek-ai/dsh-atomic-write": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-attachment": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-home-paths": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-host-webserver": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-llm": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-llm-pi-ai": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-settings": "0.2.0-rc.2",
-    "@deepseek-ai/schemastery": "^3.18.2",
+    "@deepseek-ai/cordis": "^4.0.5",
+    "@deepseek-ai/dsh-attachment": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-home-paths": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-host-webserver": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-llm": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-llm-pi-ai": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-settings": "0.2.1-rc.2",
+    "@deepseek-ai/schemastery": "^3.18.5",
     "react": "^18.2.0"
   },
   "devDependencies": {
-    "@deepseek-ai/cordis": "^4.0.4",
-    "@deepseek-ai/dsh-atomic-write": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-attachment": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-client-locale": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-client-ui-conversation": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-client-ui-model-selection": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-client-ui-plugin-manager": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-client-ui-settings-plugins": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-client-ui-renderer": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-client-ui-slots": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-credentials": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-authorization": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-fs": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-launch-environment": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-home-paths": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-host-webserver": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-llm": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-llm-pi-ai": "0.2.0-rc.2",
-    "@deepseek-ai/dsh-settings": "0.2.0-rc.2",
-    "@deepseek-ai/schemastery": "^3.18.2",
-    "@earendil-works/pi-ai": "0.87.1",
+    "@deepseek-ai/cordis": "^4.0.5",
+    "@deepseek-ai/dsh-atomic-write": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-attachment": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-client-locale": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-client-ui-conversation": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-client-ui-model-selection": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-client-ui-plugin-manager": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-client-ui-settings-plugins": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-client-ui-renderer": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-client-ui-slots": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-credentials": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-authorization": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-fs": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-launch-environment": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-home-paths": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-host-webserver": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-llm": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-llm-pi-ai": "0.2.1-rc.2",
+    "@deepseek-ai/dsh-settings": "0.2.1-rc.2",
+    "@deepseek-ai/schemastery": "^3.18.5",
+    "@earendil-works/pi-ai": "1.3.0",
     "@types/node": "^22.20.0",
     "@types/react": "~18.3.1",
     "@types/react-test-renderer": "^18.3.1",