dsh-voice-mode 0.7.19 → 0.7.21

package/README.en.md

@@ -16,7 +16,7 @@ interrupts playback and the running turn. No API key.
 
 ![Real recording: streaming transcription → auto-send → sentence-by-sentence read-aloud with live captions](https://raw.githubusercontent.com/qishuilalala/dsh-voice-mode/HEAD/plugin/dsh-voice-mode/assets/demo-voice-flow.gif)
 
-> **Runtime changes (most recent batch: v0.7.10, 2026-09-18)**: **Silent-audio causes in the output pipeline fixed** — ① a sentence whose cloud-TTS synthesis fails after 3 retries is no longer dropped silently (the status bar now says one sentence failed and was skipped; previously this showed up as "the reply is occasionally not read" with no way to tell why); ② audio playback no longer goes silent after the browser suspends the AudioContext (background tab / long idle) — the context is resumed on every enqueue and on any click/keypress; ③ dropped incomplete sentences (SSE frame loss) now leave a diagnostic trace. Otherwise as v0.7.9: **Wake-word pipeline overhaul** (Issue #10 + real-machine retest) — a harness driving the real engine pinpointed and fixed five flow defects: ① TTS echo polluting the standby segment while the agent reads (could not wake); ② only above-threshold frames uploaded, so trailing characters never flushed (wake word truncated); ③ discarding the whole segment on a wake hit (saying "wake word + command" in one breath sent only the tail); ④ the command hanging when the hit arrived after you stopped speaking; ⑤ pausing after the wake word closed the command window (command lost). **The wake word may now be said together with your command — it is stripped and never sent**, standby live-shows what it heard, and "wake word … pause … command" works. Otherwise as v0.7.7: Edge cloud TTS by default, local TTS (VITS / Kokoro) optional; silence split defaults to 1500 ms.
+> **Runtime changes (most recent batch: v0.7.17 – v0.7.19, 2026-10-02)**: (1) **Settings work on every dsh version** — fixed the settings page and persistence on dsh 0.1.7+ (dedicated Settings → Voice Mode page, writes survive restarts); (2) **i18n reworked** — the UI follows dsh's official language setting, Chinese/English are fully consistent and switch without a reload; (3) **fixed fresh-install failures with npm/pnpm and `failed to import` on dsh ≥0.1.7** (affects 0.7.17 and earlier — please upgrade to ≥0.7.18); (4) **local Kokoro fixed on the official Electron desktop host** — needs Node.js ≥18 on PATH (see the desktop note below). Earlier batches (v0.7.10: silent-audio fixes, etc.) and the full history are in [`CHANGELOG.md`](https://github.com/qishuilalala/dsh-voice-mode/blob/main/CHANGELOG.md).
 
 ---
 
@@ -303,6 +303,23 @@ flowchart LR
 
 ---
 
+## 🔐 Permissions and data flow
+
+An honest disclosure of what the plugin does (mapped to the awesome-dsh-plugin capability scan: `network / fs-read / fs-write / shell / env`):
+
+| Category | What it does | Where data goes |
+| --- | --- | --- |
+| **network** | (1) The default read-aloud engine **Edge** sends the **reply text to be spoken** over WebSocket to Microsoft `speech.platform.bing.com` (cloud synthesis — switch to local VITS / Kokoro to stay fully offline); (2) downloads local models on first use from `huggingface.co` (or your configured mirror `hf-mirror.com`) with a host allowlist and pinned SHA256; (3) registers loopback HTTP routes `/voice-mode/*` on the dsh host (loopback only by default; LAN needs `allowLan`) | Microphone audio **never leaves your machine** (recognition runs locally via sherpa-onnx); only reply text goes to Microsoft, and only with the Edge engine |
+| **fs-read / fs-write** | Model cache (Linux/macOS `~/.cache/dsh-voice-mode/models/`, Windows `%LOCALAPPDATA%\dsh-voice-mode\models\`); the dsh 0.1.7+ settings overlay `$DSH_HOME/voice-mode.settings.json`; **never touches your workspace files** | All local |
+| **shell (child processes)** | Local TTS synthesis runs in a separate child process (`child_process.fork` → `lib/tts-vits-worker.cjs`); when the host is Electron (the official desktop app), it additionally probes and launches a real Node.js for Kokoro (`node -p …` to check the version). **It never executes user input and never builds shell command strings** | Local |
+| **env** | Reads only `DSH_HOME`, `LOCALAPPDATA` (for directories) and the optional `DSHVM_NODE`; **no API key is read or needed** | — |
+
+Also: the on-device recording fixture used for debugging is **off by default**; the plugin sends no telemetry. See [`SECURITY.md`](https://github.com/qishuilalala/dsh-voice-mode/blob/main/SECURITY.md) for the threat model and how to report vulnerabilities.
+
+### Compatibility declaration (what the plugin market's version filter reads)
+
+`package.json` declares `engines.dsh = ">=0.1.1-rc.2"` (the market uses it for "filter by host version"). Verified range: **0.1.1-rc.2 → 0.2.1-alpha.1** (including an Electron host equivalent to the official desktop app). The upper bound is open; a weekly CI check watches upstream releases and opens an issue for any uncovered version. `peerDependencies` only lists `@deepseek-ai/cordis` and `react` (provided by the host).
+
 ## 🚧 Known limitations
 
 - Barge-in relies on browser echo cancellation (`echoCancellation`); loud speaker volume may leak into the mic (no JS-level AEC)

package/README.md

@@ -17,7 +17,7 @@ DeepSeek Harness 语音双工对话模式：会话内一键进入 → 边说边
 
 ![全双工对话闭环：声音 → 文字 → 声音](https://raw.githubusercontent.com/qishuilalala/dsh-voice-mode/HEAD/plugin/dsh-voice-mode/assets/duplex-banner.png)
 
-> **运行时变更（最近一批：v0.7.10，2026-09-18）**：**输出链路静默丢音根治**——① 云端 TTS 单句重试耗尽后不再静默跳句（现在状态条提示「有一句朗读失败，已跳过」，此前表现为「AI 回复偶尔不朗读」且无从判断）；② 浏览器挂起 AudioContext 后的「无声播放」（UI 显示朗读中、字幕照走）现在每次入队 + 任意点击/按键自动恢复；③ SSE 丢帧导致的坏句丢弃留诊断痕迹。其余同 v0.7.9：**唤醒词链路全面修复**（Issue #10 + 真机复测）——台架驱动真引擎逐项定位并修掉五处流程缺陷：① 朗读期 TTS 回声污染待机段（喊不醒）；② 只上传超门限帧导致尾字不 flush（唤醒词只剩半截）；③ 唤醒命中丢整段（连说「唤醒词+命令」只发出去尾部几个字）；④ 命中晚于停口时命令悬挂不定稿；⑤ 只喊唤醒词后停顿会关掉命令窗口（命令丢失）。**现在唤醒词可与命令连说、词头自动剥离不进消息**，待机态实时显示「它听到了什么」，支持「唤醒词…停顿…命令」。其余同 v0.7.7：朗读默认 Edge 云端，本地 TTS（VITS / Kokoro）可选；静音断句默认 1500 毫秒。
+> **运行时变更（最近一批：v0.7.17 ~ v0.7.19，2026-10-02）**：① **设置面全版本打通**——dsh 0.1.7+ 上设置页与持久化修复（Settings → 语音模式 专属页、写入与重启保留）；② **国际化重做**——界面语言跟随 dsh 官方语言设置，中/英文完整一致、切换无需刷新；③ **修复 npm/pnpm 全新安装失败与 dsh ≥0.1.7 已安装形态「failed to import」**（0.7.17 及更早版本受影响，请升级到 ≥0.7.18）；④ **官方桌面端（Electron）下本地 Kokoro 修复**——需系统 PATH 上有 Node.js ≥18（见下文「桌面端提示」）。此前批次（v0.7.10：输出链路静默丢音根治等）与完整历史见 [`CHANGELOG.md`](https://github.com/qishuilalala/dsh-voice-mode/blob/main/CHANGELOG.md)。
 
 ---
 
@@ -258,10 +258,27 @@ flowchart LR
 | 热词偏置 | 无 | 无（已移除，详见 v0.7.7 文档说明） |
 | 字幕 a11y | 无 | **4 档字号 + 3 档宽度 + 主题跟随** |
 | 唤醒词 | 无 | **轻量流式匹配 + 前缀语气词白名单** |
-| 兼容 dsh | — | **0.1.1-rc.2 起全版本（含 0.2.0-rc.2）** |
+| 兼容 dsh | — | **0.1.1-rc.2 起全版本（含 0.2.1-alpha.1）** |
 
 ---
 
+## 🔐 权限与数据流声明
+
+插件如实声明它会做的事（与 awesome-dsh-plugin 的能力扫描 `network / fs-read / fs-write / shell / env` 一一对应）：
+
+| 类别 | 具体行为 | 数据去向 |
+| --- | --- | --- |
+| **network** | ① 默认朗读引擎 **Edge**：把**待朗读的回复文本**经 WebSocket 发给微软 `speech.platform.bing.com`（云端合成；想完全离线请切到本地 VITS / Kokoro）；② 首次使用本地模型时从 `huggingface.co`（或你配置的镜像 `hf-mirror.com`）下载，主机白名单 + 固定 SHA256 校验；③ 在 dsh 宿主上注册回环 HTTP 路由 `/voice-mode/*`（默认仅回环，需 `allowLan` 才对局域网开放） | 麦克风音频**不出本机**（识别在本地 sherpa-onnx）；仅回复文本在 Edge 引擎下发往微软 |
+| **fs-read / fs-write** | 模型缓存（Linux/macOS `~/.cache/dsh-voice-mode/models/`，Windows `%LOCALAPPDATA%\dsh-voice-mode\models\`）；dsh 0.1.7+ 的设置覆盖层 `$DSH_HOME/voice-mode.settings.json`；**不读写你的工作区文件** | 全部在本机 |
+| **shell（子进程）** | 本地 TTS 合成跑在独立子进程（`child_process.fork` → `lib/tts-vits-worker.cjs`）；宿主为 Electron（官方桌面端）时，为 Kokoro 额外探测并启动一个真正的 Node.js（`node -p …` 探测版本，见 ADR/兼容契约）。**不执行用户输入、不经 shell 拼接命令** | 本机 |
+| **env** | 仅读取 `DSH_HOME`、`LOCALAPPDATA`（取目录）与可选的 `DSHVM_NODE`；**不读取、不需要任何 API Key** | — |
+
+补充：调试用的真机录制（fixture）默认**关闭**；插件不上传遥测。完整威胁模型与漏洞报告方式见仓库根目录 [`SECURITY.md`](https://github.com/qishuilalala/dsh-voice-mode/blob/main/SECURITY.md)。
+
+### 兼容声明（插件市场的版本筛选依据）
+
+`package.json` 声明 `engines.dsh = ">=0.1.1-rc.2"`（插件市场据此在「按宿主版本筛选」时判定兼容）。已实测范围 **0.1.1-rc.2 → 0.2.1-alpha.1**（含官方桌面端等价的 Electron 宿主）；上限开放，由每周 CI 巡检上游新版本并复验，出现未覆盖版本会自动开 issue。`peerDependencies` 仅声明 `@deepseek-ai/cordis` 与 `react`（宿主提供）。
+
 ## 🚧 已知限制
 
 - 发声打断依赖浏览器回声消除（`echoCancellation`）；扬声器音量过大时可能漏声到麦克风

package/lib/client.js

@@ -3099,7 +3099,7 @@ var TELEMETRY_VIEW = [
   { stage: "first-tts-chunk", key: "telFirstChunk" },
   { stage: "first-audio-played", key: "telFirstPlayed" }
 ];
-var BUILD_TAG = "62781d5";
+var BUILD_TAG = "b51f379";
 var TELEMETRY_FLAG = "dsh-voice-mode.telemetry";
 var telemetryEnabled = typeof localStorage !== "undefined" && localStorage.getItem(TELEMETRY_FLAG) === "1";
 console.log("[dsh-voice] build=" + BUILD_TAG);

package/package.json

@@ -2,7 +2,7 @@
   "name": "dsh-voice-mode",
   "description": "Full-duplex voice plugin for DeepSeek Harness: local zipformer2 streaming ASR (no API key) → editable draft; Edge TTS or local VITS / Kokoro read-aloud with live captions; true barge-in; hardened HTTP surface + model SHA256 pinning; compatible with all dsh releases since 0.1.1-rc.2 (verified 0.1.1-rc.2 → 0.2.0-rc.2).",
   "icon": "./icon.svg",
-  "version": "0.7.19",
+  "version": "0.7.21",
   "type": "module",
   "main": "lib/index.js",
   "repository": {
@@ -60,9 +60,7 @@
     "typecheck:dual": "bash scripts/typecheck-dual.sh",
     "smoke:runtime": "bash scripts/smoke-runtime.sh",
     "verify:dual": "bash scripts/verify-dual.sh",
-    "check:dsh-version": "bash scripts/check-dsh-version.sh",
-    "e2e:full": "bash scripts/full-e2e.sh",
-    "release": "bash scripts/release.sh"
+    "e2e:full": "bash scripts/full-e2e.sh"
   },
   "dsh": {
     "bundle": {