dsh-dafeiyu-emotion 0.2.0 → 0.2.2

package/CHANGELOG.md

@@ -1,5 +1,23 @@
 # Changelog
 
+## 0.2.2
+
+- Fixed a duplicate-send race between auto-send and the manual send button.
+- Cancelled stale deliveries when the session changes or the component unmounts.
+- Stopped overwriting a busy composer when `insertText` is refused; send failures now keep the analysis for retry.
+- Added a cancel action and timeout while waiting for microphone permission.
+- Made prosody extraction yield to the UI thread instead of blocking on long recordings.
+- Corrected the base64 request-body limit and mapped validation failures to 400/413 instead of 500.
+- Pinned speech-provider readiness checks to the provider actually used for transcription.
+- Reported missing enhanced backends in health and preserved the real worker failure reason.
+- Added regression tests for delivery gating, shortcut availability and the analyze handler.
+
+## 0.2.1
+
+- Added a native DSH voice shortcut, configurable from DSH keyboard settings.
+- The shortcut starts recording when idle and stops/analyzes recording on a second press.
+- Refreshed the project icon and README key visual with the blue whale-girl character artwork.
+
 ## 0.2.0
 
 - Added a lightweight mode using DSH Voice Input and browser-side prosody analysis.

package/README.md

@@ -1,6 +1,6 @@
 <div align="center">
 
-<img src="icon.svg" width="128" alt="BigFish Emotion Ear icon">
+<img src="assets/icon-512.png" width="128" alt="大肥鱼 · 情绪耳朵插件图标">
 
 # 大肥鱼 · 情绪耳朵
 
@@ -9,6 +9,7 @@
 本地语音转写 · 语气/情绪分析 · 情绪化回复 · 自动发送 · 音频不上传
 
 [![1024Store](https://img.shields.io/badge/1024Store-已收录-0ea5e9?style=flat-square)](https://deepseek1024.com/plugins/LiDu-china/dsh-dafeiyu-emotion)
+[![npm](https://img.shields.io/npm/v/dsh-dafeiyu-emotion?style=flat-square)](https://www.npmjs.com/package/dsh-dafeiyu-emotion)
 [![Release](https://img.shields.io/github/v/release/LiDu-china/dsh-dafeiyu-emotion?style=flat-square)](https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/latest)
 [![GitHub](https://img.shields.io/badge/GitHub-LiDu--china-181717?style=flat-square&logo=github)](https://github.com/LiDu-china/dsh-dafeiyu-emotion)
 [![license](https://img.shields.io/badge/license-MIT-blue?style=flat-square)](LICENSE)
@@ -17,9 +18,11 @@
 
 </div>
 
-![BigFish Emotion Ear preview](docs/preview.png)
+![大肥鱼 · 情绪耳朵角色主视觉](docs/preview.png)
 
 > 预览为功能示意图，实际界面以当前 DSH 版本为准。
+>
+> 角色主视觉采用蓝发鲸鱼少女的大肥鱼设定。
 
 ---
 
@@ -64,6 +67,7 @@
 ## 功能
 
 - 🎙️ 输入框内录音，不需要跳到另一个应用；
+- ⌨️ DSH 原生语音快捷键，可在快捷键设置中自行改键；
 - 🐋 大肥鱼主题的“鱼耳”入口，和普通对话共用同一个 composer；
 - 🧠 情绪标签：喜悦 / 低落 / 生气 / 紧张 / 惊讶 / 厌烦 / 平静；
 - 📈 额外给出 confidence、valence、arousal 和语气摘要；
@@ -84,26 +88,53 @@
 dsh-dafeiyu-emotion
 ```
 
-在 npm 正式包发布前，1024Store 条目为浏览态；发布后会自动出现 npm 一键安装入口，无需重新投稿。现阶段请使用下面的 GitHub 源安装。
+`dsh-dafeiyu-emotion` 已发布到 npm。1024Store 会自动刷新一键安装入口；如果客户端尚未缓存到新版本，可以直接使用下面的 npm 命令。
 
-### 通过 GitHub 源安装
+> `0.2.2` 目前通过 GitHub Release 提供；npm 上的最新版仍是 `0.2.1`，待 npm 发布完成后一键安装会自动升级。
+
+### 通过 npm
 
 ```bash
-dsh plugin --profile desktop add github:LiDu-china/dsh-dafeiyu-emotion
+dsh plugin --profile desktop add dsh-dafeiyu-emotion
 ```
 
 DSH Web 使用 `--profile web`。安装后启用 `dsh-dafeiyu-emotion`，再按提示启用 DSH 自带 **Voice Input**（轻量模式），或配置本地增强后端。
 
+### 通过 GitHub 源安装
+
+```bash
+dsh plugin --profile desktop add github:LiDu-china/dsh-dafeiyu-emotion
+```
+
 ### 通过 Release tarball
 
 ```bash
-curl -L -o dsh-dafeiyu-emotion-0.2.0.tgz \
-  https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/download/v0.2.0/dsh-dafeiyu-emotion-0.2.0.tgz
-dsh plugin --profile desktop add "D:/path/to/dsh-dafeiyu-emotion-0.2.0.tgz"
+curl -L -o dsh-dafeiyu-emotion-0.2.2.tgz \
+  https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/download/v0.2.2/dsh-dafeiyu-emotion-0.2.2.tgz
+dsh plugin --profile desktop add "D:/path/to/dsh-dafeiyu-emotion-0.2.2.tgz"
 ```
 
 > DSH Desktop 的 profile 由 Electron 应用管理；命令行安装请使用桌面应用自带的 `dsh` 命令。
 
+## 语音快捷键
+
+插件会向 DSH 注册一个原生快捷键命令。打开：
+
+```text
+设置 → 快捷键 → 搜索“大肥鱼语音”
+```
+
+即可改成你习惯的组合。默认值：
+
+| 客户端 | 默认快捷键 |
+|---|---|
+| DSH Desktop（Windows / Linux / macOS） | `Alt+V` |
+| DSH Web（Windows） | `Ctrl+Alt+V` |
+| DSH Web（macOS） | `Cmd+Option+V` |
+| DSH Web（Linux） | 默认未绑定，请在快捷键设置中指定 |
+
+空闲时按一次开始录音；录音中再按一次会结束并分析。快捷键需要包含 `Ctrl`、`Alt`、`Shift` 或 `Cmd` 等修饰键，不能设置成裸字母键。
+
 ## 启用 Voice Input（轻量模式）
 
 1. 打开 DSH 左侧 **插件**；
@@ -213,8 +244,9 @@ dsh-dafeiyu-emotion/
 ├─ lib/client.js             浏览器：录音、WAV、韵律特征、情绪 UI、自动发送
 ├─ backend/voice_worker.py   可选 VoiceSense 常驻 worker
 ├─ cordis.patch.yml          bundle 挂载：大肥鱼 + 情绪耳朵
-├─ docs/preview.svg          功能示意图
-├─ icon.svg                  插件图标
+├─ assets/icon-512.png       插件图标
+├─ assets/bigfish-*.png      角色主视觉素材
+├─ docs/preview.svg          可编辑预览源
 └─ package.json
 ```
 
@@ -250,6 +282,7 @@ dsh-dafeiyu-emotion/
 ### Highlights
 
 - One-click microphone activity inside the DSH composer;
+- A native DSH voice shortcut you can rebind in Settings → Shortcuts;
 - Local transcription via DSH Voice Input/SenseVoice;
 - Lightweight prosody + keyword emotion estimation with no Python requirement;
 - Optional high-accuracy VoiceSense backend with SenseVoice + emotion2vec;
@@ -259,13 +292,15 @@ dsh-dafeiyu-emotion/
 ### Install
 
 ```bash
-dsh plugin --profile desktop add github:LiDu-china/dsh-dafeiyu-emotion
+dsh plugin --profile desktop add dsh-dafeiyu-emotion
 ```
 
-The plugin is listed in [1024Store](https://deepseek1024.com/plugins/LiDu-china/dsh-dafeiyu-emotion). Until the first npm release it is browse-only; publishing to npm adds the one-click install automatically. Use the GitHub-source command above, or the tarball attached to [GitHub Releases](https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/latest), today.
+The package is live on [npm](https://www.npmjs.com/package/dsh-dafeiyu-emotion) and listed in [1024Store](https://deepseek1024.com/plugins/LiDu-china/dsh-dafeiyu-emotion). The store's one-click entry refreshes automatically. A GitHub-source install and a tarball on [GitHub Releases](https://github.com/LiDu-china/dsh-dafeiyu-emotion/releases/latest) are also available.
 
 Then enable **Voice Input** in DSH Plugins and prepare its local model. For the enhanced mode, set `backendRoot` to a voice-sense checkout with `.venv-ser`.
 
+The shortcut command is listed as **BigFish voice: start / stop**. Desktop defaults to `Alt+V`; Web defaults to `Ctrl+Alt+V` on Windows, `Cmd+Option+V` on macOS, and is unbound on Linux Web until you choose a binding.
+
 ### Limitations
 
 Emotion estimation is probabilistic and context-dependent. It is a communication aid, not mind reading and not a medical or psychological diagnostic tool.

package/backend/voice_worker.py

@@ -96,6 +96,7 @@ def main() -> int:
         if not line:
             continue
         request: dict[str, Any] = {}
+        request_id: Any = None
         try:
             request = json.loads(line)
             if request.get("type") == "shutdown":
@@ -175,15 +176,17 @@ def main() -> int:
             detail = traceback.format_exc()
             sys.stderr.write(detail)
             emit({
-                "id": request.get("id"),
+                "id": request_id,
                 "ok": False,
                 "error": f"{type(exc).__name__}: {exc}",
             })
 
     try:
         with contextlib.redirect_stdout(sys.stderr):
-            emotion.close()
-            asr.close()
+            try:
+                emotion.close()
+            finally:
+                asr.close()
     except Exception:
         sys.stderr.write(traceback.format_exc())
     return 0

package/docs/MARKETPLACE.md

@@ -33,12 +33,13 @@ https://raw.githubusercontent.com/LiDu-china/dsh-dafeiyu-emotion/main/assets/ico
 大肥鱼 · 情绪耳朵是一个 DeepSeek Harness 语音 companion 插件。它把一次说话变成一条有情绪上下文的对话：
 
 1. 点击 composer 旁的鱼耳，录一句话；
-2. 优先复用 DSH 自带 Voice Input/SenseVoice 做本地转写；
-3. 浏览器提取语速、停顿、基频、音高变化、能量等声学特征；
-4. 可选中英文情绪词表与声学原型融合，输出情绪、置信度、愉悦度、唤醒度；
-5. 生成“先承接情绪、再回答问题”的受控 prompt；
-6. 自动填入并发送到当前会话；
-7. 如果本机部署了 VoiceSense，可切换到 SenseVoice + emotion2vec 增强模式。
+2. 可使用 DSH 原生快捷键开始/停止录音，并在快捷键设置中自由改键；
+3. 优先复用 DSH 自带 Voice Input/SenseVoice 做本地转写；
+4. 浏览器提取语速、停顿、基频、音高变化、能量等声学特征；
+5. 可选中英文情绪词表与声学原型融合，输出情绪、置信度、愉悦度、唤醒度；
+6. 生成“先承接情绪、再回答问题”的受控 prompt；
+7. 自动填入并发送到当前会话；
+8. 如果本机部署了 VoiceSense，可切换到 SenseVoice + emotion2vec 增强模式。
 
 插件不常驻监听、不读屏幕、不声称读心。音频只进入本机临时目录，分析结束后删除。
 
@@ -64,10 +65,10 @@ https://raw.githubusercontent.com/LiDu-china/dsh-dafeiyu-emotion/main/assets/ico
 ## Install
 
 ```bash
-dsh plugin --profile desktop add github:LiDu-china/dsh-dafeiyu-emotion
+dsh plugin --profile desktop add dsh-dafeiyu-emotion
 ```
 
-The 1024Store entry is live and currently browse-only. After the npm package is published, the store adds npm one-click install automatically.
+The `0.2.2` build is attached to the GitHub Release. npm publication is pending, so npm and the 1024Store one-click entry currently serve `0.2.1`; they will refresh after the maintainer publishes `0.2.2`. GitHub-source and Release-tarball installs remain available as fallbacks.
 
 ## Privacy