Skip to content

Releases

Voice call v0.3.4 — first reply speech fix

Choose a tag to compare

@crease123 crease123 released this 14 Nov 17:20

Fixes silent first replies after starting a voice call. Runtime-context messages no longer clear ownership of an admitted voice response. Context-only requests still do not acquire playback ownership.

Validation: regression tests fail before the fix and pass after it; all 119 tests pass; full build passes. Live microphone calls were not tested.

Voice call / 语音通话 v0.3.3

Choose a tag to compare

@crease123 crease123 released this 14 Nov 12:28

语音通话直接在现有会话中使用,不再注册 Personal Agent / Voice call 独立预设。通话期间保留可编辑的提示词注入,挂断或断连后撤回,保留原会话的人设、工具和权限。

中英文 README 新增功能介绍、安装配置、通话操作和常见问题,配有裁除左侧私人信息的截图、按钮放大图和流程图。图片随主安装包分发。

验证:完整构建通过,116 项测试通过;发行包已检查图片完整性、仓库归属和私人凭据/路径。

安装:在 DSH 插件管理器添加 npm 包 voice-call,或使用本发行包的下载地址。要求 DSH 0.2.2-rc.1 或更新的 0.2.x 版本。

Voice call v0.3.2

Choose a tag to compare

@crease123 crease123 released this 14 Nov 11:41

Unifies the public npm package, source repository, documentation and installation links under voice-call.

Repository: https://github.com/crease123/voice-call
npm: https://www.npmjs.com/package/voice-call

Keeps the self-contained runtime packaging from v0.3.1 so pnpm installation works with blockExoticSubdeps enabled. Validation: full build, three Remote-generation tests, isolated pnpm 11.28.5 installation, and archive checks for old repository references, local paths and credential-shaped values.

Voice call v0.3.1 — pnpm installation fix

Choose a tag to compare

@crease123 crease123 released this 14 Nov 11:27

Fixes installation in DSH when pnpm enables blockExoticSubdeps. Version 0.3.0 referenced three GitHub URL subdependencies and was rejected with ERR_PNPM_EXOTIC_SUBDEP.

Version 0.3.1 embeds the three original prebuilt runtime components and declares their external requirements as registry dependencies. No security-policy changes are required.

Validation: isolated pnpm 11.28.5 install with blockExoticSubdeps=true, successful imports of the bundle and all three components, npm publish dry-run, and private-data/archive manifest checks. Live microphone calls were not retested; runtime code is unchanged.

Voice call / 语音通话 v0.3.0

Choose a tag to compare

@crease123 crease123 released this 14 Nov 06:17

Voice call / 语音通话 adds continuous microphone conversations to DeepSeek Harness.

  • DeepSeek ASR and TTS are the defaults and use the user's own Harness API key.
  • ASR and TTS providers can be selected independently between DeepSeek and Doubao.
  • Doubao API key or legacy App ID / Access Token can be entered in settings and saved to the Host credential store. Saved secrets are never returned to the browser or included in this release.
  • Compact five-bar microphone animation, interruption support, voice previews, and editable call instructions.

Install dsh-voice-call-0.3.0.tgz through DSH's plugin manager. The other three assets are prebuilt runtime dependencies downloaded automatically. No local build is required. Requires DSH >=0.2.2-rc.1 <0.3.0-0 and Node >=24.9.0.

Validation: full TypeScript/build checks, 116 plugin tests, three Remote-generation tests, and release credential/local-path checks. Real microphone calls against a live provider have not been tested in this release session.