Skip to content

Releases

Voice Diary Processor 0.4.0

Choose a tag to compare

@jem-code-sss jem-code-sss released this 06 Nov 20:53

Voice Diary Processor 0.4.0 adds voice tasks alongside voice diaries.

  • Record tasks separately and save them in a configurable task folder. View tomorrow, next week, items needing confirmation, or all unfinished tasks; edit dates and titles or mark items complete.
  • Interpret relative dates from the actual local recording start date, even after delayed synchronization. The diary's 04:00 day boundary remains independent.
  • Default to local SenseVoice transcription and conservative local task extraction. Text API task extraction requires both the global API switch and a separate task opt-in; neither is enabled by this update.
  • Preserve original evidence and full transcripts. Validate model candidates against the speech, retain uncertain statements for review, and protect manual checkbox changes from retries or stale editing windows.
  • Wait for matching audio and companion metadata before extracting tasks. Recover completed notes without overwriting their manual edits; keep conflicting or partial files for inspection.

For phone recording, update Voice Diary Recorder to 0.4.0 too. Synchronize both task audio and its adjacent .voice.json file, plus the task-note folder. Keep Obsidian in the foreground while recording. This feature does not create calendar events or system reminders.

Use the manual-install ZIP, or the three files main.js, manifest.json, and styles.css. Existing installations are not replaced automatically by this release process. Source code, setup instructions, privacy details and validation scope are included. No personal recordings, credentials, model weights or local device configuration are included.

See the task guide and validation scope. Automated tests use synthetic fixtures; real mobile synchronization and the user's chosen API service still require user validation.

Voice Diary Processor 0.3.2

Choose a tag to compare

@jem-code-sss jem-code-sss released this 06 Nov 11:10

Changes

  • New finished notes use readable diary titles such as 12.13日记_出游.md. Optional year prefixes and the previous source-based naming remain available. Topics are derived locally with no extra API requests.
  • Diary days begin at 04:00. Original recording start times take priority over stop, save, sync, or processing dates. Existing notes and selected in-progress paths are preserved; name collisions receive a suffix.
  • Hardened local media decoding, API configuration changes, device-state locking, and recording/final-note write verification. Failed or incomplete writes retain recoverable data and preserve existing files.
  • Remote text endpoints require HTTPS; loopback services may use HTTP. Apply endpoint, model, and credentials together, then explicitly enable the API. Changing service origins clears a reused old key.

Installation

Use the Obsidian community directory or the manual ZIP. Processor runs on Windows desktop and requires a manually prepared local SenseVoice environment. Install or update the separate Recorder companion on phones. Dependency installation remains a user-initiated external step; see the setup guide.

No API is required for local transcription. Recordings, credentials, private notes, and device settings are excluded from the public packages. See VALIDATION.md for tested behavior and remaining platform limits. Voice-derived future todos are a proposal, not part of this release.

Voice Diary Processor 0.3.1

Choose a tag to compare

@jem-code-sss jem-code-sss released this 04 Nov 10:24

Voice Diary Processor 0.3.1 adds a complete English README for the Obsidian Community directory and preserves the Chinese guide as README.zh-CN.md. Installation requirements, optional synchronization, manual SenseVoice setup, external device storage, and optional API behavior are documented in both languages.

Recording and transcription behavior is unchanged from 0.3.0. The runtime JavaScript is verified to match the previously tested 0.3.0 build.

Install main.js, manifest.json, and styles.css through Obsidian or use the manual ZIP. SenseVoice, Python, and the model are prepared manually; they are not installed or downloaded by the plugin. API use remains optional and disabled by default.

Voice Diary Processor 0.3.0

Choose a tag to compare

@jem-code-sss jem-code-sss released this 03 Nov 13:05

Voice Diary Processor 0.3.0

首个公开桌面版本:使用本机 SenseVoiceSmall INT8 将录音转成笔记,可选文字 API 整理;录音与成品路径可分别指定。

  • Windows x64 电脑使用;手机配套另装 Voice Diary Recorder。
  • 新装默认录音 Voice Diary/Recordings,成品 Voice Diary/Notes;升级不改已有目录。
  • 设置隐藏完整本机数据路径。API 默认关闭,缺少配置时只做本地识别;发布内容经过实际密钥与私人资料排除检查。
  • 插件不安装或更新依赖。模型按钮打开官方来源;Python/SenseVoice 环境由用户按说明在插件外手动准备。
  • 适配社区目录三文件安装,runner 内置在 main.js,本地按版本缓存并校验。
  • 修复全零静音误识别、单条音频失败停整队列、保存回报错误导致重复笔记,以及录音错误时丢失最终缓冲的问题。

安装与需要的软件

手动安装下载 voice-diary-unified-v0.3.0.zip,把内含插件目录复制到仓库插件目录。main.js、manifest.json、styles.css 是社区目录/兼容安装工具使用的标准附件;源码 ZIP 供构建或手动准备识别环境。

电脑需要 Obsidian 与 Python/SenseVoice 环境;手机不用安装模型。手机电脑联用需要 Syncthing 或其他可靠同步工具,纯电脑使用无需同步软件。完整步骤:安装与同步教程。

验证与状态

193 项电脑端 Node、37 项 Python、12 项 Windows PowerShell 测试通过。仅三个标准插件文件的隔离安装验证通过,实际本地模型完成两段完整多分钟录音转写,零文字 API 调用。独立新设备首次联网准备仍需设备验证。

正在准备社区目录申请,此 GitHub Release 不代表已上架或获官方审核通过。目录登录与仓库绑定由拥有者完成。详见 上架准备说明。