QuestOS / 命途OS

A Windows desktop life-progress system — evolving into a Personal Action OS — built around the Eisenhower Matrix, long-term quests, reminders, voice capture, and AI-guided execution. It does not just store tasks; it turns your real-world conversations and meetings into the next actions.

English | 简体中文

Project Status

• Roadmap: docs/ROADMAP.md
• Changelog: CHANGELOG.md
• Contributing: CONTRIBUTING.md
• Security: SECURITY.md
• Code of Conduct: CODE_OF_CONDUCT.md

Download In 30 Seconds

1. Go to Latest Release.
2. Download the TimeTask-win-x64-*.zip asset.
3. Unzip and run TimeTask.exe.

If this project is useful, please star it to help more users find it.

Why TimeTask

• Focus by priority: manage tasks in 4 quadrants.
• Close the loop: reminders + overdue nudges + review.
• Connect goals to execution: long-term goals and learning plans.
• Capture quickly: voice to task draft.
• Optional AI support: decomposition and suggestions.

Main Features

• Four-quadrant task management and quick add.
• Reminder scheduling and overdue alerts.
• Task decomposition and action suggestions (optional LLM).
• Long-term goals and learning plan management.
• Voice recognition to task drafts.
• Conversation → Action Inbox: record a meeting or chat (microphone and system audio via WASAPI Loopback), let local ASR transcribe it, auto-extract action items, and accept them into the four quadrants with one click.
• Data import/export (JSON).
• Skill management (enable/disable, import/export).

Screenshots

Main	Reminder	LLM Settings

Goals	Drafts

Conversation → Action Inbox

TimeTask treats a meeting assistant not as a separate app, but as one input sensor feeding the task system. The flow is:

现实交流 / 会议 / 电话
        │
   Audio Engine（常驻待命，按需记录）
     ├─ MicrophoneSource  ── WASAPI Capture  ──┐
     └─ SystemLoopbackSource ── WASAPI Loopback ┘
        │  （两路分别流式写盘：mic.wav / system.wav）
        │  （实时转写时再混音喂 ASR；ASR 崩了录音照常）
        ↓
   本地 ASR（Vosk，独立于录音层）
        │
   信息分类：Task / Reminder / Decision / Note
        │
   TimeTask · 四象限 / 目标 / 提醒
        │
   行动收件箱（Action Inbox）· 一键「全部接受」

Why this design (V2 audio layer)

• Two sources, two tracks. Meeting/Conversation mode opens both the microphone and the system playback audio (WASAPI Loopback). They are written to separate mic.wav / system.wav files so future "me / them" separation and speaker diarization stay easy. The live transcript is produced from a mixed stream.
• Streaming write, never all-in-memory. Recording starts the instant you click "开始"; audio is written to disk chunk-by-chunk and only finalized on stop. A 2-hour meeting is never held in a List<byte>.
• No dropped first word (pre-roll). Capture + ASR feeding begin on the first frame. If the Vosk model finishes loading a moment later, the buffered audio is back-filled into the recognizer, so the opening of the meeting is never lost.
• ASR is decoupled from recording. If Vosk is unavailable, recording and disk write continue normally; you simply get no live transcript this time (the audio is still saved for later transcription).
• Adaptive silence, not a fixed -30 dB. Quick-dictation auto-stop uses a dynamically estimated noise floor instead of a hard threshold, so it works across laptops, headsets, and offices.
• Two clearly separate entries. Quick-dictation (mic only, auto-ends on silence, feeds task drafts) is distinct from Meeting/Conversation recording (mic + system, can run for hours).

How to use:

1. TimeTask starts in standby (system tray resident). It does not record 24×7 by default.
2. Right-click the tray icon → 🎙 快捷口述 / ⏺ 开始交流记录 / 🖥 开始会议记录. While recording, the tray tooltip shows ● 录音中 mm:ss · 麦克风✓ · 系统✓.
3. Speak / hold a meeting (Tencent Meeting, Teams, Zoom…). Both your voice and the other party's audio (played through speakers/headset) are captured and written to disk.
4. Click ■ 停止记录 (or it auto-stops after the configured timeout). TimeTask transcribes locally and extracts action items.
5. The Action Inbox window shows a summary, the transcript, and each detected action with a suggested quadrant + reminder. Click 全部接受 and the tasks land in the four quadrants.

Key settings in App.config:

Key	Default	Meaning
VoiceAutoStartOnLaunch	false	If true, also auto-start the always-on quick-dictate listener on launch.
ConversationCaptureIncludeSystemAudio	true	Capture system playback audio (WASAPI Loopback) so remote meeting voices are recorded as a separate system.wav.
ConversationCaptureAutoStopMinutes	120	Auto-stop a meeting/conversation capture after this long.
ConversationCaptureQuickSilenceSeconds	8	Quick-dictate auto-stops after this much silence (adaptive threshold).
ConversationCaptureQuickMaxWaitSeconds	60	Quick-dictate also stops if no speech is detected within this long.

Design boundary: TimeTask is standby-by-default, not always-recording. Full-text search, speaker diarization, auto-meeting-detection and long-term memory are intentionally deferred.

Build From Source

• Environment: Windows + Visual Studio + .NET Framework 4.7.2 (WPF).
• Open TimeTask.sln.
• Build and run.

Optional Configuration

• LLM settings in App.config (OpenAIApiKey, LlmProvider, LlmApiBaseUrl, LlmModelName).
• Voice/FunASR can use local runtime bundle: data/funasr-runtime-bundle.zip.
• Auto update can check GitHub Releases on startup.

Contributing

• Bug reports and ideas: Issues
• Feature discussions: Discussions
• Contribution guide: CONTRIBUTING.md

License

MIT - see LICENSE.