---

🎯 What is Dayflow?

Dayflow is an AI-powered time tracking tool for Windows. It quietly captures low-frequency screen activity and window context in the background, uses a vision model to infer what you are doing, and turns the result into a timeline, statistics dashboard, and reports so you can understand where your time actually goes.

💡 Who is it for?

• People who want to know where their day went
• People who want to review focus time, distractions, and work patterns
• People who need automatic activity logs, daily reports, or weekly summaries
• People who prefer passive tracking over manual timers

🏆 Key advantages

Advantage	Description
Zero effort	Start it and let AI handle activity recognition automatically
Low overhead	1 FPS recording + smart compression to keep background usage low
Local first	Raw recordings stay on your machine and temporary chunks are cleaned up after analysis
Smart categories	Automatically detects work / study / entertainment / social / rest
Visual review	Timeline, statistics page, and HTML dashboard for different review styles

---

🔐 Privacy, upfront

Dayflow is designed around local recording + cloud analysis + local storage.

5 things you should know

1. Full videos are not uploaded
   Raw recording chunks stay on your local machine.

2. Only a limited number of keyframes are sent for analysis
   The current README describes up to 8 extracted frames per chunk being sent to the configured vision model.

3. Analysis results are stored locally
   Activities, settings, and statistics are saved in a local SQLite database.

4. Temporary chunks are cleaned up automatically
   After analysis, temporary recording files are deleted to reduce disk usage.

5. You can pause recording for sensitive moments
   Passwords, banking, private chats, or anything else sensitive can be handled by pausing recording first.

Local data location

%LOCALAPPDATA%\Dayflow\
├── dayflow.db      # activities, settings, stats
├── dayflow.log     # runtime logs
├── chunks\         # temporary recording chunks
└── updates\        # downloaded updates

💡 If privacy is your first concern, read this section before enabling continuous recording.

---

✨ Main features

Feature	Description
🎥 Low-power recording	1 FPS low-resource background capture
🪟 Window tracking	Uses Windows API to capture real app names and window titles
🤖 AI analysis	Vision LLM classifies screen activity automatically
📊 Timeline view	Clear daily activity review at a glance
📈 Statistics dashboard	Time distribution, productivity trends, week comparison
📊 Web dashboard export	Beautiful self-contained HTML report with charts
📧 Email reports	Automatic daily summaries and deeper analysis
🔄 Auto update	Check, download, and install new versions
🚀 Auto start	Launch on boot and minimize to tray
📥 CSV export	Export activity data for further analysis
⏸️ Pause recording	Pause when handling sensitive content
🎨 Theme switching	Dark / light themes with saved preferences

🆕 Recent highlights

v1.6.0 (2026-10)

• Major Architectural Overhaul: Completely migrated from desktop Qt to Tauri 2.0 (Rust) + React 18 + Vite, delivering instant startup, fluid animations, and significantly lighter resource footprint.

• Modern Responsive UI: Three-pane responsive layout, rich activity timeline cards, smooth interactive charts, and system dark theme support.

• Kernel-level Process Management (Windows Job Object): Embedded Python bridge process tree is bound to Windows Job Object, eliminating console windows and orphan processes.

• Offline Resilient Report Generation: When external AI APIs time out or fail (e.g. invalid/expired key), automatic local aggregation synthesizes a structured Markdown daily report in milliseconds.

• Interactive Statistics Dashboard: 7-day and 30-day metrics, productivity trend chart, category donut breakdown, daily activity heatmap, and hourly distribution graph.

• Multi-Monitor & Idle Pause: Panoramic multi-screen capture and automatic pause/resume based on keyboard and mouse idle detection.

• Privacy Filtering: Window title & process blacklist filtering, plus storage quota limits with automatic cleanup.

• Statistics page redesign with metric cards, donut chart, trend chart, heatmap, and week comparison

• Activity cards can now be edited or deleted

• Better visual polish across the UI

v1.5.0 (2025-12)

• Windows API window tracking for more accurate recognition
• UI refresh: efficiency indicator bar, deep work badge, live recording duration, and more
• Improved prompting and data processing logic

---

🖥️ Screenshots

Timeline page

Daily activity cards with time range, app, summary, and productivity score.

Statistics page

Dashboard-style overview with key metrics, distribution charts, heatmap, and comparisons.

Web dashboard

Date selection

HTML report

Exported HTML dashboard that can be opened in a browser and shared directly.

Email reports

Settings

Examples

Deep analysis

Auto update / auto start

---

✅ README checklist covered in this revision

• Removed garbled text, typos, and broken directory names
• Reframed the opening so the project value is clearer
• Moved privacy information near the top
• Reduced feature overload and tightened the structure
• Added a roadmap section
• Added known limitations
• Clarified the current license choice
• Aligned the English README with the Chinese version
• Added a minimal Windows CI workflow

---

🚀 Quick start

Requirements

• Windows 10 / 11 (64-bit)
• Python 3.10+
• FFmpeg added to PATH

Installation

git clone https://github.com/SeiShonagon520/Dayflow.git
cd Dayflow

conda create -n dayflow python=3.11 -y
conda activate dayflow

pip install -r requirements.txt
python main.py

Build as EXE (optional)

pip install pyinstaller
python build.py

Or simply double-click build.bat.

Download prebuilt releases

If you just want to use Dayflow without setting up Python locally:

1. Open Releases
2. Download the latest installer or archive
3. Extract it and run Dayflow.exe

💡 For the first run, configure the API first and then start recording.

---

📖 Basic usage

1. Configure your API

Go to Settings and fill in:

• API URL: any OpenAI-compatible endpoint
• API Key
• Model name: a vision-capable model

Then click Test Connection and Save Config.

2. Start recording

• Click Start Recording
• The app records at 1 FPS in the background
• A video chunk is generated every 60 seconds
• Keyframes are sent to your configured model service for analysis

3. Review your timeline

Each card shows a time range, category, app, summary, and productivity score.

4. Optional features

• Email reports: scheduled summaries and deeper analysis
• Auto start: launch with Windows and minimize to tray
• Auto update: check and install new versions
• System tray: control recording without keeping the main window open

---

📁 Project structure

Dayflow/
├── main.py
├── config.py
├── requirements.txt
├── build.py
├── build.bat
├── updater.py
│
├── core/
│   ├── types.py
│   ├── recorder.py
│   ├── window_tracker.py
│   ├── llm_provider.py
│   ├── analysis.py
│   ├── email_service.py
│   ├── updater.py
│   ├── autostart.py
│   ├── config_manager.py
│   ├── log_manager.py
│   ├── stats_collector.py
│   └── dashboard_exporter.py
│
├── database/
│   ├── schema.sql
│   ├── storage.py
│   └── connection_pool.py
│
├── ui/
│   ├── main_window.py
│   ├── timeline_view.py
│   ├── stats_view.py
│   ├── date_range_dialog.py
│   └── themes.py
│
├── templates/
│   └── dashboard.html
│
└── assets/
    └── icon.ico

---

⚙️ Configuration

Variable	Description	Default
DAYFLOW_API_URL	API endpoint	https://apis.iflow.cn/v1
DAYFLOW_API_KEY	API key	(empty)
DAYFLOW_API_MODEL	AI model	qwen3-vl-plus

---

🛠️ Tech stack

Component	Technology
GUI	PySide6 (Qt6)
Screen capture	dxcam (DirectX)
Video processing	OpenCV
HTTP client	httpx (HTTP/2)
Database	SQLite
AI analysis	OpenAI-compatible APIs

---

🗺️ Roadmap

Near term

• More explicit privacy controls and recording safeguards
• Smarter activity merging and cleaner timeline segmentation
• Richer statistics and trend analysis
• More reliable update and installation flow
• Better automated testing and CI coverage

Mid term

• Multi-monitor support
• Better website / app recognition
• Local-model or hybrid analysis mode
• Smarter daily / weekly behavioral insights
• More complete export and reporting options

---

Known limitations

• Windows 10 / 11 only
• Recognition quality depends on the vision model you configure
• Some apps may not expose stable window titles
• Full-screen games, special rendering windows, and remote desktop scenarios may behave differently
• Multi-monitor support still has room for improvement
• Analysis latency depends on your network and model provider stability

---

🧰 Troubleshooting

1. No analysis results appear

Check the following first:

• API URL, key, and model name are correct
• The configured model supports vision input
• Your network can reach the model provider
• The log contains no timeout or request errors

2. Screen capture or window tracking looks wrong

Check the following first:

• You are running on Windows 10 / 11
• FFmpeg is installed and available in PATH
• Security software is not blocking capture or window inspection
• You are not in a special case such as full-screen games, remote desktop, or unusual rendering modes

3. Email reports fail to send

Check the following first:

• QQ Mail SMTP / authorization code is correct
• The built-in test send works
• Your local network is not blocking the mail service

4. Updates or auto-start do not work

Check the following first:

• You are running the packaged EXE build
• The EXE path has not changed unexpectedly
• System permissions or security policy are not blocking startup/update behavior

---

💡 Inspiration

This project is inspired by Dayflow (macOS). The original project is macOS-only, so this repository brings a similar idea to Windows users.

Thanks to the original author for the creativity and open-source spirit.

---

📄 License

CC BY-NC-SA 4.0 © 2024-2025

This project currently uses Creative Commons Attribution-NonCommercial-ShareAlike 4.0.

• ✅ You can learn from it, modify it, and share it
• ✅ Please credit the original author
• ❌ Commercial use is not allowed

ℹ️ This is not a typical software license such as MIT or Apache-2.0. If you plan to redistribute or use the project commercially, read LICENSE carefully first.

---

⭐ Star History

---