Dayflow: Automatic Screen Recorder with AI-Generated Daily Activity Timelines 123

Dayflow is an open-source native macOS app whose main job is to know what you've been doing in front of your computer without any manual action, automatically handling time tracking for you. It relies on very low frame rate screen recording, combined with periodic AI analysis, to turn a day's real activity into an annotated, visualized timeline. Below is a detailed guide compiled from the official documentation and user feedback, covering five aspects: how it works, features, privacy, performance, and usage.

Dayflow Core Features

Fully Automatic Timeline

• Launches automatically at startup and keeps running in the background; no need to click start or stop.

• Timeline cards can be dragged, edited, and annotated with screenshots or text notes.

Time-Lapse Playback

• Can fast-play a day's 1 fps recording so you can recall what you were doing.

• Can highlight moments where you may have lost focus, like suddenly switching to Bilibili.

Dashboard and Daily Journal (Coming Soon)

• You can ask direct questions like "How many tests did I write this week?" and the answer will appear as a card on the dashboard.

• Every evening, a question is pushed to prompt you to add a quick note on why you made a key decision at the time.

Export and API

• Timelines can be exported as JSON or CSV.

• The code is MIT licensed, so you can modify the interface, add content, or deploy it inside your company's intranet.

How Dayflow Works

• Capture: Records the screen at 1 fps, saving a segment every 15 seconds; files are stored locally first.

• Cache: Every 15 minutes, the most recent 15-minute segments are combined into a small 60 fps video of 900 frames.

• Analysis: The video is sent to the AI, using either the cloud-based Gemini or a local model, which performs two passes on the footage.

• Generation: The AI produces time cards like "Review PR comments: 45 minutes" or "Research OAuth flow: 1.5 hours".

• Cleanup: Recording files are typically deleted automatically after 72 hours to save disk space.

Dayflow Technical Implementation

• The interface is built with SwiftUI, fully native.

• Supports different AIs: Gemini (requires you to provide your own API key), as well as local models (Ollama or LM Studio).

• The code structure is clean, mainly including:

• Dayflow/: The SwiftUI source code of the app, including the timeline interface, debug interface, screen recording, and analysis pipeline.

• docs/: Documentation and materials, such as screenshots and videos.

• scripts/: Automated release scripts for building DMG, notarization, updates, and signing.

Performance and Resource Usage

• The installer is 25 MB; CPU usage is under 1% at runtime, with around 100 MB of memory.

• 1 fps screen recording is roughly 20 kB/s, so about 500 MB for 8 hours a day, automatically deleted after 3 days.

• Currently only compiled for Apple Silicon (M1/2/3); Intel machines need to be built manually with Xcode.

Dayflow Quick Start

Installation

• Go to GitHub Releases, download Dayflow.dmg, drag it to the Applications folder, and grant Screen Recording permission on first launch.

Configuring AI

• Gemini: Enter GEMINI_API_KEY in settings; you can disable training usage in Google Cloud billing if billing is enabled.

• Local: First install Ollama or LM Studio, pull a 7B-11B vision model, and Dayflow will automatically detect the port.

Daily Use

• Click the menu bar icon, right-click and select Start Capture to begin; choose Stop Capture when you want to stop.

• The timeline in the main window supports mouse wheel zooming, keyword search, and filtering by app or activity type.

Data Management

• Screen recordings are cached in ~/Library/Application Support/Dayflow/records/.

• To start fresh, quit the app, delete this folder, then reopen it.

Dayflow Use Cases

• Knowledge workers: Writing daily or weekly reports no longer requires effortful recall; just drag the time cards directly.

• Freelancers: When billing by project, you can export CSV as time proof.

• Privacy-conscious companies: The code is auditable and can be deployed on an intranet, meeting compliance requirements.

• Productivity enthusiasts: The playback is like an aircraft black box, helping you find the reasons for losing focus.

Dayflow combines 1 fps screen recording with 15-minute AI batch processing to transform useless figures like "Browser: 4 hours" from traditional time tracking into a comprehensible, exportable timeline such as "Review PR 45 min, write tests 1 hour, browse Weibo 20 min". The entire process can be offline, auditable, and modifiable, making it a fairly transparent automatic time tracking solution on macOS.