Dayflow Mac App Review: Turn Screen Time Into an AI Timeline

Dayflow is a native macOS application that records your screen activity and converts it into a visual timeline with AI-generated summaries. It provides a precise record of how you spent your time without the need for manual logging.

The app captures your screen at one frame per second (FPS). Every 15 minutes, it batches these recordings and sends them to an AI model for analysis. The result is a clean timeline populated with activity cards. A time-lapse feature allows you to review your entire day in seconds. To conserve disk space, recordings are automatically deleted after three days.

You choose how the AI processes your data. You can use a Gemini API key for Google’s cloud processing or configure Dayflow to use a local server running Ollama or LM Studio. Local mode ensures all data remains on your machine. The app itself is a lightweight 25MB SwiftUI package, typically idling at under 1% CPU usage with approximately 100MB of RAM consumption.

Key Features

1. Automated Daily Timeline: Generates short activity summaries without requiring you to start or stop timers.
2. Low-Impact Recording: Captures at 1 FPS, ensuring the app runs unnoticed in the background.
3. Frequent Processing: Updates your timeline every 15 minutes to keep it current.
4. Daily Time-Lapses: Provides a fast-forward replay of your entire workday.
5. Automatic Cleanup: Deletes video files older than three days to manage storage.
6. Distraction Flagging: Highlights moments on the timeline where your focus was broken.
7. Native Experience: Built with SwiftUI for a light, responsive, and platform-consistent interface.
8. Background Updates: Uses the Sparkle framework to download and install new versions seamlessly.
9. Customizable Dashboard: Allows you to query your work week and pin important insights.
10. Daily Logs: Supports adding manual notes or screenshots to any entry on the timeline.

The Workflow

• Capture: Records 15-second video segments at 1 FPS.
• Analyze: The AI examines the latest segments every 15 minutes.
• Generate: The model creates descriptive summary cards for the timeline.
• Display: The visual timeline maps out the arc of your day.
• Clean: Records older than three days are automatically purged.

AI Pipeline: Gemini vs. Local

The choice of processing method significantly changes how the app analyzes your data:

• Local (33+ LLM calls): Dayflow extracts 30 frames from a video chunk. It describes each frame individually (30 calls), then merges those descriptions, generates a title, checks for consistency, and finalizes the card. This requires substantial local inference.
• Gemini (2 LLM calls): The video chunk is uploaded directly. Gemini understands the video natively (1 call) and generates the timeline card (1 call). This method is faster and more streamlined.

Gemini treats the input as a native video file, whereas local models currently process it as a series of still images, explaining the difference in API call volume.

Installation and Setup

Download the latest Dayflow.dmg from the GitHub Releases page and drag the app to your Applications folder. Upon launching, you must grant "Screen & System Audio Recording" permissions in System Settings > Privacy & Security.

To build from source, clone the repository, open the project in Xcode 15 or later, and run the Dayflow scheme. If you prefer cloud processing, set your GEMINI_API_KEY in the scheme's environment variables.

Data Storage Locations

• Application Support: ~/Library/Application Support/Dayflow/
• Video Clips: ~/Library/Application Support/Dayflow/recordings/
• Database: ~/Library/Application Support/Dayflow/chunks.sqlite

To wipe all recordings and start fresh, quit the app and delete the Dayflow folder.

Gemini Privacy and Billing

Google’s data usage policies contain an important distinction: if you enable cloud billing on a Gemini API project, all usage—including free tier queries—is governed by "Paid Services" rules. Under these rules, Google does not use your prompts or responses to train its models. For users in the EEA, UK, or Switzerland, Paid Service data protections generally apply by default.

Local Mode Trade-offs

Running Ollama or LM Studio ensures that all prompts and inference stay on your device. However, local models may produce less sophisticated summaries than Gemini. On Apple Silicon, local inference utilizes the GPU, which can increase battery drain; keeping your MacBook connected to power is recommended during long recording sessions.

Updates and Troubleshooting

Sparkle checks for updates daily. You can click the menu bar icon to inspect saved recordings, which is useful for debugging.

If you see blank screen captures, verify permissions in System Settings > Privacy & Security > Screen & System Audio Recording. For API errors, double-check your API key and network connection in the app's settings.