JerryZLiu/Dayflow

Dayflow is a local-first macOS application that continuously records screen activity to build a detailed, searchable timeline of a user's daily work. It synthesizes this raw data into structured insights, offering daily summaries, weekly analytics dashboards, and AI-powered chat interfaces that can query the user's history. The system also integrates with external AI coding agents via a local CLI and MCP server, allowing them to read and, with permission, modify the recorded timeline data.

22 changes.

This period focused on releasing Dayflow v2.6.0, which introduced a major rewrite of the recording engine to use HEVC segments and ScreenCaptureKit for improved privacy and performance. The update added extensive new features including a dashboard chat with AI tool execution, structured analytics via PostHog, automatic updates, and a redesigned onboarding and timeline interface. Infrastructure improvements included secure API key storage, comprehensive test coverage, and automated release tooling.

16 changes.

This period focused on integrating external AI coding agents like Codex and Claude, introducing a write bridge for timeline modifications and an Agents tab for reviewing sessions. It also delivered a comprehensive weekly dashboard with detailed visualizations and gated access, alongside a new Flow companion agent with a desktop overlay for distraction detection.

Users can now explicitly set the app language in Dayflow settings, independent of the system locale. The new AppLanguagePreferences utility manages this by reading and writing the app-specific domain in UserDefaults, supporting English, Simplified Chinese, Korean, German, Russian, French, Traditional Chinese, and Japanese, with a 'System default' option that falls back to the OS language.

Dayflow/Dayflow/Utilities · high confidence

The scripts directory now includes a comprehensive set of automation tools for building, signing, and distributing Dayflow. A new one-button release script (release.sh) handles version bumping, building, and creating a signed DMG using create-dmg, while also managing Sparkle appcast updates and GitHub release creation. The tooling integrates Sentry for crash reporting and dSYM symbolication, and injects configuration for PostHog analytics and the backend URL directly into the app bundle at release time. Supporting scripts handle Sparkle update signing via Keychain, DMG notarization, and appcast generation.

scripts · high confidence

The Dayflow CLI now provides read-only access to the application's SQLite database for timeline and status queries, while write operations (such as editing activities, managing categories, and setting goals) are routed through a Unix domain socket to the running Dayflow app and require explicit user opt-in. A new MCP server exposes these capabilities to AI agents, with write tools hidden unless edits are enabled in settings. The tool also includes stable JSON output for deterministic parsing, mirrors the app's 4 AM day boundaries, and records privacy-safe usage telemetry for both CLI and MCP interactions.

tools · high confidence

Dayflow now includes built-in crash reporting via Sentry and automatic update capabilities using Sparkle. The app also embeds the Dayflow CLI tool directly into the application bundle, allowing users to run it from within the app environment. Additionally, the project has integrated PostHog for analytics, MarkdownUI for rich text rendering, and expanded localization support to include Chinese (Simplified and Traditional), Korean, German, Russian, French, and Japanese. The app version has been bumped to 2.6.0.

Dayflow/Dayflow.xcodeproj · high confidence

Users can now receive notifications when their daily recap is ready, with the system automatically requesting permission if needed and tracking the state via a new badge manager. This change also consolidates badge logic for both the Dock icon and sidebar indicators, ensuring that unread support replies and pending daily recaps are accurately reflected in the app's visual status. The previous Journal-related notifications have been retired to clean up the notification space.

Dayflow/Dayflow/Core/Notifications · high confidence

The Flow tab now serves as the entry point for the Flow feature: users without access see a waitlist screen with an animated background and a chat interface, while whitelisted users see the remote Flow web app. A new desktop overlay displays an animated creature (HEVC video clips) that reacts to session states (nudges, breaks, session end) with hover-to-reveal speech bubbles and draggable positioning. A debug panel is available to inspect agent logs, adjust agent settings (model, tick intervals, screenshot quality), and manually trigger agent ticks.

Dayflow/Dayflow/Views/UI/Flow · high confidence

The weekly dashboard now includes several new data visualizations: a focus and distraction heatmap showing activity density in 5-minute buckets, a workflow view displaying 15-minute time slots per day, a treemap of most-used apps per category with previous-week comparisons, a sankey diagram mapping category-to-app flows, context charts tracking distribution events and day-to-day comparisons, and an application interactions graph showing app transitions and patterns. These snapshots are built from timeline card facts, excluding system and idle categories, and are localized for the user's locale.

Dayflow/Dayflow/Core/Weekly · high confidence

This change adds a new layer of support for running AI models via local command-line interfaces (specifically Codex and Claude), including process runners, configuration managers, and robust JSON parsing that handles terminal escape sequences and malformed output. It also introduces a new dashboard chat feature with a dedicated service, tool execution capabilities (such as fetching timeline data from the local SQLite database), and persistent conversation history, alongside new data models for activity cards and prompt overrides.

Dayflow/Dayflow/Core/AI · high confidence

The weekly view now enforces a 30-hour data requirement before unlocking full access. Users with insufficient recorded timeline data see a dedicated locked screen (WeeklyAccessLockedView) featuring a progress indicator, an estimated unlock date, and a notification request button. Once the threshold is met, the view transitions to the full weekly dashboard, which includes adaptive styling for light and dark modes, inline exportable graphics for workflow and focus heatmap sections, and refined header navigation. This change gates the weekly insights behind a data-accumulation milestone while providing clear feedback on progress and unlock timing.

Dayflow/Dayflow/Views/UI/Weekly · high confidence

Dayflow now supports in-app surveys delivered from the backend, allowing users to answer questions directly within the app. The new SurveyCenter manages survey eligibility based on user activity and provider, enforces pacing to avoid prompt fatigue, and integrates with the GitHub star card to prevent overlapping prompts. Responses are recorded locally and persisted to disk via SurveyEventReporter, ensuring that answers are not lost if the device is offline; they are batched and sent to the backend only when a connection is available, with failure handling to retry or drop lower-priority events if storage limits are reached.

Dayflow/Dayflow/Core/Surveys · high confidence

Users now encounter a new multi-step onboarding prototype that guides them through an intro video (with optional 2x playback speed), role selection, and paid-account preferences before reaching a provider comparison table. This table lets users choose between Dayflow Pro, ChatGPT, Claude, Gemini, OpenRouter, or Local AI, with Dayflow Pro offering a 7-day free trial via a dedicated sign-in panel that supports email, verification codes, and referral codes.

Dayflow/Dayflow/Views/Onboarding/Prototype · high confidence

Dayflow now includes an Agents tab that aggregates and summarizes your daily Codex and Claude Code activity. The feature introduces a multi-stage review flow: an overview screen displaying workstream blobs with status indicators, a briefing stage summarizing accomplishments per workstream, and a detailed card view for individual threads. These thread cards support four distinct rendering styles (messenger, transcript, brief, and milestones) to visualize conversation turns. The system also provides deep-linking capabilities, allowing users to open specific sessions directly in the Codex desktop app or focus/resume them in Ghostty, while tracking foreground usage time for analytics.

Dayflow/Dayflow/Views/UI/Agents · high confidence

Adds a new FaviconService that resolves site icons using a combination of hardcoded pattern-based overrides (for apps like ChatGPT, Apple services, and browsers) and network fetching. The service includes an in-memory cache to avoid redundant fetches, deduplicates concurrent requests, and normalizes hostnames to ensure consistent icon resolution across the application.

Dayflow/Dayflow/Core/Net · high confidence

Dayflow now tracks a comprehensive set of user events via PostHog, documented in the new AnalyticsEventDictionary, covering onboarding, agent access, timeline navigation, and chat interactions. The app supports deep links to start or stop screen recording and to save referral codes through the new AppDeepLinkRouter. A new PauseManager allows users to pause recording for timed intervals (15 minutes, 30 minutes, 1 hour) or indefinitely, with automatic resume on timer expiration. Additionally, an InactivityMonitor tracks user idle time to reset the session, and a ScreenshotShortcutTracker logs keyboard shortcut usage for analytics.

Dayflow/Dayflow · high confidence

This change adds the core components for the Flow companion: a distraction-detection agent that watches the screen via the Codex CLI, a live-updating session activity timeline persisted to JSON, and a draggable desktop overlay featuring a creature that delivers distraction nudges, break tubs, and toasts. Users can now tune agent behavior (model, reasoning effort, screenshot resolution, tick cadence) and overlay appearance (nudge variants, positions) in real time via a debug panel, while the app mirrors session state from the web UI to drive local animations and goal tracking.

Dayflow/Dayflow/Core/Flow · high confidence

The MainView now features a dedicated timeline interface that supports both day and week modes, allowing users to navigate between dates using chevron buttons or a date picker. Activity cards within the timeline can be edited directly, enabling users to change titles and categories with optimistic UI updates that persist to storage. The view includes a responsive header that adapts to available space, a screenshot slideshow player for reviewing activity frames, and logic to handle timeline data updates, review counts, and weekly tracked minutes.

Dayflow/Dayflow/Views/UI/MainView · high confidence

A new KeychainManager class has been added to the Core/Security module to handle the secure storage, retrieval, and deletion of API keys using the macOS Keychain. This component ensures that sensitive credentials are stored with appropriate accessibility attributes (WhenUnlocked) and provides thread-safe operations for managing keys associated with specific providers.

Dayflow/Dayflow/Core/Security · high confidence

Dayflow now allows connected AI agents to modify your timeline data directly through a new local Unix socket bridge. This feature introduces a permission-gated write channel (controlled by a new 'Allow edits' setting) that lets agents add, update, and delete categories and activities via a bundled CLI. The system also includes a client registration module that automatically configures supported MCP clients like Codex, Claude Code, and Claude Desktop, while ensuring that usage telemetry is safely drained and sanitized before being sent to analytics.

Dayflow/Dayflow/Core/AgentAccess · high confidence

This change introduces several new model structures in the Dayflow data layer to support upcoming features. A \ChatMessage\ model is added to handle AI conversation history, including roles (user, assistant, tool call) and tool execution statuses. A \DayGoalPlan\ model enables the persistence of daily focus and distraction targets, including category snapshots and skip states. Recording analysis now uses a \Screenshot\ model to replace the legacy \RecordingChunk\ video-based model, supporting HEVC segments and idle time tracking. Finally, \TimelineCategory\ and a \CategoryStore\ are introduced to manage user-defined categories with support for onboarding presets, color tracking, and persistence.

Dayflow/Dayflow/Models · high confidence

Users can now submit bug reports and feedback via an integrated in-app support chat (BugReportView) with options to email, join Discord, or book a call, including a button to copy debug logs. Additionally, a new dashboard chat interface (ChatPanelView) allows users to converse with AI assistants, featuring a multi-line composer, conversation history, and the ability to render rich markdown replies and various inline charts (bar, line, stacked bar, donut, heatmap, and gantt) directly in the chat bubbles.

Dayflow/Dayflow/Views/UI · high confidence

This update introduces a suite of new SwiftUI components in the Views/Components directory to power the redesigned Day Summary, category organization, and goal setup flows. Key additions include the CardsToReviewBadge for tracking review items, CategoryDonutChart for visualizing time breakdowns, and various category picker surfaces (CategoryPickerView, CategoryPickerOverlay, DayCategorySelectionEditor) that allow users to select and edit categories with dark-mode support and helper text. The update also brings interactive controls like the ColorWheelPicker for custom color assignment, DayGoalDurationPicker for setting time-based goals, and visual feedback elements such as ConfettiBurstView and CursorModifiers for hover/press states. These components collectively enable the new distraction and focus summary sections, category setup cards, and goal progress displays.

Dayflow/Dayflow/Views/Components · high confidence

The app now includes dedicated thumbnail generation logic for both video files and screenshot previews. Video thumbnails are generated on-demand from the first few seconds of the media, while screenshot thumbnails are stored in an in-memory cache to avoid redundant processing. This change introduces the underlying infrastructure for faster, more stable preview rendering in the UI.

Dayflow/Dayflow/Core/Thumbnails · high confidence

The weekly dashboard now includes several new visualization sections: an Application Interactions view that maps relationships between apps using a network graph with categorized nodes (work, personal, distraction); a Context Charts section displaying line charts comparing distractions and context shifts over the week; a Context Shift Comparison view with an interactive chart and insight card; a Donut section showing weekly time distribution with customizable ring geometry; a Focus Heatmap section visualizing focus vs. distraction patterns across days with configurable cell gaps and scrolling; and a Highlights section displaying top weekly highlights with tags. These sections use the Figtree font family and follow the new WeeklyPalette styling system.

Dayflow/Dayflow/Views/UI/Weekly/Sections · high confidence

The Settings screen has been reorganized into distinct tabs, introducing a new account management section with a dedicated email-and-code sign-in flow, an AI Tools tab for connecting external clients like Codex and Claude and managing edit permissions, and a Data tab for exporting timeline data as Markdown or reprocessing specific days. This change also establishes a unified visual design system for the settings surface, enforcing consistent typography, spacing, and button styles across all sections.

Dayflow/Dayflow/Views/UI/Settings · high confidence

A change to existing behaviour in Dayflow/Dayflow/Videos — 4 commits, 24 files.

Dayflow/Dayflow/Videos · medium confidence · unverified

A change to existing behaviour in Dayflow/Dayflow.xcodeproj/project.xcworkspace — 5 commits (1 fix), 1 file.

Dayflow/Dayflow.xcodeproj/project.xcworkspace · medium confidence · unverified

The app now uses a new configuration module to control when daily goal prompts appear and how progress is displayed. Prompts are hidden until a user has at least 3 distinct days of recorded activity (excluding today) to ensure they see real stats first, and auto-prompting stops after 5 consecutive skips. Progress text is now localized and formatted to show hours and minutes against required daily (5h) or chat (10h) targets.

Dayflow/Dayflow/Core/Access · high confidence

The analysis pipeline now automatically detects idle periods in screenshot batches and generates 'Idle' timeline cards without invoking the LLM, skipping expensive processing for periods where the user was inactive. This is powered by a new \IdleBatchClassifier\ that evaluates coverage and idle ratios against defined rules. Additionally, time parsing performance is improved by caching \DateFormatter\ instances, and the overall batch reprocessing flow has been simplified to handle completion and cleanup more efficiently.

Dayflow/Dayflow/Core/Analysis · high confidence

Dayflow now updates itself silently in the background using Sparkle, removing the previous interactive prompts so updates install immediately without user intervention. The app also introduces a centralized analytics service that tracks CPU usage and captures detailed, PII-sanitized failure logs from AI providers and validation steps, ensuring diagnostic data is sent only when the user has opted in. Additionally, a new launch-at-login manager handles the macOS login item status asynchronously to prevent blocking the app's startup, and the menu bar icon now dynamically reflects the recording state.

Dayflow/Dayflow/System · high confidence

The menu bar interface has been replaced with a new StatusMenuView that provides direct pause and resume controls for screen recording, including a duration picker for temporary pauses (15 mins, 30 mins, 1 hour, or indefinite). The menu now features localized text for all actions such as 'Open Dayflow', 'Open Recordings', 'Check for Updates', and 'Quit Completely', and uses the Dayflow logo in the open app row. The view includes refined styling with rounded corners, material backgrounds, and hover feedback on interactive elements.

Dayflow/Dayflow/Menu · high confidence

The onboarding flow has been restructured to provide a more guided and transparent setup process. Users now encounter a dedicated LLM provider selection step that clearly distinguishes between local engines (like LM Studio), cloud providers (Gemini, OpenAI-compatible), and CLI tools (Codex, Claude). The new LLMProviderSetupView breaks down configuration into a step-by-step sidebar workflow, including explicit CLI detection that checks for installed tools via login shell and offers a direct test to verify connectivity. API key input is now handled by a dedicated component with real-time validation, password visibility toggles, and reassurance about local storage. The 'How It Works' section has been refreshed with animated cards explaining privacy and local AI benefits, and the overall visual theme has been updated with a shared cloud backdrop and consistent SurfaceButton styling.

Dayflow/Dayflow/Views/Onboarding · high confidence

The recording core has been replaced to use periodic screenshots via ScreenCaptureKit instead of continuous video capture, which eliminates the system screen-recording indicator. Captured frames are now stored as hardware-encoded HEVC segments in a new FrameStore, significantly reducing disk usage and improving performance. The system now tracks the active display to auto-switch streams and handles multi-monitor setups more reliably. Additionally, users can block specific applications from being recorded; when a blocked app is active, a localized privacy placeholder is rendered instead of the app's content.

Dayflow/Dayflow/Core/Recording · high confidence

A shared Xcode scheme for the Dayflow project has been added to version control, enabling consistent build, test, launch, profile, analyze, and archive configurations across the team. This scheme includes parallelizable test execution for both unit tests (DayflowTests) and UI tests (DayflowUITests), ensuring that all developers and CI systems use the same debugging and build settings.

Dayflow/Dayflow.xcodeproj/xcshareddata · high confidence

Local builds can now configure the backend URL and PostHog API key via a new Dayflow.xcconfig file and a LocalSecrets.xcconfig.example template. Setting the POSTHOG\_API\_KEY enables analytics and the support chat, while leaving it empty disables these features in local builds.

Dayflow/Config · high confidence

The project license has been changed to the MIT License, replacing the previous Apache-2.0 license. The README has been significantly updated to highlight new product capabilities, including an automatic timeline, daily standup summaries, weekly reviews, and the ability to chat with your work journal. The documentation also clarifies that Dayflow is local-first, supports multiple AI providers (local models, Gemini, ChatGPT, Claude), and provides updated installation instructions via Homebrew and direct download.

(repo-wide) · high confidence

Expanded test coverage for AI agent integration and usage telemetry; Updated test suite for Codex fallback diagnostics and daily recap generation.

The project's Swift Package Manager resolution has been updated to include several new dependencies: NetworkImage (v6.0.1), PostHog iOS SDK (v3.31.0) for analytics, Sentry Cocoa (v8.56.2) for crash reporting, Sparkle (v2.7.1) for automatic updates, swift-cmark (v0.8.0), and Swift Markdown UI (v2.4.1). Additionally, a new standalone, read-only Dayflow CLI executable has been introduced in the tools directory, built with no external dependencies to ensure it can be compiled independently and later integrated into the main application.

(dependencies) · high confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

• CAI 42 → 53 (+10.8)
• Rubric changed (rubric-2026.08.15 → rubric-2026.09.15) — scores are not directly comparable.

• Code Health 100 → 84 (-16.0)
• Architecture 98 (new)
• Maturity 58 → 65 (+7.5)
• Readiness 17 → 33 (+16.2)
• Security 65 → 65 (+0.0)

• Dimension evaluation failed
• High: security finding (details withheld)
• High: security finding (details withheld)
• No automated tests
• No exposed public API
• No tests found
• Secret: generic-api-key (Dayflow/Dayflow/App/AppDelegate.swift)
• Test reliability not included

• ActivityCard.activityDetails (cognitive 24) (Dayflow/Dayflow/Views/UI/MainView/ActivityCard.swift)
• ActivityCard.activityDetails (cyclomatic 19) (Dayflow/Dayflow/Views/UI/MainView/ActivityCard.swift)
• ActivityCard.titleView (cognitive 16) (Dayflow/Dayflow/Views/UI/MainView/ActivityCard.swift)
• AgentCLISupporting.parseCards (cognitive 44) (Dayflow/Dayflow/Core/AI/AgentCLISupport.swift)
• AgentCLISupporting.parseCards (cyclomatic 25) (Dayflow/Dayflow/Core/AI/AgentCLISupport.swift)
• AgentCLISupporting.parseSegments (cognitive 30) (Dayflow/Dayflow/Core/AI/AgentCLISupport.swift)
• AgentCLISupporting.parseSegments (cyclomatic 22) (Dayflow/Dayflow/Core/AI/AgentCLISupport.swift)
• AgentCLISupporting.stripOSCEscapes (cognitive 46) (Dayflow/Dayflow/Core/AI/AgentCLISupport.swift)
• AgentCLISupporting.stripOSCEscapes (cyclomatic 20) (Dayflow/Dayflow/Core/AI/AgentCLISupport.swift)
• AgentPlaybackHost.receive (cognitive 17) (Dayflow/Dayflow/Views/UI/Agents/AgentPlaybackView.swift)
• AgentUsageTelemetry.operation (cyclomatic 18) (tools/dayflow-cli/Sources/dayflow/AgentUsageTelemetry.swift)
• AnalysisManager.queueLLMRequest (cognitive 22) (Dayflow/Dayflow/Core/Analysis/AnalysisManager.swift)
• AnalysisManager.queueLLMRequest (cyclomatic 16) (Dayflow/Dayflow/Core/Analysis/AnalysisManager.swift)
• AnalysisManager.reprocessDay (cognitive 19) (Dayflow/Dayflow/Core/Analysis/AnalysisManager.swift)
• AppDelegate.applicationDidFinishLaunching (cognitive 24) (Dayflow/Dayflow/App/AppDelegate.swift)
• AppDelegate.applicationDidFinishLaunching (cyclomatic 17) (Dayflow/Dayflow/App/AppDelegate.swift)
• Assertions commented out: testExample (Dayflow/DayflowUITests/DayflowUITests.swift)
• CanvasActivityCard.body (cognitive 30) (Dayflow/Dayflow/Views/UI/CanvasActivityCard.swift)
• CanvasActivityCard.body (cyclomatic 20) (Dayflow/Dayflow/Views/UI/CanvasActivityCard.swift)
• ChatCLIProcessRunner.balancedJSONEnd (cognitive 20) (Dayflow/Dayflow/Core/AI/ChatCLIProcessRunner.swift)
• …and 609 more

• 300 commits — 299 feature/other, 1 fixes

• Dayflow/Dayflow — 282 commits
• Dayflow/DayflowTests — 8 commits
• docs/appcast.xml — 7 commits
• Dayflow/Dayflow.xcodeproj — 2 commits
• (root) — 1 commit

• fix: onboarding: fix referral survey state bindings
• change: ChatCLI: use --ignore-user-config on non-streaming Codex runs too, so exec resume isn't pushed into a temporary CODEX_HOME by plugin MCP servers
• change: Flow agent: build the session activity timeline live from each tick, persist it to JSON, serve it to the summary over the bridge
• change: Flow agent: default screenshots to 1080p
• change: Flow agent: model edits timeline.json in place on timeline turns (workspace-write sandbox); native reads it back
• change: Flow agent: model revises the session timeline once a minute (verdict first, timeline second); summary takes the latest without waiting
• change: Flow agent: track goal completion from the screen and relay it to the web app as tasks completed by Flow
• change: Flow overlay: dismiss the break tub after its intro instead of for the whole break
• change: Flow overlay: gapless clip hand-offs, hover-to-reveal bubble, pointer cursor, draggable creature position
• change: Flow overlay: keep the break tub only for the 8s bath intro
• change: Flow web view: allow microphone capture for the Flow origin
• change: Flow: debug panel with live-tunable agent settings and overlay triggers
• change: access: localize recording progress
• change: accessibility: localize toggle values
• change: account: localize authentication and billing messages
• change: account: localize subscription and referral settings
• change: account: localize the referral gift message
• change: account: localize the sign-in sheet
• change: account: refresh subscription status after checkout
• change: agents: improve launch and failure diagnostics
• …and 280 more

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

JerryZLiu/Dayflow was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.