Work-Review

Work Review

A local-first personal work review tool that records context, helps you review your day, and generates daily reports.

Automatically organizes the apps you used, websites you visited, window titles, usage time, and optional screenshots into a timeline you can review, analyze, and ask questions about.

All data is stored locally by default and never uploaded to any server. AI features are entirely optional; the app works fine with them turned off.

For personal use only — all data stays on your device.

English · 简体中文 · 繁體中文

What It Solves

Work Review is designed for personal work review and helps answer questions like these:

• What did I actually do today?
• What have I been focusing on over the past few days?
• Roughly how much time did a particular task take?
• Which pages, windows, and context did I look at back then?
• How can I quickly put together today's daily report?

The focus is not "monitoring". It is helping you recall, organize, and review your own work process.

Core Capabilities

• Automatic work context recording — Records foreground apps, browser pages, window titles, usage time, optional screenshots, and OCR text, reducing manual note-taking
• Unified timeline and statistics — Overview, timeline, work assistant, and daily report share the same local records, so you can inspect trends and drill into context
• Questions over local records — Use the basic template or your configured model to answer "What did I do today?", "How long did this task take?", and "What have I been focused on?"
• Daily report generation and export — Generate structured daily reports with Markdown export, auto-export, paragraph editing, pin/hide controls, and cached AI section ordering
• Privacy-first and locally controllable — Data is stored in local SQLite by default; AI is optional, and model calls use your own API key without third-party relay
• Desktop Avatar Beta — Shows work status through a desktop avatar and is gradually expanding toward proactive reminders and context assistance
• Multilingual UI — Available in Simplified Chinese, English, Traditional Chinese, and Arabic (with RTL layout)

Interface Preview

The screenshots below are captured from the running desktop app with localized UI and representative local data, covering the main workflow and configuration surfaces.

Core Workflow

Overview

Timeline

Timeline Detail

Daily Report

Work Assistant

Integrations

Hourly Summary

General Settings

Appearance

AI Models

Desktop Avatar

Privacy

Storage

About

Privacy and Boundaries

Work Review is designed for personal use from the ground up. It is not intended for: employee monitoring · team attendance · performance evaluation · covert tracking

You can control the recording scope as needed:

• Per-app settings: Normal / Anonymize / Ignore — anonymize mode automatically skips screenshots and OCR
• Automatic sensitive keyword filtering · domain blacklist
• Auto-pause on screen lock · manual pause/resume
• AI only activates after you configure a model; disabled by default

Feature Overview

Automatic Recording

• Foreground apps, window titles, browser URLs, usage duration, and category records
• Optional screenshots and OCR with multi-display strategies
• Keyboard/mouse activity + screen-change idle detection to reduce false records
• Timeline review for pages, windows, and context from any time period

Smart Organization

• Work assistant based on local records, with basic template, AI-enhanced modes, and dynamic opening prompts after a model is configured
• Overview with answer-first insights: one-sentence summary, KPI cards with week-over-week deltas, and a "today rhythm" visual card (category composition bar + per-day investment chart)
• Top websites aggregated by domain with browser sources, page counts, duration, inline expansion, and editable semantic categorization
• Duration statistics, category filtering, trend comparison, natural-language time ranges, and hourly activity across Today, Week, Date, and Range views
• Fragments grouped into continuous work sessions
• Built-in categorization knowledge base (~140 domains + 70 apps) with AI auto-learning that backfills history

Daily Reports and Review

• Structured daily reports with TL;DR insight line, reference KPIs, and proof data panel
• Export dropdown (Markdown / range merge / copy full text) and generation settings drawer (presets / system prompt override / paragraph management)
• Skeleton screen during generation and prominent yesterday-report banner
• Historical review, auto-export, paragraph-level pin/hide/restore controls, with cached AI section ordering
• AI-enhanced prompt attachments and custom model settings
• Website semantic categorization: changing a domain category automatically backfills history
• Multi-segment work time: e.g. morning + afternoon, break time excluded

AI Modes

The core of Work Review is always local recording. AI's role is to make records easier to read and review, not a prerequisite for usage.

Supported providers: Ollama (local) / LM Studio (local) / OpenAI compatible / DeepSeek / Qwen / Zhipu / Kimi / Doubao / MiniMax / SiliconFlow / Gemini / Claude / OpenRouter / Groq / xAI Grok / Mistral / Custom endpoint

Quick Start

1. Download the installer for your platform from Releases
2. On macOS, grant Screen Recording and Accessibility permissions
3. Let it run in the background for a while
4. Check the Overview / Timeline / Daily Report to see your recorded activity

macOS: Screenshots require the "Screen Recording" permission, and avatar linkage requires "Accessibility + Input Monitoring". If you see a "damaged" warning on first launch: sudo xattr -rd com.apple.quarantine "/Applications/Work Review.app"

Windows: Depends on Microsoft Edge WebView2 Runtime.

Windows 11 Smart App Control: If Windows shows "Smart App Control blocked this app from running", it's because Work Review is signed with a self-signed certificate (not a commercial EV certificate). You can turn off Smart App Control in "Windows Security → App & browser control → Smart App Control". See #139.

Linux: Screenshots and window tracking depend on the current session type and toolchain.

# Base
sudo apt install xprintidle tesseract-ocr
# X11
sudo apt install xdotool x11-utils scrot
# Wayland: gdbus (GNOME) / kdotool (KDE) / swaymsg (Sway) / hyprctl (Hyprland)
# Screenshots: grim / gnome-screenshot / spectacle

# RPM-based distributions
sudo dnf install xprintidle tesseract xdotool xorg-x11-utils gnome-screenshot procps-ng
sudo zypper install xprintidle tesseract-ocr xdotool xorg-x11-tools gnome-screenshot procps

AppImage is not a universal package for every Linux distribution. The GLIBC compatibility gate checks only the main program ELF inside each Linux release package; it does not scan every bundled library or plugin. The current maximum allowed GLIBC version is 2.35. The actual requirement of each build is shown in the release workflow's GLIBC check output.

Ubuntu 24.04 / 24.10 Wayland (GNOME 46–47) users experiencing screenshot flash or shutter sound can use the one-click install script to fix it automatically:

bash scripts/deb/reinstall.sh           # deb method (recommended)
bash scripts/deb/reinstall.sh --dry-run # preview operations

See scripts/ubuntu-wayland-README.md for details.

KDE Plasma / Wayland startup crash (Fedora, Arch, openSUSE, …): If the app quits immediately with Gdk-Message: Error 71 (Protocol error) dispatching to Wayland display., it is an upstream webkit2gtk/GTK bug on Wayland (see tauri#10702), most common on KDE Plasma + NVIDIA. Recent builds already inject WEBKIT_DISABLE_DMABUF_RENDERER=1 at startup. If you still hit it on an older build, launch it with the same workaround manually:

WEBKIT_DISABLE_DMABUF_RENDERER=1 ./Work_Review

If that still fails, force the X11 backend as a last-resort fallback. This may render worse on some Wayland desktops:

GDK_BACKEND=x11 ./Work_Review

Extended Capabilities (Beta)

Uses a standalone desktop pet window to reflect idle / working / reading / meeting / music / video states.

Still being actively refined — interaction linkage, state expression, and preset details will continue to improve.

Query records and generate daily reports remotely from Telegram / Feishu / DingTalk / WeCom via local API + multi-device registration. Supported commands: /devices, /report, /generate, etc. WeCom prefers the intelligent-bot long connection (Bot ID + Secret, no public URL); only one live connection is allowed per bot. Restricted to personal and own multi-device use only.

After Localhost API is enabled, the app exposes a local HTTP API (default 127.0.0.1:47831), authenticated with a Bearer Token (auto-generated on first launch, saved in localhost_api_token.txt in the data directory).

Authentication

All requests (except /health and Feishu callbacks) must carry the token:

Authorization: Bearer <token>

Or via query parameter: ?token=<token>

Endpoints