TauriSight

Preview · Features · Project Structure · Installation · Development · Release

---

Tray-first Tauri desktop app for local image object detection with compact native-style React UI.

Uses TensorFlow.js Coco SSD to detect common COCO objects in uploaded images.

Preview

What It Includes

Desktop App Experience

• Tray-only desktop behavior with the main window hidden until the tray icon is clicked.
• Compact dark/light native-style interface for macOS and Windows.
• Latest upload preview in the main tray panel.
• Tauri tray menu with select file, clipboard upload, recent uploads, about, settings, and quit.
• Dynamic tray progress icon using the app icon plus a theme-aware progress bar.

Media Input & Processing

• Drag-and-drop media dropzone with file picker, clipboard image import, image support, and short MP4/MOV video support.
• Uploading and analyzing states with inline progress in the dropzone.
• TensorFlow.js Coco SSD object detection for 80 common object classes.
• Local image analysis inside the app webview; no custom upload server.
• Annotated image/video export to a user-selected local folder.

Export & File Management

• Required first-run export folder setting with guarded writes from Rust.
• Recent uploads page with expandable entries, status, confidence scores, dimensions, exported filename, reveal-in-Finder/File Explorer, and removable entries.
• Export existence check, so moved/deleted exports are shown as missing instead of silently failing.

Settings & Localization

• Settings for preferred video export format and UI language (en / pt).
• Localized about panel driven by frontend i18n (src/i18n/*.json).

Updates & Releases

• GitHub Actions release workflow for macOS, Windows, and Linux builds.
• Automatic in-app updates via Tauri Updater + GitHub Releases.
• Signed release artifacts with generated latest.json metadata.
• Silent update checks on app startup.

Technical Notes

• Built with Tauri + React + TensorFlow.js.
• Native-style lightweight desktop architecture with local-first processing.

Project Structure

• src/App.tsx: main React UI, file handling, clipboard import, and TensorFlow.js analysis.
• src/App.css: compact native-style UI theme and responsive dark/light styling.
• src/i18n/: frontend translations for supported UI languages.
• src-tauri/src/lib.rs: Tauri tray, window positioning, dock/taskbar behavior, and tray progress icon.
• src-tauri/tauri.conf.json: Tauri app, security, window, and bundle configuration.
• src-tauri/icons/: app and tray icon assets.
• .github/workflows/main.yml: tagged/manual release workflow for GitHub Releases.

Prerequisites

• Node.js: v18+ recommended. Download: https://nodejs.org/

macOS (via Homebrew):

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
brew install rust pnpm

Windows (via Chocolatey or Scoop):

• Chocolatey: https://chocolatey.org/install

  choco install rust pnpm

• Scoop: https://scoop.sh/

  scoop install rust pnpm

Linux or manual install:

• pnpm: npm install -g pnpm or https://pnpm.io/installation
• Rust: https://www.rust-lang.org/tools/install (run curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh)

Note: On Windows ensure MSVC toolchain available if building native bundles.

Local Run (development)

1. Install dependencies:

pnpm install

2. Start dev app (hot reload):

pnpm tauri dev

3. If you prefer a local desktop build instead of running the dev server, create a distributable:

pnpm tauri build

Use pnpm tauri build when you want a production binary without running the dev server each time.

Release

Automatic GitHub releases run when a version tag is pushed.

git tag v0.3.8
git push origin v0.3.8

The workflow can also be started manually from GitHub Actions with workflow_dispatch.

Installing from GitHub release (macOS / Windows notes)

• macOS: Gatekeeper may block first-run for unsigned apps. If the app is blocked:

  • Open System Settings → Privacy & Security.
  • Look for a message about the blocked app and click Open Anyway (or right-click the app → Open).

  This may be necessary after downloading from GitHub Releases on first run.

• Windows: SmartScreen or Defender may warn on first-run of an unsigned app.

  • If blocked by SmartScreen, choose More info → Run anyway.
  • To permanently unblock a downloaded installer/executable: right-click the file → Properties → check Unblock → OK.

Note: signed/notarized builds avoid these prompts; release artifacts here are unsigned unless you configure signing.

Notes

• The first analysis downloads and loads the Coco SSD model, so it can take longer than later runs.
• Video export uses browser MediaRecorder; output is WebM for broad local browser support.
• Long videos are capped for safe local processing.
• Image files are decoded locally in the Tauri webview and are not sent to a project backend.
• Coco SSD model assets are fetched from TensorFlow-hosted storage at runtime.
• Clipboard image import depends on operating system and webview permissions.
• Release builds are unsigned unless signing and notarization are configured separately.

Trademark Notice

TauriSight, the TauriSight logo, and associated branding are trademarks of no-tone.

The Apache License 2.0 applies to the source code only and does not grant permission to use project branding, logos, or product identity without permission.