Professional Desktop Workbench for Game Localization QA

AI-assisted desktop application for game localization quality assurance.

Built with Tauri v2 · Rust · React · TypeScript · SQLite · OCR · Image Analysis · Video Analysis · Knowledge Base · Enterprise Desktop UI

• Image LQA — Upload screenshots, run OCR, get AI-proposed issues matched against severity rules
• Video LQA — Import gameplay footage, extract frames, analyze through the same pipeline
• OCR Engine — Tesseract-powered text extraction with bounding geometry
• Evidence Engine — Visual red-box and red-bracket annotations over source images
• Severity Rules — Import severity matrix; issues matched deterministically
• Glossary Authority — Import terminology references; flag inconsistencies
• Reference Docs — Provide context documents for AI analysis
• Knowledge Base — Project-local semantic search across all imported data
• Batch Analysis — Queue multiple images for sequential analysis
• BYOK — Supports Ollama (local), OpenAI-compatible, Gemini, LM Studio
• CSV Export — Export reviewed issues for reporting
• Audit Trail — Append-only event log for every significant action
• Command Palette — Ctrl+K keyboard-driven navigation

┌──────────────────────────────────────────────┐
│              Tauri v2 Desktop Shell           │
├──────────────────┬───────────────────────────┤
│   Frontend       │       Backend             │
│   React 18       │       Rust                │
│   TypeScript     │       SQLite (rusqlite)   │
│   Vite           │       Image processing    │
│   Lucide Icons   │       Video extraction    │
│                  │       AI provider routing  │
│                  │       Evidence engine      │
│                  │       CSV export           │
├──────────────────┴───────────────────────────┤
│              Python Sidecar                   │
│              Glossary parser                   │
│              Severity rules parser             │
│              OCR adapter                       │
└──────────────────────────────────────────────┘

Screenshots will be added in a future release.

lqa-workbench/
├── src/                          # React frontend
│   ├── api/
│   │   └── tauri.ts              # Tauri IPC bridge
│   ├── components/               # 21 React components
│   ├── App.tsx                   # Root component + routing
│   ├── main.tsx                  # Entry point
│   ├── styles.css                # Design system (CSS custom properties)
│   └── types.ts                  # Shared TypeScript types
├── src-tauri/                    # Tauri / Rust backend
│   ├── src/                      # 11 Rust source files
│   ├── parsers/                  # Python sidecar parsers
│   ├── icons/                    # Application icons
│   ├── capabilities/             # Tauri security capabilities
│   ├── Cargo.toml                # Rust dependencies
│   └── tauri.conf.json           # Tauri configuration
├── package.json                  # NPM dependencies
├── vite.config.ts                # Vite + Vitest configuration
└── README.md

• Rust — Latest stable (backend compilation)
• Node.js — 18+ (frontend tooling)
• Ollama — Latest (local AI inference, optional)
• Tesseract OCR — 5+ (text extraction)
• FFmpeg — Latest (video processing)
• Python 3 — 3.10+ (sidecar parsers)
• pip install openpyxl — Excel parsing

# Clone
git clone https://github.com/irunna-ai/lqa-workbench.git
cd lqa-workbench

# Install frontend dependencies
npm install

# Install Python dependencies
pip install openpyxl

# Run in development mode
npm run tauri dev

# Run frontend tests
npm test

# Type check
npx tsc --noEmit

# Run backend tests
cargo test

# Build for production
npm run build

• macOS support
• Linux support
• Plugin architecture for custom analyzers
• Multi-language OCR
• Collaborative review mode
• Custom report templates
• Tracker integrations (Jira, Azure DevOps)
• Automated regression detection
• Accessibility improvements
• Localization of the workbench itself

Copyright (c) 2026 Irunna AI. All rights reserved.

See LICENSE.md for details.

Contributions are welcome. Please open an issue first to discuss proposed changes.

• Follow existing code conventions
• Keep changes focused and atomic
• Write tests for new functionality
• Never commit credentials or secrets
• Never store API keys in plaintext

• Local-first — All data stays on your machine
• No telemetry — No cloud upload, no analytics
• Secure credentials — API keys stored in OS keyring
• Project isolation — Each project's data strictly separated
• Least privilege — Tauri capabilities grant only what is needed

See SECURITY.md for the full policy.

• Tauri — Desktop application framework
• Ollama — Local AI inference
• Tesseract OCR — Optical character recognition
• FFmpeg — Video processing
• React — UI library
• Lucide — Icons
• Vite — Build tooling