Zelara Desktop App

Desktop application for Zelara built with Tauri (TypeScript + Rust).

Overview

The desktop app serves as the processing hub in the Zelara ecosystem:

• CV/ML Processing: ONNX Runtime for recycling image validation
• Device Linking Server: QR code generation, pairing with mobile devices
• Task Offloading: Receives tasks from mobile, processes, returns results
• Full Feature Access: All modules available on capable hardware

Tech Stack

• Frontend: React + TypeScript + Vite
• Backend: Rust + Tauri
• ML/CV: ONNX Runtime (Rust bindings)
• Receipt cleanup fallback: local Python helper with Qwen/Qwen3-0.6B for medium-tier OCR repair when rule parsing is uncertain
• Shared Logic: @zelara packages from core

Receipt AI Fallback

When receipt OCR comes back noisy, Zelara Core now keeps the fast rule parser as the first pass and can invoke a local Qwen3 cleanup helper on medium-tier machines to repair merchant/date/total fields before the mobile form opens.

Python helper dependencies:

python -m pip install -r scripts/requirements-qwen-receipt.txt

For detailed technical documentation, see wiki.