Cal-AI: Open Source Offline Calorie Tracker

Cal-AI is a privacy-first, fully offline, cross-platform (Windows, Android, iOS, Web) application for estimating calories from food images using local AI.

Features

• Attributes: Offline-first, Privacy-focused, Open Source.
• AI: Runs local ONNX models (MobileNetV2) for food recognition.
• Database: Embedded nutrition database for instant lookup.
• Cross-Platform: Built with Tauri v2 + React, running on Desktop, Mobile, and Web.

Prerequisites

• Node.js (v18+)
• Rust (for Tauri desktop/mobile builds)
• VS Code (recommended)

Setup

1. Clone the repository

   git clone https://github.com/agentprojects/cal-ai.git
   cd cal-ai

2. Install dependencies

   npm install

3. Install AI Model (Critical) To enable real AI features, you must place a valid ONNX model in public/model.onnx.

   RECOMMENDED: Download MobileNetV2 (ImageNet or Food-101)

   • Download Link (MobileNetV2)
   • Save as: public/model.onnx

   Note: If no model is found, the app uses a built-in Mock Inference engine for testing purposes.

4. Run Development Server

   # Desktop
   npm run tauri dev
   
   # Web Only
   npm run dev

Building for Production

Web (PWA)

npm run build
# Output in dist/

Windows (Desktop)

npm run tauri build

Android

npm run tauri android init
npm run tauri android build

Architecture

• Frontend: React + TypeScript + Vite + TailwindCSS
• Backend: Tauri (Rust) for system access (Camera/FS permissions on mobile)
• AI Engine: ONNX Runtime Web (WASM/WebGL)
• State: React State (Context/Zustand optional)

Offline Capability

This app works 100% offline.

• The Model is loaded from public/model.onnx (local asset).
• The Data is loaded from src/lib/data/nutrition.json (bundled).
• No API keys or external requests are made during usage.

License

MIT