CalAI Tracker 🥗

High-performance AI-powered calorie and nutrition tracker with local LLM support.

🚀 Key Features

• AI Food Analysis: Instant nutritional estimation from food images.
• Local Inference (Cactus Engine): Run Gemma 4 models locally on-device for privacy and offline capability.
• Hybrid Architecture: Seamlessly switch between Cloud APIs and Native Local Inference.
• Modern UI: Built with Jetpack Compose using a luxury glassmorphism aesthetic.

🛠 Technical Stack

• Frontend: Kotlin, Jetpack Compose
• Native Core: C++ (Cactus Engine)
• AI Models: Gemma 4 (Local) / OpenAI-Compatible APIs (Cloud)
• Infrastructure: Android NDK, JNI, ARM64 Optimization

📦 Native Integration

This project utilizes the Cactus Engine for high-performance local inference.

• Native Library: libcactus.so (ARM64)
• JNI Bridge: libcalai_jni.so provides a thin layer for the JVM to communicate with the C++ core.

⚙️ Local Setup

To use the local AI engine:

1. Open Settings.
2. Select a local GGUF/Cactus compatible model path on your device.
3. The app will automatically route analysis requests to the local engine.

🛠 Build Instructions

Due to the complexity of the native ARM64 build, it is recommended to build the APK using Android Studio on a PC:

1. Clone the repo.
2. Open in Android Studio (Ladybug or newer).
3. Build $\rightarrow$ Build Bundle(s) / APK(s) $\rightarrow$ Build APK(s).