Ana Sayfa / Uygulama Şablonları / LocalAI Mobile - React Native Offline AI Kit

LocalAI Mobile - React Native Offline AI Kit

llama.rn, OCR tarama ve SQLite ile GGUF LLM'lerini %100 çevrimdışı çalıştırmak için Expo 52 ve React Native anahtar teslim başlangıç kiti.

  • Yayın tarihi15 Kasım 2026
  • Paket içeriği.cpp, .css, .ts, .tsx, .jsx, .json, .kt, .gradle, .gradle.kts, .md, package.json, .xml, JavaScript .js
  • Son güncelleme15 Kasım 2026
LocalAI Mobile - React Native Offline AI Kit

Overview

LocalAI Mobile is a production-ready mobile starter kit built with Expo 52, React Native 0.76.9, and TypeScript that allows you to run local GGUF Large Language Models (LLMs) 100% offline directly on user devices using llama.rn and SQLite.

Zero cloud dependencies, zero OpenAI/Anthropic API bills, and complete privacy for your users. All inference, OCR, and document parsing happens entirely on-device.

Includes an interactive offline HTML documentation site (docs/index.html) and comprehensive reskinning guide (CUSTOMIZATION_GUIDE.md) for easy rebranding.

Tested and verified on physical hardware (including Android 14/15, Motorola Edge 50 Neo) and iOS simulators. Pre-configured for EAS builds and local signed release builds.

Features

  • 100% Local Inference: Run GGUF models (Llama 3.2, Qwen 2.5, SmolLM2) directly on the device GPU/CPU via llama.rn.
  • Adaptive C++ Duty-Cycle Throttling: Patched native decode loop prevents phone freezing, overheating, and Android OS out-of-memory (OOM) kills.
  • Background Foreground Service: Keeps streaming responses even when the user minimizes the app or locks their screen, with a persistent Stop notification and wake locks.
  • 6 AI Personas Out-of-the-Box: Instant switching between Everyday Companion, Creative Writer, Feynman Tutor, Executive Assistant, Life Coach, and Developer (with clean code formatting).
  • Multimodal OCR & Document Parsing: On-device camera & photo text scanning using expo-mlkit-ocr, plus native Android document reader for local PDF and Word DOCX files.
  • Atomic SQLite Response Checkpointing: Saves streamed tokens every 2 seconds. If generation is interrupted, chats can be resumed seamlessly.
  • Dynamic Theming: Automatically extracts accessible, high-contrast UI color palettes from custom wallpaper photos.
  • 119 Automated Tests: Fully tested and type-checked test suite included.

Requirements

  • Node.js 18 or higher (LTS recommended)
  • Expo SDK 52
  • React Native 0.76.x
  • Android Studio & Android SDK (API 29+ recommended) for Android builds
  • Xcode 15+ and macOS for iOS builds
  • Device with at least 4GB RAM (6GB+ recommended for 3B parameter models)