Skip to content
BranchesTags

Repository files navigation

WanderMind LLM - AI-Powered Offline Travel Planner

Flutter Dart Android License

Your Personal AI Travel Assistant That Works Completely Offline

Intelligent itinerary planning • Smart routing • Voice navigation • Zero internet required

About

WanderMind LLM is a revolutionary Flutter-based mobile travel companion that brings the power of Large Language Models (LLM) directly to your device. Unlike traditional travel apps that require constant internet connectivity, WanderMind runs 100% offline using on-device AI powered by Llama.cpp, providing intelligent travel planning, personalized recommendations, and smart navigation even in remote locations.

Why WanderMind?

  • Complete Privacy: Your travel data never leaves your device
  • Truly Offline: Works perfectly without internet connection
  • AI-Powered: Local LLM provides intelligent travel insights
  • Smart Navigation: Advanced routing with scenic route options
  • Budget-Friendly: No API costs, no subscription fees
  • Open Source: Free and open for everyone

Key Features

On-Device AI Intelligence

  • Local LLM Integration: Powered by Llama.cpp with native C++ implementation
  • Smart Travel Assistant: Conversational AI for travel queries and recommendations
  • Personalized Itineraries: AI-generated day-by-day travel plans
  • Context-Aware Suggestions: Understands your preferences and budget
  • Multiple Model Support: Choose between TinyLlama, Phi-2, and other optimized models

Advanced Routing & Navigation

  • Smart Routing Engine: Multi-provider support (OSRM, OpenRouteService, MapBox)
  • Scenic Route Planning: Algorithm-based scenic path optimization
  • Offline Map Support: Download and cache map tiles for offline use
  • Multiple Transportation Modes: Car, bike, walking, and public transport
  • Voice Navigation: Turn-by-turn voice guidance with Flutter TTS
  • Real-time Location Tracking: GPS-based location services

Intelligent Travel Planning

  • Day-by-Day Itineraries: Detailed plans with time slots and activities
  • Budget Management: Track expenses with cost breakdowns
  • Interest-Based Planning: Customize based on your travel interests
  • Activity Categorization: Sightseeing, food, shopping, entertainment, and more
  • Plan Storage: Local persistence with Hive database
  • Edit & Share Plans: Modify and export your travel plans

Interactive Chat Interface

  • Natural Conversations: Chat with AI about travel destinations
  • Route Guidance: "How do I get from Dubai Airport to my hotel?"
  • Local Insights: Cultural tips, safety advice, and recommendations
  • Multi-turn Context: AI remembers your conversation history
  • Rich Responses: Formatted answers with actionable suggestions

Modern UI/UX

  • Material Design 3: Beautiful, intuitive interface
  • Dark Mode Support: Comfortable viewing in any lighting
  • Custom Theming: Poppins font with carefully crafted color schemes
  • Responsive Design: Optimized for various screen sizes
  • Smooth Animations: Polished user experience

Demo

Screenshots

Add your app screenshots here to showcase the UI and features

[Home Screen] [Chat Interface] [Route Planning] [Itinerary View]

Video Demo

Add a video demonstration or GIF showcasing key features

Installation

Prerequisites

  • Flutter SDK: 3.10 or higher
  • Dart SDK: 3.0 or higher
  • Android Studio or VS Code with Flutter extensions
  • Android SDK: API level 24+ (Android 7.0+)
  • NDK: For native C++ compilation (included in Android Studio)

Quick Start

  1. Clone the repository

    git clone https://github.com/yourusername/wandermind_llm.git
cd wandermind_llm

  2. Install dependencies

    flutter pub get

  3. Set up native LLM dependency

    bash android/app/src/main/cpp/setup_llama.sh

    This clones and pins llama.cpp for the Android native build. Required before the first Android build.

  4. Generate code

    flutter pub run build_runner build --delete-conflicting-outputs

  5. Set up environment variables (optional)

    cp .env.example .env
# Edit .env only if you want Mapbox tiles/routing (OpenStreetMap works without it)

    If you use a .env file, uncomment the # - .env line under flutter: assets: in pubspec.yaml.

  6. Run the app

    flutter run

Building for Release

Android APK

bash android/app/src/main/cpp/setup_llama.sh
flutter build apk --release --target-platform android-arm64

Android App Bundle (for Play Store)

flutter build appbundle --release

Architecture

Project Structure

wandermind_llm/
├── lib/
│   ├── core/                    # Core functionality
│   │   ├── config/              # App configuration
│   │   ├── data/                # Training data & knowledge base
│   │   ├── model/               # Data models (Freezed)
│   │   ├── provider/            # Riverpod providers
│   │   ├── routing/             # GoRouter navigation
│   │   ├── service/             # Core services
│   │   ├── theme/               # App theming
│   │   └── widget/              # Reusable widgets
│   │
│   ├── feature/                 # Feature modules
│   │   ├── chat/                # AI chat interface
│   │   ├── explore/             # Map exploration
│   │   ├── home/                # Dashboard & home
│   │   ├── llm/                 # LLM integration
│   │   ├── location/            # GPS & geolocation
│   │   ├── onboarding/          # First-time user flow
│   │   ├── route/               # Routing & navigation
│   │   ├── setting/             # App settings
│   │   └── travel_plan/         # Trip planning
│   │
│   └── main.dart                # App entry point
│
├── android/                     # Android-specific code
│   └── app/src/main/
│       ├── kotlin/              # Kotlin platform channels
│       │   └── dev/abbasian/wandermind_llm/
│       │       ├── LlamaCppService.kt
│       │       ├── LlamaCppChannel.kt
│       │       └── MainActivity.kt
│       └── cpp/                 # Native C++ LLM integration
│           ├── llama_android.cpp
│           ├── llama.cpp/       # Llama.cpp library
│           └── CMakeLists.txt
│
├── assets/
│   ├── data/                    # Training data
│   ├── images/                  # App images
│   ├── fonts/                   # Custom fonts
│   └── models/                  # LLM models (downloaded)
│
└── test/                        # Unit & widget tests

Tech Stack

Frontend

  • Flutter 3.10+: Cross-platform UI framework
  • Dart 3.0+: Programming language
  • Material Design 3: Modern UI components

State Management

  • Riverpod 2.4: Reactive state management
  • Freezed: Immutable data classes
  • GoRouter: Declarative routing

AI & LLM

  • Llama.cpp: On-device LLM inference
  • Native C++ Integration: JNI bindings for Android
  • Platform Channels: Dart ↔️ Kotlin ↔️ C++ communication
  • Model Support: TinyLlama, Phi-2, Mistral (GGUF format)

Maps & Navigation

  • Flutter Map: OpenStreetMap integration
  • OpenRouteService: Routing API
  • OSRM: Open-source routing engine
  • Offline Routing: Custom A* algorithm implementation
  • Scenic Routes: Proprietary scoring algorithm

Storage & Persistence

  • Hive: Fast, lightweight NoSQL database
  • SharedPreferences: Key-value storage
  • Path Provider: File system access

Location Services

  • Geolocator: GPS positioning
  • Geocoding: Address ↔️ coordinates conversion

Networking

  • Dio: HTTP client
  • Connectivity Plus: Network state monitoring

Additional Features

  • Flutter TTS: Voice navigation
  • Logger: Structured logging
  • Image Picker: Photo integration
  • Permission Handler: Runtime permissions

Configuration

LLM Model Setup

WanderMind supports multiple LLM models optimized for mobile devices:

Model Size RAM Required Best For
TinyLlama 1.1B ~600MB 2GB+ Fast responses, basic queries
Phi-2 2.7B ~1.6GB 4GB+ Better reasoning, complex planning
Mistral 7B Q4 ~3.5GB 6GB+ Advanced conversations (high-end devices)

First Launch: The app will automatically download the selected model (TinyLlama by default). This may take 5-15 minutes depending on your connection.

Environment Variables (Optional)

Maps and routing work out of the box with OpenStreetMap and OSRM — no configuration required.

To enable Mapbox tiles and directions, create a .env file in the project root:

# Optional: Mapbox tiles and routing (https://account.mapbox.com/)
MAPBOX_ACCESS_TOKEN=your_token_here

Then uncomment # - .env under flutter: assets: in pubspec.yaml, or pass the token at build time:

flutter run --dart-define=MAPBOX_ACCESS_TOKEN=pk.your_token_here

Testing

Run All Tests

flutter test

Run Specific Tests

flutter test test/feature/llm/llm_service_test.dart

Coverage Report

flutter test --coverage
genhtml coverage/lcov.info -o coverage/html

Performance

Benchmarks (Mid-range Android Device)

  • LLM Initialization: 30-45 seconds (first launch)
  • Response Generation: 2-5 seconds for 100 tokens
  • Route Calculation: <500ms for typical routes
  • Map Rendering: 60 FPS with cached tiles
  • Memory Usage: 300-500MB (TinyLlama loaded)

Optimization Tips

  • Use TinyLlama for devices with <4GB RAM
  • Pre-download map tiles for offline areas
  • Enable route caching for frequently used paths
  • Close other apps for better LLM performance

License

This project is licensed under the MIT License - see the LICENSE file for details.

About

AI-powered offline travel planner with on-device LLM, smart routing, and voice navigation. Built with Flutter & Llama.cpp. Works 100% offline.

Topics

ai-assistantandroidartificial-intelligencedartflutterflutter-appsllama-cppllmmobile-appoffline-firston-device-airiverpodroute-planningtravel-apptravel-planner

Resources

Readme
MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Contributors

Languages