🧠 Local LLM mobile app

A modern, high-performance Flutter application that serves as a unified mobile client for local Large Language Models. It allows seamless switching between Llama.cpp, Ollama, and LM Studio servers via local network or Ngrok tunnels.

📱 App UX

local_llm_chat_mobile_app.mp4

✨ Key Features

• Multi-Provider Support: unified interface for multiple AI backends:
  • 🦙 Llama.cpp (via llama-server)
  • 🧶 Ollama
  • 🧪 LM Studio
• Modern UI: Cyberpunk-inspired dark theme with neon accents and lucide_icons.
• Smart Connectivity:
  • Supports Ngrok tunneling for remote access.
  • Supports Localhost (10.0.2.2) for Android Emulator.
  • Auto-handling of API endpoints (/v1/chat/completions).
  • Bypasses Ngrok browser warning pages via custom headers.
• Clean Architecture: Feature-first structure with separated Logic (Providers), UI, and Services.

🛠 Tech Stack

• Framework: Flutter & Dart
• State Management: Provider (ChangeNotifier)
• Networking: http (Direct API calls to control headers and timeouts)
• UI/UX: google_fonts (Space Mono/JetBrains), lucide_icons

📂 Project Structure

The project follows a Feature-First architecture to ensure scalability.

lib/
├── main.dart                   # Application Entry Point
├── core/                       # Core layer (Theme, Models)
│   ├── theme/                  # App Theme & Colors
│   └── models/                 # ChatMessage, LlmProvider Enum
├── services/                   # Network Layer
│   └── api_service.dart        # Universal HTTP Client for LLMs
└── features/                   # Feature layer
    ├── home/                   # Provider Selection Screen
    │   ├── home_screen.dart
    │   └── widgets/            # Provider Cards, Info Panels
    └── chat/                   # Chat Interface
        ├── chat_provider.dart  # State Management & Logic
        ├── chat_screen.dart
        └── widgets/            # Message Bubbles, Input Fields

🚀 Getting Started

Follow these steps to set up and run the project locally.

1️⃣ Prerequisites (Backend Setup)

Before running the app, ensure that at least one local LLM server is running.

Option A — Llama.cpp

Run the server and expose it to your local network:

llama-server -m path/to/model.gguf --port 8081 --host 0.0.0.0

Option B — Ollama

Important: Ollama must be configured to accept external requests.

OLLAMA_HOST=0.0.0.0 OLLAMA_ORIGINS="*" ollama serve

Option C — LM Studio

1. Open LM Studio
2. Go to Developer Tab (<->)
3. Start the server on Port 1234
4. Ensure Cross-Origin Resource Sharing (CORS) is enabled

2️⃣ Network Setup (Connecting Mobile to Localhost)

Since the app runs on a physical device or emulator, it cannot access localhost directly.

Android Emulator

Use the special Android host IP:

http://10.0.2.2:PORT

Example for Ollama:

http://10.0.2.2:11434

Physical Devices (iOS / Android)

Use Ngrok to expose your local server to the internet.

Example for LM Studio:

ngrok http 1234

Copy the generated https:// URL and paste it into the app.

3️⃣ Running the App

Clone the repository:

git clone https://github.com/YOUR-USERNAME/local-llm-hub.git
cd local-llm-chat

Install dependencies:

flutter pub get

Run the application:

flutter run

iOS Note:
When running on a physical iPhone (iOS 17+), use release mode to avoid memory protection issues:

flutter run --release

📄 License

This project is provided for educational purposes as part of AI Frameworks Training.