中文版 README

lmgo is a Windows system tray application that provides an easy-to-use interface for running local LLM models using llama.cpp server with ROCm GPU acceleration. It's specifically optimized for systems with AMD RYZEN AI MAX+ 395 / 8060S graphics.

This application only works on:

• Operating System: Windows 11
• Processor: AMD RYZEN AI MAX+ 395
• Graphics: Radeon 8060S
• Architecture: x86_64

The embedded llama-server is compiled specifically for ROCm GFX1151 architecture and will not work on other hardware configurations.

• System Tray Interface: Runs in the Windows system tray for easy access
• Automatic Model Discovery: Scans directories for .gguf model files
• Single Model Support: Load and run one model at a time
• Web Interface: Built-in web interface for each loaded model
• Auto-start on Boot: Option to start automatically with Windows
• Notifications: Windows toast notifications for model status
• Model-specific Configuration: Custom arguments for different models
• Automatic Web Browser Launch: Option to automatically open web interface when models load

1. Download the executable: lmgo.exe is a standalone executable
2. Create a models directory: Create a models folder in the same directory as lmgo.exe
3. Place your models: Copy your .gguf model files to the models directory

1. Run lmgo.exe: Double-click the executable
2. Configuration: On first run, a default lmgo.json configuration file will be created
3. System Tray: The application will appear in your system tray (notification area)

1. Right-click the tray icon to access the menu
2. Load Model: Select "Load Model" → choose a model from the list
3. Access Web Interface: Once loaded, select "Web Interface" to open the model's web UI
4. Unload Model: Select "Unload Current Model" to stop the currently loaded model

The application creates a lmgo.json configuration file with the following structure:

{
  "modelDir": "./models",
  "autoOpenWebEnabled": true,
  "notifications": true,
  "basePort": 8080,
  "autoLoadModels": [],
  "defaultArgs": [
    "--prio-batch", "3",
    "--no-host",
    "--ctx-size", "131072",
    "--batch-size", "4096",
    "--ubatch-size", "4096",
    "--threads", "0",
    "--threads-batch", "0",
    "-ngl", "999",
    "--flash-attn", "on",
    "--cache-type-k", "f16",
    "--cache-type-v", "f16",
    "--kv-offload",
    "--no-repack",
    "--direct-io",
    "--mlock",
    "--split-mode", "layer",
    "--main-gpu", "0"
  ],
  "modelSpecificArgs": {}
}

• modelDir: Directory containing .gguf model files
• autoOpenWebEnabled: Automatically open browser when model loads
• notifications: Enable Windows toast notifications
• basePort: Port number for the model (default: 8080)
• autoLoadModels: Model name to load automatically on startup (only one model supported)
• defaultArgs: Default arguments passed to llama-server
• modelSpecificArgs: Custom arguments for specific models

• Lists all discovered .gguf files in the models directory
• Shows sharded models as single entries
• If a model is already loaded, it will be unloaded first

• Stops the currently loaded model
• Menu item is enabled only when a model is running

• Opens browser to the loaded model's web UI
• Menu item is enabled only when a model is running

• Toggle for automatic startup with Windows
• Adds/removes registry entry for auto-start

• Stops all running models
• Cleans up temporary files
• Exits the application

• llama-server: Custom compiled version for ROCm GFX1151
• Icon: Embedded favicon.ico for tray and notifications
• Configuration: Default settings optimized for AMD hardware

• Supports both single-file and sharded (.gguf) models
• Single model at a time (loading new model unloads current one)
• Graceful cleanup on exit

• Windows registry for auto-start configuration
• System tray integration via systray library
• Windows toast notifications
• Console window hidden by default

1. "No .gguf files found"
   • Ensure models are in the correct directory (default: ./models)
   • Check file extensions are .gguf
2. Model fails to load
   • Verify model compatibility with llama.cpp
   • Check available disk space and memory
3. Web interface not accessible
   • Check firewall settings
   • Verify port is not blocked
4. Application doesn't start
   • Verify system requirements (Windows 11, AMD RYZEN AI MAX+ 395 / Radeon 8060S)

Download llama_cpp_rocm_gfx1151.tar.gz first and then

go mod tidy
go build -ldflags "-s -w -H windowsgui" -buildvcs=false .

• favicon.ico: Embedded using //go:embed
• default_config.json: Embedded default configuration
• llama_cpp_rocm_gfx1151.tar.gz: Embedded llama-server binary