LlamaServer Manager

A web-based management interface for llama.cpp. Easily configure and run llama-server instances with a simple web UI.

Features

• Web-based dashboard for managing llama-server
• Model selection and upload
• Interactive parameter configuration
• Chat interface with real-time responses
• Real-time server metrics and logging
• Integrated llama.cpp server management

Requirements

• Debian/Ubuntu (or compatible)
• Python 3.10+
• 2GB+ RAM
• Internet connection for initial setup

Installation

Run the interactive installer:

sudo bash install.sh

The installer will:

• Install system dependencies
• Set up Python virtual environment
• Configure llama.cpp server
• Create systemd service
• Set up logging

Usage

Start the service

sudo systemctl start llama-manager

Access the web interface

Open your browser to:

http://localhost:8484

View logs

sudo journalctl -u llama-manager -f

Configuration

Configuration is stored in /opt/llama-manager/config/config.json.

Key settings:

• port: Web interface port (default: 8484)
• llama_server_port: llama-server port (default: 8080)
• models_dir: Directory for downloaded models
• log_level: Logging level (debug, info, warning, error)

Architecture

• Backend: Python Flask with SocketIO for real-time updates
• Frontend: Vanilla JavaScript with real-time metrics display
• Server: gevent async server with background threading
• Integration: Manages llama.cpp subprocess

Troubleshooting

If the service fails to start:

# Check logs
sudo journalctl -u llama-manager -n 50

# Check if port is in use
sudo lsof -i :8484

# Verify installation
ls -la /opt/llama-manager

License

See LICENSE file in repository.