A Python script to install and manage the Draw Things Community gRPCServer on macOS. This installer handles downloading the latest gRPCServer binary, setting up a LaunchAgent service, and managing the server configuration.

• macOS 15
• Python 3.9+

1. Download the installer script:

git clone https://github.com/funkatron/draw-things-grpcservercli-installer.git

# Basic installation with defaults
./install-grpcservercli.py

# Install with custom model path
./install-grpcservercli.py -m /path/to/models

# Install with custom port and server name
./install-grpcservercli.py -p 7860 -n "MyServer"

# Install with security options
./install-grpcservercli.py -s "mysecret"

# Install with model browser enabled
./install-grpcservercli.py --model-browser

# Install with proxy configuration
./install-grpcservercli.py --join '{"host":"proxy.local", "port":7859}'

# Restart the service
./install-grpcservercli.py --restart

# Quiet install with defaults
./install-grpcservercli.py -q

# Uninstall
./install-grpcservercli.py --uninstall

• -m, --model-path: Custom path to store models (default: Draw Things app models directory)
• -h, --help: Show help message
• --uninstall: Uninstall gRPCServerCLI and remove all related files
• --restart: Restart the gRPCServerCLI service
• -q, --quiet: Minimize output and assume default answers to prompts

• -n, --name: Server name in local network (default: machine name)
• -p, --port: Port to run the server on (default: 7859)
• -a, --address: Address to bind to (default: 0.0.0.0)
• -g, --gpu: GPU index to use (default: 0)
• -d, --datadog-api-key: Datadog API key for logging backend
• -s, --shared-secret: Shared secret for server security
• --no-tls: Disable TLS for connections (not recommended)
• --no-response-compression: Disable response compression
• --model-browser: Enable model browsing
• --no-flash-attention: Disable Flash Attention
• --debug: Enable verbose model inference logging
• --join: JSON configuration for proxy setup

The --join option accepts a JSON string for proxy configuration:

{
  "host": "proxy.example.com",
  "port": 7859,
  "servers": [
    {
      "address": "gpu1.local",
      "port": 7859,
      "priority": 1
    }
  ]
}

Required fields:

• host: The proxy server hostname
• port: The proxy server port

Optional fields:

• servers: List of GPU servers with:
  • address: Server hostname
  • port: Server port
  • priority: Server priority (1=high, 2=low)

The installer creates a LaunchAgent service that:

• Starts automatically on login
• Restarts automatically if it crashes
• Can be managed with:

  # Stop the service
  launchctl unload ~/Library/LaunchAgents/com.drawthings.grpcserver.plist
  
  # Start the service
  launchctl load ~/Library/LaunchAgents/com.drawthings.grpcserver.plist
  
  # Restart the service
  ./install-grpcservercli.py --restart

1. Port already in use

   • The installer will detect if the port is in use and show which process is using it
   • Use a different port with -p or stop the conflicting process

2. Permission denied when installing to /usr/local/bin

   • The installer will automatically fall back to ~/.local/bin
   • You can run with sudo if you need to install in /usr/local/bin

3. Service fails to start

   • Check the system log for errors: log show --predicate 'processImagePath contains "gRPCServerCLI"' --last 5m
   • Ensure the model path exists and is accessible
   • Try restarting with --restart option

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

• Draw Things
• Draw Things Community