Author: Watson
ComfyUI node-based Stable Diffusion GUI, served at https://comfyui.beachlab.org with basic auth.
- Stack
- Install paths
- Service
- Nginx
- Change auth password
- Add/update models
- Custom nodes (ComfyUI Manager)
- Update ComfyUI
- DNS + SSL setup (one-time, after DNS propagation)
- Troubleshooting
- Service:
comfyui.service— port8188(127.0.0.1 only) - Frontend: ComfyUI frontend v1.39.14 (bundled via pip, auto-updated)
- Proxy: Nginx →
comfyui.beachlab.orgwith basic auth + HTTPS - GPU: RTX 2070 Super 8GB, CUDA 12.4, PyTorch 2.6.0
- Auth:
/etc/nginx/.htpasswd-comfyui(user:fran)
| Path | Purpose |
|---|---|
/opt/comfyui/ |
App root |
/opt/comfyui/.venv/ |
Python venv |
/opt/comfyui/models/ |
Models (checkpoints, loras, VAE, etc.) |
/opt/comfyui/output/ |
Generated images |
/opt/comfyui/input/ |
Input images |
/opt/comfyui/custom_nodes/ |
Extensions |
sudo systemctl status comfyui
sudo systemctl restart comfyui
journalctl -u comfyui -fsudo nginx -t && sudo nginx -s reload
cat /etc/nginx/sites-available/comfyui.beachlab.orgsudo htpasswd -b /etc/nginx/.htpasswd-comfyui fran NEW_PASSWORD
sudo nginx -s reloadDrop checkpoint files into /opt/comfyui/models/checkpoints/.
Refresh model list in ComfyUI UI: Manager → Refresh.
Install ComfyUI Manager for easy node/extension management:
cd /opt/comfyui/custom_nodes
git clone https://github.com/Comfy-Org/ComfyUI-Manager.git
sudo systemctl restart comfyuicd /opt/comfyui
git pull
.venv/bin/pip install -r requirements.txt
sudo systemctl restart comfyui- Add DNS A record:
comfyui.beachlab.org→85.49.212.28 - Wait for propagation, then run:
sudo /opt/comfyui/finish-ssl.shCUDA out of memory: Other GPU services (Whisper, TTS) share the 8GB VRAM.
Stop them before running heavy models:
sudo systemctl stop whisper-web qwen3-tts
# ... run ComfyUI generation ...
sudo systemctl start whisper-web qwen3-ttsNo models available: Download a checkpoint and place in models/checkpoints/.
Free options: SDXL Turbo, SD 1.5.
Port conflict on 8188: ss -tlnp | grep 8188