dcainet/forge-neo-min repository overview

⁠DCAI Forge Neo Minimal

Stable Diffusion WebUI Forge-neo + JupyterLab, tuned for RunPod with Network Volume support.

📖 Full guide: https://www.digitalcreativeai.net/en/post/install-webui-forge-neo-runpod-network-volume⁠

⁠Tags

⁠Features

• Forge-neo (port 7860): Python 3.13 / Torch 2.11.0+cu130. Triton & CivBrowser pre-installed.
• JupyterLab (port 8888): Runs alongside Forge. Password login only when Secrets are set.
• Idle auto-stop: Stops (no NV) or terminates pod (NV attached) after IDLE_TIMEOUT_MINUTES (default 30 min).
• SageAttention auto-rebuild: Compiled for your GPU architecture automatically on each launch.

⁠Quick Start (RunPod)

⁠1. Create Secret

runpod_api_key → RUNPOD_API_KEY

⁠2. Environment Variables

TZ=UTC
RUNPOD_API_KEY="{{ RUNPOD_SECRET_API_AutoStop }}"
IDLE_ENABLE=true
IDLE_TIMEOUT_MINUTES=30
JUPYTER_ENABLE=true
AUTO_UPDATE_FORGE=false
IMAGE_CLEAN_ON_START=false
MAX_AGE_HOURS=6
USE_SAGE=true
FORGE_ARGS=--api --theme dark --disable-safe-unpickle --enable-insecure-extension-access --uv --sage

⁠3. Exposed Ports

Port Service 7860 Forge-neo 8888 JupyterLab (if enabled)

⁠4. First Launch

~20 minutes. SageAttention is built automatically for your GPU.

⁠Directory Layout

├─ forge/
│  ├─ sd-webui-forge-neo/
│  ├─ venv/  (Python 3.13)
│  └─ models/
└─ outputs/

⁠Launch Flags (FORGE_ARGS)

⁠Other

Backup: run forge-backup in terminal → saved to /workspace/outputs, restored on next launch. Auth: set FORGE_USERNAME / FORGE_PASSWORD to enable Gradio login.