Add README with image/workspace usage details
Browse files
README.md
ADDED
|
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# comfyui-qwen
|
| 2 |
+
|
| 3 |
+
RunPod image for ComfyUI with Qwen/Flux workflows. Docker image: `plx1029/comfyui-qwen`
|
| 4 |
+
(current: `v4`). Everything lives in `/workspace` — do not move files around:
|
| 5 |
+
|
| 6 |
+
```
|
| 7 |
+
/workspace/
|
| 8 |
+
├── ComfyUI/ # ComfyUI itself (models under ComfyUI/models/)
|
| 9 |
+
├── workflows/ # workflow JSONs used by the download script
|
| 10 |
+
├── download_missing_models.sh
|
| 11 |
+
└── README.md
|
| 12 |
+
```
|
| 13 |
+
|
| 14 |
+
## Starting ComfyUI
|
| 15 |
+
|
| 16 |
+
ComfyUI does **not** auto-start (since v4) — start it yourself:
|
| 17 |
+
|
| 18 |
+
```bash
|
| 19 |
+
comfy # wrapper: runs ComfyUI on port $COMFY_PORT (default 7865)
|
| 20 |
+
comfy --enable-manager # extra args are passed through to main.py
|
| 21 |
+
# or directly:
|
| 22 |
+
cd /workspace/ComfyUI && python main.py --listen 0.0.0.0 --port 7865 --enable-manager
|
| 23 |
+
```
|
| 24 |
+
|
| 25 |
+
Set `COMFY_AUTOSTART=true` on the pod to re-enable auto-start at boot.
|
| 26 |
+
|
| 27 |
+
## Models + custom-node dependencies
|
| 28 |
+
|
| 29 |
+
```bash
|
| 30 |
+
download_missing_models.sh # interactive workflow picker
|
| 31 |
+
download_missing_models.sh --all # every workflow
|
| 32 |
+
download_missing_models.sh --dry-run # show the plan only
|
| 33 |
+
download_missing_models.sh --skip-nodes # models only, skip custom-node installs
|
| 34 |
+
```
|
| 35 |
+
|
| 36 |
+
For the workflows you select, the script:
|
| 37 |
+
|
| 38 |
+
1. installs every custom-node pack they need (resolved from each node's
|
| 39 |
+
`cnr_id`/`aux_id` via the Comfy Registry, with ComfyUI-Manager's node DB as
|
| 40 |
+
fallback) into `ComfyUI/custom_nodes/`, including pip requirements — restart
|
| 41 |
+
ComfyUI afterwards to load them;
|
| 42 |
+
2. downloads every model they reference from this HF repo
|
| 43 |
+
(`aleph65/ComfyUI`, which mirrors `ComfyUI/models/`) into place, in parallel.
|
| 44 |
+
|
| 45 |
+
Auth: reads `HUGGING_FACE_ACCESS_TOKEN` (falls back to `HF_TOKEN`), prompts if
|
| 46 |
+
neither is set.
|
| 47 |
+
|
| 48 |
+
## How this stays up to date
|
| 49 |
+
|
| 50 |
+
The source of truth for `download_missing_models.sh` and this README is the HF
|
| 51 |
+
repo `aleph65/ComfyUI`. At container boot, `/start-comfy.sh` pulls the latest
|
| 52 |
+
versions from there (best-effort; keeps the baked-in copies if offline). So:
|
| 53 |
+
|
| 54 |
+
- **script/README changes** → push to the HF repo, no image rebuild needed;
|
| 55 |
+
- **anything else** (start scripts, ComfyUI, system deps) → rebuild and push
|
| 56 |
+
a new `plx1029/comfyui-qwen` tag.
|
| 57 |
+
|
| 58 |
+
If a network volume is mounted over `/workspace`, missing files are restored
|
| 59 |
+
at boot from the baked seed (`/opt/workspace-seed.tar.gz`); existing files on
|
| 60 |
+
the volume are never overwritten.
|