noonghunnaandClaude Opus 4.8
092b7c7a46
Hardening from the live cutover validation: - ComfyUI pinned to a known-good commit (COMFYUI_REF default cb9f6394… — has Ideogram-4) instead of floating HEAD, so an upstream change can't silently break users. Set COMFYUI_REF=HEAD to float. Entrypoint is now mounted into the container so pin/bootstrap edits apply on `up` without a 30-min image rebuild (kept +x — a non-exec mounted entrypoint = OCI "permission denied", caught in validation). - OWUI WEBUI_SECRET_KEY no longer hardcoded to "change-this-secret-key" (a shared, forgeable secret); now empty → OWUI generates+persists a strong per-deployment key in the data volume. Override via host env only to share sessions across replicas. Live-validated on the committed compose: end-to-end gen OWUI :8080 → ComfyUI → Ideogram-4 (81 s); boot log confirms "pinned to cb9f6394" + GPU0 pin; OWUI on the persisted key file. Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
47 lines
2.1 KiB
YAML
47 lines
2.1 KiB
YAML
services:
|
|
comfyui:
|
|
build: .
|
|
image: comfyui-local:latest
|
|
container_name: comfyui
|
|
restart: "no"
|
|
ports:
|
|
- "8188:8188"
|
|
volumes:
|
|
# ComfyUI source + venv-installed custom-node code persists here
|
|
- /mnt/models/comfyui/ComfyUI:/workspace/ComfyUI
|
|
# Models tree (overrides ComfyUI/models nested mount)
|
|
- /mnt/models/comfyui/models:/workspace/ComfyUI/models
|
|
- /mnt/models/comfyui/input:/workspace/ComfyUI/input
|
|
- /mnt/models/comfyui/output:/workspace/ComfyUI/output
|
|
- /mnt/models/comfyui/user:/workspace/ComfyUI/user
|
|
# Folder-aliasing yaml so loaders find files whether they look in unet/ or diffusion_models/
|
|
- ./extra_model_paths.yaml:/workspace/ComfyUI/extra_model_paths.yaml:ro
|
|
# Mount the entrypoint (overrides the image-baked copy) so bootstrap/pin changes
|
|
# apply on `up` without a full image rebuild.
|
|
- ./entrypoint.sh:/entrypoint.sh:ro
|
|
# Pip cache so re-launches don't re-download every wheel
|
|
- /mnt/models/comfyui/pip-cache:/root/.cache/pip
|
|
# Share existing HF cache (so anything already pulled by vLLM/SGLang is reusable)
|
|
- /mnt/models/huggingface:/root/.cache/huggingface
|
|
environment:
|
|
- HUGGING_FACE_HUB_TOKEN=${HF_TOKEN:-}
|
|
- HF_HOME=/root/.cache/huggingface
|
|
- PYTHONUNBUFFERED=1
|
|
# GPU pin (opt-in): empty = all reserved GPUs visible (default — image on cuda:0,
|
|
# video can use both). Set to e.g. "0" to confine ComfyUI to one card so a chat LLM
|
|
# can run on the other (the `gpu-mode image-studio` 2-card split passes "0").
|
|
# Honored in entrypoint.sh (only exported when non-empty — empty CUDA_VISIBLE_DEVICES
|
|
# would hide ALL GPUs).
|
|
- COMFYUI_CUDA_VISIBLE_DEVICES=${COMFYUI_CUDA_VISIBLE_DEVICES:-}
|
|
# ComfyUI version pin: empty → the entrypoint's pinned commit (reproducible, has
|
|
# Ideogram-4). Set COMFYUI_REF=HEAD to float on upstream master (bleeding edge).
|
|
- COMFYUI_REF=${COMFYUI_REF:-}
|
|
shm_size: "8gb"
|
|
deploy:
|
|
resources:
|
|
reservations:
|
|
devices:
|
|
- driver: nvidia
|
|
count: all
|
|
capabilities: [gpu]
|