UNPKG

comfyui-mcp

Version:

Local-first, agent-native control plane for ComfyUI — MCP server + autonomous sidebar agent that drives your live graph in natural language on ANY LLM: Claude/ChatGPT/Gemini on your subscription (no API key), free local models via Ollama (fully offline),

62 lines (59 loc) • 2.73 kB
# comfyui-mcp RunPod image — extra model paths (FAST-RESTART design). # ============================================================================= # Baked into the image at /opt/ComfyUI/extra_model_paths.yaml and loaded at # launch via: main.py --extra-model-paths-config /opt/ComfyUI/extra_model_paths.yaml # # WHY: ComfyUI software (install + venv + custom_nodes) lives in the IMMUTABLE # image at /opt/ComfyUI. ONLY user data lives on the /workspace network volume. # This file points EVERY model category at /workspace/models/<cat> so: # * your models persist across stop/start (they're on the volume), and # * Manager's install-model writes downloads to the volume — NOT to the image. # # `is_default: true` makes these paths the PRIMARY entries (inserted at the FRONT # of each category's folder list via folder_paths.add_model_folder_path), so they # become the DEFAULT download target. Without it, ComfyUI's built-in # /opt/ComfyUI/models/<cat> (ephemeral, in the image) would stay first and # Manager downloads would NOT persist. # # NOTE: there is deliberately NO `custom_nodes:` key here. custom_nodes must stay # in the image (immutable software). Mapping custom_nodes to the volume would # defeat the fast-restart design and the image-baked-nodes contract. # # base_path is /workspace, and each value is relative to it (matching ComfyUI's # own extra_model_paths.yaml.example convention, e.g. `models/checkpoints/`). # Category KEYS match ComfyUI's folder_names_and_paths keys exactly. # ============================================================================= comfyui_mcp_volume: base_path: /workspace is_default: true checkpoints: models/checkpoints/ configs: models/configs/ loras: models/loras/ vae: models/vae/ text_encoders: | models/text_encoders/ models/clip/ diffusion_models: | models/diffusion_models/ models/unet/ clip_vision: models/clip_vision/ style_models: models/style_models/ embeddings: models/embeddings/ diffusers: models/diffusers/ vae_approx: models/vae_approx/ controlnet: | models/controlnet/ models/t2i_adapter/ gligen: models/gligen/ upscale_models: models/upscale_models/ latent_upscale_models: models/latent_upscale_models/ hypernetworks: models/hypernetworks/ photomaker: models/photomaker/ classifiers: models/classifiers/ model_patches: models/model_patches/ audio_encoders: models/audio_encoders/ background_removal: models/background_removal/ frame_interpolation: models/frame_interpolation/ geometry_estimation: models/geometry_estimation/ optical_flow: models/optical_flow/ detection: models/detection/