ComfyUI MCP Server

"""Workflow template definitions.""" from typing import Dict, List, Optional from dataclasses import dataclass @dataclass class WorkflowTemplate: """A workflow template with metadata.""" name: str description: str category: str tags: List[str] dsl_content: str parameters: Optional[Dict[str, str]] = None required_models: Optional[List[str]] = None difficulty: str = "beginner" # beginner, intermediate, advanced # Template definitions TEMPLATES = { "text2img_basic": WorkflowTemplate( name="Basic Text-to-Image", description="Simple text-to-image generation with basic prompting", category="Generation", tags=["text2img", "basic", "stable-diffusion"], difficulty="beginner", required_models=["v1-5-pruned-emaonly-fp16.safetensors"], parameters={ "prompt": "a beautiful landscape", "negative_prompt": "blurry, low quality", "width": "512", "height": "512", "steps": "20", "cfg": "7.0", "seed": "42" }, dsl_content="""## Model Loading checkpoint: CheckpointLoaderSimple ckpt_name: v1-5-pruned-emaonly-fp16.safetensors ## Text Conditioning positive: CLIPTextEncode text: {prompt} clip: @checkpoint.clip negative: CLIPTextEncode text: {negative_prompt} clip: @checkpoint.clip ## Generation latent: EmptyLatentImage width: {width} height: {height} batch_size: 1 sampler: KSampler model: @checkpoint.model positive: @positive.conditioning negative: @negative.conditioning latent_image: @latent.latent seed: {seed} steps: {steps} cfg: {cfg} sampler_name: euler scheduler: normal denoise: 1.0 ## Output decode: VAEDecode samples: @sampler.latent vae: @checkpoint.vae save: SaveImage images: @decode.image filename_prefix: text2img_basic """ ), "img2img": WorkflowTemplate( name="Image-to-Image", description="Transform existing images with text prompts", category="Generation", tags=["img2img", "transformation", "stable-diffusion"], difficulty="beginner", required_models=["v1-5-pruned-emaonly-fp16.safetensors"], parameters={ "image_path": "input.png", "prompt": "oil painting style", "negative_prompt": "blurry, low quality", "denoise": "0.7", "steps": "20", "cfg": "7.0", "seed": "42" }, dsl_content="""## Model Loading checkpoint: CheckpointLoaderSimple ckpt_name: v1-5-pruned-emaonly-fp16.safetensors ## Input Image load_image: LoadImage image: {image_path} encode: VAEEncode pixels: @load_image.image vae: @checkpoint.vae ## Text Conditioning positive: CLIPTextEncode text: {prompt} clip: @checkpoint.clip negative: CLIPTextEncode text: {negative_prompt} clip: @checkpoint.clip ## Generation sampler: KSampler model: @checkpoint.model positive: @positive.conditioning negative: @negative.conditioning latent_image: @encode.latent seed: {seed} steps: {steps} cfg: {cfg} sampler_name: euler scheduler: normal denoise: {denoise} ## Output decode: VAEDecode samples: @sampler.latent vae: @checkpoint.vae save: SaveImage images: @decode.image filename_prefix: img2img """ ), "upscaling": WorkflowTemplate( name="Image Upscaling", description="Upscale images using AI super-resolution", category="Enhancement", tags=["upscaling", "super-resolution", "enhancement"], difficulty="intermediate", required_models=["RealESRGAN_x4plus.pth"], parameters={ "image_path": "input.png", "upscale_factor": "4" }, dsl_content="""## Input Image load_image: LoadImage image: {image_path} ## Upscaling upscaler: ImageUpscaleWithModel upscale_model: RealESRGAN_x4plus.pth image: @load_image.image ## Output save: SaveImage images: @upscaler.image filename_prefix: upscaled_{upscale_factor}x """ ), "inpainting": WorkflowTemplate( name="Inpainting", description="Fill masked areas of images with AI-generated content", category="Editing", tags=["inpainting", "editing", "mask", "stable-diffusion"], difficulty="intermediate", required_models=["v1-5-inpainting.ckpt"], parameters={ "image_path": "input.png", "mask_path": "mask.png", "prompt": "beautiful garden", "negative_prompt": "blurry, artifacts", "steps": "20", "cfg": "7.0", "seed": "42" }, dsl_content="""## Model Loading checkpoint: CheckpointLoaderSimple ckpt_name: v1-5-inpainting.ckpt ## Input and Mask load_image: LoadImage image: {image_path} load_mask: LoadImageMask image: {mask_path} channel: alpha ## Text Conditioning positive: CLIPTextEncode text: {prompt} clip: @checkpoint.clip negative: CLIPTextEncode text: {negative_prompt} clip: @checkpoint.clip ## Encoding encode: VAEEncodeForInpaint pixels: @load_image.image vae: @checkpoint.vae mask: @load_mask.mask ## Generation sampler: KSampler model: @checkpoint.model positive: @positive.conditioning negative: @negative.conditioning latent_image: @encode.latent seed: {seed} steps: {steps} cfg: {cfg} sampler_name: euler scheduler: normal denoise: 1.0 ## Output decode: VAEDecode samples: @sampler.latent vae: @checkpoint.vae save: SaveImage images: @decode.image filename_prefix: inpainted """ ), "controlnet_pose": WorkflowTemplate( name="ControlNet Pose Control", description="Generate images following pose guidance from reference image", category="Controlled Generation", tags=["controlnet", "pose", "guided-generation"], difficulty="advanced", required_models=["v1-5-pruned-emaonly-fp16.safetensors", "control_v11p_sd15_openpose.pth"], parameters={ "pose_image": "pose_reference.png", "prompt": "professional dancer in elegant attire", "negative_prompt": "blurry, deformed, low quality", "width": "512", "height": "512", "steps": "20", "cfg": "7.0", "seed": "42", "control_strength": "1.0" }, dsl_content="""## Model Loading checkpoint: CheckpointLoaderSimple ckpt_name: v1-5-pruned-emaonly-fp16.safetensors controlnet: ControlNetLoader control_net_name: control_v11p_sd15_openpose.pth ## Control Input control_image: LoadImage image: {pose_image} ## Text Conditioning positive: CLIPTextEncode text: {prompt} clip: @checkpoint.clip negative: CLIPTextEncode text: {negative_prompt} clip: @checkpoint.clip ## ControlNet Application control_apply: ControlNetApply conditioning: @positive.conditioning control_net: @controlnet.control_net image: @control_image.image strength: {control_strength} ## Generation latent: EmptyLatentImage width: {width} height: {height} batch_size: 1 sampler: KSampler model: @checkpoint.model positive: @control_apply.conditioning negative: @negative.conditioning latent_image: @latent.latent seed: {seed} steps: {steps} cfg: {cfg} sampler_name: euler scheduler: normal denoise: 1.0 ## Output decode: VAEDecode samples: @sampler.latent vae: @checkpoint.vae save: SaveImage images: @decode.image filename_prefix: controlnet_pose """ ), "batch_processing": WorkflowTemplate( name="Batch Image Processing", description="Process multiple images with the same workflow", category="Batch Operations", tags=["batch", "automation", "processing"], difficulty="intermediate", required_models=["v1-5-pruned-emaonly-fp16.safetensors"], parameters={ "input_directory": "input_images/", "prompt": "enhanced and improved", "negative_prompt": "blurry, artifacts", "steps": "15", "cfg": "7.0", "denoise": "0.5" }, dsl_content="""## Model Loading checkpoint: CheckpointLoaderSimple ckpt_name: v1-5-pruned-emaonly-fp16.safetensors ## Batch Input batch_loader: LoadImageBatch mode: incremental index: 0 label: batch_input path: {input_directory} pattern: * ## Text Conditioning positive: CLIPTextEncode text: {prompt} clip: @checkpoint.clip negative: CLIPTextEncode text: {negative_prompt} clip: @checkpoint.clip ## Encoding encode: VAEEncode pixels: @batch_loader.image vae: @checkpoint.vae ## Generation sampler: KSampler model: @checkpoint.model positive: @positive.conditioning negative: @negative.conditioning latent_image: @encode.latent seed: 42 steps: {steps} cfg: {cfg} sampler_name: euler scheduler: normal denoise: {denoise} ## Output decode: VAEDecode samples: @sampler.latent vae: @checkpoint.vae save: SaveImage images: @decode.image filename_prefix: batch_processed """ ), "style_transfer": WorkflowTemplate( name="Style Transfer", description="Apply artistic style from one image to another", category="Artistic", tags=["style-transfer", "artistic", "neural-style"], difficulty="advanced", required_models=["v1-5-pruned-emaonly-fp16.safetensors"], parameters={ "content_image": "content.png", "style_image": "style.png", "style_strength": "0.8", "steps": "25", "cfg": "7.5", "seed": "42" }, dsl_content="""## Model Loading checkpoint: CheckpointLoaderSimple ckpt_name: v1-5-pruned-emaonly-fp16.safetensors ## Input Images content: LoadImage image: {content_image} style: LoadImage image: {style_image} ## Style Analysis style_model: StyleModelLoader style_model_name: style_transfer_model.safetensors style_transfer: StyleModelApply conditioning: @positive.conditioning style_model: @style_model.style_model clip_vision_output: @style.image ## Content Encoding encode: VAEEncode pixels: @content.image vae: @checkpoint.vae ## Text Conditioning positive: CLIPTextEncode text: apply artistic style clip: @checkpoint.clip negative: CLIPTextEncode text: blurry, artifacts, low quality clip: @checkpoint.clip ## Generation sampler: KSampler model: @checkpoint.model positive: @style_transfer.conditioning negative: @negative.conditioning latent_image: @encode.latent seed: {seed} steps: {steps} cfg: {cfg} sampler_name: euler scheduler: normal denoise: {style_strength} ## Output decode: VAEDecode samples: @sampler.latent vae: @checkpoint.vae save: SaveImage images: @decode.image filename_prefix: style_transfer """ ), } def get_template_by_name(name: str) -> Optional[WorkflowTemplate]: """Get a template by name.""" return TEMPLATES.get(name) def get_templates_by_category(category: str) -> List[WorkflowTemplate]: """Get all templates in a category.""" return [t for t in TEMPLATES.values() if t.category == category] def get_templates_by_tag(tag: str) -> List[WorkflowTemplate]: """Get all templates with a specific tag.""" return [t for t in TEMPLATES.values() if tag in t.tags] def get_templates_by_difficulty(difficulty: str) -> List[WorkflowTemplate]: """Get all templates of a specific difficulty level.""" return [t for t in TEMPLATES.values() if t.difficulty == difficulty] def list_all_categories() -> List[str]: """Get all unique categories.""" return list(set(t.category for t in TEMPLATES.values())) def list_all_tags() -> List[str]: """Get all unique tags.""" tags = set() for template in TEMPLATES.values(): tags.update(template.tags) return sorted(list(tags))