What problem does it solve? Running ComfyUI generation jobs from OpenMontage requires correct server configuration, API-format workflow exports, output node selection, and model placement, and mistakes in any of these cause failed or unreproducible generations. ## Core Features & Use Cases - Workflow Selection and Execution: Choose between bundled workflows and custom workflow_json/workflow_path inputs for comfyui_image, comfyui_video, and comfyui_music, including low-VRAM options like Wan 2.1 1.3B, LTXV FP8, and GGUF quantized workflows for 8GB-12GB GPUs. - Output Node and Model Management: Identify the correct terminal saver node (SaveImage, SaveVideo, VHS_VideoCombine, SaveAudioMP3), place diffusion models, text encoders, VAEs, and LoRAs in the expected ComfyUI model folders, and handle structured missing_models responses. - Failure Recovery and Provenance: Resume timed-out jobs with resume_prompt_id, record workflow_name, workflow_model, and workflow_model_stack for reproducibility, and distinguish hosted Partner Nodes from local open-weight paths. - Use Case: A user imports a community Wan 2.2 GGUF video workflow, passes it as workflow_path with output_node set to the VHS_VideoCombine node ID, and the tool reports any missing models with download hints before generating. ## Quick Start Use the comfyui skill to generate a short video with the bundled Wan workflow, or run my custom workflow file with the correct output node selected.
