dsh-comfyui-commander

Manually operating ComfyUI means dragging nodes in a browser, tweaking parameters one by one, staring at the render queue, and grabbing output files by hand. Multi-stage workflows and batch storyboard runs make the process even more fragmented. dsh-comfyui-commander hands the entire pipeline to a DSH Agent: you describe what you want in plain language, and the Agent chains workflow selection, parameter overrides, generation submission, progress tracking, and result download automatically, freeing you from repetitive clicking.\n\nCore capabilities span the full ComfyUI control surface: UI-to-API workflow conversion grounded in the official graphToPrompt algorithm, node-level parameter overrides, workflow reconstruction (adding or removing nodes, rewiring edges, cross-workflow splicing via workflow_json), image and video generation with real-time progress monitoring (queue position, sampling percentage, WebSocket push), and automatic result download. Serial batch orchestration through comfy_wait_task turns a multi-stage request like \"render five storyboard frames, then compose them into a video\" into a single instruction. Model file management is built in as well—upload, listing, disk-usage reporting, archival to secondary storage, and intra-directory moves. Deletion is not registered by default; it requires an explicit safety-config opt-in plus a per-call confirm flag, keeping destructive actions firmly behind a gate.\n\nIdeal for creation teams or independent developers who already use DSH Agents and run ComfyUI locally or on an internal network. If your day-to-day involves frequent workflow swaps, prompt and sampler tuning, batch image generation, or short-video storyboard pipelines, this plugin cuts down window-switching and idle wait time significantly. It depends only on Node 22 or later with zero extra runtime installs, and a single JSON config file is all that is needed to get started.