dsh-vidgen-plugin

Building AI video and image pipelines inside DSH Web GUI used to mean hopping between browser tabs, hand-writing API calls, and babysitting task-polling responses. dsh-vidgen-plugin pulls the entire workflow into a single session: a ComfyUI-style node canvas studio with six node types (text-to-image, image-to-image and multi-image composition, image source, text-to-video, image-to-video, and keyframes) that you wire together by dragging connect points. Nodes execute in parallel by dependency, the canvas auto-persists across restarts, and you can save, load, or delete manual snapshots at any time. A dedicated multi-provider settings page keeps video and image platform lists independent, letting you add, remove, or switch providers, fetch available models with one click, and test connections without ever leaving the page.

The plugin also exposes model tools that agents can call from within a conversation, such as listing video models, generating video from text, image, or keyframe sequences, and polling task status by ID, so an agent can kick off generation and return a finished asset link. Everything runs as a dynamic Cordis plugin injected at runtime, requires no host configuration changes, and keeps API keys strictly local. It is built for creators and developers who want a unified visual pipeline inside DSH, who need side-by-side management of multiple API platforms for comparison or failover, or who simply want their conversational agent to trigger visual generation and deliver finished results directly.