Add ComfyUI backend for Stable Diffusion image generation

[Copilot]

Adds ComfyUI as a modern alternative to the unmaintained A1111 backend for Stable Diffusion image generation.

Implementation

• Backend (backend/comfyui.py): Workflow-based architecture using ComfyUI's native JSON format. Polls history API for completion, auto-injects prompts into CLIPTextEncode nodes. Supports txt2img and img2img via URL.

• Entry point (dreambot_backend_comfyui.py): CLI launcher following existing backend patterns.

• Configuration: Workflows exported from ComfyUI UI can be used directly. Multiple named workflows supported per instance.

Configuration Example

{
  "comfyui": {
    "host": "comfyui",
    "port": "8188",
    "default_workflow": "txt2img",
    "workflows": {
      "txt2img": {
        "workflow": {
          "6": {
            "class_type": "CLIPTextEncode",
            "inputs": {
              "text": "beautiful scenery",
              "clip": ["4", 1]
            }
          },
          ...
        }
      }
    }
  }
}

The backend automatically finds and updates the first CLIPTextEncode node with user prompts. Includes workflow validation and graceful error handling.

Testing

Added comprehensive test coverage for boot/shutdown, argument parsing, workflow listing, and validation.

Original prompt

In this repo you will find a chatbot that has various backend services it can talk to. One of these is A1111, an old web UI for Stable Diffusion image generation. Since A1111 is no longer maintained, I would like to replace this with ComfyUI. Please propose a Pull Request that adds a ComfyUI backend to dreambot.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.