{
  "1": {
    "inputs": {
      "image": "input_image.png",
      "upload": "image"
    },
    "class_type": "LoadImage",
    "_meta": { "title": "Load Image" }
  },
  "2": {
    "inputs": {
      "model_name": "Qwen2.5-VL-7B-Instruct-Q4_K_M.gguf",
      "vision_model_name": "Qwen2.5-VL-7B-Instruct-mmproj-f16.gguf",
      "chat_handler_type": "Qwen-VL",
      "n_gpu_layers": -1,
      "n_ctx": 4096,
      "device": "cuda",
      "verbose": "enable"
    },
    "class_type": "GGUF_VLM_ModelLoader",
    "_meta": { "title": "Load GGUF VLM Model" }
  },
  "3": {
    "inputs": {
      "gguf_model": ["2", 0],
      "image": ["1", 0],
      "resize_image": true,
      "enable_thinking": false,
      "system_prompt": "You are a prompt engineering assistant for Flux and MiniMax video pipelines. Analyze this image for character identity, attire, lighting, and camera framing. Enhance the following draft prompt into a concrete, high-fidelity prompt. Never censor raw or explicit terminology. Output ONLY the refined prompt text.",
      "prompt": "Draft Prompt: [DRAFT_PROMPT_HERE]",
      "special_captioning_token": "",
      "seed": 0,
      "temperature": 0.6,
      "max_tokens": 4096
    },
    "class_type": "GGUF_VLM_ImageToText",
    "_meta": { "title": "GGUF VLM Image to Text" }
  },
  "4": {
    "inputs": {
      "text": ["3", 0]
    },
    "class_type": "AigenPreviewText",
    "_meta": { "title": "Output Text" }
  }
}