dify's comfyui tool cant read file variable.

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

I use dify's comfyui tool to call a img2img workflow, I expect it works normally and return an edited image.

### Actual Behavior

tool report error, and I guess it is because comfyui tool cant read file variable, the reason why I assume is in "Steps to Reproduce":

>
> Run failed: Failed to transform tool message: PluginInvokeError: {"args":{},"error_type":"ConnectTimeout","message":"timed out"}
>

### Steps to Reproduce

### I have used text2img workflow, it works, so the comfyui configuration is right.

### 1. use img2img "workflow" node of comfyui tool

# 2. input comfyui's img2img json api

> {
>   "3": {
>     "inputs": {
>       "seed": 601685180812477,
>       "steps": 20,
>       "cfg": 2.5,
>       "sampler_name": "euler",
>       "scheduler": "simple",
>       "denoise": 1,
>       "model": [
>         "75",
>         0
>       ],
>       "positive": [
>         "76",
>         0
>       ],
>       "negative": [
>         "77",
>         0
>       ],
>       "latent_image": [
>         "88",
>         0
>       ]
>     },
>     "class_type": "KSampler",
>     "_meta": {
>       "title": "KSampler"
>     }
>   },
>   "8": {
>     "inputs": {
>       "samples": [
>         "3",
>         0
>       ],
>       "vae": [
>         "39",
>         0
>       ]
>     },
>     "class_type": "VAEDecode",
>     "_meta": {
>       "title": "VAE Decode"
>     }
>   },
>   "37": {
>     "inputs": {
>       "unet_name": "qwen_image_edit_fp8_e4m3fn.safetensors",
>       "weight_dtype": "default"
>     },
>     "class_type": "UNETLoader",
>     "_meta": {
>       "title": "Load Diffusion Model"
>     }
>   },
>   "38": {
>     "inputs": {
>       "clip_name": "qwen_2.5_vl_7b_fp8_scaled.safetensors",
>       "type": "qwen_image",
>       "device": "default"
>     },
>     "class_type": "CLIPLoader",
>     "_meta": {
>       "title": "Load CLIP"
>     }
>   },
>   "39": {
>     "inputs": {
>       "vae_name": "qwen_image_vae.safetensors"
>     },
>     "class_type": "VAELoader",
>     "_meta": {
>       "title": "Load VAE"
>     }
>   },
>   "60": {
>     "inputs": {
>       "filename_prefix": "ComfyUI",
>       "images": [
>         "8",
>         0
>       ]
>     },
>     "class_type": "SaveImage",
>     "_meta": {
>       "title": "Save Image"
>     }
>   },
>   "66": {
>     "inputs": {
>       "shift": 3,
>       "model": [
>         "37",
>         0
>       ]
>     },
>     "class_type": "ModelSamplingAuraFlow",
>     "_meta": {
>       "title": "ModelSamplingAuraFlow"
>     }
>   },
>   "75": {
>     "inputs": {
>       "strength": 1,
>       "model": [
>         "66",
>         0
>       ]
>     },
>     "class_type": "CFGNorm",
>     "_meta": {
>       "title": "CFGNorm"
>     }
>   },
>   "76": {
>     "inputs": {
>       "prompt": "{{#llm.text#}}",
>       "clip": [
>         "38",
>         0
>       ],
>       "vae": [
>         "39",
>         0
>       ],
>       "image": [
>         "93",
>         0
>       ]
>     },
>     "class_type": "TextEncodeQwenImageEdit",
>     "_meta": {
>       "title": "TextEncodeQwenImageEdit"
>     }
>   },
>   "77": {
>     "inputs": {
>       "prompt": "",
>       "clip": [
>         "38",
>         0
>       ],
>       "vae": [
>         "39",
>         0
>       ],
>       "image": [
>         "93",
>         0
>       ]
>     },
>     "class_type": "TextEncodeQwenImageEdit",
>     "_meta": {
>       "title": "TextEncodeQwenImageEdit"
>     }
>   },
>   "78": {
>     "inputs": {
>       "image": ""
>     },
>     "class_type": "LoadImage",
>     "_meta": {
>       "title": "Load Image"
>     }
>   },
>   "88": {
>     "inputs": {
>       "pixels": [
>         "93",
>         0
>       ],
>       "vae": [
>         "39",
>         0
>       ]
>     },
>     "class_type": "VAEEncode",
>     "_meta": {
>       "title": "VAE Encode"
>     }
>   },
>   "93": {
>     "inputs": {
>       "upscale_method": "lanczos",
>       "megapixels": 1,
>       "image": [
>         "78",
>         0
>       ]
>     },
>     "class_type": "ImageScaleToTotalPixels",
>     "_meta": {
>       "title": "Scale Image to Total Pixels"
>     }
>   }
> }

### 3. select sys.files variable and basic settings

### 4. run work flow and fail beacuse the sys.files has values
Run failed: Failed to transform tool message: PluginInvokeError: {"args":{},"error_type":"ConnectTimeout","message":"timed out"}

### 5. run work flow and success after reset sys.files, but it doest meet my requirements of edit image in this situation.

### Debug Logs

```powershell
Run failed: Failed to transform tool message: PluginInvokeError: {"args":{},"error_type":"ConnectTimeout","message":"timed out"}
```

### Other

_No response_

The report points to Dify's ComfyUI tool rather than a named ComfyUI source file or test. Start by reproducing the img2img workflow with sys.files and compare it with the working text2img workflow; determine whether the timeout originates in the integration or the ComfyUI API, then add a focused reproduction or test once the responsible component is identified.

Written by the indexing model from the issue text.