Issues with gradio client

Spaces
1 
from gradio_client import Client

# Connect to the Gradio client
client = Client("black-forest-labs/FLUX.1-schnell")

# Call the inference endpoint
result = client.predict(
    "/infer",  # Endpoint
    { 
        "prompt": "Hello!!",  # Input parameters
        "seed": 0,
        "randomize_seed": True,
        "width": 256,
        "height": 256,
        "num_inference_steps": 1
    }
)

result

this is the code and the output of code is

Loaded as API: https://black-forest-labs-flux-1-schnell.hf.space âś”
('/tmp/gradio/489613c1bc4fe4fcf1b0e5a1d10b5977449d157f9eae4ccd9b8117765eabca12/image.webp',
 653764749)

after that i tried to get the actual generated image and it’s not showing

the full url look like

https://black-forest-labs-flux-1-schnell.hf.space/file=/tmp/gradio/489613c1bc4fe4fcf1b0e5a1d10b5977449d157f9eae4ccd9b8117765eabca12/image.webp

{"detail":"File not found: /tmp/gradio/489613c1bc4fe4fcf1b0e5a1d10b5977449d157f9eae4ccd9b8117765eabca12/image.webp."}

tell me what to do , this happen only while using through api ,

and if i use it in space , there’s no issue.

please tell me what to do

1 Like
2

The image is saved normally, but is there something wrong with the settings?

from gradio_client import Client
from PIL import Image

# Connect to the Gradio client
client = Client("black-forest-labs/FLUX.1-schnell")

# Call the inference endpoint
result = client.predict(
    "/infer",  # Endpoint
    { 
        "prompt": "Hello!!",  # Input parameters
        "seed": 0,
        "randomize_seed": True,
        "width": 256,
        "height": 256,
        "num_inference_steps": 1
    }
)

imagepath, seed = result
image = Image.open(imagepath)
image.save("test1.webp")
1 Like
3

The issue you’re encountering stems from the fact that Gradio Spaces temporarily generate files in a /tmp directory on the server. These files may not persist after the request is processed, which can cause a “File not found” error when accessed via the API.

When using the API, the file path provided (e.g., /tmp/gradio/...) is typically not a public-facing URL but a server-side temporary location. It works fine in the Gradio Space interface because the file is served directly by the UI for a limited time.

 ```python
 import base64

 def inference(prompt, seed, randomize_seed, width, height, num_inference_steps):
     # Existing logic
     image_path = "/tmp/gradio/489613c1bc4fe4fcf1b0e5a1d10b5977449d157f9eae4ccd9b8117765eabca12/image.webp"
     with open(image_path, "rb") as image_file:
         image_data = base64.b64encode(image_file.read()).decode("utf-8")
     return {"image_data": image_data}

 iface = gr.Interface(inference, inputs=[...], outputs=["json"])
 iface.launch()
 ```

This way, the response will include the encoded image content, which can be decoded and saved locally.

1 Like