app.py

import gradio as gr
from transformers import pipeline

# Load your model
pipe = pipeline("image-classification", model="ander-machine/autotrain-u2mob-eufcd")

def predict_safe(image):
    try:
        # Run the pipeline
        preds = pipe(image)

        # preds normalmente es algo como:
        # [{'label': 'cat', 'score': 0.97}, {'label': 'dog', 'score': 0.03}]
        if isinstance(preds, list) and len(preds) > 0:
            best = preds[0]
            return {
                "label": best["label"],
                "confidences": [(p["label"], float(p["score"])) for p in preds]
            }
        else:
            # fallback si la lista viene vacía
            return {"label": "unknown", "confidences": [("unknown", 0.0)]}

    except Exception as e:
        # fallback en caso de error (como el HTTP 404 que viste)
        return {"label": "error", "confidences": [("error", 0.0)]}

# Interfaz Gradio
demo = gr.Interface(
    fn=predict_safe,
    inputs=gr.Image(type="pil"),
    outputs=gr.Label(num_top_classes=3),
    title="Image Classifier",
    description="Clasificador con fallback seguro"
)

if __name__ == "__main__":
    demo.launch()