Create app.py

app.py

@@ -0,0 +1,34 @@
+import os, time, io
+import gradio as gr
+from huggingface_hub import InferenceClient
+from PIL import Image
+
+HF_MODEL_ID = os.environ.get("HF_MODEL_ID", "stabilityai/stable-diffusion-x4-upscaler")
+HF_TOKEN = os.environ.get("HF_TOKEN", None)
+client = InferenceClient(token=HF_TOKEN)
+
+def call_inference_image(fileobj, scale: int = 2):
+    start = time.time()
+    if hasattr(fileobj, "name"):
+        img_bytes = open(fileobj.name, "rb").read()
+    else:
+        img_bytes = fileobj.read()
+    params = {"scale": int(scale)} if scale else {}
+    res = client.invoke(HF_MODEL_ID, inputs=img_bytes, params=params)
+    latency = time.time() - start
+    if isinstance(res, (bytes, bytearray)):
+        return Image.open(io.BytesIO(res)), f"{latency:.3f}s"
+    return str(res), f"{latency:.3f}s"
+
+with gr.Blocks() as demo:
+    gr.Markdown("# API-based Upscaler (calls HF Inference API)")
+    with gr.Row():
+        inp = gr.Image(type="file", label="Upload image")
+        scale = gr.Slider(2, 8, value=2, step=1, label="Scale (requested)")
+    out_img = gr.Image(label="Result")
+    latency = gr.Textbox(label="Latency")
+    btn = gr.Button("Upscale via API")
+    btn.click(call_inference_image, inputs=[inp, scale], outputs=[out_img, latency])
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", server_port=7860)