Update app.py

app.py

@@ -28,16 +28,21 @@ LORA_STYLES = [
     'Studio-DeLight',
     'Cinematic-FlatLog',
 ]
-MAX_SEED = 2**32-1
+MAX_SEED = 2**32 - 1
 
-def encode_image_to_base64(image_path):
-    """Read image file and encode as base64 string."""
+def encode_image_to_gallery_dict(image_path):
+    """
+    Read image file and encode as a dictionary suitable for the Hugging Face Space GalleryData API.
+    It must be {"image": {"data": <base64 str>}} for image input.
+    """
     if not image_path or not isinstance(image_path, str):
         return None
     try:
         with open(image_path, "rb") as f:
-            encoded = base64.b64encode(f.read()).decode("utf-8")
-        return encoded
+            img_bytes = f.read()
+            b64_str = base64.b64encode(img_bytes).decode("utf-8")
+        # API expects: {"image": {"data": <base64 str>}}
+        return {"image": {"data": b64_str}}
     except Exception as e:
         print(f"无法读取图片或编码失败: {image_path}: {e}")
         return None
@@ -52,20 +57,20 @@ def infer(
     steps,
     progress=gr.Progress(track_tqdm=True),
 ):
-    # Process input image(s) as base64 for API (instead of file path)
+    # Process input image(s) as a list of dict, each of form {"image": {"data": ...}}
     images_input = []
     if image is not None:
         if isinstance(image, list):  # Gradio Gallery
             for im in image:
-                img_b64 = encode_image_to_base64(im)
-                if img_b64:
-                    images_input.append({"image": img_b64, "caption": None})
+                img_obj = encode_image_to_gallery_dict(im)
+                if img_obj:
+                    images_input.append(img_obj)
                 else:
                     print(f"警告: 路径无效或无法读取: {im}")
         else:
-            img_b64 = encode_image_to_base64(image)
-            if img_b64:
-                images_input.append({"image": img_b64, "caption": None})
+            img_obj = encode_image_to_gallery_dict(image)
+            if img_obj:
+                images_input.append(img_obj)
             else:
                 print(f"警告: 路径无效或无法读取: {image}")
 
@@ -77,7 +82,7 @@ def infer(
         seed = random.randint(0, MAX_SEED)
 
     print("[调用API] space_client.predict 输入参数：")
-    print(f"  images (base64): {[len(i['image']) for i in images_input]}")
+    print(f"  images (base64 bytes): {[len(i['image']['data']) for i in images_input]}")
     print(f"  prompt: {prompt}")
     print(f"  lora_adapter: {lora_adapter}")
     print(f"  seed: {seed}")
@@ -86,8 +91,7 @@ def infer(
     print(f"  steps: {steps}")
 
     try:
-        # The API might expect images as list of base64 dicts instead of {path: ...}
-        # If error persists, check the API interface definition.
+        # Gradio Space expects list of {"image": {"data": <b64 str>}} elements (see AppError in prompt)
         result = space_client.predict(
             images=images_input,
             prompt=prompt,
@@ -100,6 +104,7 @@ def infer(
         )
         print(f"[调用API] space_client.predict 返回值: {result}")
         image_info, seed_used = result
+        # API may return url, path, or other keys for the resulting image
         img_url = image_info.get("url") or image_info.get("path") or None
         return img_url, seed_used
     except Exception as e:
@@ -113,7 +118,6 @@ def infer(
             )
         return None, seed
 
-# For examples, use None for image input to avoid Gradio directory/File errors on startup.
 examples = [
     [None, "Astronaut in jungle, anime style", "Photo-to-Anime", 0, True, 1.0, 4],
     [None, "A delicious ceviche cheesecake slice", "Style-Transfer", 0, True, 1.0, 4],