Update app.py

app.py

@@ -12,7 +12,19 @@ HF_TOKEN = os.getenv("HF_TOKEN")
 if not HF_TOKEN:
     raise ValueError("HF_TOKEN environment variable is not set")
 
-def query(prompt, model, custom_lora, is_negative=False, steps=35, cfg_scale=7, sampler="DPM++ 2M Karras", seed=-1, strength=0.7, width=1024, height=1024):
+def query(
+    prompt,
+    model,
+    custom_lora,
+    negative_prompt="",       # ← 기존 is_negative=False → negative_prompt="" 로 변경
+    steps=35,
+    cfg_scale=7,
+    sampler="DPM++ 2M Karras",
+    seed=-1,
+    strength=0.7,
+    width=1024,
+    height=1024
+):
     print("Starting query function...")
     
     if not prompt:
@@ -235,18 +247,22 @@ def query(prompt, model, custom_lora, is_negative=False, steps=35, cfg_scale=7,
             else:
                 API_URL = "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-xl-base-1.0"
 
-        # Prepare payload
+        # Prepare payload in Hugging Face Inference API style
+        # (negative_prompt, steps, cfg_scale, seed, strength 등은 parameters 안에 배치)
         payload = {
             "inputs": prompt,
-            "is_negative": is_negative,
-            "steps": steps,
-            "cfg_scale": cfg_scale,
-            "seed": seed if seed != -1 else random.randint(1, 1000000000),
-            "strength": strength,
             "parameters": {
+                "negative_prompt": negative_prompt,
+                "num_inference_steps": steps,
+                "guidance_scale": cfg_scale,
                 "width": width,
-                "height": height
-            }
+                "height": height,
+                "strength": strength,
+                # seed를 지원하는 모델/엔드포인트에 따라 무시될 수도 있음
+                "seed": seed if seed != -1 else random.randint(1, 1000000000),
+            },
+            # 모델이 로딩 중일 경우 기다리도록 설정
+            "options": {"wait_for_model": True}
         }
 
         # Improved retry logic with exponential backoff
@@ -256,20 +272,27 @@ def query(prompt, model, custom_lora, is_negative=False, steps=35, cfg_scale=7,
         
         while current_retry < max_retries:
             try:
-                response = requests.post(API_URL, headers=headers, json=payload, timeout=180)  # 3-minute timeout
-                response.raise_for_status()
+                response = requests.post(API_URL, headers=headers, json=payload, timeout=180)
                 
+                # 디버깅용 정보 출력
+                print("Response Content-Type:", response.headers.get("Content-Type"))
+                print("Response Text (snippet):", response.text[:500])
+
+                response.raise_for_status()  # HTTP 에러 코드 시 예외 발생
                 image = Image.open(io.BytesIO(response.content))
+                
                 print(f'Generation {key} completed successfully')
                 return image
                 
-            except (requests.exceptions.Timeout, requests.exceptions.ConnectionError, 
-                    requests.exceptions.HTTPError, requests.exceptions.RequestException) as e:
+            except (requests.exceptions.Timeout,
+                    requests.exceptions.ConnectionError, 
+                    requests.exceptions.HTTPError,
+                    requests.exceptions.RequestException) as e:
                 current_retry += 1
                 if current_retry < max_retries:
                     wait_time = backoff_factor ** current_retry  # Exponential backoff
                     print(f"Network error occurred: {str(e)}. Retrying in {wait_time} seconds... (Attempt {current_retry + 1}/{max_retries})")
-                    time.sleep(wait_time)  # Add delay before retry
+                    time.sleep(wait_time)
                     continue
                 else:
                     # Detailed error message based on exception type
@@ -513,4 +536,4 @@ with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", css=css) as dalle:
     dalle.load(fn=update_network_status, outputs=network_status)
 
 if __name__ == "__main__":
-    dalle.launch(show_api=False, share=False)
+    dalle.launch(show_api=False, share=False)