Update app.py

app.py

@@ -6,8 +6,11 @@ import modelscope_studio.components.base as ms
 import modelscope_studio.components.pro as pro
 from openai import OpenAI
 from config import API_KEY, MODEL, SYSTEM_PROMPT, ENDPOINT, EXAMPLES, DEFAULT_LOCALE, DEFAULT_THEME
+from huggingface_hub import InferenceClient
 
-client = OpenAI(api_key=API_KEY, base_url=ENDPOINT)
+# Initialize both clients
+hf_client = InferenceClient(model=MODEL, token=API_KEY)
+openai_client = OpenAI(api_key=API_KEY, base_url=ENDPOINT + MODEL)
 
 react_imports = {
     "lucide-react": "https://esm.sh/lucide-react@0.525.0",
@@ -66,25 +69,85 @@ class GradioEvents:
             # 'content': system_prompt_input_value
         }] + state_value["history"]
 
-        messages.append({'role': "user", 'content': input_value})
+        messages.append({'role': 'user', 'content': input_value})
+
+        try:
+            # Try using HuggingFace InferenceClient first
+            formatted_messages = [{"role": msg["role"], "content": msg["content"]} for msg in messages]
+            response = ""
+            for chunk in hf_client.chat_completion(
+                messages=formatted_messages,
+                stream=True,
+                max_tokens=4000,
+                temperature=0.7
+            ):
+                if chunk.choices[0].delta.content:
+                    response += chunk.choices[0].delta.content
+                    yield {
+                        output: gr.update(value=response),
+                        output_loading: gr.update(spinning=False),
+                    }
+            
+            # Process the final response
+            state_value["history"] = messages + [{
+                'role': 'assistant',
+                'content': response
+            }]
+            generated_files = get_generated_files(response)
+            react_code = generated_files.get(
+                "index.tsx") or generated_files.get("index.jsx")
+            html_code = generated_files.get("index.html")
+            
+            yield {
+                output:
+                gr.update(value=response),
+                download_content:
+                gr.update(value=react_code or html_code),
+                state_tab:
+                gr.update(active_key="render"),
+                output_loading:
+                gr.update(spinning=False),
+                sandbox:
+                gr.update(
+                    template="react" if react_code else "html",
+                    imports=react_imports if react_code else {},
+                    value={
+                        "./index.tsx": """import Demo from './demo.tsx'
+import "@tailwindcss/browser"
 
-        generator = client.chat.completions.create(model=MODEL,
-                                                   messages=messages,
-                                                   stream=True)
-        response = ""
-        for chunk in generator:
-            content = chunk.choices[0].delta.content
-            response += content
-            if chunk.choices[0].finish_reason == 'stop':
+export default Demo
+""",
+                            "./demo.tsx": react_code
+                        } if react_code else {"./index.html": html_code}),
+                state:
+                gr.update(value=state_value)
+            }
+        except Exception as e:
+            # Fallback to OpenAI client if HuggingFace client fails
+            try:
+                generator = openai_client.chat.completions.create(model=MODEL,
+                                                               messages=messages,
+                                                               stream=True)
+                response = ""
+                for chunk in generator:
+                    content = chunk.choices[0].delta.content
+                    if content:
+                        response += content
+                        yield {
+                            output: gr.update(value=response),
+                            output_loading: gr.update(spinning=False),
+                        }
+                
+                # Process the final response
                 state_value["history"] = messages + [{
-                    'role': "assistant",
+                    'role': 'assistant',
                     'content': response
                 }]
                 generated_files = get_generated_files(response)
                 react_code = generated_files.get(
                     "index.tsx") or generated_files.get("index.jsx")
                 html_code = generated_files.get("index.html")
-                # Completed
+                
                 yield {
                     output:
                     gr.update(value=response),
@@ -104,17 +167,18 @@ import "@tailwindcss/browser"
 
 export default Demo
 """,
-                            "./demo.tsx": react_code
-                        } if react_code else {"./index.html": html_code}),
+                                "./demo.tsx": react_code
+                            } if react_code else {"./index.html": html_code}),
                     state:
                     gr.update(value=state_value)
                 }
-
-            else:
-                # Generating
+            except Exception as e2:
+                # If both methods fail, show an error message
+                error_message = f"Error generating code: {str(e2)}"
                 yield {
-                    output: gr.update(value=response),
+                    output: gr.update(value=error_message),
                     output_loading: gr.update(spinning=False),
+                    state_tab: gr.update(active_key="render"),
                 }
 
     @staticmethod
@@ -215,7 +279,7 @@ with gr.Blocks(css=css) as demo:
                                     height=200,
                                     preview=False)
                                 antd.Typography.Title(
-                                    "Qwen3-Coder-WebDev",
+                                    "GLM-4.5-Coder",  # Changed from "Qwen3-Coder-WebDev" to "GLM-4.5-Coder"
                                     level=1,
                                     elem_style=dict(fontSize=24))
                             # Input
@@ -455,4 +519,4 @@ with gr.Blocks(css=css) as demo:
 
 if __name__ == "__main__":
     demo.queue(default_concurrency_limit=100,
-               max_size=100).launch(ssr_mode=False, max_threads=100)
+               max_size=100).launch(ssr_mode=False, max_threads=100)