Update app.py

app.py

@@ -7,6 +7,8 @@ from huggingface_hub.inference._generated.types.chat_completion import ChatCompl
 MODEL = "HuggingFaceH4/zephyr-7b-beta"
 client = InferenceClient(MODEL)
 
+from huggingface_hub.inference._generated.types.chat_completion import ChatCompletionStreamOutput
+
 def respond(
     message,
     history: list[tuple[str, str]],
@@ -40,12 +42,32 @@ def respond(
                 # Check if this is the last message in the stream
                 if message.choices[0].finish_reason == 'stop':
                     break
+            elif isinstance(message, dict):
+                # Handle dictionary responses (possible API format change)
+                content = message.get('choices', [{}])[0].get('delta', {}).get('content')
+                if content:
+                    response += content
+                    yield response
+                if message.get('choices', [{}])[0].get('finish_reason') == 'stop':
+                    break
+            elif isinstance(message, str):
+                # Handle string responses (possible raw content)
+                response += message
+                yield response
             else:
                 print(f"Unexpected message type: {type(message)}")
                 print(f"Message content: {message}")
+        
+        # Final yield to ensure all content is returned
+        if response:
+            yield response
+
     except Exception as e:
         print(f"An error occurred: {e}")
-        yield f"An error occurred: {e}"
+        if response:
+            yield response + f"\n\nAn error occurred: {e}"
+        else:
+            yield f"An error occurred: {e}"
 
 # Gradio interface setup
 demo = gr.ChatInterface(