Update app.py

app.py

@@ -5,8 +5,6 @@ from huggingface_hub import InferenceClient
 For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
 """
 client = InferenceClient("swiss-ai/Apertus-8B-Instruct-2509")
-#client = InferenceClient("swiss-ai/Apertus-8B-Instruct-2509")
-
 
 def respond(
     message,
@@ -28,53 +26,39 @@ def respond(
 
     response = ""
 
-
     try:
-        for message in client.chat_completion(
+        for chunk in client.chat_completion(
             messages,
             max_tokens=max_tokens,
             stream=True,
             temperature=temperature,
             top_p=top_p,
         ):
-            # Ensure the message has a valid structure
-            if not message or not isinstance(message, dict):
-                continue
-            
             try:
-                # Extract content and finish reason
-                content = message.choices[0].delta.content
-                finish_reason = message.choices[0].finish_reason
-
-                # Check if the content is empty
-                if content.strip() == "":
-                    # If the finish reason is 'stop', it's expected and we can break the loop
-                    if finish_reason == "stop":
-                        print("Stream ended normally.")
-                        break
-                    else:
-                        print("Received unexpected empty content, skipping...")
-                        continue
+                # Primeiro tenta pegar via delta (alguns modelos usam)
+                if hasattr(chunk.choices[0], "delta") and chunk.choices[0].delta and getattr(chunk.choices[0].delta, "content", None):
+                    content = chunk.choices[0].delta.content
+                # Se não tiver, tenta via message.content (outros modelos usam)
+                elif hasattr(chunk.choices[0], "message") and chunk.choices[0].message and getattr(chunk.choices[0].message, "content", None):
+                    content = chunk.choices[0].message.content
+                else:
+                    continue  # não há conteúdo válido, pula
 
                 response += content
                 yield response
 
-            except (AttributeError, IndexError, KeyError) as e:
-                print(f"Error processing message: {e}")
+            except Exception as e:
+                print(f"Erro ao processar chunk: {e}")
                 continue
 
     except Exception as e:
-        print(f"Unexpected error: {e}")
-        yield "An error occurred while generating the response."
+        yield f"Erro inesperado: {e}"
 
-    # Final check if the response is empty
     if response.strip() == "":
-        yield "No response generated. Please try again or adjust the settings."
+        yield "⚠️ O modelo não retornou resposta. Tente ajustar max_tokens, temperature ou escolha outro modelo."
 
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
+# Interface do Gradio
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
@@ -91,6 +75,5 @@ demo = gr.ChatInterface(
     ],
 )
 
-
 if __name__ == "__main__":
     demo.launch()