Update app.py

app.py

@@ -120,8 +120,6 @@ def process_and_summarize(audio_file, translate, model_size, do_summarize=True):
         logger.error(traceback.format_exc())
         raise gr.Error(f"Processing failed: {str(e)}")
 
-        
-
 
 @spaces.GPU(duration=60)
 def answer_question(context, question):
@@ -138,25 +136,31 @@ def answer_question(context, question):
         
         outputs = qa_pipeline(messages, max_new_tokens=256)
         
-        # Extract the answer from the output
-        if isinstance(outputs, list) and len(outputs) > 0 and isinstance(outputs[0], dict):
-            for message in outputs[0]:
-                if message.get('role') == 'assistant':
-                    answer = message.get('content', '')
-                    break
+        logger.info(f"Raw model output: {outputs}")
+
+        # Parse the string output into a list of dictionaries
+        try:
+            parsed_output = ast.literal_eval(outputs)
+            if isinstance(parsed_output, list):
+                for item in parsed_output:
+                    if isinstance(item, dict) and item.get('role') == 'assistant':
+                        answer = item.get('content', '')
+                        break
+                else:
+                    answer = "No assistant response found in the model's output."
             else:
-                answer = "No answer found in the model's response."
-        else:
-            answer = str(outputs)  # Fallback to string representation of outputs
-        
-        logger.info("Q&A process completed successfully")
+                answer = str(parsed_output)
+        except (ValueError, SyntaxError):
+            # If parsing fails, return the raw output
+            answer = str(outputs)
+
+        logger.info(f"Extracted answer: {answer}")
         return answer
     except Exception as e:
         logger.error(f"Q&A process failed: {str(e)}")
         logger.error(traceback.format_exc())
         return f"Error occurred during Q&A process. Please try again. Error: {str(e)}"
 
-
 # Main interface
 with gr.Blocks() as iface:
     gr.Markdown("# WhisperX Audio Transcription, Translation, Summarization, and Q&A (with ZeroGPU support)")