Update app.py

app.py

@@ -737,7 +737,6 @@ def process_segment_with_gpt(segment, source_lang, target_lang, model="gpt-4", o
     """
     Processes a single text segment: restores punctuation and translates using an OpenAI GPT model.
     """
-    # Essential check: Ensure the OpenAI client is provided
     if openai_client is None:
         segment_identifier = f"{segment.get('start', 'N/A')}-{segment.get('end', 'N/A')}"
         logger.error(f"❌ OpenAI client was not provided for segment {segment_identifier}. Cannot process.")
@@ -767,31 +766,46 @@ def process_segment_with_gpt(segment, source_lang, target_lang, model="gpt-4", o
 
     try:
         logger.debug(f"Sending request to OpenAI model '{model}' for segment {segment_id}...")
-        response = openai_client.chat.completions.create( # Using the passed 'openai_client'
+        response = openai_client.chat.completions.create(
             model=model,
             messages=[{"role": "user", "content": prompt}],
             temperature=0.3
         )
         content = response.choices[0].message.content.strip()
+
+        # --- NEW LOGIC: Clean markdown code block fences from the response ---
+        cleaned_content = content
+        if content.startswith("```") and content.endswith("```"):
+            # Attempt to find the actual JSON object within the markdown fence
+            json_start_index = content.find('{')
+            json_end_index = content.rfind('}')
+
+            if json_start_index != -1 and json_end_index != -1 and json_end_index > json_start_index:
+                cleaned_content = content[json_start_index : json_end_index + 1]
+                logger.debug(f"Removed markdown fences for segment {segment_id}. Extracted JSON portion.")
+            else:
+                logger.warning(
+                    f"⚠️ Content starts/ends with '```' but a valid JSON object ({{...}}) was not found within "
+                    f"fences for segment {segment_id}. Attempting to parse raw content. Raw content: '{content}'"
+                )
+        # --- END NEW LOGIC ---
+
         logger.debug(
-            f"Received raw response from model for segment {segment_id}: "
-            f"'{content[:200]}{'...' if len(content) > 200 else ''}'" # Truncate for log readability
+            f"Attempting to parse JSON for segment {segment_id}. "
+            f"Content for parsing preview: '{cleaned_content[:200]}{'...' if len(cleaned_content) > 200 else ''}'"
         )
 
         result_json = {}
         try:
-            # Use json.loads for safer and standard JSON parsing compared to eval()
-            result_json = json.loads(content)
+            result_json = json.loads(cleaned_content)
         except json.JSONDecodeError as e:
             logger.warning(
                 f"⚠️ Failed to parse JSON response for segment {segment_id}. Error: {e}. "
-                f"Raw content received: '{content}'"
+                f"Content attempted to parse: '{cleaned_content}'" # Log cleaned content here
             )
-            # Fallback behavior if JSON parsing fails: use original text, empty translation
             punctuated_text = original_text
-            translated_text = ""
+            translated_text = "" # Return empty translated text on parsing failure
         else:
-            # If JSON parsing was successful
             punctuated_text = result_json.get("punctuated", original_text)
             translated_text = result_json.get("translated", "")
 
@@ -808,12 +822,10 @@ def process_segment_with_gpt(segment, source_lang, target_lang, model="gpt-4", o
             "translated": translated_text
         }
     except Exception as e:
-        # Log the full traceback using exc_info=True for better debugging
         logger.error(
-            f"❌ An unexpected error occurred while processing segment {segment_id}: {e}",
-            exc_info=True
+            f"❌ An unexpected error occurred for segment {segment_id}: {e}",
+            exc_info=True # This logs the full traceback
         )
-        # Return the original segment with an empty translated text on error
         return {
             "start": segment["start"],
             "end": segment["end"],