Update app.py

app.py

@@ -585,6 +585,7 @@ def gemini_generate_content(prompt_text, file_upload_obj=None, image_obj=None, m
     """ 
     Send prompt_text and optionally an uploaded file (or an image object/list) to the model using NEW SDK.
     Automatically rotates through available API keys on RESOURCE_EXHAUSTED errors.
+    Note: Files uploaded with one API key cannot be accessed by another API key.
     Returns textual response and prints progress. 
     """ 
     contents = [prompt_text] 
@@ -617,6 +618,7 @@ def gemini_generate_content(prompt_text, file_upload_obj=None, image_obj=None, m
         current_client = client_manager.get_current_client()
         current_key_num = client_manager.current_key_index + 1
         
+        # Try primary model first
         try:
             print(f"🔑 Using API key #{current_key_num} with model {model_name}")
             response = current_client.models.generate_content(
@@ -632,35 +634,58 @@ def gemini_generate_content(prompt_text, file_upload_obj=None, image_obj=None, m
             
         except Exception as e:
             error_str = str(e)
-            print(f"❌ Generation failed with API key #{current_key_num}: {e}")
+            print(f"❌ Generation failed with API key #{current_key_num} and model {model_name}: {e}")
             
             # Check if it's a RESOURCE_EXHAUSTED error
             if "429" in error_str or "RESOURCE_EXHAUSTED" in error_str:
-                print(f"⚠️ Quota exhausted for API key #{current_key_num}")
+                print(f"⚠️ Quota exhausted for API key #{current_key_num} with model {model_name}")
                 
-                # Try to rotate to next key
-                if client_manager.rotate_to_next_key():
-                    attempt += 1
-                    print(f"🔄 Retrying with next API key (attempt {attempt + 1}/{max_attempts})...")
-                    continue
-                else:
-                    # Only one key available, try fallback model
-                    print(f"⚡ Trying fallback model: {fallback_model}")
-                    try:
-                        response = current_client.models.generate_content(
-                            model=fallback_model,
-                            contents=contents
-                        )
-                        raw_text = response.text
-                        print(f"📥 Received response (chars): {len(raw_text)}") 
-                        client_manager.reset_to_primary()
-                        return raw_text
-                    except Exception as e2:
-                        print(f"❌ Fallback also failed: {e2}")
-                        raise Exception(f"All API keys exhausted. Error: {e2}")
+                # Try fallback model with SAME API key before switching keys
+                print(f"⚡ Trying fallback model {fallback_model} with same API key #{current_key_num}")
+                try:
+                    response = current_client.models.generate_content(
+                        model=fallback_model,
+                        contents=contents
+                    )
+                    raw_text = response.text
+                    print(f"📥 Received response (chars): {len(raw_text)}") 
+                    client_manager.reset_to_primary()
+                    return raw_text
+                except Exception as e_fallback:
+                    error_fallback_str = str(e_fallback)
+                    print(f"❌ Fallback model {fallback_model} also failed: {e_fallback}")
+                    
+                    # Check if fallback also exhausted
+                    if "429" in error_fallback_str or "RESOURCE_EXHAUSTED" in error_fallback_str:
+                        print(f"⚠️ Fallback model also exhausted for API key #{current_key_num}")
+                        
+                        # Now try next API key if available
+                        if attempt < max_attempts - 1:
+                            # Check if we have file uploads - they won't work with different keys
+                            if file_upload_obj:
+                                print("⚠️ WARNING: File uploads cannot be shared across API keys!")
+                                print("   You need to re-upload files with each new API key.")
+                                print("   Stopping rotation to avoid PERMISSION_DENIED errors.")
+                                raise Exception(f"All models exhausted for API key #{current_key_num}. Cannot rotate keys with file uploads.")
+                            
+                            client_manager.rotate_to_next_key()
+                            attempt += 1
+                            print(f"🔄 Trying next API key (attempt {attempt + 1}/{max_attempts})...")
+                            continue
+                        else:
+                            raise Exception(f"All {max_attempts} API key(s) exhausted with both models.")
+                    else:
+                        # Fallback failed with different error
+                        raise Exception(f"Fallback model failed: {e_fallback}")
+            
+            elif "403" in error_str or "PERMISSION_DENIED" in error_str:
+                # This happens when trying to access a file uploaded with a different API key
+                print(f"⚠️ Permission denied - likely due to file uploaded with different API key")
+                raise Exception(f"File access denied. Files uploaded with one API key cannot be accessed by another. Error: {e}")
+            
             else:
-                # Not a quota error, try fallback model with same key
-                print(f"⚡ Trying fallback model: {fallback_model}")
+                # Other error - try fallback model with same key
+                print(f"⚡ Trying fallback model {fallback_model} with same API key #{current_key_num}")
                 try:
                     response = current_client.models.generate_content(
                         model=fallback_model,
@@ -672,19 +697,24 @@ def gemini_generate_content(prompt_text, file_upload_obj=None, image_obj=None, m
                     return raw_text
                 except Exception as e2:
                     print(f"❌ Fallback also failed: {e2}")
-                    # If we have more keys, try them
-                    if attempt < max_attempts - 1:
+                    
+                    # If we have more keys and no file uploads, try them
+                    if attempt < max_attempts - 1 and not file_upload_obj:
                         client_manager.rotate_to_next_key()
                         attempt += 1
                         print(f"🔄 Trying next API key (attempt {attempt + 1}/{max_attempts})...")
                         continue
                     else:
-                        raise Exception(f"All attempts failed. Last error: {e2}")
+                        if file_upload_obj:
+                            raise Exception(f"All models failed. Cannot rotate keys with file uploads. Last error: {e2}")
+                        else:
+                            raise Exception(f"All attempts failed. Last error: {e2}")
     
     # If we exhausted all attempts
     raise Exception(f"❌ All {max_attempts} API key(s) exhausted. Please check your quota or try again later.")
 
 
+
 # ---------------- PARSERS ---------------- 
 def extract_question_ids_from_qpms(text: str):
     """Extract question IDs from QP+MS transcript."""