refactor: render from output and derive content at save

Author: Algorithm5838

backend/open_webui/utils/middleware.py

@@ -4069,7 +4069,6 @@ async def flush_pending_delta_data(threshold: int = 0):
 
                                     processed_data = {
                                         'output': full_output(),
-                                        'content': serialize_output(full_output()),
                                     }
 
                                     # print(data)
@@ -4235,7 +4234,7 @@ async def flush_pending_delta_data(threshold: int = 0):
                                                 {
                                                     'type': 'chat:completion',
                                                     'data': {
-                                                        'content': serialize_output(full_output() + pending_fc_items),
+                                                        'output': full_output() + pending_fc_items,
                                                     },
                                                 }
                                             )
@@ -4313,7 +4312,7 @@ async def flush_pending_delta_data(threshold: int = 0):
                                             _pending_reasoning_details.extend(items)
 
                                     if reasoning_content or reasoning_details_chunk:
-                                        data = {'content': serialize_output(full_output())}
+                                        data = {'output': full_output()}
 
                                     if value:
                                         if (
@@ -4470,7 +4469,7 @@ async def flush_pending_delta_data(threshold: int = 0):
                                             )
                                         else:
                                             data = {
-                                                'content': serialize_output(full_output()),
+                                                'output': full_output(),
                                             }
 
                                 if delta:

src/lib/components/chat/Chat.svelte

@@ -1813,7 +1813,9 @@
 			message.sources = sources;
 		}
 
-		if (choices) {
+		// Only accumulate content from choices when there's no structured output
+		// (output-based rendering takes priority — avoids redundant serialize/parse round-trip)
+		if (choices && !output) {
 			if (choices[0]?.message?.content) {
 				// Non-stream response
 				message.content += choices[0]?.message?.content;