From 58b36765a7c20f5943a3180bd289de48876d0878 Mon Sep 17 00:00:00 2001
From: Timothy Jaeryang Baek <tim@openwebui.com>
Date: Wed, 16 Sep 2026 22:47:37 -0400
Subject: [PATCH] refac

---
 backend/open_webui/socket/main.py             |  4 +++
 backend/open_webui/utils/middleware.py        | 29 +++++++++++++++----
 .../chat/Messages/structuredOutput.ts         |  5 ++++
 3 files changed, 33 insertions(+), 5 deletions(-)

diff --git a/backend/open_webui/socket/main.py b/backend/open_webui/socket/main.py
index 3f79bd77db..d4ac18f655 100644
--- a/backend/open_webui/socket/main.py
+++ b/backend/open_webui/socket/main.py
@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import asyncio
+import copy
 import logging
 import random
 import sys
@@ -1010,6 +1011,9 @@ async def _make_channel_emitter(request_info):
             if not content and not output and not done:
                 return
 
+            if isinstance(output, list):
+                state['output'] = copy.deepcopy(output)
+
             now = time.time()
             if done or (now - state['last_emit_at']) >= THROTTLE_INTERVAL:
                 state['last_emit_at'] = now
diff --git a/backend/open_webui/utils/middleware.py b/backend/open_webui/utils/middleware.py
index e0337fe553..5b59fbd7ad 100644
--- a/backend/open_webui/utils/middleware.py
+++ b/backend/open_webui/utils/middleware.py
@@ -4401,6 +4401,9 @@ async def streaming_chat_response_handler(response, ctx):
 
                 Uses the text from the output items themselves for tag detection,
                 eliminating state divergence between accumulated content and items.
+
+                Mutates output in place; returns the rewritten output (or None
+                if no tag was consumed) and whether a block ended.
                 """
                 end_flag = False
 
@@ -4577,7 +4580,7 @@ async def streaming_chat_response_handler(response, ctx):
                                 if recursive_end:
                                     end_flag = True
 
-                            break
+                            return output, end_flag
                     else:
                         save_scanned_length(item, item_text)
 
@@ -4666,10 +4669,11 @@ async def streaming_chat_response_handler(response, ctx):
                                     ],
                                 }
                             )
+                        return output, end_flag
                     else:
                         save_scanned_length(item, block_content)
 
-                return output, end_flag
+                return None, False
 
             message = (
                 ctx.get('assistant_message')
@@ -5563,25 +5567,31 @@ async def streaming_chat_response_handler(response, ctx):
                                                     }
                                                 ]
 
+                                        tag_output = None
+
                                         if DETECT_REASONING_TAGS:
-                                            output, _ = tag_output_handler(
+                                            tag_output, _ = tag_output_handler(
                                                 'reasoning',
                                                 reasoning_tags,
                                                 output,
                                             )
 
-                                            output, _ = tag_output_handler(
+                                            solution_output, _ = tag_output_handler(
                                                 'solution',
                                                 DEFAULT_SOLUTION_TAGS,
                                                 output,
                                             )
+                                            if solution_output is not None:
+                                                tag_output = solution_output
 
                                         if DETECT_CODE_INTERPRETER:
-                                            output, end = tag_output_handler(
+                                            code_output, end = tag_output_handler(
                                                 'code_interpreter',
                                                 DEFAULT_CODE_INTERPRETER_TAGS,
                                                 output,
                                             )
+                                            if code_output is not None:
+                                                tag_output = code_output
 
                                             if end:
                                                 break
@@ -5604,6 +5614,15 @@ async def streaming_chat_response_handler(response, ctx):
                                         }
                                         delta_type = delta_event_type
 
+                                        # the raw chunk still carries the tag text: resend the cleaned output instead
+                                        if tag_output is not None:
+                                            await flush_pending_delta_data()
+                                            await event_emitter(
+                                                {'type': 'chat:completion', 'data': {'output': full_output()}}
+                                            )
+                                            await save_current_response_stream()
+                                            data = None
+
                                 if delta and data:
                                     await queue_pending_delta_data(data, delta_type)
                                 elif data:
diff --git a/src/lib/components/chat/Messages/structuredOutput.ts b/src/lib/components/chat/Messages/structuredOutput.ts
index 55ef655a2b..bca4a0ff08 100644
--- a/src/lib/components/chat/Messages/structuredOutput.ts
+++ b/src/lib/components/chat/Messages/structuredOutput.ts
@@ -626,6 +626,11 @@ export function applyResponseStreamEvent(
 			return nextOutput;
 		}
 
+		if (item.type === 'open_webui:code_interpreter') {
+			item.code = `${item.code ?? ''}${event.delta ?? ''}`;
+			return nextOutput;
+		}
+
 		const key = deltaType === 'output_text' || deltaType === 'reasoning_text' ? 'text' : deltaType;
 		item.content = [...(item.content ?? [])];
 		const part = ensurePart(item.content, event.content_index ?? 0);