scriptoriumadmin/open-webui
1
0
Fork 0
mirror of https://github.com/open-webui/open-webui.git synced 2026-11-14 09:21:22 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

refac

Browse source
This commit is contained in:
Timothy Jaeryang Baek 2026-11-05 21:59:46 +09:00
parent 1632222f39
commit 8366e28725
1 changed files with 22 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

38
backend/open_webui/utils/middleware.py
View file

@ -4550,15 +4550,18 @@ async def streaming_chat_response_handler(response, ctx):
finish_reason = None
last_response_id = None
def full_output():
if (
def is_message_continuation() -> bool:
return bool(
continuing
and prior_output
and output
and prior_output[-1].get('type') == output[0].get('type') == 'message'
and prior_output[-1].get('role', 'assistant') == 'assistant'
and output[0].get('role', 'assistant') == 'assistant'
):
)
def full_output():
if is_message_continuation():
return [
*prior_output[:-1],
{
@ -4588,11 +4591,9 @@ async def streaming_chat_response_handler(response, ctx):
nonlocal stream_size
if not stream_size or not request.app.state.redis:
return
# Continue merges items; code blocks write to `code`, not the event's text field.
if (
continuing
or not isinstance(delta.get('delta'), str)
or any(item.get('type') == 'open_webui:code_interpreter' for item in output)
# Code blocks write to `code`, not the event's text field.
if not isinstance(delta.get('delta'), str) or any(
item.get('type') == 'open_webui:code_interpreter' for item in output
):
stream_size = 0
return
@ -4661,6 +4662,10 @@ async def streaming_chat_response_handler(response, ctx):
reasoning_tags = DEFAULT_REASONING_TAGS
try:
# deltas of a continued reply assume the client already holds the prior output
if prior_output:
await event_emitter({'type': 'chat:completion', 'data': {'output': full_output()}})
for event in events:
await event_emitter(
{
@ -4704,6 +4709,14 @@ async def streaming_chat_response_handler(response, ctx):
def get_response_data_with_full_output_index(response_data: dict):
if prior_output and isinstance(response_data.get('output_index'), int):
if is_message_continuation():
output_index = response_data['output_index']
content_offset = len(prior_output[-1].get('content', [])) if output_index == 0 else 0
return {
**response_data,
'output_index': output_index + len(prior_output) - 1,
'content_index': response_data.get('content_index', 0) + content_offset,
}
return {
**response_data,
'output_index': response_data['output_index'] + len(prior_output),
@ -4725,14 +4738,7 @@ async def streaming_chat_response_handler(response, ctx):
nonlocal last_delta_key
if delta_count >= threshold and last_delta_data:
await event_emitter(
{
'type': 'chat:completion' if continuing else 'response:completion',
'data': {'output': full_output(), 'type': last_delta_data.get('type')}
if continuing
else last_delta_data,
}
)
await event_emitter({'type': 'response:completion', 'data': last_delta_data})
delta_count = 0
last_delta_data = None
last_delta_type = None