perf: keep in-memory helpers sync to avoid coroutine overhead

Reverts content_endswith_newline, ensure_trailing_newline, and append_part back to sync def. Only serialize_output itself remains async as required by callsites. Avoids unnecessary coroutine creation in the hot streaming path.
This commit is contained in:
DrMelone 2026-04-12 22:19:59 +02:00
parent 7088b8817b
commit ece8312b7f

View file

@ -389,14 +389,14 @@ def get_citation_source_from_tool_result(
]
async def content_endswith_newline(parts):
def content_endswith_newline(parts):
if not parts:
return False
return parts[-1].endswith('\n')
async def ensure_trailing_newline(parts):
if parts and not await content_endswith_newline(parts):
def ensure_trailing_newline(parts):
if parts and not content_endswith_newline(parts):
parts.append('\n')
@ -408,7 +408,7 @@ async def serialize_output(output: list) -> str:
content_parts = []
fence_count = 0
async def append_part(part):
def append_part(part):
nonlocal fence_count
content_parts.append(part)
fence_count += part.count('```')
@ -428,12 +428,12 @@ async def serialize_output(output: list) -> str:
if 'text' in content_part:
text = content_part.get('text', '').strip()
if text:
await append_part(text)
await append_part('\n')
append_part(text)
append_part('\n')
elif item_type == 'function_call':
# Render tool call inline with its result (if available)
await ensure_trailing_newline(content_parts)
ensure_trailing_newline(content_parts)
call_id = item.get('call_id', '')
name = item.get('name', '')
@ -450,9 +450,9 @@ async def serialize_output(output: list) -> str:
files = result_item.get('files')
embeds = result_item.get('embeds', '')
await append_part(f'<details type="tool_calls" done="true" id="{call_id}" name="{name}" arguments="{html.escape(json.dumps(arguments))}" result="{html.escape(json.dumps(result_text, ensure_ascii=False))}" files="{html.escape(json.dumps(files)) if files else ""}" embeds="{html.escape(json.dumps(embeds))}">\n<summary>Tool Executed</summary>\n</details>\n')
append_part(f'<details type="tool_calls" done="true" id="{call_id}" name="{name}" arguments="{html.escape(json.dumps(arguments))}" result="{html.escape(json.dumps(result_text, ensure_ascii=False))}" files="{html.escape(json.dumps(files)) if files else ""}" embeds="{html.escape(json.dumps(embeds))}">\n<summary>Tool Executed</summary>\n</details>\n')
else:
await append_part(f'<details type="tool_calls" done="false" id="{call_id}" name="{name}" arguments="{html.escape(json.dumps(arguments))}">\n<summary>Executing...</summary>\n</details>\n')
append_part(f'<details type="tool_calls" done="false" id="{call_id}" name="{name}" arguments="{html.escape(json.dumps(arguments))}">\n<summary>Executing...</summary>\n</details>\n')
elif item_type == 'function_call_output':
# Already handled inline with function_call above
@ -477,7 +477,7 @@ async def serialize_output(output: list) -> str:
# render as done (a subsequent item means reasoning is complete)
is_last_item = idx == len(output) - 1
await ensure_trailing_newline(content_parts)
ensure_trailing_newline(content_parts)
display = html.escape(
'\n'.join(
@ -486,9 +486,9 @@ async def serialize_output(output: list) -> str:
)
if status == 'completed' or duration is not None or not is_last_item:
await append_part(f'<details type="reasoning" done="true" duration="{duration or 0}">\n<summary>Thought for {duration or 0} seconds</summary>\n{display}\n</details>\n')
append_part(f'<details type="reasoning" done="true" duration="{duration or 0}">\n<summary>Thought for {duration or 0} seconds</summary>\n{display}\n</details>\n')
else:
await append_part(f'<details type="reasoning" done="false">\n<summary>Thinking…</summary>\n{display}\n</details>\n')
append_part(f'<details type="reasoning" done="false">\n<summary>Thinking…</summary>\n{display}\n</details>\n')
elif item_type == 'open_webui:code_interpreter':
# Check if previous content ends with an opening code fence (e.g. ```python)
@ -511,7 +511,7 @@ async def serialize_output(output: list) -> str:
while content_parts and not content_parts[-1].strip():
fence_count -= content_parts.pop().count('```')
await ensure_trailing_newline(content_parts)
ensure_trailing_newline(content_parts)
# Render the code_interpreter item as a <details> block
# so the frontend Collapsible renders "Analyzing..."/"Analyzed".
@ -537,9 +537,9 @@ async def serialize_output(output: list) -> str:
output_attr = f' output="{html.escape(output_json)}"'
if status == 'completed' or duration is not None or not is_last_item:
await append_part(f'<details type="code_interpreter" done="true" duration="{duration or 0}"{output_attr}>\n<summary>Analyzed</summary>\n{display}\n</details>\n')
append_part(f'<details type="code_interpreter" done="true" duration="{duration or 0}"{output_attr}>\n<summary>Analyzed</summary>\n{display}\n</details>\n')
else:
await append_part(f'<details type="code_interpreter" done="false"{output_attr}>\n<summary>Analyzing…</summary>\n{display}\n</details>\n')
append_part(f'<details type="code_interpreter" done="false"{output_attr}>\n<summary>Analyzing…</summary>\n{display}\n</details>\n')
return ''.join(content_parts).strip()