mirror of
https://github.com/open-webui/open-webui.git
synced 2026-08-28 05:27:35 +00:00
perf: batch and deduplicate per-request DB reads in the chat middleware (#27223)
Several spots in the chat pipeline issued sequential single-key config SELECTs, or fetched the same key twice back-to-back, on every request: - chat_completion_tools_handler: task model default/external and the tools prompt template were four sequential Config round trips (the template was fetched twice). One batched Config.get_many now serves all of them. - chat_completion_files_handler: the six RAG settings (top_k, top_k_reranker, relevance_threshold, hybrid_bm25_weight, enable_hybrid_search, full_context) were six sequential round trips inside the retrieval call. Batched into one get_many. - Voice and code-interpreter prompt templates were each fetched twice within one conditional; fetch once and reuse. The code-interpreter engine was likewise fetched twice per execution. - Skill resolution fetched the accessible-skills list, kept only the ids, then re-fetched each mentioned skill by id (N+1). Reuse the rows from the access query. Value semantics are identical: get_many applies the same defaults as the individual gets, and the pre-existing truthiness/empty-string checks on templates are preserved exactly. Claude-Session: https://claude.ai/code/session_01MHg5zs1VBjvRWQ54qHpfYD Co-authored-by: Claude <noreply@anthropic.com>
This commit is contained in:
Classic298
GitHub
parent
ca2d7c9deb
commit
e64acf1c0a
1 changed files with 39 additions and 25 deletions
|
|
@ -1172,10 +1172,16 @@ async def chat_completion_tools_handler(
|
|||
event_emitter = extra_params['__event_emitter__']
|
||||
metadata = extra_params['__metadata__']
|
||||
|
||||
# One batched SELECT instead of four sequential round trips.
|
||||
task_config = await Config.get_many(
|
||||
'task.model.default',
|
||||
'task.model.external',
|
||||
'task.tools.prompt_template',
|
||||
)
|
||||
task_model_id = get_task_model_id(
|
||||
body['model'],
|
||||
await Config.get('task.model.default'),
|
||||
await Config.get('task.model.external'),
|
||||
task_config.get('task.model.default'),
|
||||
task_config.get('task.model.external'),
|
||||
models,
|
||||
)
|
||||
|
||||
|
|
@ -1185,8 +1191,9 @@ async def chat_completion_tools_handler(
|
|||
specs = [tool['spec'] for tool in tools.values()]
|
||||
tools_specs = json.dumps(specs, ensure_ascii=False)
|
||||
|
||||
if await Config.get('task.tools.prompt_template') != '':
|
||||
template = await Config.get('task.tools.prompt_template')
|
||||
tools_prompt_template = task_config.get('task.tools.prompt_template')
|
||||
if tools_prompt_template != '':
|
||||
template = tools_prompt_template
|
||||
else:
|
||||
template = DEFAULT_TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE
|
||||
|
||||
|
|
@ -1859,6 +1866,15 @@ async def chat_completion_files_handler(
|
|||
queries = [get_last_user_message(body['messages']) or '']
|
||||
|
||||
try:
|
||||
# One batched SELECT instead of six sequential round trips.
|
||||
rag_config = await Config.get_many(
|
||||
'rag.top_k',
|
||||
'rag.top_k_reranker',
|
||||
'rag.relevance_threshold',
|
||||
'rag.hybrid_bm25_weight',
|
||||
'rag.enable_hybrid_search',
|
||||
'rag.full_context',
|
||||
)
|
||||
# Directly await async get_sources_from_items (no thread needed - fully async now)
|
||||
sources = await get_sources_from_items(
|
||||
request=request,
|
||||
|
|
@ -1867,17 +1883,17 @@ async def chat_completion_files_handler(
|
|||
embedding_function=lambda query, prefix: request.app.state.EMBEDDING_FUNCTION(
|
||||
query, prefix=prefix, user=user
|
||||
),
|
||||
k=await Config.get('rag.top_k'),
|
||||
k=rag_config.get('rag.top_k'),
|
||||
reranking_function=(
|
||||
(lambda query, documents: request.app.state.RERANKING_FUNCTION(query, documents, user=user))
|
||||
if request.app.state.RERANKING_FUNCTION
|
||||
else None
|
||||
),
|
||||
k_reranker=await Config.get('rag.top_k_reranker'),
|
||||
r=await Config.get('rag.relevance_threshold'),
|
||||
hybrid_bm25_weight=await Config.get('rag.hybrid_bm25_weight'),
|
||||
hybrid_search=await Config.get('rag.enable_hybrid_search'),
|
||||
full_context=all_full_context or await Config.get('rag.full_context'),
|
||||
k_reranker=rag_config.get('rag.top_k_reranker'),
|
||||
r=rag_config.get('rag.relevance_threshold'),
|
||||
hybrid_bm25_weight=rag_config.get('rag.hybrid_bm25_weight'),
|
||||
hybrid_search=rag_config.get('rag.enable_hybrid_search'),
|
||||
full_context=all_full_context or rag_config.get('rag.full_context'),
|
||||
user=user,
|
||||
)
|
||||
except Exception as e:
|
||||
|
|
@ -2492,9 +2508,8 @@ async def process_chat_payload(request, form_data, user, metadata, model):
|
|||
if features:
|
||||
if 'voice' in features and features['voice']:
|
||||
if await Config.get('task.voice.prompt.enable'):
|
||||
if await Config.get('task.voice.prompt_template'):
|
||||
template = await Config.get('task.voice.prompt_template')
|
||||
else:
|
||||
template = await Config.get('task.voice.prompt_template')
|
||||
if not template:
|
||||
template = DEFAULT_VOICE_MODE_PROMPT_TEMPLATE
|
||||
|
||||
form_data['messages'] = add_or_update_system_message(
|
||||
|
|
@ -2521,11 +2536,8 @@ async def process_chat_payload(request, form_data, user, metadata, model):
|
|||
# Skip XML-tag prompt injection when native FC is enabled —
|
||||
# execute_code will be injected as a builtin tool instead
|
||||
if metadata.get('params', {}).get('function_calling') == 'legacy':
|
||||
prompt = (
|
||||
await Config.get('code_interpreter.prompt_template')
|
||||
if await Config.get('code_interpreter.prompt_template') != ''
|
||||
else DEFAULT_CODE_INTERPRETER_PROMPT
|
||||
)
|
||||
ci_prompt_template = await Config.get('code_interpreter.prompt_template')
|
||||
prompt = ci_prompt_template if ci_prompt_template != '' else DEFAULT_CODE_INTERPRETER_PROMPT
|
||||
|
||||
# Append filesystem awareness only for pyodide engine
|
||||
if engine != 'jupyter':
|
||||
|
|
@ -2605,12 +2617,13 @@ async def process_chat_payload(request, form_data, user, metadata, model):
|
|||
if skill_ids:
|
||||
from open_webui.models.skills import Skills as SkillsModel
|
||||
|
||||
accessible_skill_ids = {s.id for s in await SkillsModel.get_skills_by_user_id(user.id, 'read')}
|
||||
# Reuse the rows from the access query instead of re-fetching each
|
||||
# skill by id.
|
||||
accessible_skills = {s.id: s for s in await SkillsModel.get_skills_by_user_id(user.id, 'read')}
|
||||
for sid in skill_ids:
|
||||
if sid in accessible_skill_ids:
|
||||
s = await SkillsModel.get_skill_by_id(sid)
|
||||
if s and s.is_active:
|
||||
available_skills.append(s)
|
||||
s = accessible_skills.get(sid)
|
||||
if s and s.is_active:
|
||||
available_skills.append(s)
|
||||
|
||||
skill_manifest = ''
|
||||
for skill in available_skills:
|
||||
|
|
@ -5192,7 +5205,8 @@ async def streaming_chat_response_handler(response, ctx):
|
|||
""")
|
||||
code = blocking_code + '\n' + code
|
||||
|
||||
if await Config.get('code_interpreter.engine') == 'pyodide':
|
||||
ci_engine = await Config.get('code_interpreter.engine')
|
||||
if ci_engine == 'pyodide':
|
||||
ci_output = await event_caller(
|
||||
{
|
||||
'type': 'execute:python',
|
||||
|
|
@ -5204,7 +5218,7 @@ async def streaming_chat_response_handler(response, ctx):
|
|||
},
|
||||
}
|
||||
)
|
||||
elif await Config.get('code_interpreter.engine') == 'jupyter':
|
||||
elif ci_engine == 'jupyter':
|
||||
ci_output = await execute_code_jupyter(
|
||||
await Config.get('code_interpreter.jupyter.url'),
|
||||
code,
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue