xinyin025/LangBot
1
0
Fork 0
mirror of https://github.com/langbot-app/LangBot.git synced 2026-07-23 12:56:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

perf: ruff format & remove stream params in requester

Browse Source
This commit is contained in:
Junyan Qin
2025-08-03 13:08:51 +08:00
parent 52280d7a05
commit 377d455ec1
39 changed files with 685 additions and 856 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pkg/provider/modelmgr/requesters/ppiochatcmpl.py
+55 -56
View File
@@ -5,8 +5,8 @@ import typing
from . import chatcmpl
import openai.types.chat.chat_completion as chat_completion
from .. import errors, requester
from ....core import entities as core_entities, app
from .. import requester
from ....core import entities as core_entities
from ... import entities as llm_entities
from ...tools import entities as tools_entities
import re
@@ -25,9 +25,9 @@ class PPIOChatCompletions(chatcmpl.OpenAIChatCompletions):
is_think: bool = False
async def _make_msg(
self,
chat_completion: chat_completion.ChatCompletion,
pipeline_config: dict[str, typing.Any] = {'trigger': {'misc': {'remove_think': False}}},
self,
chat_completion: chat_completion.ChatCompletion,
pipeline_config: dict[str, typing.Any] = {'trigger': {'misc': {'remove_think': False}}},
) -> llm_entities.Message:
chatcmpl_message = chat_completion.choices[0].message.model_dump()
# print(chatcmpl_message.keys(), chatcmpl_message.values())
@@ -40,21 +40,24 @@ class PPIOChatCompletions(chatcmpl.OpenAIChatCompletions):
# deepseek的reasoner模型
if pipeline_config['trigger'].get('misc', '').get('remove_think'):
chatcmpl_message['content'] = re.sub(r'<think>.*?</think>', '', chatcmpl_message['content'], flags=re.DOTALL)
chatcmpl_message['content'] = re.sub(
r'<think>.*?</think>', '', chatcmpl_message['content'], flags=re.DOTALL
)
else:
if reasoning_content is not None:
chatcmpl_message['content'] = '<think>\n' + reasoning_content + '\n</think>\n' + chatcmpl_message['content']
chatcmpl_message['content'] = (
'<think>\n' + reasoning_content + '\n</think>\n' + chatcmpl_message['content']
)
message = llm_entities.Message(**chatcmpl_message)
return message
async def _make_msg_chunk(
self,
pipeline_config: dict[str, typing.Any],
chat_completion: chat_completion.ChatCompletion,
idx: int,
self,
pipeline_config: dict[str, typing.Any],
chat_completion: chat_completion.ChatCompletion,
idx: int,
) -> llm_entities.MessageChunk:
# 处理流式chunk和完整响应的差异
# print(chat_completion.choices[0])
@@ -80,7 +83,7 @@ class PPIOChatCompletions(chatcmpl.OpenAIChatCompletions):
if '<think>' in delta['content']:
self.is_think = True
delta['content'] = ''
if rf'</think>' in delta['content']:
if r'</think>' in delta['content']:
self.is_think = False
delta['content'] = ''
if not self.is_think:
@@ -95,15 +98,13 @@ class PPIOChatCompletions(chatcmpl.OpenAIChatCompletions):
return message
async def _closure_stream(
self,
query: core_entities.Query,
req_messages: list[dict],
use_model: requester.RuntimeLLMModel,
use_funcs: list[tools_entities.LLMFunction] = None,
stream: bool = False,
extra_args: dict[str, typing.Any] = {},
self,
query: core_entities.Query,
req_messages: list[dict],
use_model: requester.RuntimeLLMModel,
use_funcs: list[tools_entities.LLMFunction] = None,
extra_args: dict[str, typing.Any] = {},
) -> llm_entities.Message | typing.AsyncGenerator[llm_entities.MessageChunk, None]:
self.client.api_key = use_model.token_mgr.get_token()
@@ -130,40 +131,38 @@ class PPIOChatCompletions(chatcmpl.OpenAIChatCompletions):
args['messages'] = messages
if stream:
current_content = ''
args["stream"] = True
chunk_idx = 0
self.is_content = False
tool_calls_map: dict[str, llm_entities.ToolCall] = {}
pipeline_config = query.pipeline_config
async for chunk in self._req_stream(args, extra_body=extra_args):
# 处理流式消息
delta_message = await self._make_msg_chunk(pipeline_config, chunk, chunk_idx)
if delta_message.content:
current_content += delta_message.content
delta_message.content = current_content
# delta_message.all_content = current_content
if delta_message.tool_calls:
for tool_call in delta_message.tool_calls:
if tool_call.id not in tool_calls_map:
tool_calls_map[tool_call.id] = llm_entities.ToolCall(
id=tool_call.id,
type=tool_call.type,
function=llm_entities.FunctionCall(
name=tool_call.function.name if tool_call.function else '',
arguments=''
),
)
if tool_call.function and tool_call.function.arguments:
# 流式处理中，工具调用参数可能分多个chunk返回，需要追加而不是覆盖
tool_calls_map[tool_call.id].function.arguments += tool_call.function.arguments
current_content = ''
args['stream'] = True
chunk_idx = 0
self.is_content = False
tool_calls_map: dict[str, llm_entities.ToolCall] = {}
pipeline_config = query.pipeline_config
async for chunk in self._req_stream(args, extra_body=extra_args):
# 处理流式消息
delta_message = await self._make_msg_chunk(pipeline_config, chunk, chunk_idx)
if delta_message.content:
current_content += delta_message.content
delta_message.content = current_content
# delta_message.all_content = current_content
if delta_message.tool_calls:
for tool_call in delta_message.tool_calls:
if tool_call.id not in tool_calls_map:
tool_calls_map[tool_call.id] = llm_entities.ToolCall(
id=tool_call.id,
type=tool_call.type,
function=llm_entities.FunctionCall(
name=tool_call.function.name if tool_call.function else '', arguments=''
),
)
if tool_call.function and tool_call.function.arguments:
# 流式处理中，工具调用参数可能分多个chunk返回，需要追加而不是覆盖
tool_calls_map[tool_call.id].function.arguments += tool_call.function.arguments
chunk_idx += 1
chunk_choices = getattr(chunk, 'choices', None)
if chunk_choices and getattr(chunk_choices[0], 'finish_reason', None):
delta_message.is_final = True
delta_message.content = current_content
chunk_idx += 1
chunk_choices = getattr(chunk, 'choices', None)
if chunk_choices and getattr(chunk_choices[0], 'finish_reason', None):
delta_message.is_final = True
delta_message.content = current_content
if chunk_idx % 64 == 0 or delta_message.is_final:
yield delta_message
if chunk_idx % 64 == 0 or delta_message.is_final:
yield delta_message