diff --git a/pkg/pipeline/process/handlers/chat.py b/pkg/pipeline/process/handlers/chat.py
index c88d3ce1..4b190370 100644
--- a/pkg/pipeline/process/handlers/chat.py
+++ b/pkg/pipeline/process/handlers/chat.py
@@ -63,7 +63,6 @@ class ChatMessageHandler(handler.MessageHandler):
is_stream = await query.adapter.is_stream_output_supported()
except AttributeError:
is_stream = False
- print(is_stream)
try:
for r in runner_module.preregistered_runners:
@@ -74,7 +73,7 @@ class ChatMessageHandler(handler.MessageHandler):
raise ValueError(f'未找到请求运行器: {query.pipeline_config["ai"]["runner"]["runner"]}')
if is_stream:
resp_message_id = uuid.uuid4()
- if await query.adapter.create_message_card(resp_message_id,query.message_event):
+ if await query.adapter.create_message_card(resp_message_id, query.message_event):
async for result in runner.run(query):
result.resp_message_id = resp_message_id
if query.resp_messages:
@@ -82,43 +81,21 @@ class ChatMessageHandler(handler.MessageHandler):
if query.resp_message_chain:
query.resp_message_chain.pop()
-
query.resp_messages.append(result)
- self.ap.logger.info(f'对话({query.query_id})流式响应: {self.cut_str(result.readable_str())}')
+ self.ap.logger.info(
+ f'对话({query.query_id})流式响应: {self.cut_str(result.readable_str())}'
+ )
if result.content is not None:
text_length += len(result.content)
yield entities.StageProcessResult(result_type=entities.ResultType.CONTINUE, new_query=query)
else:
- yield entities.StageProcessResult(result_type=entities.ResultType.INTERRUPT,
- new_query=query)
- # for result in results:
- #
- # query.resp_messages.append(result)
- # print(result)
- #
- # self.ap.logger.info(f'对话({query.query_id})流式响应: {self.cut_str(result.content)}')
- #
- # if result.content is not None:
- # text_length += len(result.content)
- #
- # # current_chain = platform_message.MessageChain([])
- # # for msg in accumulated_messages:
- # # if msg.content is not None:
- # # current_chain.append(platform_message.Plain(msg.content))
- # # query.resp_message_chain = [current_chain]
- #
- # yield entities.StageProcessResult(result_type=entities.ResultType.CONTINUE, new_query=query)
- # query.resp_messages.append(results)
- # self.ap.logger.info(f'对话({query.query_id})响应')
- # yield entities.StageProcessResult(result_type=entities.ResultType.CONTINUE, new_query=query)
+ yield entities.StageProcessResult(result_type=entities.ResultType.INTERRUPT, new_query=query)
else:
- print("非流式")
async for result in runner.run(query):
query.resp_messages.append(result)
- print(result)
self.ap.logger.info(f'对话({query.query_id})响应: {self.cut_str(result.readable_str())}')
@@ -128,7 +105,7 @@ class ChatMessageHandler(handler.MessageHandler):
yield entities.StageProcessResult(result_type=entities.ResultType.CONTINUE, new_query=query)
query.session.using_conversation.messages.append(query.user_message)
-
+
query.session.using_conversation.messages.extend(query.resp_messages)
except Exception as e:
self.ap.logger.error(f'对话({query.query_id})请求失败: {type(e).__name__} {str(e)}')
diff --git a/pkg/pipeline/respback/respback.py b/pkg/pipeline/respback/respback.py
index 9a410b3f..f4153218 100644
--- a/pkg/pipeline/respback/respback.py
+++ b/pkg/pipeline/respback/respback.py
@@ -39,11 +39,9 @@ class SendResponseBackStage(stage.PipelineStage):
quote_origin = query.pipeline_config['output']['misc']['quote-origin']
- has_chunks = any(isinstance(msg, llm_entities.MessageChunk) for msg in query.resp_messages)
- print(has_chunks)
- if has_chunks and hasattr(query.adapter,'reply_message_chunk'):
+ # has_chunks = any(isinstance(msg, llm_entities.MessageChunk) for msg in query.resp_messages)
+ if await query.adapter.is_stream_output_supported():
is_final = [msg.is_final for msg in query.resp_messages][0]
- print(is_final)
await query.adapter.reply_message_chunk(
message_source=query.message_event,
message_id=query.resp_messages[-1].resp_message_id,
@@ -58,10 +56,6 @@ class SendResponseBackStage(stage.PipelineStage):
quote_origin=quote_origin,
)
- # await query.adapter.reply_message(
- # message_source=query.message_event,
- # message=query.resp_message_chain[-1],
- # quote_origin=quote_origin,
- # )
+
return entities.StageProcessResult(result_type=entities.ResultType.CONTINUE, new_query=query)
diff --git a/pkg/platform/adapter.py b/pkg/platform/adapter.py
index d4b48ef6..e4369efb 100644
--- a/pkg/platform/adapter.py
+++ b/pkg/platform/adapter.py
@@ -25,7 +25,6 @@ class MessagePlatformAdapter(metaclass=abc.ABCMeta):
logger: EventLogger
- is_stream: bool
def __init__(self, config: dict, ap: app.Application, logger: EventLogger):
"""初始化适配器
@@ -62,26 +61,31 @@ class MessagePlatformAdapter(metaclass=abc.ABCMeta):
quote_origin (bool, optional): 是否引用原消息. Defaults to False.
"""
raise NotImplementedError
-
+
async def reply_message_chunk(
self,
- message_source: platform_events.MessageEvent,
+ message_source: platform_events.MessageEvent,
message_id: int,
message: platform_message.MessageChain,
quote_origin: bool = False,
is_final: bool = False,
- ):
+ ):
"""回复消息（流式输出）
Args:
message_source (platform.types.MessageEvent): 消息源事件
message_id (int): 消息ID
message (platform.types.MessageChain): 消息链
quote_origin (bool, optional): 是否引用原消息. Defaults to False.
+ is_final (bool, optional): 流式是否结束. Defaults to False.
"""
raise NotImplementedError
- async def create_message_card(self,message_id,event):
- '''创建卡片消息'''
+ async def create_message_card(self, message_id:typing.Type[str,int], event:platform_events.MessageEvent) -> bool:
+ """创建卡片消息
+ Args:
+ message_id (str): 消息ID
+ event (platform_events.MessageEvent): 消息源事件
+ """
return False
async def is_muted(self, group_id: int) -> bool:
@@ -117,11 +121,9 @@ class MessagePlatformAdapter(metaclass=abc.ABCMeta):
async def run_async(self):
"""异步运行"""
raise NotImplementedError
-
async def is_stream_output_supported(self) -> bool:
"""是否支持流式输出"""
- self.is_stream = False
return False
async def kill(self) -> bool:
diff --git a/pkg/platform/sources/dingtalk.py b/pkg/platform/sources/dingtalk.py
index 7de5975c..187bafb0 100644
--- a/pkg/platform/sources/dingtalk.py
+++ b/pkg/platform/sources/dingtalk.py
@@ -148,7 +148,7 @@ class DingTalkAdapter(adapter.MessagePlatformAdapter):
message: platform_message.MessageChain,
quote_origin: bool = False,
is_final: bool = False,
- ):
+ ):
event = await DingTalkEventConverter.yiri2target(
message_source,
)
@@ -158,13 +158,12 @@ class DingTalkAdapter(adapter.MessagePlatformAdapter):
content, at = await DingTalkMessageConverter.yiri2target(message)
- card_instance,card_instance_id = self.card_instance_id_dict[message_id]
+ card_instance, card_instance_id = self.card_instance_id_dict[message_id]
# print(card_instance_id)
- await self.bot.send_card_message(card_instance,card_instance_id,content,is_final)
+ await self.bot.send_card_message(card_instance, card_instance_id, content, is_final)
if is_final:
self.card_instance_id_dict.pop(message_id)
-
async def send_message(self, target_type: str, target_id: str, message: platform_message.MessageChain):
content = await DingTalkMessageConverter.yiri2target(message)
if target_type == 'person':
@@ -174,11 +173,11 @@ class DingTalkAdapter(adapter.MessagePlatformAdapter):
async def is_stream_output_supported(self) -> bool:
is_stream = False
- if self.config.get("enable-stream-reply", None):
+ if self.config.get('enable-stream-reply', None):
is_stream = True
return is_stream
- async def create_message_card(self,message_id,event):
+ async def create_message_card(self, message_id, event):
card_template_id = self.config['card_template_id']
incoming_message = event.source_platform_object.incoming_message
# message_id = incoming_message.message_id
@@ -186,7 +185,6 @@ class DingTalkAdapter(adapter.MessagePlatformAdapter):
self.card_instance_id_dict[message_id] = (card_instance, card_instance_id)
return True
-
def register_listener(
self,
event_type: typing.Type[platform_events.Event],
@@ -194,7 +192,6 @@ class DingTalkAdapter(adapter.MessagePlatformAdapter):
):
async def on_message(event: DingTalkEvent):
try:
- await self.is_stream_output_supported()
return await callback(
await self.event_converter.target2yiri(event, self.config['robot_name']),
self,
@@ -208,15 +205,7 @@ class DingTalkAdapter(adapter.MessagePlatformAdapter):
self.bot.on_message('GroupMessage')(on_message)
async def run_async(self):
- # config = self.config
- # self.bot = DingTalkClient(
- # client_id=config['client_id'],
- # client_secret=config['client_secret'],
- # robot_name=config['robot_name'],
- # robot_code=config['robot_code'],
- # markdown_card=config['markdown_card'],
- # logger=self.logger,
- # )
+
await self.bot.start()
async def kill(self) -> bool:
diff --git a/pkg/platform/sources/lark.py b/pkg/platform/sources/lark.py
index 1e7f00b2..dcafbf9f 100644
--- a/pkg/platform/sources/lark.py
+++ b/pkg/platform/sources/lark.py
@@ -344,7 +344,7 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
config: dict
quart_app: quart.Quart
ap: app.Application
-
+
message_id_to_card_id: typing.Dict[str, typing.Tuple[str, int]]
card_id_dict: dict[str, str]
@@ -395,25 +395,17 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
try:
event = await self.event_converter.target2yiri(p2v1, self.api_client)
except Exception as e:
- await self.logger.error(f"Error in lark callback: {traceback.format_exc()}")
+ await self.logger.error(f'Error in lark callback: {traceback.format_exc()}')
if event.__class__ in self.listeners:
await self.listeners[event.__class__](event, self)
return {'code': 200, 'message': 'ok'}
except Exception as e:
- await self.logger.error(f"Error in lark callback: {traceback.format_exc()}")
+ await self.logger.error(f'Error in lark callback: {traceback.format_exc()}')
return {'code': 500, 'message': 'error'}
-
-
-
-
-
-
-
async def on_message(event: lark_oapi.im.v1.P2ImMessageReceiveV1):
-
lb_event = await self.event_converter.target2yiri(event, self.api_client)
await self.listeners[type(lb_event)](lb_event, self)
@@ -435,29 +427,28 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
async def is_stream_output_supported(self) -> bool:
is_stream = False
- if self.config.get("enable-stream-reply", None):
+ if self.config.get('enable-stream-reply', None):
is_stream = True
return is_stream
- async def create_card_id(self,message_id):
+ async def create_card_id(self, message_id):
try:
is_stream = await self.is_stream_output_supported()
if is_stream:
self.ap.logger.debug('飞书支持stream输出,创建卡片......')
- card_data = {"schema": "2.0", "header": {"title": {"content": "bot", "tag": "plain_text"}},
- "body": {"elements": [
- {"tag": "markdown", "content": "[思考中.....]", "element_id": "markdown_1"}]},
- "config": {"streaming_mode": True,
- "streaming_config": {"print_strategy": "delay"}}} # delay / fast
+ card_data = {
+ 'schema': '2.0',
+ 'header': {'title': {'content': 'bot', 'tag': 'plain_text'}},
+ 'body': {'elements': [{'tag': 'markdown', 'content': '[思考中.....]', 'element_id': 'markdown_1'}]},
+ 'config': {'streaming_mode': True, 'streaming_config': {'print_strategy': 'delay'}},
+ } # delay / fast
- request: CreateCardRequest = CreateCardRequest.builder() \
- .request_body(
- CreateCardRequestBody.builder()
- .type("card_json")
- .data(json.dumps(card_data)) \
+ request: CreateCardRequest = (
+ CreateCardRequest.builder()
+ .request_body(CreateCardRequestBody.builder().type('card_json').data(json.dumps(card_data)).build())
.build()
- ).build()
+ )
# 发起请求
response: CreateCardResponse = self.api_client.cardkit.v1.card.create(request)
@@ -465,7 +456,8 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
# 处理失败返回
if not response.success():
raise Exception(
- f"client.cardkit.v1.card.create failed, code: {response.code}, msg: {response.msg}, log_id: {response.get_log_id()}, resp: \n{json.dumps(json.loads(response.raw.content), indent=4, ensure_ascii=False)}")
+ f'client.cardkit.v1.card.create failed, code: {response.code}, msg: {response.msg}, log_id: {response.get_log_id()}, resp: \n{json.dumps(json.loads(response.raw.content), indent=4, ensure_ascii=False)}'
+ )
self.ap.logger.debug(f'飞书卡片创建成功,卡片ID: {response.data.card_id}')
self.card_id_dict[message_id] = response.data.card_id
@@ -476,7 +468,7 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
except Exception as e:
self.ap.logger.error(f'飞书卡片创建失败,错误信息: {e}')
- async def create_message_card(self,message_id,event) -> str:
+ async def create_message_card(self, message_id, event) -> str:
"""
创建卡片消息。
使用卡片消息是因为普通消息更新次数有限制，而大模型流式返回结果可能很多而超过限制，而飞书卡片没有这个限制
@@ -545,7 +537,6 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
f'client.im.v1.message.reply failed, code: {response.code}, msg: {response.msg}, log_id: {response.get_log_id()}, resp: \n{json.dumps(json.loads(response.raw.content), indent=4, ensure_ascii=False)}'
)
-
async def reply_message_chunk(
self,
message_source: platform_events.MessageEvent,
@@ -559,33 +550,33 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
"""
lark_message = await self.message_converter.yiri2target(message, self.api_client)
-
self.seq += 1
-
-
text_message = ''
for ele in lark_message[0]:
if ele['tag'] == 'text':
text_message += ele['text']
elif ele['tag'] == 'md':
text_message += ele['text']
- print(text_message)
content = {
'type': 'card_json',
'data': {'card_id': self.card_id_dict[message_id], 'elements': {'content': text_message}},
}
- request: ContentCardElementRequest = ContentCardElementRequest.builder() \
- .card_id(self.card_id_dict[message_id]) \
- .element_id("markdown_1") \
- .request_body(ContentCardElementRequestBody.builder()
- # .uuid("a0d69e20-1dd1-458b-k525-dfeca4015204")
- .content(text_message)
- .sequence(self.seq)
- .build()) \
+ request: ContentCardElementRequest = (
+ ContentCardElementRequest.builder()
+ .card_id(self.card_id_dict[message_id])
+ .element_id('markdown_1')
+ .request_body(
+ ContentCardElementRequestBody.builder()
+ # .uuid("a0d69e20-1dd1-458b-k525-dfeca4015204")
+ .content(text_message)
+ .sequence(self.seq)
+ .build()
+ )
.build()
+ )
if is_final:
self.seq = 1
@@ -593,7 +584,6 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
# 发起请求
response: ContentCardElementResponse = self.api_client.cardkit.v1.card_element.content(request)
-
# 处理失败返回
if not response.success():
raise Exception(
@@ -601,13 +591,6 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
)
return
-
-
-
-
-
-
-
async def is_muted(self, group_id: int) -> bool:
return False
diff --git a/pkg/platform/sources/telegram.py b/pkg/platform/sources/telegram.py
index 363ad702..e021c7b7 100644
--- a/pkg/platform/sources/telegram.py
+++ b/pkg/platform/sources/telegram.py
@@ -167,7 +167,7 @@ class TelegramAdapter(adapter.MessagePlatformAdapter):
await self.listeners[type(lb_event)](lb_event, self)
await self.is_stream_output_supported()
except Exception as e:
- await self.logger.error(f"Error in telegram callback: {traceback.format_exc()}")
+ await self.logger.error(f'Error in telegram callback: {traceback.format_exc()}')
self.application = ApplicationBuilder().token(self.config['token']).build()
self.bot = self.application.bot
@@ -206,7 +206,6 @@ class TelegramAdapter(adapter.MessagePlatformAdapter):
await self.bot.send_message(**args)
-
async def reply_message_chunk(
self,
message_source: platform_events.MessageEvent,
@@ -214,8 +213,7 @@ class TelegramAdapter(adapter.MessagePlatformAdapter):
message: platform_message.MessageChain,
quote_origin: bool = False,
is_final: bool = False,
- ):
-
+ ):
assert isinstance(message_source.source_platform_object, Update)
components = await TelegramMessageConverter.yiri2target(message, self.bot)
args = {}
@@ -240,7 +238,6 @@ class TelegramAdapter(adapter.MessagePlatformAdapter):
if self.config['markdown_card'] is True:
args['parse_mode'] = 'MarkdownV2'
-
send_msg = await self.bot.send_message(**args)
send_msg_id = send_msg.message_id
self.msg_stream_id[message_id] = send_msg_id
@@ -264,16 +261,12 @@ class TelegramAdapter(adapter.MessagePlatformAdapter):
if is_final:
self.msg_stream_id.pop(message_id)
-
async def is_stream_output_supported(self) -> bool:
is_stream = False
- if self.config.get("enable-stream-reply", None):
+ if self.config.get('enable-stream-reply', None):
is_stream = True
- self.is_stream = is_stream
-
return is_stream
-
async def is_muted(self, group_id: int) -> bool:
return False
diff --git a/pkg/provider/modelmgr/requesters/chatcmpl.py b/pkg/provider/modelmgr/requesters/chatcmpl.py
index f30bfd4e..6e72d78e 100644
--- a/pkg/provider/modelmgr/requesters/chatcmpl.py
+++ b/pkg/provider/modelmgr/requesters/chatcmpl.py
@@ -17,14 +17,13 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
"""OpenAI ChatCompletion API 请求器"""
client: openai.AsyncClient
- is_content:bool
+ is_content: bool
default_config: dict[str, typing.Any] = {
'base_url': 'https://api.openai.com/v1',
'timeout': 120,
}
-
async def initialize(self):
self.client = openai.AsyncClient(
api_key='',
@@ -46,7 +45,6 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
args: dict,
extra_body: dict = {},
) -> chat_completion.ChatCompletion:
-
async for chunk in await self.client.chat.completions.create(**args, extra_body=extra_body):
yield chunk
@@ -66,23 +64,23 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
# deepseek的reasoner模型
if pipeline_config['trigger'].get('misc', '').get('remove_think'):
-
pass
else:
- if reasoning_content is not None :
- chatcmpl_message['content'] = '\n' + reasoning_content + '\n\n' + chatcmpl_message['content']
+ if reasoning_content is not None:
+ chatcmpl_message['content'] = (
+ '\n' + reasoning_content + '\n\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,
) -> llm_entities.MessageChunk:
-
# 处理流式chunk和完整响应的差异
# print(chat_completion.choices[0])
if hasattr(chat_completion, 'choices'):
@@ -98,7 +96,6 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
if 'role' not in delta or delta['role'] is None:
delta['role'] = 'assistant'
-
reasoning_content = delta['reasoning_content'] if 'reasoning_content' in delta else None
delta['content'] = '' if delta['content'] is None else delta['content']
@@ -106,13 +103,13 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
# deepseek的reasoner模型
if pipeline_config['trigger'].get('misc', '').get('remove_think'):
- if reasoning_content is not None :
+ if reasoning_content is not None:
pass
else:
delta['content'] = delta['content']
else:
if reasoning_content is not None and idx == 0:
- delta['content'] += f'\n{reasoning_content}'
+ delta['content'] += f'\n{reasoning_content}'
elif reasoning_content is None:
if self.is_content:
delta['content'] = delta['content']
@@ -122,7 +119,6 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
else:
delta['content'] += reasoning_content
-
message = llm_entities.MessageChunk(**delta)
return message
@@ -135,9 +131,10 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
use_funcs: list[tools_entities.LLMFunction] = None,
stream: bool = False,
extra_args: dict[str, typing.Any] = {},
- ) -> llm_entities.Message | typing.AsyncGenerator[llm_entities.MessageChunk, None]:
+ ) ->llm_entities.MessageChunk:
self.client.api_key = use_model.token_mgr.get_token()
+
args = {}
args['model'] = use_model.model_entity.name
@@ -163,14 +160,14 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
if stream:
current_content = ''
- args["stream"] = True
+ 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)
+ 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
@@ -182,15 +179,13 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
id=tool_call.id,
type=tool_call.type,
function=llm_entities.FunctionCall(
- name=tool_call.function.name if tool_call.function else '',
- arguments=''
+ 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):
@@ -198,11 +193,9 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
delta_message.content = current_content
if chunk_idx % 64 == 0 or delta_message.is_final:
-
yield delta_message
# return
-
async def _closure(
self,
query: core_entities.Query,
@@ -211,7 +204,7 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
use_funcs: list[tools_entities.LLMFunction] = None,
stream: bool = False,
extra_args: dict[str, typing.Any] = {},
- ) -> llm_entities.Message | typing.AsyncGenerator[llm_entities.MessageChunk, None]:
+ ) -> llm_entities.Message:
self.client.api_key = use_model.token_mgr.get_token()
args = {}
@@ -237,22 +230,15 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
args['messages'] = messages
-
-
# 发送请求
resp = await self._req(args, extra_body=extra_args)
# 处理请求结果
pipeline_config = query.pipeline_config
- message = await self._make_msg(resp,pipeline_config)
-
+ message = await self._make_msg(resp, pipeline_config)
return message
-
-
-
-
async def invoke_llm(
self,
query: core_entities.Query,
@@ -273,7 +259,6 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
req_messages.append(msg_dict)
try:
-
msg = await self._closure(
query=query,
req_messages=req_messages,
@@ -306,7 +291,7 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
funcs: typing.List[tools_entities.LLMFunction] = None,
stream: bool = False,
extra_args: dict[str, typing.Any] = {},
- ) -> llm_entities.MessageChunk:
+ ) -> llm_entities.MessageChunk:
req_messages = [] # req_messages 仅用于类内，外部同步由 query.messages 进行
for m in messages:
msg_dict = m.dict(exclude_none=True)
@@ -343,4 +328,4 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
except openai.RateLimitError as e:
raise errors.RequesterError(f'请求过于频繁或余额不足: {e.message}')
except openai.APIError as e:
- raise errors.RequesterError(f'请求错误: {e.message}')
\ No newline at end of file
+ raise errors.RequesterError(f'请求错误: {e.message}')
diff --git a/pkg/provider/modelmgr/requesters/deepseekchatcmpl.py b/pkg/provider/modelmgr/requesters/deepseekchatcmpl.py
index f57f624f..d75d0fb6 100644
--- a/pkg/provider/modelmgr/requesters/deepseekchatcmpl.py
+++ b/pkg/provider/modelmgr/requesters/deepseekchatcmpl.py
@@ -55,6 +55,6 @@ class DeepseekChatCompletions(chatcmpl.OpenAIChatCompletions):
raise errors.RequesterError('接口返回为空，请确定模型提供商服务是否正常')
pipeline_config = query.pipeline_config
# 处理请求结果
- message = await self._make_msg(resp,pipeline_config)
+ message = await self._make_msg(resp, pipeline_config)
return message
diff --git a/pkg/provider/runners/dashscopeapi.py b/pkg/provider/runners/dashscopeapi.py
index fe72b0a8..9bb5824c 100644
--- a/pkg/provider/runners/dashscopeapi.py
+++ b/pkg/provider/runners/dashscopeapi.py
@@ -185,8 +185,6 @@ class DashScopeAPIRunner(runner.RequestRunner):
# 将参考资料替换到文本中
pending_content = self._replace_references(pending_content, references_dict)
-
-
yield llm_entities.Message(
role='assistant',
content=pending_content,
@@ -261,13 +259,11 @@ class DashScopeAPIRunner(runner.RequestRunner):
role='assistant',
content=pending_content,
is_final=is_final,
-
)
# 保存当前会话的session_id用于下次对话的语境
query.session.using_conversation.uuid = stream_output.get('session_id')
-
else:
for chunk in response:
if chunk.get('status_code') != 200:
diff --git a/pkg/provider/runners/difysvapi.py b/pkg/provider/runners/difysvapi.py
index 7c7d81ad..8182cc54 100644
--- a/pkg/provider/runners/difysvapi.py
+++ b/pkg/provider/runners/difysvapi.py
@@ -148,7 +148,6 @@ class DifyServiceAPIRunner(runner.RequestRunner):
if mode == 'workflow':
if chunk['event'] == 'node_finished':
if not is_stream:
-
if chunk['data']['node_type'] == 'answer':
yield llm_entities.Message(
role='assistant',
@@ -274,7 +273,6 @@ class DifyServiceAPIRunner(runner.RequestRunner):
content=self._try_convert_thinking(pending_agent_message),
)
-
if chunk['event'] == 'agent_thought':
if chunk['tool'] != '' and chunk['observation'] != '': # 工具调用结果，跳过
continue
diff --git a/pkg/provider/runners/localagent.py b/pkg/provider/runners/localagent.py
index 92522f27..b70d4157 100644
--- a/pkg/provider/runners/localagent.py
+++ b/pkg/provider/runners/localagent.py
@@ -1,7 +1,6 @@
from __future__ import annotations
import json
-from ssl import ALERT_DESCRIPTION_BAD_CERTIFICATE_HASH_VALUE
import typing
from .. import runner
@@ -15,26 +14,27 @@ class LocalAgentRunner(runner.RequestRunner):
class ToolCallTracker:
"""工具调用追踪器"""
+
def __init__(self):
- self.active_calls: dict[str,dict] = {}
+ self.active_calls: dict[str, dict] = {}
self.completed_calls: list[llm_entities.ToolCall] = []
- async def run(self, query: core_entities.Query) -> typing.AsyncGenerator[llm_entities.Message | llm_entities.MessageChunk, None]:
+ async def run(
+ self, query: core_entities.Query
+ ) -> typing.AsyncGenerator[llm_entities.Message | llm_entities.MessageChunk, None]:
"""运行请求"""
pending_tool_calls = []
req_messages = query.prompt.messages.copy() + query.messages.copy() + [query.user_message]
try:
- # print(await query.adapter.is_stream_output_supported())
is_stream = await query.adapter.is_stream_output_supported()
except AttributeError:
is_stream = False
- # while True:
- # pass
+
if not is_stream:
# 非流式输出，直接请求
- # print(123)
+
msg = await query.use_llm_model.requester.invoke_llm(
query,
query.use_llm_model,
@@ -44,7 +44,6 @@ class LocalAgentRunner(runner.RequestRunner):
)
yield msg
final_msg = msg
- print(final_msg)
else:
# 流式输出，需要处理工具调用
tool_calls_map: dict[str, llm_entities.ToolCall] = {}
@@ -55,29 +54,28 @@ class LocalAgentRunner(runner.RequestRunner):
query.use_funcs,
stream=is_stream,
extra_args=query.use_llm_model.model_entity.extra_args,
- ):
+ ):
yield msg
- # if msg.tool_calls:
- # for tool_call in msg.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
+ if msg.tool_calls:
+ for tool_call in msg.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
final_msg = llm_entities.Message(
role=msg.role,
content=msg.all_content,
tool_calls=list(tool_calls_map.values()),
)
-
pending_tool_calls = final_msg.tool_calls
req_messages.append(final_msg)
@@ -117,8 +115,8 @@ class LocalAgentRunner(runner.RequestRunner):
req_messages,
query.use_funcs,
stream=is_stream,
- extra_args=query.use_llm_model.model_entity.extra_args,
- ):
+ extra_args=query.use_llm_model.model_entity.extra_args,
+ ):
yield msg
if msg.tool_calls:
for tool_call in msg.tool_calls:
@@ -127,8 +125,7 @@ class LocalAgentRunner(runner.RequestRunner):
id=tool_call.id,
type=tool_call.type,
function=llm_entities.FunctionCall(
- name=tool_call.function.name if tool_call.function else '',
- arguments=''
+ name=tool_call.function.name if tool_call.function else '', arguments=''
),
)
if tool_call.function and tool_call.function.arguments:
@@ -140,7 +137,6 @@ class LocalAgentRunner(runner.RequestRunner):
tool_calls=list(tool_calls_map.values()),
)
else:
- print("非流式")
# 处理完所有调用，再次请求
msg = await query.use_llm_model.requester.invoke_llm(
query,