243 lines
9.3 KiB
Python
Raw Normal View History

import json
from typing import Optional, Tuple, Union
from azure.ai.inference.models import (
CompletionsFinishReason,
ImageContentItem,
ImageUrl,
TextContentItem,
ToolMessage,
UserMessage,
)
from nonebot.adapters import Bot, Event
from nonebot.log import logger
from nonebot.matcher import (
Matcher,
current_bot,
current_event,
current_matcher,
)
from nonebot_plugin_alconna.uniseg import UniMessage, UniMsg
from openai import AsyncOpenAI
from openai.types.chat import ChatCompletion, ChatCompletionMessage
from .config import config
from .constants import SUPPORT_IMAGE_MODELS
from .instances import target_list
from .models import MarshoContext
from .plugin.func_call.caller import get_function_calls
from .plugin.func_call.models import SessionContext
from .util import (
extract_content_and_think,
get_image_b64,
get_nickname_by_user_id,
get_prompt,
make_chat_openai,
parse_richtext,
)
class MarshoHandler:
def __init__(
self,
client: AsyncOpenAI,
context: MarshoContext,
):
self.client = client
self.context = context
self.bot: Bot = current_bot.get()
self.event: Event = current_event.get()
# self.state: T_State = current_handler.get().state
self.matcher: Matcher = current_matcher.get()
self.message_id: str = UniMessage.get_message_id(self.event)
self.target = UniMessage.get_target(self.event)
async def process_user_input(
self, user_input: UniMsg, model_name: str
) -> Union[str, list]:
"""
处理用户输入为可输入 API 的格式并添加昵称提示
"""
is_support_image_model = (
model_name.lower()
in SUPPORT_IMAGE_MODELS + config.marshoai_additional_image_models
)
usermsg = [] if is_support_image_model else ""
user_nickname = await get_nickname_by_user_id(self.event.get_user_id())
if user_nickname:
nickname_prompt = f"\n此消息的说话者为: {user_nickname}"
else:
nickname_prompt = ""
for i in user_input: # type: ignore
if i.type == "text":
if is_support_image_model:
usermsg += [TextContentItem(text=i.data["text"] + nickname_prompt).as_dict()] # type: ignore
else:
usermsg += str(i.data["text"] + nickname_prompt) # type: ignore
elif i.type == "image":
if is_support_image_model:
usermsg.append( # type: ignore
ImageContentItem(
image_url=ImageUrl( # type: ignore
url=str(await get_image_b64(i.data["url"])) # type: ignore
) # type: ignore
).as_dict() # type: ignore
) # type: ignore
logger.info(f"输入图片 {i.data['url']}")
elif config.marshoai_enable_support_image_tip:
await UniMessage(
"*此模型不支持图片处理或管理员未启用此模型的图片支持。图片将被忽略。"
).send()
return usermsg # type: ignore
async def handle_single_chat(
self,
user_message: Union[str, list],
model_name: str,
tools_list: list,
tool_message: Optional[list] = None,
) -> ChatCompletion:
"""
处理单条聊天
"""
context_msg = get_prompt(model_name) + (
self.context.build(self.target.id, self.target.private)
)
response = await make_chat_openai(
client=self.client,
msg=context_msg + [UserMessage(content=user_message).as_dict()] + (tool_message if tool_message else []), # type: ignore
model_name=model_name,
tools=tools_list if tools_list else None,
)
return response
async def handle_function_call(
self,
completion: ChatCompletion,
user_message: Union[str, list],
model_name: str,
tools_list: list,
):
# function call
# 需要获取额外信息,调用函数工具
tool_msg = []
choice = completion.choices[0]
# await UniMessage(str(response)).send()
tool_calls = choice.message.tool_calls
# try:
# if tool_calls[0]["function"]["name"].startswith("$"):
# choice.message.tool_calls[0][
# "type"
# ] = "builtin_function" # 兼容 moonshot AI 内置函数的临时方案
# except:
# pass
tool_msg.append(choice.message)
for tool_call in tool_calls: # type: ignore
try:
function_args = json.loads(tool_call.function.arguments)
except json.JSONDecodeError:
function_args = json.loads(
tool_call.function.arguments.replace("'", '"')
)
# 删除args的placeholder参数
if "placeholder" in function_args:
del function_args["placeholder"]
logger.info(
f"调用函数 {tool_call.function.name.replace('-', '.')}\n参数:"
+ "\n".join([f"{k}={v}" for k, v in function_args.items()])
)
await UniMessage(
f"调用函数 {tool_call.function.name.replace('-', '.')}\n参数:"
+ "\n".join([f"{k}={v}" for k, v in function_args.items()])
).send()
if caller := get_function_calls().get(tool_call.function.name):
logger.debug(f"调用插件函数 {caller.full_name}")
# 权限检查,规则检查 TODO
# 实现依赖注入检查函数参数及参数注解类型对Event类型的参数进行注入
func_return = await caller.with_ctx(
SessionContext(
bot=self.bot,
event=self.event,
matcher=self.matcher,
state=None,
)
).call(**function_args)
else:
logger.error(f"未找到函数 {tool_call.function.name.replace('-', '.')}")
func_return = f"未找到函数 {tool_call.function.name.replace('-', '.')}"
tool_msg.append(
ToolMessage(tool_call_id=tool_call.id, content=func_return).as_dict() # type: ignore
)
# tool_msg[0]["tool_calls"][0]["type"] = "builtin_function"
# await UniMessage(str(tool_msg)).send()
return await self.handle_common_chat(
user_message=user_message,
model_name=model_name,
tools_list=tools_list,
tool_message=tool_msg,
)
async def handle_common_chat(
self,
user_message: Union[str, list],
model_name: str,
tools_list: list,
stream: bool = False,
tool_message: Optional[list] = None,
) -> Optional[Tuple[UserMessage, ChatCompletionMessage]]:
"""
处理一般聊天
"""
global target_list
if stream:
raise NotImplementedError
response = await self.handle_single_chat(
user_message=user_message,
model_name=model_name,
tools_list=tools_list,
tool_message=tool_message,
)
choice = response.choices[0]
# Sprint(choice)
# 当tool_calls非空时将finish_reason设置为TOOL_CALLS
if choice.message.tool_calls is not None and config.marshoai_fix_toolcalls:
choice.finish_reason = "tool_calls"
logger.info(f"完成原因:{choice.finish_reason}")
if choice.finish_reason == CompletionsFinishReason.STOPPED:
##### DeepSeek-R1 兼容部分 #####
choice_msg_content, choice_msg_thinking, choice_msg_after = (
extract_content_and_think(choice.message)
)
if choice_msg_thinking and config.marshoai_send_thinking:
await UniMessage("思维链:\n" + choice_msg_thinking).send()
##### 兼容部分结束 #####
if [self.target.id, self.target.private] not in target_list:
target_list.append([self.target.id, self.target.private])
# 对话成功发送消息
if config.marshoai_enable_richtext_parse:
await (await parse_richtext(str(choice_msg_content))).send(
reply_to=True
)
else:
await UniMessage(str(choice_msg_content)).send(reply_to=True)
return UserMessage(content=user_message), choice_msg_after
elif choice.finish_reason == CompletionsFinishReason.CONTENT_FILTERED:
# 对话失败,消息过滤
await UniMessage("*已被内容过滤器过滤。请调整聊天内容后重试。").send(
reply_to=True
)
return None
elif choice.finish_reason == CompletionsFinishReason.TOOL_CALLS:
return await self.handle_function_call(
response, user_message, model_name, tools_list
)
else:
await UniMessage(f"意外的完成原因:{choice.finish_reason}").send()
return None