343 lines
15 KiB
Python
343 lines
15 KiB
Python
import json
|
|
import os
|
|
import sys
|
|
from typing import Optional, Dict, Any, List, Callable
|
|
import asyncio
|
|
from openai import AsyncOpenAI
|
|
from openai.types.chat import ChatCompletion
|
|
import httpx
|
|
from .error_util import format_api_error
|
|
|
|
class LLMService:
|
|
_provider_base_urls = {
|
|
'openai': None, # 使用默认
|
|
'siliconflow': 'https://api.siliconflow.cn/v1',
|
|
'zhipu': 'https://open.bigmodel.cn/api/paas/v4',
|
|
'custom': None # 使用配置中的自定义URL
|
|
}
|
|
|
|
@classmethod
|
|
def get_openai_client(cls, api_key: str, provider: str) -> AsyncOpenAI:
|
|
"""获取OpenAI客户端"""
|
|
# 否则创建新客户端
|
|
base_url = None
|
|
|
|
# 如果是自定义提供商，从配置中获取base_url
|
|
if provider == 'custom':
|
|
from ..config_manager import config_manager
|
|
config = config_manager.get_llm_config()
|
|
if 'providers' in config and 'custom' in config['providers']:
|
|
base_url = config['providers']['custom'].get('base_url')
|
|
# 确保base_url不以/chat/completions结尾，避免路径重复
|
|
if base_url and base_url.endswith('/chat/completions'):
|
|
base_url = base_url.rstrip('/chat/completions')
|
|
else:
|
|
base_url = cls._provider_base_urls.get(provider)
|
|
|
|
# 创建简化的httpx客户端，不使用HTTP/2，避免额外依赖
|
|
http_client = httpx.AsyncClient(
|
|
timeout=httpx.Timeout(15.0) # 设置超时时间，稍微增加以适应网络延迟
|
|
)
|
|
|
|
kwargs = {
|
|
"api_key": api_key,
|
|
"http_client": http_client,
|
|
"max_retries": 2 # 设置最大重试次数
|
|
}
|
|
if base_url:
|
|
# 确保base_url末尾没有斜杠
|
|
base_url = base_url.rstrip('/')
|
|
kwargs["base_url"] = base_url
|
|
|
|
# 添加调试日志
|
|
from ..server import PREFIX
|
|
print(f"{PREFIX} 创建OpenAI客户端 | 提供商:{provider} | 基础URL:{base_url}")
|
|
|
|
# 创建客户端并缓存
|
|
client = AsyncOpenAI(**kwargs)
|
|
return client
|
|
|
|
@staticmethod
|
|
def _get_config() -> Dict[str, Any]:
|
|
"""获取LLM配置"""
|
|
from ..config_manager import config_manager
|
|
config = config_manager.get_llm_config()
|
|
current_provider = config.get('provider')
|
|
|
|
# 获取实际配置
|
|
if 'providers' in config and current_provider in config['providers']:
|
|
provider_config = config['providers'][current_provider]
|
|
return {
|
|
'provider': current_provider,
|
|
'model': provider_config.get('model', ''),
|
|
'base_url': provider_config.get('base_url', ''),
|
|
'api_key': provider_config.get('api_key', ''),
|
|
'temperature': provider_config.get('temperature', 0.7),
|
|
'top_p': provider_config.get('top_p', 0.9),
|
|
'max_tokens': provider_config.get('max_tokens', 2000)
|
|
}
|
|
else:
|
|
# 兼容旧版配置格式
|
|
return config
|
|
|
|
@staticmethod
|
|
async def expand_prompt(prompt: str, request_id: Optional[str] = None, stream_callback: Optional[Callable[[str], None]] = None) -> Dict[str, Any]:
|
|
"""
|
|
使用大语言模型扩写提示词，自动判断用户输入语言，并设置大语言模型回答语言。
|
|
支持流式输出以提高响应速度。
|
|
|
|
参数:
|
|
prompt: 要扩写的提示词
|
|
request_id: 请求ID
|
|
stream_callback: 流式输出的回调函数
|
|
|
|
返回:
|
|
包含扩写结果的字典
|
|
"""
|
|
try:
|
|
# 获取配置
|
|
config = LLMService._get_config()
|
|
api_key = config.get('api_key')
|
|
model = config.get('model')
|
|
provider = config.get('provider', 'unknown')
|
|
temperature = config.get('temperature', 0.7)
|
|
top_p = config.get('top_p', 0.9)
|
|
max_tokens = config.get('max_tokens', 2000)
|
|
|
|
if not api_key:
|
|
return {"success": False, "error": "请先配置大语言模型 API密钥"}
|
|
if not model:
|
|
return {"success": False, "error": "未配置模型名称"}
|
|
|
|
# 从server.py导入颜色常量和前缀
|
|
from ..server import PREFIX, ERROR_PREFIX
|
|
|
|
# 获取提供商显示名称
|
|
provider_display_name = {
|
|
'zhipu': '智谱',
|
|
'siliconflow': '硅基流动',
|
|
'openai': 'OpenAI',
|
|
'custom': '自定义'
|
|
}.get(provider, provider)
|
|
|
|
print(f"{PREFIX} LLM扩写请求 | 服务:{provider_display_name} | ID:{request_id} | 内容:{prompt[:30]}...")
|
|
|
|
# 加载系统提示词
|
|
from ..config_manager import config_manager
|
|
system_prompts = config_manager.get_system_prompts()
|
|
|
|
if not system_prompts or 'expand_prompts' not in system_prompts:
|
|
return {"success": False, "error": "扩写系统提示词加载失败"}
|
|
|
|
# 获取激活的提示词ID
|
|
active_prompt_id = system_prompts.get('active_prompts', {}).get('expand', 'expand_default')
|
|
|
|
# 获取对应的提示词
|
|
if active_prompt_id not in system_prompts['expand_prompts']:
|
|
# 如果找不到激活的提示词，尝试使用第一个可用的提示词
|
|
if len(system_prompts['expand_prompts']) > 0:
|
|
active_prompt_id = list(system_prompts['expand_prompts'].keys())[0]
|
|
else:
|
|
return {"success": False, "error": "未找到可用的扩写系统提示词"}
|
|
|
|
system_message = system_prompts['expand_prompts'][active_prompt_id]
|
|
|
|
# 输出使用的提示词名称
|
|
prompt_name = system_message.get('name', active_prompt_id)
|
|
print(f"{PREFIX} 使用扩写提示词: {prompt_name} | ID:{active_prompt_id}")
|
|
|
|
# 判断用户输入语言
|
|
def is_chinese(text):
|
|
return any('\u4e00' <= char <= '\u9fff' for char in text)
|
|
if is_chinese(prompt):
|
|
lang_message = {"role": "system", "content": "请用中文回答"}
|
|
else:
|
|
lang_message = {"role": "system", "content": "Please answer in English."}
|
|
|
|
# 构建消息
|
|
messages = [
|
|
lang_message,
|
|
system_message,
|
|
{"role": "user", "content": prompt}
|
|
]
|
|
|
|
# 使用OpenAI SDK
|
|
client = LLMService.get_openai_client(api_key, provider)
|
|
try:
|
|
# 添加调试信息
|
|
print(f"{PREFIX} 调用LLM API | 服务:{provider_display_name} | 模型:{model}")
|
|
|
|
# 使用配置中的参数
|
|
stream = await client.chat.completions.create(
|
|
model=model,
|
|
messages=[{"role": m["role"], "content": m["content"]} for m in messages],
|
|
temperature=temperature,
|
|
top_p=top_p,
|
|
max_tokens=max_tokens,
|
|
stream=True,
|
|
# 添加响应格式参数，减少不必要的token
|
|
response_format={"type": "text"}
|
|
)
|
|
|
|
full_content = ""
|
|
async for chunk in stream:
|
|
if chunk.choices[0].delta.content:
|
|
content = chunk.choices[0].delta.content
|
|
full_content += content
|
|
if stream_callback:
|
|
stream_callback(content)
|
|
|
|
# 输出结构化成功日志
|
|
print(f"{PREFIX} LLM扩写成功 | 服务:{provider_display_name} | 请求ID:{request_id} | 结果字符数:{len(full_content)}")
|
|
|
|
return {
|
|
"success": True,
|
|
"data": {
|
|
"original": prompt,
|
|
"expanded": full_content
|
|
}
|
|
}
|
|
except asyncio.CancelledError:
|
|
print(f"{PREFIX} LLM扩写任务在服务层被取消 | ID:{request_id}")
|
|
return {"success": False, "error": "请求已取消", "cancelled": True}
|
|
except Exception as e:
|
|
return {"success": False, "error": format_api_error(e, provider_display_name)}
|
|
|
|
except asyncio.CancelledError:
|
|
from ..server import PREFIX
|
|
print(f"{PREFIX} LLM扩写任务在服务层被取消 | ID:{request_id}")
|
|
return {"success": False, "error": "请求已取消", "cancelled": True}
|
|
except Exception as e:
|
|
# 从server.py导入颜色常量和前缀
|
|
from ..server import ERROR_PREFIX
|
|
print(f"{ERROR_PREFIX} LLM扩写请求失败 | 错误:{str(e)}")
|
|
return {"success": False, "error": str(e)}
|
|
|
|
@staticmethod
|
|
async def translate(text: str, from_lang: str = 'auto', to_lang: str = 'zh', request_id: Optional[str] = None, is_auto: bool = False, stream_callback: Optional[Callable[[str], None]] = None) -> Dict[str, Any]:
|
|
"""
|
|
使用大语言模型翻译文本，自动设置提示词语言和输出语言。
|
|
支持流式输出以提高响应速度。
|
|
|
|
参数:
|
|
text: 要翻译的文本
|
|
from_lang: 源语言 (默认为auto自动检测)
|
|
to_lang: 目标语言 (默认为zh中文)
|
|
request_id: 请求ID
|
|
is_auto: 是否为工作流自动翻译
|
|
stream_callback: 流式输出的回调函数
|
|
|
|
返回:
|
|
包含翻译结果的字典
|
|
"""
|
|
try:
|
|
# 获取配置
|
|
config = LLMService._get_config()
|
|
api_key = config.get('api_key')
|
|
model = config.get('model')
|
|
provider = config.get('provider', 'unknown')
|
|
temperature = config.get('temperature', 0.7)
|
|
top_p = config.get('top_p', 0.9)
|
|
max_tokens = config.get('max_tokens', 2000)
|
|
|
|
if not api_key:
|
|
return {"success": False, "error": "请先配置大语言模型 API密钥"}
|
|
if not model:
|
|
return {"success": False, "error": "未配置模型名称"}
|
|
|
|
# 从server.py导入颜色常量和前缀
|
|
from ..server import PREFIX, AUTO_TRANSLATE_PREFIX
|
|
|
|
# 获取提供商显示名称
|
|
provider_display_name = {
|
|
'zhipu': '智谱',
|
|
'siliconflow': '硅基流动',
|
|
'openai': 'OpenAI',
|
|
'custom': '自定义'
|
|
}.get(provider, provider)
|
|
|
|
# 使用统一的前缀
|
|
prefix = AUTO_TRANSLATE_PREFIX if is_auto else PREFIX
|
|
print(f"{prefix} {'工作流自动翻译' if is_auto else '翻译请求'} | 服务:{provider_display_name}翻译 | 请求ID:{request_id} | 原文长度:{len(text)} | 方向:{from_lang}->{to_lang}")
|
|
|
|
# 加载系统提示词
|
|
from ..config_manager import config_manager
|
|
system_prompts = config_manager.get_system_prompts()
|
|
|
|
if not system_prompts or 'translate_prompts' not in system_prompts or 'ZH' not in system_prompts['translate_prompts']:
|
|
return {"success": False, "error": "翻译系统提示词加载失败"}
|
|
|
|
system_message = system_prompts['translate_prompts']['ZH']
|
|
|
|
# 动态替换提示词中的{src_lang}和{dst_lang}
|
|
lang_map = {'zh': '中文', 'en': '英文', 'auto': '原文'}
|
|
src_lang = lang_map.get(from_lang, from_lang)
|
|
dst_lang = lang_map.get(to_lang, to_lang)
|
|
sys_msg_content = system_message['content'].replace('{src_lang}', src_lang).replace('{dst_lang}', dst_lang)
|
|
sys_msg = {"role": "system", "content": sys_msg_content}
|
|
|
|
# 设置输出语言
|
|
if to_lang == 'en':
|
|
lang_message = {"role": "system", "content": "Please answer in English."}
|
|
else:
|
|
lang_message = {"role": "system", "content": "请用中文回答"}
|
|
|
|
# 构建消息
|
|
messages = [
|
|
lang_message,
|
|
sys_msg,
|
|
{"role": "user", "content": text}
|
|
]
|
|
|
|
# 使用OpenAI SDK
|
|
client = LLMService.get_openai_client(api_key, provider)
|
|
try:
|
|
# 添加调试信息
|
|
print(f"{PREFIX} 调用LLM API | 服务:{provider_display_name} | 模型:{model}")
|
|
|
|
# 使用配置中的参数
|
|
stream = await client.chat.completions.create(
|
|
model=model,
|
|
messages=[{"role": m["role"], "content": m["content"]} for m in messages],
|
|
temperature=temperature,
|
|
top_p=top_p,
|
|
max_tokens=max_tokens,
|
|
stream=True,
|
|
# 添加响应格式参数，减少不必要的token
|
|
response_format={"type": "text"}
|
|
)
|
|
|
|
full_content = ""
|
|
async for chunk in stream:
|
|
if chunk.choices[0].delta.content:
|
|
content = chunk.choices[0].delta.content
|
|
full_content += content
|
|
if stream_callback:
|
|
stream_callback(content)
|
|
|
|
# 输出结构化成功日志
|
|
prefix = AUTO_TRANSLATE_PREFIX if is_auto else PREFIX
|
|
print(f"{prefix} {'工作流翻译完成' if is_auto else '翻译完成'} | 服务:{provider_display_name}翻译 | 请求ID:{request_id} | 结果字符数:{len(full_content)}")
|
|
|
|
return {
|
|
"success": True,
|
|
"data": {
|
|
"from": from_lang,
|
|
"to": to_lang,
|
|
"original": text,
|
|
"translated": full_content
|
|
}
|
|
}
|
|
except asyncio.CancelledError:
|
|
print(f"{prefix} {'工作流翻译' if is_auto else '翻译'}任务在服务层被取消 | ID:{request_id}")
|
|
return {"success": False, "error": "请求已取消", "cancelled": True}
|
|
except Exception as e:
|
|
return {"success": False, "error": format_api_error(e, provider_display_name)}
|
|
|
|
except asyncio.CancelledError:
|
|
from ..server import PREFIX, AUTO_TRANSLATE_PREFIX
|
|
prefix = AUTO_TRANSLATE_PREFIX if is_auto else PREFIX
|
|
print(f"{prefix} {'工作流翻译' if is_auto else '翻译'}任务在服务层被取消 | ID:{request_id}")
|
|
return {"success": False, "error": "请求已取消", "cancelled": True}
|
|
except Exception as e:
|
|
return {"success": False, "error": str(e)} |