优化对话转工单功能，添加重试机制以提高稳定性，限制处理会话数量为前2000个，更新示例查询和文件路径，增强代码可读性和维护性。同时新增数据库客户端功能，支持批量处理会话数据并导出至Excel。
This commit is contained in:
@@ -231,7 +231,7 @@ class DialogueToWorkorder:
|
||||
output_format = self.user_question_and_solution_parser.get_format_instructions()
|
||||
llm_prompt = prompt.format(output_format=output_format, dialogue_str=dialogue_str)
|
||||
|
||||
response = self.llm.invoke(user_prompt=llm_prompt)
|
||||
response = self.llm.invoke(user_prompt=llm_prompt, need_retry=False)
|
||||
|
||||
try:
|
||||
if response.content.count('user_question') == 1:
|
||||
@@ -261,7 +261,7 @@ class DialogueToWorkorder:
|
||||
except Exception as e:
|
||||
output_format = self.user_question_and_solution_list_parser.get_format_instructions()
|
||||
llm_prompt = prompt.format(output_format=output_format, dialogue_str=dialogue_str)
|
||||
response = self.llm.invoke(user_prompt=llm_prompt)
|
||||
response = self.llm.invoke(user_prompt=llm_prompt, need_retry=False)
|
||||
user_question_and_solution_temp = self.user_question_and_solution_list_parser.parse(response.content)
|
||||
return user_question_and_solution_temp.user_question_list
|
||||
|
||||
@@ -293,7 +293,7 @@ class DialogueToWorkorder:
|
||||
{dialogue_str}
|
||||
"""
|
||||
|
||||
response = self.llm.invoke(user_prompt=prompt)
|
||||
response = self.llm.invoke(user_prompt=prompt, need_retry=False)
|
||||
product_name_and_module_name = self.product_name_and_module_name_parser.parse(response.content)
|
||||
|
||||
return product_name_and_module_name.product_name, product_name_and_module_name.module_name
|
||||
@@ -322,7 +322,7 @@ class DialogueToWorkorder:
|
||||
{dialogue_str}
|
||||
"""
|
||||
|
||||
response = self.llm.invoke(user_prompt=prompt)
|
||||
response = self.llm.invoke(user_prompt=prompt, need_retry=False)
|
||||
product_line = self.product_line_parser.parse(response.content)
|
||||
|
||||
return product_line.product_line
|
||||
@@ -358,7 +358,7 @@ class DialogueToWorkorder:
|
||||
{dialogue_str}
|
||||
"""
|
||||
|
||||
response = self.llm.invoke(user_prompt=prompt)
|
||||
response = self.llm.invoke(user_prompt=prompt, need_retry=False)
|
||||
question_type = self.question_type_parser.parse(response.content)
|
||||
|
||||
return question_type.question_type
|
||||
@@ -394,7 +394,7 @@ class DialogueToWorkorder:
|
||||
|
||||
"""
|
||||
|
||||
response = self.llm.invoke(user_prompt=prompt)
|
||||
response = self.llm.invoke(user_prompt=prompt, need_retry=False)
|
||||
is_complaint = self.is_complaint_parser.parse(response.content)
|
||||
|
||||
return (is_complaint.is_dissatisfaction,
|
||||
@@ -479,7 +479,19 @@ class DialogueToWorkorder:
|
||||
|
||||
# 按会话ID分组
|
||||
conversation_dict = self.group_conversations_by_id(df)
|
||||
|
||||
# 限制处理的会话数量为前2000个
|
||||
if len(conversation_dict) > 2000:
|
||||
print(f"会话总数为 {len(conversation_dict)}，限制处理前2000个会话")
|
||||
# 获取所有会话ID
|
||||
conversation_ids = list(conversation_dict.keys())
|
||||
# 只保留前2000个会话
|
||||
limited_conversation_dict = {
|
||||
conversation_id: conversation_dict[conversation_id]
|
||||
for conversation_id in conversation_ids[:2000]
|
||||
}
|
||||
conversation_dict = limited_conversation_dict
|
||||
else:
|
||||
print(f"会话总数为 {len(conversation_dict)}，处理全部会话")
|
||||
# 使用线程池处理每个会话
|
||||
workorder_dict_list = []
|
||||
with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
|
||||
@@ -593,7 +605,7 @@ def main():
|
||||
args = parse_arguments()
|
||||
|
||||
# 设置默认文件路径
|
||||
conversation_excel_path = args.conversation_file or os.path.join('data', 'excel', '会话内容详情20250528110230.xlsx')
|
||||
conversation_excel_path = args.conversation_file or os.path.join('data', 'excel', '2025年1月到6月12号所有对话记录.xlsx')
|
||||
product_detail_excel_path = args.product_detail_file or os.path.join('data', 'excel', '产品详情_工单.xlsx')
|
||||
|
||||
# 创建处理实例
|
||||
|
||||
@@ -0,0 +1,537 @@
|
||||
#!/usr/bin/env python
|
||||
# -*- coding: utf-8 -*-
|
||||
|
||||
from __future__ import annotations
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import configparser
|
||||
import logging
|
||||
from datetime import datetime
|
||||
from typing import Any, Dict, List, Optional, Tuple, Union
|
||||
from dataclasses import dataclass
|
||||
from contextlib import contextmanager
|
||||
import threading
|
||||
import time
|
||||
from queue import Queue, Empty, Full
|
||||
|
||||
import pandas as pd
|
||||
import pymysql
|
||||
from pymysql.connections import Connection
|
||||
from pymysql.cursors import Cursor
|
||||
from tqdm import tqdm
|
||||
import concurrent.futures
|
||||
import sys
|
||||
|
||||
# 配置日志
|
||||
logging.basicConfig(
|
||||
level=logging.INFO,
|
||||
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
|
||||
handlers=[
|
||||
logging.FileHandler('./data/log/mariadb_client.log'),
|
||||
logging.StreamHandler()
|
||||
]
|
||||
)
|
||||
logger = logging.getLogger(__name__)
|
||||
os.makedirs('./data/log', exist_ok=True)
|
||||
|
||||
@dataclass
|
||||
class DatabaseConfig:
|
||||
"""数据库配置类"""
|
||||
host: str = '192.168.0.123'
|
||||
port: int = 3307
|
||||
user: str = 'fuzhimei'
|
||||
password: str = 'fuzhimei@135'
|
||||
charset: str = 'utf8mb4'
|
||||
connect_timeout: int = 10
|
||||
read_timeout: int = 300
|
||||
write_timeout: int = 300
|
||||
|
||||
@classmethod
|
||||
def from_config_file(cls, config_file: str = 'config.ini') -> 'DatabaseConfig':
|
||||
"""从配置文件加载配置"""
|
||||
if not os.path.exists(config_file):
|
||||
logger.warning(f"配置文件 {config_file} 不存在，使用默认配置")
|
||||
return cls()
|
||||
|
||||
config = configparser.ConfigParser()
|
||||
config.read(config_file, encoding='utf-8')
|
||||
|
||||
if 'database' not in config:
|
||||
logger.warning("配置文件中没有 [database] 部分，使用默认配置")
|
||||
return cls()
|
||||
|
||||
db_config = config['database']
|
||||
return cls(
|
||||
host=db_config.get('host', cls.host),
|
||||
port=int(db_config.get('port', cls.port)),
|
||||
user=db_config.get('user', cls.user),
|
||||
password=db_config.get('password', cls.password),
|
||||
charset=db_config.get('charset', cls.charset),
|
||||
connect_timeout=int(db_config.get('connect_timeout', cls.connect_timeout)),
|
||||
read_timeout=int(db_config.get('read_timeout', cls.read_timeout)),
|
||||
write_timeout=int(db_config.get('write_timeout', cls.write_timeout))
|
||||
)
|
||||
|
||||
|
||||
class ConnectionPool:
|
||||
"""数据库连接池"""
|
||||
|
||||
def __init__(self, config: DatabaseConfig, max_connections: int = 10):
|
||||
self.config = config
|
||||
self.max_connections = max_connections
|
||||
self.pool = Queue(maxsize=max_connections)
|
||||
self.active_connections = 0
|
||||
self.lock = threading.Lock()
|
||||
|
||||
# 预创建一些连接
|
||||
self._initialize_pool()
|
||||
|
||||
def _initialize_pool(self) -> None:
|
||||
"""初始化连接池，预创建一些连接"""
|
||||
initial_connections = min(3, self.max_connections)
|
||||
for _ in range(initial_connections):
|
||||
try:
|
||||
conn = self._create_connection()
|
||||
if conn:
|
||||
self.pool.put_nowait(conn)
|
||||
self.active_connections += 1
|
||||
except Full:
|
||||
break
|
||||
except Exception as e:
|
||||
logger.error(f"初始化连接池时创建连接失败: {e}")
|
||||
|
||||
def _create_connection(self) -> Optional[Connection]:
|
||||
"""创建新的数据库连接"""
|
||||
try:
|
||||
conn = pymysql.connect(
|
||||
host=self.config.host,
|
||||
port=self.config.port,
|
||||
user=self.config.user,
|
||||
password=self.config.password,
|
||||
charset=self.config.charset,
|
||||
connect_timeout=self.config.connect_timeout,
|
||||
read_timeout=self.config.read_timeout,
|
||||
write_timeout=self.config.write_timeout,
|
||||
autocommit=True
|
||||
)
|
||||
return conn
|
||||
except Exception as e:
|
||||
logger.error(f"创建数据库连接失败: {e}")
|
||||
return None
|
||||
|
||||
@contextmanager
|
||||
def get_connection(self):
|
||||
"""获取连接的上下文管理器"""
|
||||
conn = None
|
||||
try:
|
||||
# 尝试从池中获取连接
|
||||
try:
|
||||
conn = self.pool.get_nowait()
|
||||
except Empty:
|
||||
# 池中没有连接，尝试创建新连接
|
||||
with self.lock:
|
||||
if self.active_connections < self.max_connections:
|
||||
conn = self._create_connection()
|
||||
if conn:
|
||||
self.active_connections += 1
|
||||
else:
|
||||
raise Exception("无法创建新的数据库连接")
|
||||
else:
|
||||
# 等待可用连接
|
||||
logger.info("等待可用连接...")
|
||||
conn = self.pool.get(timeout=30)
|
||||
|
||||
# 检查连接是否仍然有效
|
||||
if conn and not self._is_connection_alive(conn):
|
||||
logger.warning("连接已失效，重新创建")
|
||||
try:
|
||||
conn.close()
|
||||
except:
|
||||
pass
|
||||
conn = self._create_connection()
|
||||
if not conn:
|
||||
raise Exception("重新创建连接失败")
|
||||
|
||||
yield conn
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"获取数据库连接时出错: {e}")
|
||||
if conn:
|
||||
try:
|
||||
conn.close()
|
||||
except:
|
||||
pass
|
||||
with self.lock:
|
||||
self.active_connections -= 1
|
||||
raise
|
||||
else:
|
||||
# 归还连接到池中
|
||||
if conn:
|
||||
try:
|
||||
self.pool.put_nowait(conn)
|
||||
except Full:
|
||||
# 池已满，关闭连接
|
||||
try:
|
||||
conn.close()
|
||||
except:
|
||||
pass
|
||||
with self.lock:
|
||||
self.active_connections -= 1
|
||||
|
||||
def _is_connection_alive(self, conn: Connection) -> bool:
|
||||
"""检查连接是否仍然有效"""
|
||||
try:
|
||||
conn.ping(reconnect=False)
|
||||
return True
|
||||
except:
|
||||
return False
|
||||
|
||||
def close_all(self) -> None:
|
||||
"""关闭所有连接"""
|
||||
logger.info("正在关闭连接池中的所有连接...")
|
||||
while not self.pool.empty():
|
||||
try:
|
||||
conn = self.pool.get_nowait()
|
||||
conn.close()
|
||||
except (Empty, Exception):
|
||||
break
|
||||
|
||||
self.active_connections = 0
|
||||
logger.info("连接池已关闭")
|
||||
|
||||
|
||||
class DataProcessor:
|
||||
"""数据处理器"""
|
||||
|
||||
@staticmethod
|
||||
def clean_html_tags(text: str) -> str:
|
||||
"""清除文本中的HTML标签"""
|
||||
if not isinstance(text, str):
|
||||
return str(text) if text is not None else ""
|
||||
|
||||
# 使用正则表达式移除HTML标签
|
||||
clean_text = re.sub(r'<[^>]+>', '', text)
|
||||
# 处理HTML实体
|
||||
html_entities = {
|
||||
' ': ' ',
|
||||
'<': '<',
|
||||
'>': '>',
|
||||
'&': '&',
|
||||
'"': '"',
|
||||
''': "'"
|
||||
}
|
||||
for entity, char in html_entities.items():
|
||||
clean_text = clean_text.replace(entity, char)
|
||||
|
||||
return clean_text.strip()
|
||||
|
||||
@staticmethod
|
||||
def messages_df_to_list(messages_df: pd.DataFrame) -> List[Dict[str, Any]]:
|
||||
"""将消息DataFrame转换为字典列表，使用高效的向量化操作"""
|
||||
if messages_df.empty:
|
||||
return []
|
||||
|
||||
# 过滤掉系统消息
|
||||
mask = (messages_df["MODE"] != "system") & (messages_df["SYSTEM_MODE_MESSAGE_TYPE"].isna())
|
||||
filtered_df = messages_df[mask].copy()
|
||||
|
||||
if filtered_df.empty:
|
||||
return []
|
||||
|
||||
# 向量化操作
|
||||
filtered_df['message_sender'] = filtered_df["MODE"].map({'reply': '坐席', 'receive': '访客'}).fillna('未知')
|
||||
|
||||
# 处理发送者昵称
|
||||
filtered_df['sender_nickname'] = filtered_df.apply(
|
||||
lambda row: row["AGENT_NAME"] if row["message_sender"] == "坐席" else row["CUS_NICK_NAME"],
|
||||
axis=1
|
||||
)
|
||||
|
||||
# 处理内容
|
||||
def process_content(row):
|
||||
content = row["CONTENT"]
|
||||
if row["MSG_TYPE"] == "attachment":
|
||||
return f"附件：{DataProcessor.clean_html_tags(content)}"
|
||||
elif row["MSG_TYPE"] == "image":
|
||||
return f"图片：{DataProcessor.clean_html_tags(content)}"
|
||||
else:
|
||||
return content
|
||||
|
||||
filtered_df['processed_content'] = filtered_df.apply(process_content, axis=1)
|
||||
|
||||
# 过滤掉空昵称
|
||||
filtered_df = filtered_df[filtered_df['sender_nickname'].notna() & (filtered_df['sender_nickname'] != '')]
|
||||
|
||||
# 转换为字典列表
|
||||
result = []
|
||||
for record in filtered_df.to_dict('records'):
|
||||
result.append({
|
||||
"账号id": record["ACCOUNT"],
|
||||
"会话id": record["SESSION_ID"],
|
||||
"消息内容": record["processed_content"],
|
||||
"消息发送者": record["message_sender"],
|
||||
"发送者昵称": record["sender_nickname"],
|
||||
"创建时间": record["CREATE_TIME"],
|
||||
})
|
||||
|
||||
return result
|
||||
|
||||
|
||||
class MariaDBClient:
|
||||
"""优化后的MariaDB数据库客户端"""
|
||||
|
||||
def __init__(self, config: DatabaseConfig, max_connections: int = 10):
|
||||
self.config = config
|
||||
self.connection_pool = ConnectionPool(config, max_connections)
|
||||
self.data_processor = DataProcessor()
|
||||
|
||||
def __enter__(self) -> 'MariaDBClient':
|
||||
return self
|
||||
|
||||
def __exit__(self, exc_type, exc_val, exc_tb) -> None:
|
||||
self.close()
|
||||
|
||||
def close(self) -> None:
|
||||
"""关闭客户端"""
|
||||
self.connection_pool.close_all()
|
||||
|
||||
def execute_query(self, sql: str, params: Optional[Tuple] = None) -> Tuple[Optional[pd.DataFrame], List[str]]:
|
||||
"""执行SQL查询"""
|
||||
try:
|
||||
with self.connection_pool.get_connection() as conn:
|
||||
with conn.cursor() as cursor:
|
||||
cursor.execute(sql, params)
|
||||
results = cursor.fetchall()
|
||||
|
||||
# 获取列名
|
||||
column_names = [desc[0] for desc in cursor.description] if cursor.description else []
|
||||
|
||||
if results:
|
||||
df = pd.DataFrame(results, columns=column_names)
|
||||
return df, column_names
|
||||
else:
|
||||
return pd.DataFrame(), column_names
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"执行查询时出错: {e}")
|
||||
logger.error(f"SQL: {sql}")
|
||||
return None, []
|
||||
|
||||
def query_sessions(self, start_date: str, end_date: str) -> Optional[pd.DataFrame]:
|
||||
"""查询指定日期范围内的会话数据"""
|
||||
sql = """
|
||||
SELECT ACCOUNT, BEGIN_TIME, END_TIME, CUST_SEND_MESSAGE_COUNT,
|
||||
AGENT_SEND_MESSAGE_COUNT, STATUS, CHANNEL_NAME, SESSION_ID, SESSION_TAG_NAME
|
||||
FROM crm_hlyj.crm_hlyj_dsri
|
||||
WHERE BEGIN_TIME >= %s
|
||||
AND BEGIN_TIME < %s
|
||||
AND STATUS = 'assign'
|
||||
ORDER BY BEGIN_TIME DESC
|
||||
"""
|
||||
|
||||
df, _ = self.execute_query(sql, (start_date, end_date))
|
||||
return df
|
||||
|
||||
def query_messages_by_session_id(self, session_id: str) -> Optional[pd.DataFrame]:
|
||||
"""根据会话ID查询消息详情"""
|
||||
sql = """
|
||||
SELECT CREATE_TIME, CUS_NICK_NAME, MODE, MSG_TYPE, AGENT_NAME, CONTENT,
|
||||
SESSION_ID, ACCOUNT, SYSTEM_MODE_MESSAGE_TYPE
|
||||
FROM crm_hlyj.crm_hlyj_dmri
|
||||
WHERE SESSION_ID = %s
|
||||
ORDER BY CREATE_TIME
|
||||
"""
|
||||
|
||||
df, _ = self.execute_query(sql, (session_id,))
|
||||
return df
|
||||
|
||||
def export_to_excel(self, data: List[Dict[str, Any]], filename: str, output_dir: str = "output") -> Optional[str]:
|
||||
"""导出数据到Excel文件"""
|
||||
if not data:
|
||||
logger.warning(f"没有数据可导出到 {filename}")
|
||||
return None
|
||||
|
||||
try:
|
||||
# 创建输出目录
|
||||
os.makedirs(output_dir, exist_ok=True)
|
||||
|
||||
# 生成文件路径
|
||||
# timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
|
||||
file_path = os.path.join(output_dir, f"{filename}.xlsx")
|
||||
|
||||
# 准备数据：不同对话之间添加空行
|
||||
all_rows = []
|
||||
current_session_id = None
|
||||
|
||||
for conversation in data:
|
||||
if not conversation: # 跳过空对话
|
||||
continue
|
||||
|
||||
# 如果是新的会话，添加空行（除了第一个会话）
|
||||
if current_session_id and current_session_id != conversation[0]["会话id"]:
|
||||
empty_row = {key: "" for key in conversation[0].keys()}
|
||||
all_rows.append(empty_row)
|
||||
|
||||
# 更新当前会话ID
|
||||
current_session_id = conversation[0]["会话id"]
|
||||
|
||||
# 添加当前会话的所有消息
|
||||
all_rows.extend(conversation)
|
||||
|
||||
# 创建DataFrame并导出
|
||||
if all_rows:
|
||||
df = pd.DataFrame(all_rows)
|
||||
with pd.ExcelWriter(file_path, engine='openpyxl') as writer:
|
||||
df.to_excel(writer, sheet_name='对话记录', index=False)
|
||||
|
||||
logger.info(f"数据已导出到 {file_path}")
|
||||
return file_path
|
||||
else:
|
||||
logger.warning("没有有效数据可导出")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"导出到Excel时出错: {e}")
|
||||
return None
|
||||
|
||||
|
||||
def process_session_batch(db_client: MariaDBClient, session_batch: pd.DataFrame) -> List[List[Dict[str, Any]]]:
|
||||
"""批量处理会话数据"""
|
||||
conversations = []
|
||||
|
||||
for _, session_row in session_batch.iterrows():
|
||||
try:
|
||||
session_id = session_row['SESSION_ID']
|
||||
messages_df = db_client.query_messages_by_session_id(session_id)
|
||||
|
||||
if messages_df is not None and not messages_df.empty:
|
||||
conversation = db_client.data_processor.messages_df_to_list(messages_df)
|
||||
if conversation:
|
||||
conversations.append(conversation)
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"处理会话 {session_row.get('SESSION_ID', 'unknown')} 时出错: {e}")
|
||||
continue
|
||||
|
||||
return conversations
|
||||
|
||||
|
||||
class SessionProcessor:
|
||||
"""会话处理器，负责批量和并发处理"""
|
||||
|
||||
def __init__(self, db_client: MariaDBClient, max_workers: int = None, batch_size: int = 50):
|
||||
self.db_client = db_client
|
||||
self.max_workers = max_workers if max_workers is not None else os.cpu_count()
|
||||
self.batch_size = batch_size
|
||||
self.temp_save_lock = threading.Lock() # 添加锁用于保护临时保存操作
|
||||
|
||||
logger.info(f"初始化会话处理器: max_workers={self.max_workers}, batch_size={self.batch_size}")
|
||||
|
||||
def process_sessions(self, sessions_df: pd.DataFrame) -> List[List[Dict[str, Any]]]:
|
||||
"""处理所有会话数据"""
|
||||
if sessions_df.empty:
|
||||
logger.warning("没有会话数据需要处理")
|
||||
return []
|
||||
|
||||
total_sessions = len(sessions_df)
|
||||
logger.info(f"开始处理 {total_sessions} 个会话...")
|
||||
|
||||
# 分批处理
|
||||
all_conversations = []
|
||||
batch_count = (total_sessions + self.batch_size - 1) // self.batch_size
|
||||
# 使用线程池处理批次
|
||||
with concurrent.futures.ThreadPoolExecutor(max_workers=self.max_workers) as executor:
|
||||
# 提交所有批次任务
|
||||
future_to_batch = {}
|
||||
|
||||
for i in range(0, total_sessions, self.batch_size):
|
||||
batch = sessions_df.iloc[i:i + self.batch_size]
|
||||
future = executor.submit(process_session_batch, self.db_client, batch)
|
||||
future_to_batch[future] = i // self.batch_size + 1
|
||||
|
||||
# 收集结果
|
||||
with tqdm(total=batch_count, desc="处理批次进度") as pbar:
|
||||
for future in concurrent.futures.as_completed(future_to_batch):
|
||||
try:
|
||||
batch_conversations = future.result()
|
||||
all_conversations.extend(batch_conversations)
|
||||
|
||||
# 使用锁保护临时列表的操作
|
||||
with self.temp_save_lock:
|
||||
# 每处理100个对话临时保存一次
|
||||
logger.info(f"临时保存 {len(all_conversations)} 个对话")
|
||||
temp_output_file = self.db_client.export_to_excel(
|
||||
all_conversations,
|
||||
f"客服对话记录_临时保存",
|
||||
output_dir="/data/QueryRewrite/data/excel"
|
||||
)
|
||||
if temp_output_file:
|
||||
logger.info(f"临时保存完成: {temp_output_file}")
|
||||
|
||||
batch_num = future_to_batch[future]
|
||||
logger.debug(f"批次 {batch_num} 完成，获得 {len(batch_conversations)} 个对话")
|
||||
|
||||
except Exception as e:
|
||||
batch_num = future_to_batch[future]
|
||||
logger.error(f"处理批次 {batch_num} 时出错: {e}")
|
||||
|
||||
pbar.update(1)
|
||||
|
||||
logger.info(f"处理完成，共获得 {len(all_conversations)} 个有效对话")
|
||||
return all_conversations
|
||||
|
||||
|
||||
def main() -> None:
|
||||
"""主函数"""
|
||||
try:
|
||||
# 加载配置
|
||||
config = DatabaseConfig.from_config_file()
|
||||
logger.info(f"使用数据库配置: {config.host}:{config.port}")
|
||||
|
||||
# 创建数据库客户端
|
||||
with MariaDBClient(config, max_connections=12) as db_client:
|
||||
# 查询会话数据
|
||||
start_date = '2025-01-01 00:00:00'
|
||||
end_date = '2025-06-12 00:00:00'
|
||||
|
||||
logger.info(f"查询时间范围: {start_date} 到 {end_date}")
|
||||
# 创建会话处理器
|
||||
processor = SessionProcessor(db_client, batch_size=100)
|
||||
is_debug = hasattr(sys, 'gettrace') and sys.gettrace() is not None
|
||||
if is_debug:
|
||||
messages_df = db_client.query_messages_by_session_id("86c919e0-09f1-11f0-84ae-2daf59566989")
|
||||
print(db_client.data_processor.messages_df_to_list(messages_df))
|
||||
return []
|
||||
|
||||
sessions_df = db_client.query_sessions(start_date, end_date)
|
||||
|
||||
if sessions_df is None or sessions_df.empty:
|
||||
logger.warning("没有找到符合条件的会话数据")
|
||||
return
|
||||
|
||||
# 处理会话数据
|
||||
all_conversations = processor.process_sessions(sessions_df)
|
||||
# 导出结果
|
||||
if all_conversations:
|
||||
output_file = db_client.export_to_excel(
|
||||
all_conversations,
|
||||
"客服对话记录",
|
||||
output_dir="/data/QueryRewrite/data/excel"
|
||||
)
|
||||
|
||||
if output_file:
|
||||
logger.info(f"处理完成！共导出 {len(all_conversations)} 个对话到文件: {output_file}")
|
||||
else:
|
||||
logger.error("导出文件失败")
|
||||
else:
|
||||
logger.warning("没有有效的对话数据可导出")
|
||||
|
||||
except KeyboardInterrupt:
|
||||
logger.info("用户中断程序")
|
||||
except Exception as e:
|
||||
logger.error(f"程序执行出错: {e}", exc_info=True)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -175,7 +175,7 @@ def save_results_to_excel(results, output_file, is_final=False):
|
||||
logging.info(f"已保存{len(valid_results)}条结果至: {temp_output_file}")
|
||||
|
||||
# 示例查询
|
||||
examples_query = """那西藏软件呢"""
|
||||
examples_query = """那储能软件如何操作"""
|
||||
conversation_context=""
|
||||
chat_history=[
|
||||
{
|
||||
@@ -214,8 +214,8 @@ def main():
|
||||
|
||||
# 读取提问数据
|
||||
current_dir = os.path.dirname(os.path.abspath(__file__))
|
||||
data_file = os.path.join(current_dir, "..", "..", "data", "excel", "历史提问数据(like)_提问明确.xlsx")
|
||||
output_file = os.path.join(current_dir, "..", "..", "data", "excel", "测试提问数据_槽位填充结果.xlsx")
|
||||
data_file = os.path.join(current_dir, "..", "..", "data", "excel", "200条点踩数据测试.xlsx")
|
||||
output_file = os.path.join(current_dir, "..", "..", "data", "excel", "200条点踩数据测试_槽位填充结果.xlsx")
|
||||
|
||||
# 检测是否为调试模式，调试模式下使用examples_query，否则从Excel读取
|
||||
is_debug = hasattr(sys, 'gettrace') and sys.gettrace() is not None
|
||||
@@ -226,7 +226,7 @@ def main():
|
||||
examples = load_questions_from_excel(data_file)
|
||||
|
||||
if not is_debug:
|
||||
max_workers = 40 # 减少并发数以避免API限制
|
||||
max_workers = 20 # 减少并发数以避免API限制
|
||||
logging.info(f"共有 {len(examples)} 个问题需要处理，使用 {max_workers} 个并发线程")
|
||||
|
||||
# 创建一个与输入顺序相同的结果列表
|
||||
@@ -260,9 +260,10 @@ def main():
|
||||
logging.info(f"所有处理完成，最终结果已保存至: {output_file}")
|
||||
else:
|
||||
for idx, query in enumerate(examples):
|
||||
if query.strip() == "":
|
||||
continue
|
||||
process_query(recognizer, query, conversation_context, chat_history, previous_slots)
|
||||
if query.strip() == "":
|
||||
continue
|
||||
process_query(recognizer, query, conversation_context, chat_history, previous_slots)
|
||||
# print(json.dumps(process_query(recognizer, query), ensure_ascii=False, indent=2))
|
||||
|
||||
def setup_logging():
|
||||
# 配置日志输出到控制台
|
||||
|
||||
@@ -6,10 +6,23 @@ import json
|
||||
import time
|
||||
import threading
|
||||
import datetime
|
||||
import logging
|
||||
|
||||
# 加载环境变量
|
||||
load_dotenv()
|
||||
|
||||
logging.basicConfig(
|
||||
level=logging.INFO,
|
||||
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
|
||||
handlers=[
|
||||
logging.StreamHandler()
|
||||
]
|
||||
)
|
||||
logging.getLogger('httpx').setLevel(logging.WARNING)
|
||||
logging.getLogger('openai').setLevel(logging.WARNING)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
app = Flask(__name__)
|
||||
|
||||
# 创建线程锁，用于保护共享资源
|
||||
@@ -50,8 +63,8 @@ def intent_recognize():
|
||||
|
||||
end_time = time.time()
|
||||
current_time = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S %z")
|
||||
print(f"[{current_time}] [{os.getpid()}] [INFO] 意图识别耗时: {end_time - start_time:.2f}秒")
|
||||
|
||||
logger.info(f"[{os.getpid()}] 意图识别耗时: {end_time - start_time:.2f}秒")
|
||||
|
||||
# 提取分类信息
|
||||
classification = result["classification"]
|
||||
|
||||
|
||||
@@ -150,12 +150,14 @@ class SoftwareFunctionSlots(SlotBase):
|
||||
software_name: str = Field(default="", description="软件名称")
|
||||
function_name: str = Field(default="", description="具体功能名称")
|
||||
operation: str = Field(default="", description="用户操作意图(如何使用功能、功能入口、功能使用场景)")
|
||||
project_type: Optional[str] = Field(default="单工程", description="工程类型(单工程、多工程、批次工程)")
|
||||
project_type: Optional[str] = Field(default="单工程", description="工程类型(单工程、多工程、批次工程), 未明确提及则默认下是(单工程)")
|
||||
software_version: Optional[str] = Field(default="", description="软件版本")
|
||||
operation_steps: Optional[str] = Field(default="", description="操作步骤描述")
|
||||
|
||||
def check_required_slots(self) -> Tuple[bool, Dict[str, str]]:
|
||||
"""检查必填槽位是否都存在"""
|
||||
if self.project_type is None or len(self.project_type) == 0:
|
||||
self.project_type="单工程"
|
||||
missing_slots = {}
|
||||
if not self.software_name:
|
||||
missing_slots["software_name"] = f"{SoftwareFunctionSlots.model_fields['software_name'].description}，可选值：{', '.join([name.value for name in SoftwareName if name not in [SoftwareName.UNKNOWN, SoftwareName.ALIASES]])}"
|
||||
|
||||
@@ -14,6 +14,8 @@ import json
|
||||
from typing import List, Tuple, Dict, Any, Optional
|
||||
import re
|
||||
import jieba
|
||||
import time
|
||||
|
||||
from .PromptTemplates import (classification_prompt, query_rewrite_prompt,
|
||||
extract_nouns_prompt, classification_info,
|
||||
slot_filling_prompt)
|
||||
@@ -95,7 +97,9 @@ class IntentRecognizer:
|
||||
except Exception as e:
|
||||
raise RuntimeError(f"加载后缀关键词失败: {e}") from e
|
||||
|
||||
def _classify_intent(self, query: str) -> Classification:
|
||||
def _classify_intent(self, query: str, conversation_context: str = "",
|
||||
chat_history: List[Dict[str, str]] = None,
|
||||
previous_slots: Dict[str, Any] = None) -> Classification:
|
||||
"""
|
||||
对用户输入进行意图分类
|
||||
|
||||
@@ -109,7 +113,9 @@ class IntentRecognizer:
|
||||
classification_parser = PydanticOutputParser(pydantic_object=Classification)
|
||||
formatted_prompt = classification_prompt.format(user_input=query,
|
||||
classification_info=classification_info,
|
||||
output_format=classification_parser.get_format_instructions())
|
||||
output_format=classification_parser.get_format_instructions(),
|
||||
conversation_context=conversation_context,
|
||||
chat_history=json.dumps(chat_history, ensure_ascii=False))
|
||||
|
||||
# 调用LLM
|
||||
response = self._llm.invoke(formatted_prompt, False)
|
||||
@@ -208,7 +214,7 @@ class IntentRecognizer:
|
||||
term_texts = ["名称:" + term.name + "|" + "同义词:" + ";".join(term.synonymous) for term in matched_terms]
|
||||
|
||||
# 使用重排序模型
|
||||
xinference_reranker = SiliconFlowReRankerModel()
|
||||
xinference_reranker = XinferenceReRankerModel()
|
||||
rerank_results = xinference_reranker.rerank(query_key, term_texts, top_k=top_k)
|
||||
|
||||
# 将matched_terms转换为列表以便按索引访问
|
||||
@@ -220,7 +226,7 @@ class IntentRecognizer:
|
||||
return reranked_terms
|
||||
|
||||
except Exception as e:
|
||||
raise RuntimeError(f"SiliconFlowReRankerModel重排失败：{e}") from e
|
||||
raise RuntimeError(f"_rerank_matched_terms重排失败：{e}") from e
|
||||
|
||||
def _match_keywords(self, query: str, use_jieba: bool = False) -> Tuple[TermList, List[str]]:
|
||||
"""
|
||||
@@ -233,18 +239,23 @@ class IntentRecognizer:
|
||||
Returns:
|
||||
匹配到的关键词列表
|
||||
"""
|
||||
start_time = time.time()
|
||||
query_keys=[]
|
||||
# 步骤1: 使用LLM提取查询中的关键词
|
||||
try:
|
||||
llm_start_time = time.time()
|
||||
extracted_terms = self._extract_keywords_with_llm(query, use_jieba)
|
||||
for term in extracted_terms:
|
||||
query_keys.append(term.name)
|
||||
llm_end_time = time.time()
|
||||
llm_time = llm_end_time - llm_start_time
|
||||
except Exception as e:
|
||||
raise RuntimeError(f"LLM关键词提取失败: {e}") from e
|
||||
|
||||
matched_terms = [] # 存储匹配到的Term对象
|
||||
# 步骤2: 使用向量检索找到相似的专业名词
|
||||
try:
|
||||
vector_start_time = time.time()
|
||||
# 对matched_terms中的每个关键字进行向量检索
|
||||
for current_key in query_keys:
|
||||
vector_results = self._noun_retriever.query(current_key, top_k=5, use_intersection=False)
|
||||
@@ -262,12 +273,20 @@ class IntentRecognizer:
|
||||
if len(current_key_terms) > 0:
|
||||
reranked_terms = self._rerank_matched_terms(current_key, current_key_terms)
|
||||
matched_terms.extend(reranked_terms)
|
||||
vector_end_time = time.time()
|
||||
vector_time = vector_end_time - vector_start_time
|
||||
except Exception as e:
|
||||
raise RuntimeError(f"向量检索关键词时出错: {e}") from e
|
||||
|
||||
# 提取所有Term对象的名称并排序
|
||||
# 将set类型的matched_terms转换为TermList类型
|
||||
term_list = TermList(terms=list(matched_terms))
|
||||
end_time = time.time()
|
||||
total_time = end_time - start_time
|
||||
|
||||
# 输出整合的时间日志
|
||||
logging.info(f"关键词匹配耗时统计 - 总耗时: {total_time:.2f}秒, 问题关键词提取: {llm_time:.2f}秒, 向量检索+重排序: {vector_time:.2f}秒")
|
||||
|
||||
return term_list, query_keys
|
||||
|
||||
def _rewrite_query(self, query: str, keywords: TermList, query_keys:List[str], chat_history: List[Dict[str, str]] = None, context: str = "") -> QueryRewrite:
|
||||
@@ -282,6 +301,8 @@ class IntentRecognizer:
|
||||
Returns:
|
||||
改写结果
|
||||
"""
|
||||
|
||||
rewrite_start_time = time.time()
|
||||
# 准备问题改写提示
|
||||
# terms_dict = [term.model_dump(exclude={"description"}) for term in keywords.terms]
|
||||
terms_dict = [term.model_dump() for term in keywords.terms]
|
||||
@@ -295,7 +316,7 @@ class IntentRecognizer:
|
||||
keywords=keywords_str,
|
||||
chat_history=chat_history,
|
||||
context=context)
|
||||
|
||||
|
||||
# 调用LLM
|
||||
response = self._llm.invoke(formatted_prompt, False)
|
||||
|
||||
@@ -303,6 +324,9 @@ class IntentRecognizer:
|
||||
try:
|
||||
# 尝试直接解析JSON响应
|
||||
parsed_output = query_rewrite_parser.parse(response.content)
|
||||
rewrite_end_time = time.time()
|
||||
rewrite_time = rewrite_end_time - rewrite_start_time
|
||||
logging.info(f"问题改写耗时统计 - 总耗时: {rewrite_time:.2f}秒")
|
||||
return parsed_output
|
||||
except Exception as e:
|
||||
raise RuntimeError(f"解析问题改写结果时出错: {e}") from e
|
||||
@@ -360,7 +384,10 @@ class IntentRecognizer:
|
||||
# suffix_terms.append(suffix_term)
|
||||
|
||||
# return Classification(vertical_classification="安装下载", sub_classification="查询"), TermList(terms=suffix_terms), QueryRewrite(rewrite=query), matched_suffixes
|
||||
|
||||
if chat_history is None:
|
||||
chat_history = []
|
||||
if previous_slots is None:
|
||||
previous_slots = {}
|
||||
# 步骤1: 匹配关键词
|
||||
keywords_terms, query_keys = self._match_keywords(query, use_jieba)
|
||||
|
||||
@@ -397,7 +424,9 @@ class IntentRecognizer:
|
||||
# }
|
||||
|
||||
|
||||
def _fill_slots(self, query: str, classification: Classification) -> Dict[str, Any]:
|
||||
def _fill_slots(self, query: str, classification: Classification, conversation_context: str = "",
|
||||
chat_history: List[Dict[str, str]] = None,
|
||||
previous_slots: Dict[str, Any] = None,) -> Dict[str, Any]:
|
||||
"""
|
||||
根据分类结果对问题进行槽位填充
|
||||
|
||||
@@ -415,7 +444,7 @@ class IntentRecognizer:
|
||||
raise RuntimeError("未找到匹配的槽位模型")
|
||||
|
||||
# 使用LLM进行槽位填充
|
||||
filled_slots = self._fill_slots_with_llm(query, classification, slot_model)
|
||||
filled_slots = self._fill_slots_with_llm(query, classification, slot_model, conversation_context, chat_history, previous_slots)
|
||||
|
||||
# 检查必填槽位是否都已填充
|
||||
is_complete, missing_slots = filled_slots.check_required_slots()
|
||||
@@ -467,7 +496,12 @@ class IntentRecognizer:
|
||||
|
||||
return None
|
||||
|
||||
def _fill_slots_with_llm(self, query: str, classification: Classification, slot_model_class: type) -> Any:
|
||||
def _fill_slots_with_llm(self, query: str,
|
||||
classification: Classification,
|
||||
slot_model_class: type,
|
||||
conversation_context: str = "",
|
||||
chat_history: List[Dict[str, str]] = None,
|
||||
previous_slots: Dict[str, Any] = None) -> Any:
|
||||
"""
|
||||
使用LLM进行槽位填充
|
||||
|
||||
@@ -486,7 +520,10 @@ class IntentRecognizer:
|
||||
query=query,
|
||||
vertical_classification=classification.vertical_classification,
|
||||
sub_classification=classification.sub_classification,
|
||||
output_format=slot_parser.get_format_instructions()
|
||||
output_format=slot_parser.get_format_instructions(),
|
||||
conversation_context=conversation_context,
|
||||
chat_history=json.dumps(chat_history,ensure_ascii=False),
|
||||
previous_slots=json.dumps(previous_slots,ensure_ascii=False),
|
||||
)
|
||||
|
||||
# 调用LLM
|
||||
@@ -537,9 +574,14 @@ class IntentRecognizer:
|
||||
output_format=parser.get_format_instructions(),
|
||||
classification_info=classification_info
|
||||
)
|
||||
|
||||
# 调用LLM
|
||||
llm_start_time = time.time()
|
||||
response = self._llm.invoke(formatted_prompt + output_example, False)
|
||||
llm_end_time = time.time()
|
||||
llm_time = llm_end_time - llm_start_time
|
||||
|
||||
|
||||
try:
|
||||
# 解析LLM响应为JSON
|
||||
result_json = parser.parse(response.content)
|
||||
@@ -552,8 +594,19 @@ class IntentRecognizer:
|
||||
if expected_slot_model is None:
|
||||
# 添加容错处理，应对LLM返回错误分类信息,一级分类跟二级分类错乱
|
||||
# 重新分类
|
||||
classification = self._classify_intent(user_input)
|
||||
fill_slots = self._fill_slots(user_input, classification)
|
||||
classify_start_time = time.time()
|
||||
classification = self._classify_intent(user_input, conversation_context, chat_history, previous_slots)
|
||||
classify_end_time = time.time()
|
||||
classify_time = classify_end_time - classify_start_time
|
||||
# logging.info(f"重新分类耗时: {classify_time:.2f}秒")
|
||||
|
||||
fill_start_time = time.time()
|
||||
fill_slots = self._fill_slots(user_input, classification, conversation_context, chat_history, previous_slots)
|
||||
fill_end_time = time.time()
|
||||
fill_time = fill_end_time - fill_start_time
|
||||
all_time=fill_end_time-llm_start_time
|
||||
logging.info(f"总耗时：{all_time:.2f}秒,首次槽位+分类:{llm_time:.2f}秒, 重新分类耗时: {classify_time:.2f}秒, 重新槽位填充耗时: {fill_time:.2f}秒")
|
||||
|
||||
result = {
|
||||
"classification": classification.model_dump(),
|
||||
"slot_filling": fill_slots
|
||||
@@ -562,13 +615,21 @@ class IntentRecognizer:
|
||||
return result
|
||||
elif expected_slot_model.__name__ != type(slot_filling).__name__:
|
||||
# 添加容错处理，应对LLM槽位与分类不匹配。重新填充槽位
|
||||
fill_start_time = time.time()
|
||||
slot_filling = self._fill_slots(user_input, classification)
|
||||
fill_end_time = time.time()
|
||||
fill_time = fill_end_time - fill_start_time
|
||||
all_time=fill_end_time-llm_start_time
|
||||
logging.info(f"总耗时：{all_time:.2f}秒,首次槽位+分类:{llm_time:.2f}秒, 重新槽位填充耗时: {fill_time:.2f}秒")
|
||||
|
||||
result = {
|
||||
"classification": classification.model_dump(),
|
||||
"slot_filling": slot_filling
|
||||
}
|
||||
logging.warning(f"重新填充槽点")
|
||||
return result
|
||||
|
||||
logging.info(f"意图识别+槽位LLM调用耗时: {llm_time:.2f}秒")
|
||||
|
||||
# 构建最终结果
|
||||
result = {
|
||||
|
||||
@@ -126,7 +126,7 @@ query_rewrite_prompt_pro_old="""
|
||||
|
||||
query_rewrite_prompt_pro="""
|
||||
# 电力造价问答优化工程师（精简版）
|
||||
**角色**：基于历史对话和专业术语库重构问题，提升知识库检索准确率。
|
||||
**角色**：基于历史对话和术语库重构问题，提升知识库检索准确率。
|
||||
|
||||
## 核心原则
|
||||
1. 语义保真 → 保持问题核心意图
|
||||
@@ -135,8 +135,14 @@ query_rewrite_prompt_pro="""
|
||||
|
||||
## 处理流程
|
||||
### 一、输入解析
|
||||
- 原始问题(需保留核心语义)：{query}
|
||||
- 关键词集合：{keywords}
|
||||
- 原始问题(需保留核心语义):
|
||||
<query>
|
||||
{query}
|
||||
</query>
|
||||
- 术语库集合：
|
||||
<keywords>
|
||||
{keywords}
|
||||
</keywords>
|
||||
- 历史对话记录：
|
||||
<history>
|
||||
{chat_history}
|
||||
@@ -159,14 +165,14 @@ graph TD
|
||||
|
||||
### 三、重构优先级
|
||||
1. **背景补充**
|
||||
- 历史对话中确定的背景信息需要保留（例："这软件"→"【配网工程D3】"）
|
||||
- 历史对话中确定的背景信息需要保留（例："这软件"→"【配网工程计价通D3软件】"）
|
||||
|
||||
2. **术语处理**
|
||||
- 同义词转标准词 → 批量设置定额
|
||||
- 同义词转标准词 → 将提问中的同义词(synonymous)替换为标准词(name)
|
||||
- 存在即标记 → 【计算式】
|
||||
|
||||
3. **结构优化**
|
||||
- 保持原问题的5W2H特征
|
||||
- 保持原问题的5W2H特征,确保问题意图不发生改变。
|
||||
- 明确指代关系（"该功能"→"【批量导入】功能"）
|
||||
|
||||
## 输出规范
|
||||
@@ -184,7 +190,7 @@ graph TD
|
||||
- [] 背景信息是否合理补充？
|
||||
- [] 术语标记是否完整【】？
|
||||
- [] 语句是否自然流畅？
|
||||
- [] 避免过度补充无关信息
|
||||
- [] 避免补充无关信息
|
||||
"""
|
||||
|
||||
|
||||
@@ -349,7 +355,7 @@ def generate_slot_mapping_doc() -> str:
|
||||
doc.append(f"- {sub_class} -> {slot_model}")
|
||||
|
||||
doc.append("\n## 【注意事项】")
|
||||
doc.append("1. 分类与槽位模型必须严格对应")
|
||||
doc.append("1. 分类与槽位模型必须严格对应。严格遵守，不得违背")
|
||||
doc.append("2. 每个分类只能使用其对应的槽位模型")
|
||||
doc.append("3. 不允许混用不同分类的槽位模型")
|
||||
|
||||
|
||||
@@ -58,6 +58,12 @@ classification_prompt="""
|
||||
用户正在使用电力造价软件或想询问电力造价领域相关知识，你需要根据用户的输入内容，将其归类为以下垂直领域之一：
|
||||
{classification_info}
|
||||
|
||||
## 【会话背景信息】
|
||||
{conversation_context}
|
||||
|
||||
## 【历史对话记录】
|
||||
{chat_history}
|
||||
|
||||
【用户输入】：
|
||||
{user_input}
|
||||
|
||||
@@ -154,6 +160,15 @@ slot_filling_prompt = """
|
||||
【用户问题】
|
||||
{query}
|
||||
|
||||
## 【会话背景信息】
|
||||
{conversation_context}
|
||||
|
||||
## 【历史对话记录】
|
||||
{chat_history}
|
||||
|
||||
## 【历史槽位信息】
|
||||
{previous_slots}
|
||||
|
||||
【问题分类】
|
||||
垂直领域分类: {vertical_classification}
|
||||
子分类: {sub_classification}
|
||||
|
||||
@@ -23,16 +23,6 @@ API_KEY_LIST=[
|
||||
"sk-kzhxlqvqcxlnbdgnpalqnzumkmspepkttkgbophnkqanainw",
|
||||
"sk-bzttugqtlskrvguvhckwamdssvgmgnrqpsialpdbskfsyyak",
|
||||
"sk-tovmogiablsoeabwgqyvevpcfichyjpuzqdymmvksspdrtqt",
|
||||
"sk-wqdpapdkisovziexgcyxvumpwzbjnhqbxvcqcspzctjhyhjk",
|
||||
"sk-bbntrnifrtdzhhgrtlrhvwbnaysuszviemshdakxonnnymnb",
|
||||
"sk-vmpnwjxersrwybmfhfxgsvbmhsmpjldxseiyxovnysrlbuzi",
|
||||
"sk-nscsxwfqigkfpfqfzebkmaickxjzbhtfwywdppmmobrrbfnw",
|
||||
"sk-irbxuakhntsrusrympiubkkjbkabbfbdgpstqnxbztzdtxdq",
|
||||
"sk-hcfojzczbgwgcuhzxkicxqrhadurtakwbawiesyxyvksmcoz",
|
||||
"sk-wiyosqgyutjypgzibveiwkgqwfkfsnonrmvjfbvrbkoicciv",
|
||||
"sk-ocglenyvxkkvzupzumoypnyndjpjqhivyqpedusunboglspz",
|
||||
"sk-dtbawdwajkhdctrukundbkqwswzfzihqbebfuvqnfnounbuc",
|
||||
"sk-zqiyiqtbwqgyeenkvppymfbkspriolwbnxnjakugzxyvcuql",
|
||||
"sk-wtnjpejveiobtvzsmnuaefqkocsafbfyrtqkkyqardndtxcs",
|
||||
"sk-gqdvtrwvzxewnagwsfakrvajtzwgcknatpflkesyqhzjrlal",
|
||||
"sk-plivglrkxahodgtgjlaqdjusdoerxspjbcbizaybicarfyuk",
|
||||
@@ -96,6 +86,26 @@ API_KEY_LIST=[
|
||||
"sk-jrdzerhmvrtvzawkksowbgkggkubwfquplmrxbdhespqgtis",
|
||||
"sk-jjbpnkbeupsxyclcivbhizcfpfjrppddunbqynyjkqhtmpwu",
|
||||
"sk-oqehupcveovkjqqtxypqyifidcdissuyehwrkdwgruoyjkpq",
|
||||
"sk-jnnmltwtqwuoyagoogzzeraczmyfxhoairiddgayksqdfnbr",
|
||||
"sk-eghuepxnbcollzrjwbzqvbnhiiwagkejaclyhvaodeqgwrog",
|
||||
"sk-poszkbjdmamimconjustnrxxqusuzlryxkrzkpronlenrmen",
|
||||
"sk-zolvcegarsrwqhwgvwzgtqupodsdmckjiocyvoyldbkusbzc",
|
||||
"sk-ywfafulcniaqdgdcsnbtqquaqeuiqlkcnknkaflwxyuemcow",
|
||||
"sk-hhedmocgtfpywbbpwamgfkygrahiqsuurntlbqqbmjwfipmm",
|
||||
"sk-gzdqfoyvulrqscdpjlwlufdecrsyjpmwpkknuhnjsvtyftox",
|
||||
"sk-bkcufidsebujopqqwexwxwpmevrpelmvxzdymncvllcyojce",
|
||||
"sk-olabhscekudzkyudypkcjvehwqunagubwdmtppugrjmcptwv",
|
||||
"sk-zpdqyocliebhqpkuwvebpgcnfjdkvavdltimllmgkthwnwph",
|
||||
"sk-gvhchlfelocjniuydusyhhwacnomxnvucjonzkhtqoplnbcr",
|
||||
"sk-lzneagvdxhisodndnxnpkntghpkimjmjsebiqdzaoqzuhbla",
|
||||
"sk-xotcfdkigykevngedupitbcatjqppxmcibjtcebyoglykuxz",
|
||||
"sk-ufydqsdqnwsegaqwtappzwdyzqnoblyunfvslomnnmykedgk",
|
||||
"sk-jwasykftbkyjzdqlwcxuicrwzxsbhttilxfefbrozrznpwlv",
|
||||
"sk-xngteojwkxmftyaabjdwwgyoadspsowmcpcqobteutdcfmnr",
|
||||
"sk-akzkgniebruqrtuqskvlibkpcxjuazhcatysptkfyqivldfn",
|
||||
"sk-vpqkxtmcgkggllexchzysuewyfaoexzasoumxngdplzgwksw",
|
||||
"sk-fvcsqdbqmdlwxzjyofrilusqcypbfyczogaqwqrjrwvojmer",
|
||||
"sk-htjprscvfgskjtjzpxxxjhyymshagogykpawxekrrfbgftyx",
|
||||
]
|
||||
|
||||
class APIKeyManager:
|
||||
|
||||
+13
-48
@@ -100,10 +100,10 @@ class XinferenceReRankerModel:
|
||||
Returns:
|
||||
List[dict]: 重排序后的文档列表，每个元素包含document内容、相关性分数和原始索引
|
||||
"""
|
||||
url = "http://10.1.16.39:9995/v1/rerank"
|
||||
url = "http://172.20.0.145:9995/v1/rerank"
|
||||
|
||||
|
||||
params = {"documents": documents, "query": query, "top_n": top_k, "return_documents": True, "model": os.getenv("RERANKER_MODEL_NAME")}
|
||||
params = {"documents": documents, "query": query, "top_n": top_k, "return_documents": True, "model": "bge-reranker-v2-m3"}
|
||||
headers = {
|
||||
"Authorization": "Bearer <token>", # 这里需要替换为实际的token
|
||||
"Content-Type": "application/json"
|
||||
@@ -140,8 +140,7 @@ class OpenAiLLM:
|
||||
|
||||
def invoke(self, user_prompt="你是谁？", need_retry=True):
|
||||
# 初始化 OpenAI 客户端
|
||||
api_key = APIKeyManager.get_api_key()
|
||||
client = OpenAI(api_key=api_key, base_url=self._url)
|
||||
|
||||
|
||||
max_retries = 3
|
||||
retry_count = 0
|
||||
@@ -149,6 +148,8 @@ class OpenAiLLM:
|
||||
if need_retry:
|
||||
while retry_count < max_retries:
|
||||
try:
|
||||
api_key = APIKeyManager.get_api_key()
|
||||
client = OpenAI(api_key=api_key, base_url=self._url)
|
||||
# 创建 Completion 请求. 超时120s
|
||||
completion = client.chat.completions.create(
|
||||
model=self._model,
|
||||
@@ -162,11 +163,13 @@ class OpenAiLLM:
|
||||
retry_count += 1
|
||||
if retry_count == max_retries:
|
||||
logging.error(f"LLM 重试{max_retries}次后仍然失败: {e}")
|
||||
return ""
|
||||
raise e
|
||||
else:
|
||||
time.sleep(5*retry_count) # 重试前等待1秒
|
||||
else:
|
||||
# 创建 Completion 请求. 超时120s
|
||||
api_key = APIKeyManager.get_api_key()
|
||||
client = OpenAI(api_key=api_key, base_url=self._url)
|
||||
completion = client.chat.completions.create(
|
||||
model=self._model,
|
||||
messages=[{'role': 'user', 'content': user_prompt}],
|
||||
@@ -180,53 +183,15 @@ if __name__ == "__main__":
|
||||
reranker = SiliconFlowReRankerModel()
|
||||
|
||||
# 测试用例1：简单问题
|
||||
query = "他想做什么"
|
||||
documents = ["她想去公园跑步", "她想换一个新手机", "明天她想出去旅游"]
|
||||
query = "如何通过【电力经济评价软件】的【打开】功能加载工程文件?"
|
||||
documents = ["\n# (电力建设计价通软件) （概预算工程）工程备份管理\n## 操作步骤\n**方法一：** \n\n1、查找工程：输入工程文件名称的关键字，点击“查找”按钮，可以快速定位需查找的工程；\n\n\n\n2、根据时间点找备份工程：选中对应工程文件，在右侧选中“备份时间”的备份记录，点击“还原工程”或者“另存为工程”；\n\n **还原工程：** 将工程还原保存在原路径下；\n\n **另存为工程：** 另存为一个新工程，可选择保存路径，保存后，可点击文件——打开，浏览到另存的新工程打开。\n\n注：不确定备份是否是需要时，优先建议另存为工程。\n\n\n\n **方法二：** \n\n1、点击桌面软件快捷图标，右键属性—打开文件位置，直接定位软件安装根目录。 \n\n\n\n2、在软件安装根目录，点击“数据备份”文件夹，进入到文件夹内，根据修改日期找到对应工程，右键复制粘贴至桌面。\n\n\n\n\n\n3、定位桌面复制粘贴出来的数据工程，右键\"重命名\"，将bak修改成相应的文件后缀（概预算工程及施工图预算工程后缀为zwzj，招标工程及投标工程后缀为zwqd），然后点击“确定”，再通过软件的“文件”——“打开”按钮去浏览工程打开。\n",
|
||||
"\n# （配网计价通D3）插件管理/全国版和专版切换\n## 使用场景\n1.打开软件提示“当前工程文件为全国版文件，请使用全国版软件打开！”，该如何打开这个工程呢？\n\n\n\n2.打开软件提示“当前工程文件为辽宁版文件，请确认是否要在全国版软件中打开？”，这是什么意思？点击“确定”又可以打开工程？\n\n\n## 知识原理\n\n## 费用去向\n\n",
|
||||
"\n(电力建设计价通软件) 云造价--停用\n# 工程文件管理\n\n## 【主页】中点击“云端工程管理”，进入博微服务大厅；\n\n## 工程文件管理界面中显示云端备份的工程列表，可支持\n\n## 高级设置：可对历史版本数量进行设置，默认数量为10，可设置（5-15）；\n\n## 历史版本：勾选单个工程，点击“历史版本”可查看该工程保存的不同时间节点的历史工程；\n\n## 在线查阅：可查看工程数据，仅为只读模式不支持任何编辑；\n\n## 下载；选择需要的工程点击“下载”，可下载软件版本工程；\n\n",
|
||||
"\n（配网D3软件）打开工程\n\n# （配网D3软件）打开工程\n\n## 功能入口\n各界面点击“文件”按钮——“打开”按钮 \n\n\n## 操作步骤\n**打开工程：** \n\n点击“打开”按钮，浏览到工程存放位置，选中工程文件，点击“打开”即可。"]
|
||||
results = reranker.rerank(query, documents)
|
||||
print(f"测试用例1 - 查询:{query}")
|
||||
for idx, item in enumerate(results):
|
||||
print(f"{idx+1}. 文档: {item['document']}, 分数: {item['score']}")
|
||||
print("-" * 50)
|
||||
|
||||
# 测试用例2：技术问题
|
||||
query = "Python如何处理JSON数据"
|
||||
documents = [
|
||||
"Python中可以使用json模块来处理JSON数据，例如json.loads()将JSON字符串转换为字典",
|
||||
"Java提供了多种处理JSON的库，比如Jackson和Gson",
|
||||
"在Python中，可以使用pandas库来分析CSV数据",
|
||||
"JavaScript可以使用JSON.parse()方法解析JSON字符串"
|
||||
]
|
||||
results = reranker.rerank(query, documents)
|
||||
print(f"测试用例2 - 查询:{query}")
|
||||
for idx, item in enumerate(results):
|
||||
print(f"{idx+1}. 文档: {item['document']}, 分数: {item['score']}")
|
||||
print("-" * 50)
|
||||
|
||||
# 测试用例3：医疗问题
|
||||
query = "高血压的症状有哪些"
|
||||
documents = [
|
||||
"高血压的常见症状包括头痛、头晕、耳鸣和视力模糊",
|
||||
"糖尿病的症状包括多饮、多尿和体重减轻",
|
||||
"心脏病的症状通常包括胸痛、呼吸急促和疲劳",
|
||||
"高血压患者应该定期监测血压，保持健康的生活方式"
|
||||
]
|
||||
results = reranker.rerank(query, documents)
|
||||
print(f"测试用例3 - 查询:{query}")
|
||||
for idx, item in enumerate(results):
|
||||
print(f"{idx+1}. 文档: {item['document']}, 分数: {item['score']}")
|
||||
print("-" * 50)
|
||||
|
||||
# 测试用例4：长文本查询和文档
|
||||
query = "人工智能在医疗领域的应用及其伦理问题"
|
||||
documents = [
|
||||
"人工智能在医疗诊断中的应用已经显示出良好的效果，例如通过分析医学影像来检测疾病。然而，这也引发了关于医生角色和责任的伦理问题。",
|
||||
"在教育领域，人工智能可以提供个性化学习体验，适应不同学生的学习进度和风格。",
|
||||
"医疗伦理问题主要包括患者隐私保护、知情同意和医疗资源分配等方面。",
|
||||
"人工智能技术在金融领域的应用主要集中在风险评估、欺诈检测和算法交易等方面。"
|
||||
]
|
||||
results = reranker.rerank(query, documents)
|
||||
print(f"测试用例4 - 查询:{query}")
|
||||
for idx, item in enumerate(results):
|
||||
print(f"{idx+1}. 文档: {item['document']}, 分数: {item['score']}")
|
||||
print("-" * 50)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user