Files
assist/src/dialogue/conversation_history.py

378 lines
14 KiB
Python
Raw Normal View History

# -*- coding: utf-8 -*-
"""
对话历史管理器
支持Redis缓存和数据库持久化
"""
import json
import logging
from typing import Dict, List, Optional, Any, Tuple
from datetime import datetime, timedelta
from sqlalchemy.orm import Session
from ..core.database import db_manager
from ..core.models import Conversation
from ..core.redis_manager import redis_manager
from ..config.config import Config
logger = logging.getLogger(__name__)
class ConversationHistoryManager:
"""对话历史管理器"""
def __init__(self):
self.max_history_length = 20 # 最大历史记录数
self.cache_ttl = 3600 * 24 # 缓存24小时
def _get_redis_client(self):
"""获取Redis客户端"""
return redis_manager.get_connection()
def _get_cache_key(self, user_id: str, work_order_id: Optional[int] = None) -> str:
"""生成缓存键"""
if work_order_id:
return f"conversation_history:work_order:{work_order_id}"
return f"conversation_history:user:{user_id}"
def save_conversation(
self,
user_id: str,
user_message: str,
assistant_response: str,
work_order_id: Optional[int] = None,
confidence_score: Optional[float] = None,
response_time: Optional[float] = None,
knowledge_used: Optional[List[int]] = None
) -> int:
"""保存对话记录到数据库和Redis"""
conversation_id = 0
try:
# 保存到数据库
with db_manager.get_session() as session:
conversation = Conversation(
work_order_id=work_order_id,
user_message=user_message,
assistant_response=assistant_response,
confidence_score=confidence_score,
response_time=response_time,
knowledge_used=json.dumps(knowledge_used or [], ensure_ascii=False),
timestamp=datetime.now()
)
session.add(conversation)
session.commit()
conversation_id = conversation.id
# 保存到Redis缓存
self._save_to_cache(
user_id=user_id,
work_order_id=work_order_id,
user_message=user_message,
assistant_response=assistant_response,
conversation_id=conversation_id,
confidence_score=confidence_score,
response_time=response_time
)
logger.info(f"对话记录保存成功: ID={conversation_id}")
return conversation_id
except Exception as e:
logger.error(f"保存对话记录失败: {e}")
return conversation_id
def _save_to_cache(
self,
user_id: str,
work_order_id: Optional[int],
user_message: str,
assistant_response: str,
conversation_id: int,
confidence_score: Optional[float] = None,
response_time: Optional[float] = None
):
"""保存对话到Redis缓存"""
redis_client = self._get_redis_client()
if not redis_client:
return
try:
cache_key = self._get_cache_key(user_id, work_order_id)
# 构建对话记录
conversation_record = {
"id": conversation_id,
"user_message": user_message,
"assistant_response": assistant_response,
"timestamp": datetime.now().isoformat(),
"confidence_score": confidence_score,
"response_time": response_time
}
# 添加到Redis列表
redis_client.lpush(cache_key, json.dumps(conversation_record, ensure_ascii=False))
# 限制列表长度
redis_client.ltrim(cache_key, 0, self.max_history_length - 1)
# 设置过期时间
redis_client.expire(cache_key, self.cache_ttl)
except Exception as e:
logger.error(f"保存到Redis缓存失败: {e}")
def get_conversation_history(
self,
user_id: str,
work_order_id: Optional[int] = None,
limit: int = 10,
offset: int = 0
) -> List[Dict[str, Any]]:
"""获取对话历史优先从Redis获取"""
try:
# 先尝试从Redis获取
if redis_client:
cached_history = self._get_from_cache(user_id, work_order_id, limit, offset)
if cached_history:
return cached_history
# 从数据库获取
return self._get_from_database(user_id, work_order_id, limit, offset)
except Exception as e:
logger.error(f"获取对话历史失败: {e}")
return []
def _get_from_cache(
self,
user_id: str,
work_order_id: Optional[int],
limit: int,
offset: int
) -> List[Dict[str, Any]]:
"""从Redis缓存获取对话历史"""
redis_client = self._get_redis_client()
if not redis_client:
return []
try:
cache_key = self._get_cache_key(user_id, work_order_id)
# 获取指定范围的记录
start = offset
end = offset + limit - 1
cached_data = redis_client.lrange(cache_key, start, end)
history = []
for data in cached_data:
try:
record = json.loads(data)
history.append(record)
except json.JSONDecodeError:
continue
return history
except Exception as e:
logger.error(f"从Redis获取对话历史失败: {e}")
return []
def _get_from_database(
self,
user_id: str,
work_order_id: Optional[int],
limit: int,
offset: int
) -> List[Dict[str, Any]]:
"""从数据库获取对话历史"""
try:
with db_manager.get_session() as session:
query = session.query(Conversation)
if work_order_id:
query = query.filter(Conversation.work_order_id == work_order_id)
conversations = query.order_by(Conversation.timestamp.desc()).offset(offset).limit(limit).all()
history = []
for conv in conversations:
history.append({
"id": conv.id,
"user_message": conv.user_message,
"assistant_response": conv.assistant_response,
"timestamp": conv.timestamp.isoformat(),
"confidence_score": conv.confidence_score,
"response_time": conv.response_time,
"knowledge_used": json.loads(conv.knowledge_used) if conv.knowledge_used else []
})
return history
except Exception as e:
logger.error(f"从数据库获取对话历史失败: {e}")
return []
def get_conversation_context(
self,
user_id: str,
work_order_id: Optional[int] = None,
context_length: int = 6
) -> str:
"""获取对话上下文用于LLM"""
try:
history = self.get_conversation_history(user_id, work_order_id, context_length)
if not history:
return ""
context_parts = []
for record in reversed(history): # 按时间正序
context_parts.append(f"用户: {record['user_message']}")
context_parts.append(f"助手: {record['assistant_response']}")
return "\n".join(context_parts)
except Exception as e:
logger.error(f"获取对话上下文失败: {e}")
return ""
def delete_conversation(self, conversation_id: int) -> bool:
"""删除对话记录"""
try:
with db_manager.get_session() as session:
conversation = session.query(Conversation).filter(
Conversation.id == conversation_id
).first()
if not conversation:
return False
# 从数据库删除
session.delete(conversation)
session.commit()
# 从Redis缓存删除需要重建缓存
self._invalidate_cache(conversation.work_order_id)
logger.info(f"对话记录删除成功: ID={conversation_id}")
return True
except Exception as e:
logger.error(f"删除对话记录失败: {e}")
return False
def delete_user_conversations(self, user_id: str, work_order_id: Optional[int] = None) -> int:
"""删除用户的所有对话记录"""
try:
with db_manager.get_session() as session:
query = session.query(Conversation)
if work_order_id:
query = query.filter(Conversation.work_order_id == work_order_id)
conversations = query.all()
count = len(conversations)
# 删除数据库记录
for conv in conversations:
session.delete(conv)
session.commit()
# 清除Redis缓存
self._invalidate_cache(work_order_id)
logger.info(f"删除用户对话记录成功: 用户={user_id}, 数量={count}")
return count
except Exception as e:
logger.error(f"删除用户对话记录失败: {e}")
return 0
def _invalidate_cache(self, work_order_id: Optional[int] = None):
"""清除相关缓存"""
redis_client = self._get_redis_client()
if not redis_client:
return
try:
# 清除工单相关缓存
if work_order_id:
cache_key = f"conversation_history:work_order:{work_order_id}"
redis_client.delete(cache_key)
# 清除所有用户缓存(简单粗暴的方式)
pattern = "conversation_history:user:*"
keys = redis_client.keys(pattern)
if keys:
redis_client.delete(*keys)
except Exception as e:
logger.error(f"清除缓存失败: {e}")
def get_conversation_stats(self, user_id: str, work_order_id: Optional[int] = None) -> Dict[str, Any]:
"""获取对话统计信息"""
try:
with db_manager.get_session() as session:
query = session.query(Conversation)
if work_order_id:
query = query.filter(Conversation.work_order_id == work_order_id)
total_count = query.count()
# 计算平均响应时间
conversations_with_time = query.filter(Conversation.response_time.isnot(None)).all()
avg_response_time = 0
if conversations_with_time:
total_time = sum(conv.response_time for conv in conversations_with_time)
avg_response_time = total_time / len(conversations_with_time)
# 计算平均置信度
conversations_with_confidence = query.filter(Conversation.confidence_score.isnot(None)).all()
avg_confidence = 0
if conversations_with_confidence:
total_confidence = sum(conv.confidence_score for conv in conversations_with_confidence)
avg_confidence = total_confidence / len(conversations_with_confidence)
return {
"total_conversations": total_count,
"avg_response_time": round(avg_response_time, 2),
"avg_confidence": round(avg_confidence, 2),
"cache_status": "connected" if redis_manager.test_connection() else "disconnected"
}
except Exception as e:
logger.error(f"获取对话统计失败: {e}")
return {
"total_conversations": 0,
"avg_response_time": 0,
"avg_confidence": 0,
"cache_status": "error"
}
def cleanup_old_conversations(self, days: int = 30) -> int:
"""清理旧对话记录"""
try:
cutoff_date = datetime.now() - timedelta(days=days)
with db_manager.get_session() as session:
old_conversations = session.query(Conversation).filter(
Conversation.timestamp < cutoff_date
).all()
count = len(old_conversations)
for conv in old_conversations:
session.delete(conv)
session.commit()
logger.info(f"清理旧对话记录成功: 数量={count}")
return count
except Exception as e:
logger.error(f"清理旧对话记录失败: {e}")
return 0