mirror of https://github.com/Mai-with-u/MaiBot.git
🤖 自动格式化代码 [skip ci]
github-actions[bot]
parent
facc4bbef0
commit
37822fb347
|
|
@ -1,12 +1,12 @@
|
|||
import traceback
|
||||
import re
|
||||
from typing import Any
|
||||
from datetime import datetime # 确保导入 datetime
|
||||
from maim_message import UserInfo # UserInfo 来自 maim_message 包 # 从 maim_message 导入 MessageRecv
|
||||
from src.plugins.chat.message import MessageRecv # MessageRecv 来自message.py
|
||||
from datetime import datetime # 确保导入 datetime
|
||||
from maim_message import UserInfo # UserInfo 来自 maim_message 包 # 从 maim_message 导入 MessageRecv
|
||||
from src.plugins.chat.message import MessageRecv # MessageRecv 来自message.py
|
||||
from src.config.config import global_config
|
||||
from src.common.logger_manager import get_logger
|
||||
from ..chat.chat_stream import ChatStream, chat_manager
|
||||
from ..chat.chat_stream import ChatStream, chat_manager
|
||||
from src.plugins.chat.utils import get_embedding
|
||||
from src.common.database import db
|
||||
from .pfc_manager import PFCManager
|
||||
|
|
@ -14,18 +14,22 @@ from .pfc_manager import PFCManager
|
|||
logger = get_logger("pfc_processor")
|
||||
|
||||
|
||||
async def _handle_error(error: Exception, context: str, message: MessageRecv | None = None) -> None: # 明确 message 类型
|
||||
async def _handle_error(
|
||||
error: Exception, context: str, message: MessageRecv | None = None
|
||||
) -> None: # 明确 message 类型
|
||||
"""统一的错误处理函数
|
||||
# ... (方法注释不变) ...
|
||||
"""
|
||||
logger.error(f"{context}: {error}")
|
||||
logger.error(traceback.format_exc())
|
||||
# 检查 message 是否 None 以及是否有 raw_message 属性
|
||||
if message and hasattr(message, 'message_info') and hasattr(message.message_info, 'raw_message'): # MessageRecv 结构可能没有直接的 raw_message
|
||||
raw_msg_content = getattr(message.message_info, 'raw_message', None) # 安全获取
|
||||
if (
|
||||
message and hasattr(message, "message_info") and hasattr(message.message_info, "raw_message")
|
||||
): # MessageRecv 结构可能没有直接的 raw_message
|
||||
raw_msg_content = getattr(message.message_info, "raw_message", None) # 安全获取
|
||||
if raw_msg_content:
|
||||
logger.error(f"相关消息原始内容: {raw_msg_content}")
|
||||
elif message and hasattr(message, 'raw_message'): # 如果 MessageRecv 直接有 raw_message
|
||||
elif message and hasattr(message, "raw_message"): # 如果 MessageRecv 直接有 raw_message
|
||||
logger.error(f"相关消息原始内容: {message.raw_message}")
|
||||
|
||||
|
||||
|
|
@ -35,21 +39,22 @@ class PFCProcessor:
|
|||
# MessageStorage() 的实例化位置和具体类是什么？
|
||||
# 我们假设它来自 src.plugins.storage.storage
|
||||
# 但由于我们不能修改那个文件，所以这里的 self.storage 将按原样使用
|
||||
from src.plugins.storage.storage import MessageStorage # 明确导入，以便类型提示和理解
|
||||
from src.plugins.storage.storage import MessageStorage # 明确导入，以便类型提示和理解
|
||||
|
||||
self.storage: MessageStorage = MessageStorage()
|
||||
self.pfc_manager = PFCManager.get_instance()
|
||||
|
||||
async def process_message(self, message_data: dict[str, Any]) -> None: # 使用 dict[str, Any] 替代 Dict
|
||||
async def process_message(self, message_data: dict[str, Any]) -> None: # 使用 dict[str, Any] 替代 Dict
|
||||
"""处理接收到的原始消息数据
|
||||
# ... (方法注释不变) ...
|
||||
"""
|
||||
message_obj: MessageRecv | None = None # 初始化为 None，并明确类型
|
||||
message_obj: MessageRecv | None = None # 初始化为 None，并明确类型
|
||||
try:
|
||||
# 1. 消息解析与初始化
|
||||
message_obj = MessageRecv(message_data) # 使用你提供的 message.py 中的 MessageRecv
|
||||
message_obj = MessageRecv(message_data) # 使用你提供的 message.py 中的 MessageRecv
|
||||
|
||||
groupinfo = getattr(message_obj.message_info, 'group_info', None)
|
||||
userinfo = getattr(message_obj.message_info, 'user_info', None)
|
||||
groupinfo = getattr(message_obj.message_info, "group_info", None)
|
||||
userinfo = getattr(message_obj.message_info, "user_info", None)
|
||||
|
||||
logger.trace(f"准备为{userinfo.user_id}创建/获取聊天流")
|
||||
chat = await chat_manager.get_or_create_stream(
|
||||
|
|
@ -57,12 +62,13 @@ class PFCProcessor:
|
|||
user_info=userinfo,
|
||||
group_info=groupinfo,
|
||||
)
|
||||
message_obj.update_chat_stream(chat) # message.py 中 MessageRecv 有此方法
|
||||
message_obj.update_chat_stream(chat) # message.py 中 MessageRecv 有此方法
|
||||
|
||||
# 2. 过滤检查
|
||||
await message_obj.process() # 调用 MessageRecv 的异步 process 方法
|
||||
if self._check_ban_words(message_obj.processed_plain_text, userinfo) or \
|
||||
self._check_ban_regex(message_obj.raw_message, userinfo): # MessageRecv 有 raw_message 属性
|
||||
await message_obj.process() # 调用 MessageRecv 的异步 process 方法
|
||||
if self._check_ban_words(message_obj.processed_plain_text, userinfo) or self._check_ban_regex(
|
||||
message_obj.raw_message, userinfo
|
||||
): # MessageRecv 有 raw_message 属性
|
||||
return
|
||||
|
||||
# 3. 消息存储 (保持原有调用)
|
||||
|
|
@ -71,7 +77,7 @@ class PFCProcessor:
|
|||
await self.storage.store_message(message_obj, chat)
|
||||
logger.trace(f"存储成功 (初步): {message_obj.processed_plain_text}")
|
||||
|
||||
await self._update_embedding_vector(message_obj) # 明确传递 message_obj
|
||||
await self._update_embedding_vector(message_obj) # 明确传递 message_obj
|
||||
|
||||
# 4. 创建 PFC 聊天流
|
||||
await self._create_pfc_chat(message_obj, chat)
|
||||
|
|
@ -81,43 +87,41 @@ class PFCProcessor:
|
|||
current_time_display = datetime.fromtimestamp(float(message_obj.message_info.time)).strftime("%H:%M:%S")
|
||||
|
||||
# 确保 userinfo.user_nickname 存在
|
||||
user_nickname_display = getattr(userinfo, 'user_nickname', '未知用户')
|
||||
user_nickname_display = getattr(userinfo, "user_nickname", "未知用户")
|
||||
|
||||
logger.info(
|
||||
f"[{current_time_display}][私聊]{user_nickname_display}: {message_obj.processed_plain_text}"
|
||||
)
|
||||
logger.info(f"[{current_time_display}][私聊]{user_nickname_display}: {message_obj.processed_plain_text}")
|
||||
|
||||
except Exception as e:
|
||||
await _handle_error(e, "消息处理失败", message_obj) # 传递 message_obj
|
||||
await _handle_error(e, "消息处理失败", message_obj) # 传递 message_obj
|
||||
|
||||
async def _create_pfc_chat(self, message: MessageRecv): # 明确 message 类型
|
||||
async def _create_pfc_chat(self, message: MessageRecv): # 明确 message 类型
|
||||
try:
|
||||
chat_id = str(message.chat_stream.stream_id)
|
||||
private_name = str(message.message_info.user_info.user_nickname) # 假设 UserInfo 有 user_nickname
|
||||
private_name = str(message.message_info.user_info.user_nickname) # 假设 UserInfo 有 user_nickname
|
||||
|
||||
if global_config.enable_pfc_chatting:
|
||||
await self.pfc_manager.get_or_create_conversation(chat_id, private_name)
|
||||
|
||||
except Exception as e:
|
||||
logger.error(f"创建PFC聊天失败: {e}", exc_info=True) # 添加 exc_info=True
|
||||
logger.error(f"创建PFC聊天失败: {e}", exc_info=True) # 添加 exc_info=True
|
||||
|
||||
@staticmethod
|
||||
def _check_ban_words(text: str, userinfo: UserInfo) -> bool: # 明确 userinfo 类型
|
||||
def _check_ban_words(text: str, userinfo: UserInfo) -> bool: # 明确 userinfo 类型
|
||||
"""检查消息中是否包含过滤词"""
|
||||
for word in global_config.ban_words:
|
||||
if word in text:
|
||||
logger.info(f"[私聊]{userinfo.user_nickname}:{text}") # 假设 UserInfo 有 user_nickname
|
||||
logger.info(f"[私聊]{userinfo.user_nickname}:{text}") # 假设 UserInfo 有 user_nickname
|
||||
logger.info(f"[过滤词识别]消息中含有{word}，filtered")
|
||||
return True
|
||||
return False
|
||||
|
||||
@staticmethod
|
||||
def _check_ban_regex(text: str, userinfo: UserInfo) -> bool: # 明确 userinfo 类型
|
||||
def _check_ban_regex(text: str, userinfo: UserInfo) -> bool: # 明确 userinfo 类型
|
||||
"""检查消息是否匹配过滤正则表达式"""
|
||||
for pattern in global_config.ban_msgs_regex:
|
||||
if pattern.search(text): # 假设 ban_msgs_regex 中的元素是已编译的正则对象
|
||||
logger.info(f"[私聊]{userinfo.user_nickname}:{text}") # _nickname
|
||||
logger.info(f"[正则表达式过滤]消息匹配到{pattern.pattern}，filtered") # .pattern 获取原始表达式字符串
|
||||
if pattern.search(text): # 假设 ban_msgs_regex 中的元素是已编译的正则对象
|
||||
logger.info(f"[私聊]{userinfo.user_nickname}:{text}") # _nickname
|
||||
logger.info(f"[正则表达式过滤]消息匹配到{pattern.pattern}，filtered") # .pattern 获取原始表达式字符串
|
||||
return True
|
||||
return False
|
||||
|
||||
|
|
@ -125,7 +129,7 @@ class PFCProcessor:
|
|||
"""更新消息的嵌入向量"""
|
||||
# === 新增：为已存储的消息生成嵌入并更新数据库文档 ===
|
||||
embedding_vector = None
|
||||
text_for_embedding = message_obj.processed_plain_text # 使用处理后的纯文本
|
||||
text_for_embedding = message_obj.processed_plain_text # 使用处理后的纯文本
|
||||
|
||||
# 在 storage.py 中，会对 processed_plain_text 进行一次过滤
|
||||
# 为了保持一致，我们也在这里应用相同的过滤逻辑
|
||||
|
|
@ -148,18 +152,25 @@ class PFCProcessor:
|
|||
# 确保你有权限访问和操作 db 对象
|
||||
update_result = db.messages.update_one(
|
||||
{"message_id": message_obj.message_info.message_id, "chat_id": chat.stream_id},
|
||||
{"$set": {"embedding_vector": embedding_vector}}
|
||||
{"$set": {"embedding_vector": embedding_vector}},
|
||||
)
|
||||
if update_result.modified_count > 0:
|
||||
logger.info(f"成功为消息 ID '{message_obj.message_info.message_id}' 更新嵌入向量到数据库。")
|
||||
elif update_result.matched_count > 0:
|
||||
logger.warning(f"消息 ID '{message_obj.message_info.message_id}' 已存在嵌入向量或未作修改。")
|
||||
else:
|
||||
logger.error(f"未能找到消息 ID '{message_obj.message_info.message_id}' (chat_id: {chat.stream_id}) 来更新嵌入向量。可能是存储和更新之间存在延迟或问题。")
|
||||
logger.error(
|
||||
f"未能找到消息 ID '{message_obj.message_info.message_id}' (chat_id: {chat.stream_id}) 来更新嵌入向量。可能是存储和更新之间存在延迟或问题。"
|
||||
)
|
||||
else:
|
||||
logger.warning(f"未能为消息 ID '{message_obj.message_info.message_id}' 的文本 '{filtered_text_for_embedding[:30]}...' 生成嵌入向量。")
|
||||
logger.warning(
|
||||
f"未能为消息 ID '{message_obj.message_info.message_id}' 的文本 '{filtered_text_for_embedding[:30]}...' 生成嵌入向量。"
|
||||
)
|
||||
except Exception as e_embed_update:
|
||||
logger.error(f"为消息 ID '{message_obj.message_info.message_id}' 生成嵌入或更新数据库时发生异常: {e_embed_update}", exc_info=True)
|
||||
logger.error(
|
||||
f"为消息 ID '{message_obj.message_info.message_id}' 生成嵌入或更新数据库时发生异常: {e_embed_update}",
|
||||
exc_info=True,
|
||||
)
|
||||
else:
|
||||
logger.debug(f"消息 ID '{message_obj.message_info.message_id}' 的过滤后纯文本为空，不生成或更新嵌入。")
|
||||
# === 新增结束 ===
|
||||
# === 新增结束 ===
|
||||
|
|
|
|||
|
|
@ -3,35 +3,36 @@ import json
|
|||
import re
|
||||
import time
|
||||
from datetime import datetime
|
||||
from typing import Dict, Any, Optional, Tuple, List, Union # 确保导入这些类型
|
||||
from typing import Dict, Any, Optional, Tuple, List, Union # 确保导入这些类型
|
||||
|
||||
from src.common.logger_manager import get_logger
|
||||
from src.config.config import global_config
|
||||
from src.common.database import db # << 确认此路径
|
||||
from src.common.database import db # << 确认此路径
|
||||
|
||||
# --- 依赖于你项目结构的导入，请务必仔细检查并根据你的实际情况调整 ---
|
||||
from src.plugins.memory_system.Hippocampus import HippocampusManager # << 确认此路径
|
||||
from src.plugins.heartFC_chat.heartflow_prompt_builder import prompt_builder # << 确认此路径
|
||||
from src.plugins.chat.utils import get_embedding # << 确认此路径
|
||||
from src.plugins.utils.chat_message_builder import build_readable_messages # << 确认此路径
|
||||
from src.plugins.memory_system.Hippocampus import HippocampusManager # << 确认此路径
|
||||
from src.plugins.heartFC_chat.heartflow_prompt_builder import prompt_builder # << 确认此路径
|
||||
from src.plugins.chat.utils import get_embedding # << 确认此路径
|
||||
from src.plugins.utils.chat_message_builder import build_readable_messages # << 确认此路径
|
||||
# --- 依赖导入结束 ---
|
||||
|
||||
from src.plugins.chat.chat_stream import ChatStream # 来自原始 pfc_utils.py
|
||||
from ..person_info.person_info import person_info_manager # 来自原始 pfc_utils.py (相对导入)
|
||||
import math # 来自原始 pfc_utils.py
|
||||
from .observation_info import ObservationInfo # 来自原始 pfc_utils.py (相对导入)
|
||||
from src.plugins.chat.chat_stream import ChatStream # 来自原始 pfc_utils.py
|
||||
from ..person_info.person_info import person_info_manager # 来自原始 pfc_utils.py (相对导入)
|
||||
import math # 来自原始 pfc_utils.py
|
||||
from .observation_info import ObservationInfo # 来自原始 pfc_utils.py (相对导入)
|
||||
|
||||
|
||||
logger = get_logger("pfc_utils")
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 新增：专门用于检索 PFC 私聊历史对话上下文的函数
|
||||
# ==============================================================================
|
||||
async def find_most_relevant_historical_message(
|
||||
chat_id: str,
|
||||
query_text: str,
|
||||
similarity_threshold: float = 0.3, # 相似度阈值，可以根据效果调整
|
||||
absolute_search_time_limit: Optional[float] = None # 新增参数：排除最近多少秒内的消息（例如5分钟）
|
||||
similarity_threshold: float = 0.3, # 相似度阈值，可以根据效果调整
|
||||
absolute_search_time_limit: Optional[float] = None, # 新增参数：排除最近多少秒内的消息（例如5分钟）
|
||||
) -> Optional[Dict[str, Any]]:
|
||||
"""
|
||||
根据查询文本，在指定 chat_id 的历史消息中查找最相关的消息。
|
||||
|
|
@ -51,20 +52,22 @@ async def find_most_relevant_historical_message(
|
|||
|
||||
effective_search_upper_limit: float
|
||||
log_source_of_limit: str = ""
|
||||
|
||||
|
||||
if absolute_search_time_limit is not None:
|
||||
effective_search_upper_limit = absolute_search_time_limit
|
||||
log_source_of_limit = "传入的绝对时间上限"
|
||||
else:
|
||||
# 如果没有传入绝对时间上限，可以设置一个默认的回退逻辑
|
||||
fallback_exclude_seconds = getattr(global_config, "pfc_historical_fallback_exclude_seconds", 7200) # 默认2小时
|
||||
fallback_exclude_seconds = getattr(global_config, "pfc_historical_fallback_exclude_seconds", 7200) # 默认2小时
|
||||
effective_search_upper_limit = time.time() - fallback_exclude_seconds
|
||||
log_source_of_limit = f"回退逻辑 (排除最近 {fallback_exclude_seconds} 秒)"
|
||||
|
||||
logger.debug(f"[{chat_id}] (私聊历史) find_most_relevant_historical_message: "
|
||||
f"将使用时间上限 {effective_search_upper_limit} "
|
||||
f"(可读: {datetime.fromtimestamp(effective_search_upper_limit).strftime('%Y-%m-%d %H:%M:%S')}) "
|
||||
f"进行历史消息锚点搜索。来源: {log_source_of_limit}")
|
||||
|
||||
logger.debug(
|
||||
f"[{chat_id}] (私聊历史) find_most_relevant_historical_message: "
|
||||
f"将使用时间上限 {effective_search_upper_limit} "
|
||||
f"(可读: {datetime.fromtimestamp(effective_search_upper_limit).strftime('%Y-%m-%d %H:%M:%S')}) "
|
||||
f"进行历史消息锚点搜索。来源: {log_source_of_limit}"
|
||||
)
|
||||
# --- [新代码结束] ---
|
||||
|
||||
pipeline = [
|
||||
|
|
@ -72,14 +75,46 @@ async def find_most_relevant_historical_message(
|
|||
"$match": {
|
||||
"chat_id": chat_id,
|
||||
"embedding_vector": {"$exists": True, "$ne": None, "$not": {"$size": 0}},
|
||||
"time": {"$lt": effective_search_upper_limit} # <--- 使用新的 effective_search_upper_limit
|
||||
"time": {"$lt": effective_search_upper_limit}, # <--- 使用新的 effective_search_upper_limit
|
||||
}
|
||||
},
|
||||
{
|
||||
"$addFields": {
|
||||
"dotProduct": {"$reduce": {"input": {"$range": [0, {"$size": "$embedding_vector"}]}, "initialValue": 0, "in": {"$add": ["$$value", {"$multiply": [{"$arrayElemAt": ["$embedding_vector", "$$this"]}, {"$arrayElemAt": [query_embedding, "$$this"]}]}]}}},
|
||||
"queryVecMagnitude": {"$sqrt": {"$reduce": {"input": query_embedding, "initialValue": 0, "in": {"$add": ["$$value", {"$multiply": ["$$this", "$$this"]}]}}}},
|
||||
"docVecMagnitude": {"$sqrt": {"$reduce": {"input": "$embedding_vector", "initialValue": 0, "in": {"$add": ["$$value", {"$multiply": ["$$this", "$$this"]}]}}}}
|
||||
"dotProduct": {
|
||||
"$reduce": {
|
||||
"input": {"$range": [0, {"$size": "$embedding_vector"}]},
|
||||
"initialValue": 0,
|
||||
"in": {
|
||||
"$add": [
|
||||
"$$value",
|
||||
{
|
||||
"$multiply": [
|
||||
{"$arrayElemAt": ["$embedding_vector", "$$this"]},
|
||||
{"$arrayElemAt": [query_embedding, "$$this"]},
|
||||
]
|
||||
},
|
||||
]
|
||||
},
|
||||
}
|
||||
},
|
||||
"queryVecMagnitude": {
|
||||
"$sqrt": {
|
||||
"$reduce": {
|
||||
"input": query_embedding,
|
||||
"initialValue": 0,
|
||||
"in": {"$add": ["$$value", {"$multiply": ["$$this", "$$this"]}]},
|
||||
}
|
||||
}
|
||||
},
|
||||
"docVecMagnitude": {
|
||||
"$sqrt": {
|
||||
"$reduce": {
|
||||
"input": "$embedding_vector",
|
||||
"initialValue": 0,
|
||||
"in": {"$add": ["$$value", {"$multiply": ["$$this", "$$this"]}]},
|
||||
}
|
||||
}
|
||||
},
|
||||
}
|
||||
},
|
||||
{
|
||||
|
|
@ -88,7 +123,7 @@ async def find_most_relevant_historical_message(
|
|||
"$cond": [
|
||||
{"$and": [{"$gt": ["$queryVecMagnitude", 0]}, {"$gt": ["$docVecMagnitude", 0]}]},
|
||||
{"$divide": ["$dotProduct", {"$multiply": ["$queryVecMagnitude", "$docVecMagnitude"]}]},
|
||||
0
|
||||
0,
|
||||
]
|
||||
}
|
||||
}
|
||||
|
|
@ -96,26 +131,44 @@ async def find_most_relevant_historical_message(
|
|||
{"$match": {"similarity": {"$gte": similarity_threshold}}},
|
||||
{"$sort": {"similarity": -1}},
|
||||
{"$limit": 1},
|
||||
{"$project": {"_id": 0, "message_id": 1, "time": 1, "chat_id": 1, "user_info": 1, "processed_plain_text": 1, "similarity": 1}} # 可以不返回 embedding_vector 节省带宽
|
||||
{
|
||||
"$project": {
|
||||
"_id": 0,
|
||||
"message_id": 1,
|
||||
"time": 1,
|
||||
"chat_id": 1,
|
||||
"user_info": 1,
|
||||
"processed_plain_text": 1,
|
||||
"similarity": 1,
|
||||
}
|
||||
}, # 可以不返回 embedding_vector 节省带宽
|
||||
]
|
||||
|
||||
try:
|
||||
# --- 确定性修改：同步执行聚合和结果转换 ---
|
||||
cursor = db.messages.aggregate(pipeline) # PyMongo 的 aggregate 返回一个 CommandCursor
|
||||
results = list(cursor) # 直接将 CommandCursor 转换为列表
|
||||
cursor = db.messages.aggregate(pipeline) # PyMongo 的 aggregate 返回一个 CommandCursor
|
||||
results = list(cursor) # 直接将 CommandCursor 转换为列表
|
||||
if not results:
|
||||
logger.info(f"[{chat_id}] (私聊历史) find_most_relevant_historical_message: 在时间点 {effective_search_upper_limit} 之前，未能找到任何与 '{query_text[:30]}...' 相关的历史消息。")
|
||||
logger.info(
|
||||
f"[{chat_id}] (私聊历史) find_most_relevant_historical_message: 在时间点 {effective_search_upper_limit} 之前，未能找到任何与 '{query_text[:30]}...' 相关的历史消息。"
|
||||
)
|
||||
else:
|
||||
logger.info(f"[{chat_id}] (私聊历史) find_most_relevant_historical_message: 在时间点 {effective_search_upper_limit} 之前，找到了 {len(results)} 条候选历史消息。最相关的一条是：")
|
||||
for res_msg in results:
|
||||
msg_time_readable = datetime.fromtimestamp(res_msg.get('time',0)).strftime('%Y-%m-%d %H:%M:%S')
|
||||
logger.info(f" - MsgID: {res_msg.get('message_id')}, Time: {msg_time_readable} (原始: {res_msg.get('time')}), Sim: {res_msg.get('similarity'):.4f}, Text: '{res_msg.get('processed_plain_text','')[:50]}...'")
|
||||
logger.info(
|
||||
f"[{chat_id}] (私聊历史) find_most_relevant_historical_message: 在时间点 {effective_search_upper_limit} 之前，找到了 {len(results)} 条候选历史消息。最相关的一条是："
|
||||
)
|
||||
for res_msg in results:
|
||||
msg_time_readable = datetime.fromtimestamp(res_msg.get("time", 0)).strftime("%Y-%m-%d %H:%M:%S")
|
||||
logger.info(
|
||||
f" - MsgID: {res_msg.get('message_id')}, Time: {msg_time_readable} (原始: {res_msg.get('time')}), Sim: {res_msg.get('similarity'):.4f}, Text: '{res_msg.get('processed_plain_text', '')[:50]}...'"
|
||||
)
|
||||
# --- [修改结束] ---
|
||||
|
||||
# --- 修改结束 ---
|
||||
if results and len(results) > 0:
|
||||
most_similar_message = results[0]
|
||||
logger.info(f"[{chat_id}] (私聊历史)找到最相关消息 ID: {most_similar_message.get('message_id')}, 相似度: {most_similar_message.get('similarity'):.4f}")
|
||||
logger.info(
|
||||
f"[{chat_id}] (私聊历史)找到最相关消息 ID: {most_similar_message.get('message_id')}, 相似度: {most_similar_message.get('similarity'):.4f}"
|
||||
)
|
||||
return most_similar_message
|
||||
else:
|
||||
logger.debug(f"[{chat_id}] (私聊历史)未找到相似度超过 {similarity_threshold} 的相关消息。")
|
||||
|
|
@ -124,13 +177,14 @@ async def find_most_relevant_historical_message(
|
|||
logger.error(f"[{chat_id}] (私聊历史)在数据库中检索时出错: {e}", exc_info=True)
|
||||
return None
|
||||
|
||||
|
||||
async def retrieve_chat_context_window(
|
||||
chat_id: str,
|
||||
anchor_message_id: str,
|
||||
anchor_message_time: float,
|
||||
excluded_time_threshold_for_window: float,
|
||||
window_size_before: int = 7,
|
||||
window_size_after: int = 7
|
||||
window_size_after: int = 7,
|
||||
) -> List[Dict[str, Any]]:
|
||||
"""
|
||||
以某条消息为锚点，获取其前后的聊天记录形成一个上下文窗口。
|
||||
|
|
@ -138,33 +192,50 @@ async def retrieve_chat_context_window(
|
|||
if not anchor_message_id or anchor_message_time is None:
|
||||
return []
|
||||
|
||||
context_messages: List[Dict[str, Any]] = [] # 明确类型
|
||||
logger.debug(f"[{chat_id}] (私聊历史)准备以消息 ID '{anchor_message_id}' (时间: {anchor_message_time}) 为锚点，获取上下文窗口...")
|
||||
context_messages: List[Dict[str, Any]] = [] # 明确类型
|
||||
logger.debug(
|
||||
f"[{chat_id}] (私聊历史)准备以消息 ID '{anchor_message_id}' (时间: {anchor_message_time}) 为锚点，获取上下文窗口..."
|
||||
)
|
||||
|
||||
try:
|
||||
# --- 同步执行 find_one 和 find ---
|
||||
anchor_message = db.messages.find_one({"message_id": anchor_message_id, "chat_id": chat_id})
|
||||
|
||||
messages_before_cursor = db.messages.find(
|
||||
{"chat_id": chat_id, "time": {"$lt": anchor_message_time}}
|
||||
).sort("time", -1).limit(window_size_before)
|
||||
messages_before_cursor = (
|
||||
db.messages.find({"chat_id": chat_id, "time": {"$lt": anchor_message_time}})
|
||||
.sort("time", -1)
|
||||
.limit(window_size_before)
|
||||
)
|
||||
messages_before = list(messages_before_cursor)
|
||||
messages_before.reverse()
|
||||
# --- 新增日志 ---
|
||||
logger.debug(f"[{chat_id}] (私聊历史) retrieve_chat_context_window: Anchor Time: {anchor_message_time}, Excluded Window End Time: {excluded_time_threshold_for_window}")
|
||||
logger.debug(f"[{chat_id}] (私聊历史) retrieve_chat_context_window: Messages BEFORE anchor ({len(messages_before)}):")
|
||||
logger.debug(
|
||||
f"[{chat_id}] (私聊历史) retrieve_chat_context_window: Anchor Time: {anchor_message_time}, Excluded Window End Time: {excluded_time_threshold_for_window}"
|
||||
)
|
||||
logger.debug(
|
||||
f"[{chat_id}] (私聊历史) retrieve_chat_context_window: Messages BEFORE anchor ({len(messages_before)}):"
|
||||
)
|
||||
for msg_b in messages_before:
|
||||
logger.debug(f" - Time: {datetime.fromtimestamp(msg_b.get('time',0)).strftime('%Y-%m-%d %H:%M:%S')}, Text: '{msg_b.get('processed_plain_text','')[:30]}...'")
|
||||
logger.debug(
|
||||
f" - Time: {datetime.fromtimestamp(msg_b.get('time', 0)).strftime('%Y-%m-%d %H:%M:%S')}, Text: '{msg_b.get('processed_plain_text', '')[:30]}...'"
|
||||
)
|
||||
|
||||
messages_after_cursor = db.messages.find(
|
||||
{"chat_id": chat_id, "time": {"$gt": anchor_message_time, "$lt": excluded_time_threshold_for_window}}
|
||||
).sort("time", 1).limit(window_size_after)
|
||||
messages_after_cursor = (
|
||||
db.messages.find(
|
||||
{"chat_id": chat_id, "time": {"$gt": anchor_message_time, "$lt": excluded_time_threshold_for_window}}
|
||||
)
|
||||
.sort("time", 1)
|
||||
.limit(window_size_after)
|
||||
)
|
||||
messages_after = list(messages_after_cursor)
|
||||
# --- 新增日志 ---
|
||||
logger.debug(f"[{chat_id}] (私聊历史) retrieve_chat_context_window: Messages AFTER anchor ({len(messages_after)}):")
|
||||
logger.debug(
|
||||
f"[{chat_id}] (私聊历史) retrieve_chat_context_window: Messages AFTER anchor ({len(messages_after)}):"
|
||||
)
|
||||
for msg_a in messages_after:
|
||||
logger.debug(f" - Time: {datetime.fromtimestamp(msg_a.get('time',0)).strftime('%Y-%m-%d %H:%M:%S')}, Text: '{msg_a.get('processed_plain_text','')[:30]}...'")
|
||||
|
||||
logger.debug(
|
||||
f" - Time: {datetime.fromtimestamp(msg_a.get('time', 0)).strftime('%Y-%m-%d %H:%M:%S')}, Text: '{msg_a.get('processed_plain_text', '')[:30]}...'"
|
||||
)
|
||||
|
||||
if messages_before:
|
||||
context_messages.extend(messages_before)
|
||||
|
|
@ -173,32 +244,35 @@ async def retrieve_chat_context_window(
|
|||
context_messages.append(anchor_message)
|
||||
if messages_after:
|
||||
context_messages.extend(messages_after)
|
||||
|
||||
final_window: List[Dict[str, Any]] = [] # 明确类型
|
||||
seen_ids: set[str] = set() # 明确类型
|
||||
|
||||
final_window: List[Dict[str, Any]] = [] # 明确类型
|
||||
seen_ids: set[str] = set() # 明确类型
|
||||
for msg in context_messages:
|
||||
msg_id = msg.get("message_id")
|
||||
if msg_id and msg_id not in seen_ids: # 确保 msg_id 存在
|
||||
if msg_id and msg_id not in seen_ids: # 确保 msg_id 存在
|
||||
final_window.append(msg)
|
||||
seen_ids.add(msg_id)
|
||||
|
||||
|
||||
final_window.sort(key=lambda m: m.get("time", 0))
|
||||
logger.info(f"[{chat_id}] (私聊历史)为锚点 '{anchor_message_id}' 构建了包含 {len(final_window)} 条消息的上下文窗口。")
|
||||
logger.info(
|
||||
f"[{chat_id}] (私聊历史)为锚点 '{anchor_message_id}' 构建了包含 {len(final_window)} 条消息的上下文窗口。"
|
||||
)
|
||||
return final_window
|
||||
except Exception as e:
|
||||
logger.error(f"[{chat_id}] (私聊历史)获取消息 ID '{anchor_message_id}' 的上下文窗口时出错: {e}", exc_info=True)
|
||||
return []
|
||||
|
||||
|
||||
# ==============================================================================
|
||||
# 修改后的 retrieve_contextual_info 函数
|
||||
# ==============================================================================
|
||||
async def retrieve_contextual_info(
|
||||
text: str, # 用于全局记忆和知识检索的主查询文本 (通常是短期聊天记录)
|
||||
private_name: str, # 用于日志
|
||||
chat_id: str, # 用于特定私聊历史的检索
|
||||
text: str, # 用于全局记忆和知识检索的主查询文本 (通常是短期聊天记录)
|
||||
private_name: str, # 用于日志
|
||||
chat_id: str, # 用于特定私聊历史的检索
|
||||
historical_chat_query_text: Optional[str] = None,
|
||||
current_short_term_history_earliest_time: Optional[float] = None # <--- 新增参数
|
||||
) -> Tuple[str, str, str]: # 返回: 全局记忆, 知识, 私聊历史回忆
|
||||
current_short_term_history_earliest_time: Optional[float] = None, # <--- 新增参数
|
||||
) -> Tuple[str, str, str]: # 返回: 全局记忆, 知识, 私聊历史回忆
|
||||
"""
|
||||
检索三种类型的上下文信息：全局压缩记忆、知识库知识、当前私聊的特定历史对话。
|
||||
|
||||
|
|
@ -222,9 +296,9 @@ async def retrieve_contextual_info(
|
|||
related_memory = await HippocampusManager.get_instance().get_memory_from_text(
|
||||
text=text,
|
||||
max_memory_num=2,
|
||||
max_memory_length=2,
|
||||
max_memory_length=2,
|
||||
max_depth=3,
|
||||
fast_retrieval=False,
|
||||
fast_retrieval=False,
|
||||
)
|
||||
if related_memory:
|
||||
temp_global_memory_info = ""
|
||||
|
|
@ -233,7 +307,7 @@ async def retrieve_contextual_info(
|
|||
temp_global_memory_info += str(memory_item[1]) + "\n"
|
||||
elif isinstance(memory_item, str):
|
||||
temp_global_memory_info += memory_item + "\n"
|
||||
|
||||
|
||||
if temp_global_memory_info.strip():
|
||||
retrieved_global_memory_str = f"你回忆起一些相关的全局记忆：\n{temp_global_memory_info.strip()}\n(以上是你的全局记忆，供参考)\n"
|
||||
global_memory_log_msg = f"自动检索到全局压缩记忆: {temp_global_memory_info.strip()[:100]}..."
|
||||
|
|
@ -250,7 +324,6 @@ async def retrieve_contextual_info(
|
|||
else:
|
||||
logger.debug(f"[私聊][{private_name}] (retrieve_contextual_info) 无有效主查询文本，跳过全局压缩记忆检索。")
|
||||
|
||||
|
||||
# --- 2. 相关知识检索 (来自 prompt_builder) ---
|
||||
# (保持你原始 pfc_utils.py 中这部分的逻辑基本不变)
|
||||
knowledge_log_msg = f"开始知识检索 (基于文本: '{text[:30]}...')"
|
||||
|
|
@ -260,8 +333,8 @@ async def retrieve_contextual_info(
|
|||
message=text,
|
||||
threshold=0.38,
|
||||
)
|
||||
if knowledge_result and knowledge_result.strip(): # 确保结果不为空
|
||||
retrieved_knowledge_str = knowledge_result # 直接使用返回结果，如果需要也可以包装
|
||||
if knowledge_result and knowledge_result.strip(): # 确保结果不为空
|
||||
retrieved_knowledge_str = knowledge_result # 直接使用返回结果，如果需要也可以包装
|
||||
knowledge_log_msg = f"自动检索到相关知识: {knowledge_result[:100]}..."
|
||||
else:
|
||||
knowledge_log_msg = "知识检索返回为空。"
|
||||
|
|
@ -274,9 +347,10 @@ async def retrieve_contextual_info(
|
|||
else:
|
||||
logger.debug(f"[私聊][{private_name}] (retrieve_contextual_info) 无有效主查询文本，跳过知识检索。")
|
||||
|
||||
|
||||
# --- 3. 当前私聊的特定历史对话上下文检索 ---
|
||||
query_for_historical_chat = historical_chat_query_text if historical_chat_query_text and historical_chat_query_text.strip() else None
|
||||
query_for_historical_chat = (
|
||||
historical_chat_query_text if historical_chat_query_text and historical_chat_query_text.strip() else None
|
||||
)
|
||||
# historical_chat_log_msg 的初始化可以移到 try 块之后，根据实际情况赋值
|
||||
|
||||
if query_for_historical_chat:
|
||||
|
|
@ -284,13 +358,15 @@ async def retrieve_contextual_info(
|
|||
# ---- [新代码] 计算最终的、严格的搜索时间上限 ----
|
||||
# 1. 设置一个基础的、较大的时间回溯窗口，例如2小时 (7200秒)
|
||||
# 这个值可以从全局配置读取，如果没配置则使用默认值
|
||||
default_search_exclude_seconds = getattr(global_config, "pfc_historical_search_default_exclude_seconds", 7200) # 默认2小时
|
||||
default_search_exclude_seconds = getattr(
|
||||
global_config, "pfc_historical_search_default_exclude_seconds", 7200
|
||||
) # 默认2小时
|
||||
base_excluded_time_limit = time.time() - default_search_exclude_seconds
|
||||
|
||||
|
||||
final_search_upper_limit_time = base_excluded_time_limit
|
||||
if current_short_term_history_earliest_time is not None:
|
||||
# 我们希望找到的消息严格早于 short_term_history 的开始，减去一个小量确保不包含边界
|
||||
limit_from_short_term = current_short_term_history_earliest_time - 0.001
|
||||
limit_from_short_term = current_short_term_history_earliest_time - 0.001
|
||||
final_search_upper_limit_time = min(base_excluded_time_limit, limit_from_short_term)
|
||||
log_earliest_time_str = "未提供"
|
||||
if current_short_term_history_earliest_time is not None:
|
||||
|
|
@ -298,55 +374,60 @@ async def retrieve_contextual_info(
|
|||
log_earliest_time_str = f"{current_short_term_history_earliest_time} (即 {datetime.fromtimestamp(current_short_term_history_earliest_time).strftime('%Y-%m-%d %H:%M:%S')})"
|
||||
except Exception:
|
||||
log_earliest_time_str = str(current_short_term_history_earliest_time)
|
||||
|
||||
logger.debug(f"[{private_name}] (私聊历史) retrieve_contextual_info: "
|
||||
f"最终用于历史搜索的时间上限: {final_search_upper_limit_time} "
|
||||
f"(可读: {datetime.fromtimestamp(final_search_upper_limit_time).strftime('%Y-%m-%d %H:%M:%S')}). "
|
||||
f"基于默认排除 {default_search_exclude_seconds}s 和 '最近记录'片段开始时间: {log_earliest_time_str}")
|
||||
|
||||
|
||||
logger.debug(
|
||||
f"[{private_name}] (私聊历史) retrieve_contextual_info: "
|
||||
f"最终用于历史搜索的时间上限: {final_search_upper_limit_time} "
|
||||
f"(可读: {datetime.fromtimestamp(final_search_upper_limit_time).strftime('%Y-%m-%d %H:%M:%S')}). "
|
||||
f"基于默认排除 {default_search_exclude_seconds}s 和 '最近记录'片段开始时间: {log_earliest_time_str}"
|
||||
)
|
||||
|
||||
most_relevant_message_doc = await find_most_relevant_historical_message(
|
||||
chat_id=chat_id,
|
||||
query_text=query_for_historical_chat,
|
||||
similarity_threshold=0.5, # 您可以调整这个
|
||||
similarity_threshold=0.5, # 您可以调整这个
|
||||
# exclude_recent_seconds 不再直接使用，而是传递计算好的绝对时间上限
|
||||
absolute_search_time_limit=final_search_upper_limit_time # <--- 传递计算好的绝对时间上限
|
||||
absolute_search_time_limit=final_search_upper_limit_time, # <--- 传递计算好的绝对时间上限
|
||||
)
|
||||
|
||||
|
||||
if most_relevant_message_doc:
|
||||
anchor_id = most_relevant_message_doc.get("message_id")
|
||||
anchor_time = most_relevant_message_doc.get("time")
|
||||
|
||||
anchor_time = most_relevant_message_doc.get("time")
|
||||
|
||||
# 校验锚点时间是否真的符合我们的硬性上限 (理论上 find_most_relevant_historical_message 内部已保证)
|
||||
if anchor_time is not None and anchor_time >= final_search_upper_limit_time:
|
||||
logger.warning(f"[{private_name}] (私聊历史) find_most_relevant_historical_message 返回的锚点时间 {anchor_time} "
|
||||
f"并未严格小于最终搜索上限 {final_search_upper_limit_time}。可能导致重叠。跳过构建上下文。")
|
||||
logger.warning(
|
||||
f"[{private_name}] (私聊历史) find_most_relevant_historical_message 返回的锚点时间 {anchor_time} "
|
||||
f"并未严格小于最终搜索上限 {final_search_upper_limit_time}。可能导致重叠。跳过构建上下文。"
|
||||
)
|
||||
historical_chat_log_msg = "检索到的锚点不符合最终时间要求，可能导致重叠。"
|
||||
# 直接进入下一个分支 (else)，使得 retrieved_historical_chat_str 保持默认值
|
||||
elif anchor_id and anchor_time is not None:
|
||||
# 构建上下文窗口时，其“未来”消息的上限也应该是 final_search_upper_limit_time
|
||||
# 因为我们不希望历史回忆的上下文窗口延伸到“最近聊天记录”的范围内或更近
|
||||
time_limit_for_context_window_after = final_search_upper_limit_time
|
||||
|
||||
logger.debug(f"[{private_name}] (私聊历史) 调用 retrieve_chat_context_window "
|
||||
f"with anchor_time: {anchor_time}, "
|
||||
f"excluded_time_threshold_for_window: {time_limit_for_context_window_after}")
|
||||
time_limit_for_context_window_after = final_search_upper_limit_time
|
||||
|
||||
logger.debug(
|
||||
f"[{private_name}] (私聊历史) 调用 retrieve_chat_context_window "
|
||||
f"with anchor_time: {anchor_time}, "
|
||||
f"excluded_time_threshold_for_window: {time_limit_for_context_window_after}"
|
||||
)
|
||||
|
||||
context_window_messages = await retrieve_chat_context_window(
|
||||
chat_id=chat_id,
|
||||
anchor_message_id=anchor_id,
|
||||
anchor_message_time=anchor_time,
|
||||
anchor_message_time=anchor_time,
|
||||
excluded_time_threshold_for_window=time_limit_for_context_window_after,
|
||||
window_size_before=7,
|
||||
window_size_after=7
|
||||
window_size_after=7,
|
||||
)
|
||||
if context_window_messages:
|
||||
formatted_window_str = await build_readable_messages(
|
||||
context_window_messages,
|
||||
replace_bot_name=False, # 在回忆中，保留原始发送者名称
|
||||
replace_bot_name=False, # 在回忆中，保留原始发送者名称
|
||||
merge_messages=False,
|
||||
timestamp_mode="relative", # 可以选择 'absolute' 或 'none'
|
||||
read_mark=0.0
|
||||
timestamp_mode="relative", # 可以选择 'absolute' 或 'none'
|
||||
read_mark=0.0,
|
||||
)
|
||||
if formatted_window_str and formatted_window_str.strip():
|
||||
retrieved_historical_chat_str = f"你回忆起一段与当前对话相关的历史聊天：\n------\n{formatted_window_str.strip()}\n------\n(以上是针对本次私聊的回忆，供参考)\n"
|
||||
|
|
@ -359,14 +440,18 @@ async def retrieve_contextual_info(
|
|||
historical_chat_log_msg = "检索到的最相关私聊历史消息文档缺少 message_id 或 time。"
|
||||
else:
|
||||
historical_chat_log_msg = "未找到足够相关的私聊历史对话消息。"
|
||||
logger.debug(f"[私聊][{private_name}] (retrieve_contextual_info) 私聊历史对话检索: {historical_chat_log_msg}")
|
||||
logger.debug(
|
||||
f"[私聊][{private_name}] (retrieve_contextual_info) 私聊历史对话检索: {historical_chat_log_msg}"
|
||||
)
|
||||
except Exception as e:
|
||||
logger.error(
|
||||
f"[私聊][{private_name}] (retrieve_contextual_info) 检索私聊历史对话时出错: {e}\n{traceback.format_exc()}"
|
||||
)
|
||||
retrieved_historical_chat_str = "[检索私聊历史对话时出错]\n"
|
||||
else:
|
||||
logger.debug(f"[私聊][{private_name}] (retrieve_contextual_info) 无专门的私聊历史查询文本，跳过私聊历史对话检索。")
|
||||
logger.debug(
|
||||
f"[私聊][{private_name}] (retrieve_contextual_info) 无专门的私聊历史查询文本，跳过私聊历史对话检索。"
|
||||
)
|
||||
|
||||
return retrieved_global_memory_str, retrieved_knowledge_str, retrieved_historical_chat_str
|
||||
|
||||
|
|
@ -410,13 +495,13 @@ def get_items_from_json(
|
|||
json_array = json.loads(cleaned_content)
|
||||
if isinstance(json_array, list):
|
||||
valid_items_list: List[Dict[str, Any]] = []
|
||||
for item_json in json_array: # Renamed item to item_json to avoid conflict
|
||||
for item_json in json_array: # Renamed item to item_json to avoid conflict
|
||||
if not isinstance(item_json, dict):
|
||||
logger.warning(f"[私聊][{private_name}] JSON数组中的元素不是字典: {item_json}")
|
||||
continue
|
||||
current_item_result = default_result.copy()
|
||||
valid_item = True
|
||||
for field in items: # items is args from function signature
|
||||
for field in items: # items is args from function signature
|
||||
if field in item_json:
|
||||
current_item_result[field] = item_json[field]
|
||||
elif field not in default_result:
|
||||
|
|
@ -427,15 +512,25 @@ def get_items_from_json(
|
|||
continue
|
||||
if required_types:
|
||||
for field, expected_type in required_types.items():
|
||||
if field in current_item_result and not isinstance(current_item_result[field], expected_type):
|
||||
logger.warning(f"[私聊][{private_name}] JSON数组元素字段 '{field}' 类型错误 (应为 {expected_type.__name__}, 实际为 {type(current_item_result[field]).__name__}): {item_json}")
|
||||
if field in current_item_result and not isinstance(
|
||||
current_item_result[field], expected_type
|
||||
):
|
||||
logger.warning(
|
||||
f"[私聊][{private_name}] JSON数组元素字段 '{field}' 类型错误 (应为 {expected_type.__name__}, 实际为 {type(current_item_result[field]).__name__}): {item_json}"
|
||||
)
|
||||
valid_item = False
|
||||
break
|
||||
if not valid_item:
|
||||
continue
|
||||
for field in items:
|
||||
if field in current_item_result and isinstance(current_item_result[field], str) and not current_item_result[field].strip():
|
||||
logger.warning(f"[私聊][{private_name}] JSON数组元素字段 '{field}' 不能为空字符串: {item_json}")
|
||||
if (
|
||||
field in current_item_result
|
||||
and isinstance(current_item_result[field], str)
|
||||
and not current_item_result[field].strip()
|
||||
):
|
||||
logger.warning(
|
||||
f"[私聊][{private_name}] JSON数组元素字段 '{field}' 不能为空字符串: {item_json}"
|
||||
)
|
||||
valid_item = False
|
||||
break
|
||||
if valid_item:
|
||||
|
|
@ -472,12 +567,14 @@ def get_items_from_json(
|
|||
logger.error(f"[私聊][{private_name}] 正则提取的部分 '{potential_json_str[:100]}...' 无法解析为JSON。")
|
||||
return False, default_result
|
||||
else:
|
||||
logger.error(f"[私聊][{private_name}] 无法在返回内容中找到有效的JSON对象部分。原始内容: {cleaned_content[:100]}...")
|
||||
logger.error(
|
||||
f"[私聊][{private_name}] 无法在返回内容中找到有效的JSON对象部分。原始内容: {cleaned_content[:100]}..."
|
||||
)
|
||||
return False, default_result
|
||||
if not isinstance(result, dict):
|
||||
result = default_result.copy()
|
||||
valid_single_object = True
|
||||
for item_field in items: # Renamed item to item_field
|
||||
for item_field in items: # Renamed item to item_field
|
||||
if item_field in json_data:
|
||||
result[item_field] = json_data[item_field]
|
||||
elif item_field not in default_result:
|
||||
|
|
@ -489,7 +586,9 @@ def get_items_from_json(
|
|||
if required_types:
|
||||
for field, expected_type in required_types.items():
|
||||
if field in result and not isinstance(result[field], expected_type):
|
||||
logger.error(f"[私聊][{private_name}] JSON对象字段 '{field}' 类型错误 (应为 {expected_type.__name__}, 实际为 {type(result[field]).__name__})")
|
||||
logger.error(
|
||||
f"[私聊][{private_name}] JSON对象字段 '{field}' 类型错误 (应为 {expected_type.__name__}, 实际为 {type(result[field]).__name__})"
|
||||
)
|
||||
valid_single_object = False
|
||||
break
|
||||
if not valid_single_object:
|
||||
|
|
@ -507,7 +606,7 @@ def get_items_from_json(
|
|||
|
||||
|
||||
async def get_person_id(private_name: str, chat_stream: ChatStream):
|
||||
""" (保持你原始 pfc_utils.py 中的此函数代码不变) """
|
||||
"""(保持你原始 pfc_utils.py 中的此函数代码不变)"""
|
||||
private_user_id_str: Optional[str] = None
|
||||
private_platform_str: Optional[str] = None
|
||||
# private_nickname_str = private_name # 这行在你提供的代码中没有被使用，可以考虑移除
|
||||
|
|
@ -516,7 +615,7 @@ async def get_person_id(private_name: str, chat_stream: ChatStream):
|
|||
private_user_id_str = str(chat_stream.user_info.user_id)
|
||||
private_platform_str = chat_stream.user_info.platform
|
||||
logger.debug(
|
||||
f"[私聊][{private_name}] 从 ChatStream 获取到私聊对象信息: ID={private_user_id_str}, Platform={private_platform_str}, Name={private_name}" # 使用 private_name
|
||||
f"[私聊][{private_name}] 从 ChatStream 获取到私聊对象信息: ID={private_user_id_str}, Platform={private_platform_str}, Name={private_name}" # 使用 private_name
|
||||
)
|
||||
# elif chat_stream.group_info is None and private_name: # 这个 elif 条件体为空，可以移除
|
||||
# pass
|
||||
|
|
@ -547,7 +646,7 @@ async def get_person_id(private_name: str, chat_stream: ChatStream):
|
|||
|
||||
|
||||
async def adjust_relationship_value_nonlinear(old_value: float, raw_adjustment: float) -> float:
|
||||
""" (保持你原始 pfc_utils.py 中的此函数代码不变) """
|
||||
"""(保持你原始 pfc_utils.py 中的此函数代码不变)"""
|
||||
old_value = max(-1000, min(1000, old_value))
|
||||
value = raw_adjustment
|
||||
if old_value >= 0:
|
||||
|
|
@ -555,7 +654,9 @@ async def adjust_relationship_value_nonlinear(old_value: float, raw_adjustment:
|
|||
value = value * math.cos(math.pi * old_value / 2000)
|
||||
if old_value > 500:
|
||||
# 确保 person_info_manager.get_specific_value_list 是异步的，如果是同步则需要调整
|
||||
rdict = await person_info_manager.get_specific_value_list("relationship_value", lambda x: x > 700 if isinstance(x, (int, float)) else False)
|
||||
rdict = await person_info_manager.get_specific_value_list(
|
||||
"relationship_value", lambda x: x > 700 if isinstance(x, (int, float)) else False
|
||||
)
|
||||
high_value_count = len(rdict)
|
||||
if old_value > 700:
|
||||
value *= 3 / (high_value_count + 2)
|
||||
|
|
@ -564,7 +665,7 @@ async def adjust_relationship_value_nonlinear(old_value: float, raw_adjustment:
|
|||
elif value < 0:
|
||||
value = value * math.exp(old_value / 2000)
|
||||
# else: value = 0 # 你原始代码中没有这句，如果value为0，保持为0
|
||||
else: # old_value < 0
|
||||
else: # old_value < 0
|
||||
if value >= 0:
|
||||
value = value * math.exp(old_value / 2000)
|
||||
elif value < 0:
|
||||
|
|
@ -586,12 +687,12 @@ async def build_chat_history_text(observation_info: ObservationInfo, private_nam
|
|||
)
|
||||
else:
|
||||
chat_history_text = "还没有聊天记录。\n"
|
||||
|
||||
|
||||
unread_count = getattr(observation_info, "new_messages_count", 0)
|
||||
unread_messages = getattr(observation_info, "unprocessed_messages", [])
|
||||
if unread_count > 0 and unread_messages:
|
||||
bot_qq_str = str(global_config.BOT_QQ) if global_config.BOT_QQ else None # 安全获取
|
||||
if bot_qq_str: # 仅当 bot_qq_str 有效时进行过滤
|
||||
bot_qq_str = str(global_config.BOT_QQ) if global_config.BOT_QQ else None # 安全获取
|
||||
if bot_qq_str: # 仅当 bot_qq_str 有效时进行过滤
|
||||
other_unread_messages = [
|
||||
msg for msg in unread_messages if msg.get("user_info", {}).get("user_id") != bot_qq_str
|
||||
]
|
||||
|
|
@ -599,12 +700,12 @@ async def build_chat_history_text(observation_info: ObservationInfo, private_nam
|
|||
if other_unread_count > 0:
|
||||
new_messages_str = await build_readable_messages(
|
||||
other_unread_messages,
|
||||
replace_bot_name=True, # 这里是未处理消息，可能不需要替换机器人名字
|
||||
replace_bot_name=True, # 这里是未处理消息，可能不需要替换机器人名字
|
||||
merge_messages=False,
|
||||
timestamp_mode="relative",
|
||||
read_mark=0.0,
|
||||
)
|
||||
chat_history_text += f"\n{new_messages_str}\n------\n" # 原始代码是加在末尾的
|
||||
chat_history_text += f"\n{new_messages_str}\n------\n" # 原始代码是加在末尾的
|
||||
else:
|
||||
logger.warning(f"[私聊][{private_name}] BOT_QQ 未配置，无法准确过滤未读消息中的机器人自身消息。")
|
||||
|
||||
|
|
@ -614,4 +715,4 @@ async def build_chat_history_text(observation_info: ObservationInfo, private_nam
|
|||
except Exception as e:
|
||||
logger.error(f"[私聊][{private_name}] 处理聊天记录时发生未知错误: {e}")
|
||||
chat_history_text = "[处理聊天记录时出错]\n"
|
||||
return chat_history_text
|
||||
return chat_history_text
|
||||
|
|
|
|||
|
|
@ -226,7 +226,7 @@ class ReplyGenerator:
|
|||
|
||||
chat_history_for_prompt_builder: list = []
|
||||
recent_history_start_time_for_exclusion: Optional[float] = None
|
||||
|
||||
|
||||
# 我们需要知道 build_chat_history_text 函数大致会用 observation_info.chat_history 的多少条记录
|
||||
# 或者 build_chat_history_text 内部的逻辑。
|
||||
# 假设 build_chat_history_text 主要依赖 observation_info.chat_history_str，
|
||||
|
|
@ -238,7 +238,7 @@ class ReplyGenerator:
|
|||
# 如果 observation_info.chat_history_str 是由 observation_info.py 中的 update_from_message 等方法维护的，
|
||||
# 并且总是代表一个固定长度（比如最后30条）的聊天记录字符串，那么我们就需要从 observation_info.chat_history
|
||||
# 取出这部分原始消息来确定起始时间。
|
||||
|
||||
|
||||
# 我们先做一个合理的假设： “最近聊天记录” 字符串 chat_history_text 是基于
|
||||
# observation_info.chat_history 的一个有限的尾部片段生成的。
|
||||
# 假设这个片段的长度由 global_config.pfc_recent_history_display_count 控制，默认为20条。
|
||||
|
|
@ -249,25 +249,29 @@ class ReplyGenerator:
|
|||
# 如果 observation_info.chat_history 长度小于 display_count，则取全部
|
||||
start_index = max(0, len(observation_info.chat_history) - recent_history_display_count)
|
||||
chat_history_for_prompt_builder = observation_info.chat_history[start_index:]
|
||||
|
||||
if chat_history_for_prompt_builder: # 如果片段不为空
|
||||
|
||||
if chat_history_for_prompt_builder: # 如果片段不为空
|
||||
try:
|
||||
first_message_in_display_slice = chat_history_for_prompt_builder[0]
|
||||
recent_history_start_time_for_exclusion = first_message_in_display_slice.get('time')
|
||||
recent_history_start_time_for_exclusion = first_message_in_display_slice.get("time")
|
||||
if recent_history_start_time_for_exclusion:
|
||||
# 导入 datetime (如果 reply_generator.py 文件顶部没有的话)
|
||||
# from datetime import datetime # 通常建议放在文件顶部
|
||||
logger.debug(f"[{self.private_name}] (ReplyGenerator) “最近聊天记录”片段(共{len(chat_history_for_prompt_builder)}条)的最早时间戳: "
|
||||
f"{recent_history_start_time_for_exclusion} "
|
||||
f"(即 {datetime.fromtimestamp(recent_history_start_time_for_exclusion).strftime('%Y-%m-%d %H:%M:%S')})")
|
||||
logger.debug(
|
||||
f"[{self.private_name}] (ReplyGenerator) “最近聊天记录”片段(共{len(chat_history_for_prompt_builder)}条)的最早时间戳: "
|
||||
f"{recent_history_start_time_for_exclusion} "
|
||||
f"(即 {datetime.fromtimestamp(recent_history_start_time_for_exclusion).strftime('%Y-%m-%d %H:%M:%S')})"
|
||||
)
|
||||
else:
|
||||
logger.warning(f"[{self.private_name}] (ReplyGenerator) “最近聊天记录”片段的首条消息无时间戳。")
|
||||
except (IndexError, KeyError, TypeError) as e:
|
||||
logger.warning(f"[{self.private_name}] (ReplyGenerator) 获取“最近聊天记录”起始时间失败: {e}")
|
||||
recent_history_start_time_for_exclusion = None
|
||||
recent_history_start_time_for_exclusion = None
|
||||
else:
|
||||
logger.debug(f"[{self.private_name}] (ReplyGenerator) observation_info.chat_history 为空，无法确定“最近聊天记录”起始时间。")
|
||||
# --- [新代码结束] ---
|
||||
logger.debug(
|
||||
f"[{self.private_name}] (ReplyGenerator) observation_info.chat_history 为空，无法确定“最近聊天记录”起始时间。"
|
||||
)
|
||||
# --- [新代码结束] ---
|
||||
|
||||
chat_history_text = await build_chat_history_text(observation_info, self.private_name)
|
||||
|
||||
|
|
@ -278,28 +282,32 @@ class ReplyGenerator:
|
|||
|
||||
persona_text = f"你的名字是{self.name}，{self.personality_info}。"
|
||||
historical_chat_query = ""
|
||||
num_recent_messages_for_query = 3 # 例如，取最近3条作为查询引子
|
||||
num_recent_messages_for_query = 3 # 例如，取最近3条作为查询引子
|
||||
if observation_info.chat_history and len(observation_info.chat_history) > 0:
|
||||
# 从 chat_history (已处理并存入 ObservationInfo 的历史) 中取最新N条
|
||||
# 或者，如果 observation_info.unprocessed_messages 更能代表“当前上下文”，也可以考虑用它
|
||||
# 我们先用 chat_history，因为它包含了双方的对话历史，可能更稳定
|
||||
recent_messages_for_query_list = observation_info.chat_history[-num_recent_messages_for_query:]
|
||||
|
||||
|
||||
# 将这些消息的文本内容合并
|
||||
query_texts_list = []
|
||||
for msg_dict in recent_messages_for_query_list:
|
||||
text_content = msg_dict.get("processed_plain_text", "")
|
||||
if text_content.strip(): # 只添加有内容的文本
|
||||
if text_content.strip(): # 只添加有内容的文本
|
||||
# 可以选择是否添加发送者信息到查询文本中，例如：
|
||||
# sender_nickname = msg_dict.get("user_info", {}).get("user_nickname", "用户")
|
||||
# query_texts_list.append(f"{sender_nickname}: {text_content}")
|
||||
query_texts_list.append(text_content) # 简单合并文本内容
|
||||
|
||||
query_texts_list.append(text_content) # 简单合并文本内容
|
||||
|
||||
if query_texts_list:
|
||||
historical_chat_query = " ".join(query_texts_list).strip()
|
||||
logger.debug(f"[私聊][{self.private_name}] (ReplyGenerator) 生成的私聊历史查询文本 (最近{num_recent_messages_for_query}条): '{historical_chat_query[:100]}...'")
|
||||
logger.debug(
|
||||
f"[私聊][{self.private_name}] (ReplyGenerator) 生成的私聊历史查询文本 (最近{num_recent_messages_for_query}条): '{historical_chat_query[:100]}...'"
|
||||
)
|
||||
else:
|
||||
logger.debug(f"[私聊][{self.private_name}] (ReplyGenerator) 最近{num_recent_messages_for_query}条消息无有效文本内容，不进行私聊历史查询。")
|
||||
logger.debug(
|
||||
f"[私聊][{self.private_name}] (ReplyGenerator) 最近{num_recent_messages_for_query}条消息无有效文本内容，不进行私聊历史查询。"
|
||||
)
|
||||
else:
|
||||
logger.debug(f"[私聊][{self.private_name}] (ReplyGenerator) 无聊天历史可用于生成私聊历史查询文本。")
|
||||
|
||||
|
|
@ -316,13 +324,13 @@ class ReplyGenerator:
|
|||
(
|
||||
retrieved_global_memory_str,
|
||||
retrieved_knowledge_str,
|
||||
retrieved_historical_chat_str # << 新增接收私聊历史回忆
|
||||
retrieved_historical_chat_str, # << 新增接收私聊历史回忆
|
||||
) = await retrieve_contextual_info(
|
||||
text=retrieval_context_for_global_and_knowledge, # 用于全局记忆和知识
|
||||
text=retrieval_context_for_global_and_knowledge, # 用于全局记忆和知识
|
||||
private_name=self.private_name,
|
||||
chat_id=current_chat_id, # << 传递 chat_id
|
||||
historical_chat_query_text=historical_chat_query, # << 传递专门的查询文本
|
||||
current_short_term_history_earliest_time=recent_history_start_time_for_exclusion # <--- 新增传递的参数
|
||||
chat_id=current_chat_id, # << 传递 chat_id
|
||||
historical_chat_query_text=historical_chat_query, # << 传递专门的查询文本
|
||||
current_short_term_history_earliest_time=recent_history_start_time_for_exclusion, # <--- 新增传递的参数
|
||||
)
|
||||
# === 调用修改结束 ===
|
||||
|
||||
|
|
@ -394,10 +402,18 @@ class ReplyGenerator:
|
|||
base_format_params = {
|
||||
"persona_text": persona_text,
|
||||
"goals_str": goals_str,
|
||||
"chat_history_text": chat_history_text if chat_history_text.strip() else "还没有聊天记录。", # 当前短期历史
|
||||
"retrieved_global_memory_str": retrieved_global_memory_str if retrieved_global_memory_str.strip() else "无相关全局记忆。",
|
||||
"retrieved_knowledge_str": retrieved_knowledge_str if retrieved_knowledge_str.strip() else "无相关知识。",
|
||||
"retrieved_historical_chat_str": retrieved_historical_chat_str if retrieved_historical_chat_str.strip() else "无相关私聊历史回忆。", # << 新增
|
||||
"chat_history_text": chat_history_text
|
||||
if chat_history_text.strip()
|
||||
else "还没有聊天记录。", # 当前短期历史
|
||||
"retrieved_global_memory_str": retrieved_global_memory_str
|
||||
if retrieved_global_memory_str.strip()
|
||||
else "无相关全局记忆。",
|
||||
"retrieved_knowledge_str": retrieved_knowledge_str
|
||||
if retrieved_knowledge_str.strip()
|
||||
else "无相关知识。",
|
||||
"retrieved_historical_chat_str": retrieved_historical_chat_str
|
||||
if retrieved_historical_chat_str.strip()
|
||||
else "无相关私聊历史回忆。", # << 新增
|
||||
"last_rejection_info": last_rejection_info_str,
|
||||
"current_time_str": current_time_value,
|
||||
"sender_name": sender_name_str,
|
||||
|
|
|
|||
Loading…
Reference in New Issue