gardel/Mofox-Core
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix：优化关系构建频率

Browse Source
This commit is contained in:
SengokuCola
2025-06-16 16:33:09 +08:00
parent 06fbf125d5
commit 9b7066109d
5 changed files with 456 additions and 108 deletions
Download Patch File Download Diff File Expand all files Collapse all files

483
src/chat/focus_chat/info_processors/relationship_processor.py
View File

@@ -17,7 +17,9 @@ from json_repair import repair_json
from src.person_info.person_info import get_person_info_manager
import json
import asyncio
from src.chat.utils.chat_message_builder import get_raw_msg_by_timestamp_with_chat
from src.chat.utils.chat_message_builder import get_raw_msg_by_timestamp_with_chat, get_raw_msg_by_timestamp_with_chat_inclusive, get_raw_msg_before_timestamp_with_chat, num_new_messages_since
import os
import pickle
# 配置常量：是否启用小模型即时信息提取
@@ -25,6 +27,14 @@ from src.chat.utils.chat_message_builder import get_raw_msg_by_timestamp_with_ch
# 关闭时：使用原来的异步模式，精度更高但速度较慢
ENABLE_INSTANT_INFO_EXTRACTION = True
# 消息段清理配置
SEGMENT_CLEANUP_CONFIG = {
"enable_cleanup": True, # 是否启用清理
"max_segment_age_days": 7, # 消息段最大保存天数
"max_segments_per_user": 10, # 每用户最大消息段数
"cleanup_interval_hours": 1, # 清理间隔（小时）
}
logger = get_logger("processor")
@@ -95,8 +105,19 @@ class RelationshipProcessor(BaseProcessor):
self.info_fetched_cache: Dict[
str, Dict[str, any]
] = {} # {person_id: {"info": str, "ttl": int, "start_time": float}}
self.person_engaged_cache: List[Dict[str, any]] = [] # [{person_id: str, start_time: float, rounds: int}]
self.grace_period_rounds = 5
# 新的消息段缓存结构：
# {person_id: [{"start_time": float, "end_time": float, "last_msg_time": float, "message_count": int}, ...]}
self.person_engaged_cache: Dict[str, List[Dict[str, any]]] = {}
# 持久化存储文件路径
self.cache_file_path = os.path.join("data", f"relationship_cache_{self.subheartflow_id}.pkl")
# 最后处理的消息时间，避免重复处理相同消息
self.last_processed_message_time = 0.0
# 最后清理时间，用于定期清理老消息段
self.last_cleanup_time = 0.0
self.llm_model = LLMRequest(
model=global_config.model.relation,
@@ -112,7 +133,293 @@ class RelationshipProcessor(BaseProcessor):
name = get_chat_manager().get_stream_name(self.subheartflow_id)
self.log_prefix = f"[{name}] "
# 加载持久化的缓存
self._load_cache()
# ================================
# 缓存管理模块
# 负责持久化存储、状态管理、缓存读写
# ================================
def _load_cache(self):
"""从文件加载持久化的缓存"""
if os.path.exists(self.cache_file_path):
try:
with open(self.cache_file_path, 'rb') as f:
cache_data = pickle.load(f)
if isinstance(cache_data, dict) and 'person_engaged_cache' in cache_data:
# 新格式：包含额外信息的缓存
self.person_engaged_cache = cache_data.get('person_engaged_cache', {})
self.last_processed_message_time = cache_data.get('last_processed_message_time', 0.0)
self.last_cleanup_time = cache_data.get('last_cleanup_time', 0.0)
else:
# 旧格式：仅包含person_engaged_cache
self.person_engaged_cache = cache_data
self.last_processed_message_time = 0.0
self.last_cleanup_time = 0.0
logger.info(f"{self.log_prefix} 成功加载关系缓存，包含 {len(self.person_engaged_cache)} 个用户，最后处理时间：{time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(self.last_processed_message_time)) if self.last_processed_message_time > 0 else '未设置'}")
except Exception as e:
logger.error(f"{self.log_prefix} 加载关系缓存失败: {e}")
self.person_engaged_cache = {}
self.last_processed_message_time = 0.0
else:
logger.info(f"{self.log_prefix} 关系缓存文件不存在，使用空缓存")
def _save_cache(self):
"""保存缓存到文件"""
try:
os.makedirs(os.path.dirname(self.cache_file_path), exist_ok=True)
cache_data = {
'person_engaged_cache': self.person_engaged_cache,
'last_processed_message_time': self.last_processed_message_time,
'last_cleanup_time': self.last_cleanup_time
}
with open(self.cache_file_path, 'wb') as f:
pickle.dump(cache_data, f)
logger.debug(f"{self.log_prefix} 成功保存关系缓存")
except Exception as e:
logger.error(f"{self.log_prefix} 保存关系缓存失败: {e}")
# ================================
# 消息段管理模块
# 负责跟踪用户消息活动、管理消息段、清理过期数据
# ================================
def _update_message_segments(self, person_id: str, message_time: float):
"""更新用户的消息段
Args:
person_id: 用户ID
message_time: 消息时间戳
"""
if person_id not in self.person_engaged_cache:
self.person_engaged_cache[person_id] = []
segments = self.person_engaged_cache[person_id]
current_time = time.time()
# 获取该消息前5条消息的时间作为潜在的开始时间
before_messages = get_raw_msg_before_timestamp_with_chat(self.subheartflow_id, message_time, limit=5)
if before_messages:
# 由于get_raw_msg_before_timestamp_with_chat返回按时间升序排序的消息，最后一个是最接近message_time的
# 我们需要第一个消息作为开始时间，但应该确保至少包含5条消息或该用户之前的消息
potential_start_time = before_messages[0]['time']
else:
# 如果没有前面的消息，就从当前消息开始
potential_start_time = message_time
# 如果没有现有消息段，创建新的
if not segments:
new_segment = {
"start_time": potential_start_time,
"end_time": message_time,
"last_msg_time": message_time,
"message_count": self._count_messages_in_timerange(potential_start_time, message_time)
}
segments.append(new_segment)
logger.info(f"{self.log_prefix} 为用户 {person_id} 创建新消息段: 时间范围 {time.strftime('%H:%M:%S', time.localtime(potential_start_time))} - {time.strftime('%H:%M:%S', time.localtime(message_time))}, 消息数: {new_segment['message_count']}")
self._save_cache()
return
# 获取最后一个消息段
last_segment = segments[-1]
# 计算从最后一条消息到当前消息之间的消息数量（不包含边界）
messages_between = self._count_messages_between(last_segment["last_msg_time"], message_time)
if messages_between <= 10:
# 在10条消息内，延伸当前消息段
last_segment["end_time"] = message_time
last_segment["last_msg_time"] = message_time
# 重新计算整个消息段的消息数量
last_segment["message_count"] = self._count_messages_in_timerange(
last_segment["start_time"], last_segment["end_time"]
)
logger.info(f"{self.log_prefix} 延伸用户 {person_id} 的消息段: {last_segment}")
else:
# 超过10条消息，结束当前消息段并创建新的
# 结束当前消息段：延伸到原消息段最后一条消息后5条消息的时间
after_messages = get_raw_msg_by_timestamp_with_chat(
self.subheartflow_id, last_segment["last_msg_time"], current_time, limit=5, limit_mode="earliest"
)
if after_messages and len(after_messages) >= 5:
# 如果有足够的后续消息，使用第5条消息的时间作为结束时间
last_segment["end_time"] = after_messages[4]['time']
else:
# 如果没有足够的后续消息，保持原有的结束时间
pass
# 重新计算当前消息段的消息数量
last_segment["message_count"] = self._count_messages_in_timerange(
last_segment["start_time"], last_segment["end_time"]
)
# 创建新的消息段
new_segment = {
"start_time": potential_start_time,
"end_time": message_time,
"last_msg_time": message_time,
"message_count": self._count_messages_in_timerange(potential_start_time, message_time)
}
segments.append(new_segment)
logger.info(f"{self.log_prefix} 为用户 {person_id} 创建新消息段（超过10条消息间隔）: {new_segment}")
self._save_cache()
def _count_messages_in_timerange(self, start_time: float, end_time: float) -> int:
"""计算指定时间范围内的消息数量（包含边界）"""
messages = get_raw_msg_by_timestamp_with_chat_inclusive(self.subheartflow_id, start_time, end_time)
return len(messages)
def _count_messages_between(self, start_time: float, end_time: float) -> int:
"""计算两个时间点之间的消息数量（不包含边界），用于间隔检查"""
return num_new_messages_since(self.subheartflow_id, start_time, end_time)
def _get_total_message_count(self, person_id: str) -> int:
"""获取用户所有消息段的总消息数量"""
if person_id not in self.person_engaged_cache:
return 0
total_count = 0
for segment in self.person_engaged_cache[person_id]:
total_count += segment["message_count"]
return total_count
def _cleanup_old_segments(self) -> bool:
"""清理老旧的消息段
Returns:
bool: 是否执行了清理操作
"""
if not SEGMENT_CLEANUP_CONFIG["enable_cleanup"]:
return False
current_time = time.time()
# 检查是否需要执行清理（基于时间间隔）
cleanup_interval_seconds = SEGMENT_CLEANUP_CONFIG["cleanup_interval_hours"] * 3600
if current_time - self.last_cleanup_time < cleanup_interval_seconds:
return False
logger.info(f"{self.log_prefix} 开始执行老消息段清理...")
cleanup_stats = {
"users_cleaned": 0,
"segments_removed": 0,
"total_segments_before": 0,
"total_segments_after": 0
}
max_age_seconds = SEGMENT_CLEANUP_CONFIG["max_segment_age_days"] * 24 * 3600
max_segments_per_user = SEGMENT_CLEANUP_CONFIG["max_segments_per_user"]
users_to_remove = []
for person_id, segments in self.person_engaged_cache.items():
cleanup_stats["total_segments_before"] += len(segments)
original_segment_count = len(segments)
# 1. 按时间清理：移除过期的消息段
segments_after_age_cleanup = []
for segment in segments:
segment_age = current_time - segment["end_time"]
if segment_age <= max_age_seconds:
segments_after_age_cleanup.append(segment)
else:
cleanup_stats["segments_removed"] += 1
logger.debug(f"{self.log_prefix} 移除用户 {person_id} 的过期消息段: {time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(segment['start_time']))} - {time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(segment['end_time']))}")
# 2. 按数量清理：如果消息段数量仍然过多，保留最新的
if len(segments_after_age_cleanup) > max_segments_per_user:
# 按end_time排序，保留最新的
segments_after_age_cleanup.sort(key=lambda x: x["end_time"], reverse=True)
segments_removed_count = len(segments_after_age_cleanup) - max_segments_per_user
cleanup_stats["segments_removed"] += segments_removed_count
segments_after_age_cleanup = segments_after_age_cleanup[:max_segments_per_user]
logger.debug(f"{self.log_prefix} 用户 {person_id} 消息段数量过多，移除 {segments_removed_count} 个最老的消息段")
# 使用清理后的消息段
# 更新缓存
if len(segments_after_age_cleanup) == 0:
# 如果没有剩余消息段，标记用户为待移除
users_to_remove.append(person_id)
else:
self.person_engaged_cache[person_id] = segments_after_age_cleanup
cleanup_stats["total_segments_after"] += len(segments_after_age_cleanup)
if original_segment_count != len(segments_after_age_cleanup):
cleanup_stats["users_cleaned"] += 1
# 移除没有消息段的用户
for person_id in users_to_remove:
del self.person_engaged_cache[person_id]
logger.debug(f"{self.log_prefix} 移除用户 {person_id}：没有剩余消息段")
# 更新最后清理时间
self.last_cleanup_time = current_time
# 保存缓存
if cleanup_stats["segments_removed"] > 0 or len(users_to_remove) > 0:
self._save_cache()
logger.info(f"{self.log_prefix} 清理完成 - 影响用户: {cleanup_stats['users_cleaned']}, 移除消息段: {cleanup_stats['segments_removed']}, 移除用户: {len(users_to_remove)}")
logger.info(f"{self.log_prefix} 消息段统计 - 清理前: {cleanup_stats['total_segments_before']}, 清理后: {cleanup_stats['total_segments_after']}")
else:
logger.debug(f"{self.log_prefix} 清理完成 - 无需清理任何内容")
return cleanup_stats["segments_removed"] > 0 or len(users_to_remove) > 0
def force_cleanup_user_segments(self, person_id: str) -> bool:
"""强制清理指定用户的所有消息段
Args:
person_id: 用户ID
Returns:
bool: 是否成功清理
"""
if person_id in self.person_engaged_cache:
segments_count = len(self.person_engaged_cache[person_id])
del self.person_engaged_cache[person_id]
self._save_cache()
logger.info(f"{self.log_prefix} 强制清理用户 {person_id} 的 {segments_count} 个消息段")
return True
return False
def get_cache_status(self) -> str:
"""获取缓存状态信息，用于调试和监控"""
if not self.person_engaged_cache:
return f"{self.log_prefix} 关系缓存为空"
status_lines = [f"{self.log_prefix} 关系缓存状态："]
status_lines.append(f"最后处理消息时间：{time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(self.last_processed_message_time)) if self.last_processed_message_time > 0 else '未设置'}")
status_lines.append(f"最后清理时间：{time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(self.last_cleanup_time)) if self.last_cleanup_time > 0 else '未执行'}")
status_lines.append(f"总用户数：{len(self.person_engaged_cache)}")
status_lines.append(f"清理配置：{'启用' if SEGMENT_CLEANUP_CONFIG['enable_cleanup'] else '禁用'} (最大保存{SEGMENT_CLEANUP_CONFIG['max_segment_age_days']}天, 每用户最多{SEGMENT_CLEANUP_CONFIG['max_segments_per_user']}段)")
status_lines.append("")
for person_id, segments in self.person_engaged_cache.items():
total_count = self._get_total_message_count(person_id)
status_lines.append(f"用户 {person_id}:")
status_lines.append(f" 总消息数：{total_count} ({total_count}/45)")
status_lines.append(f" 消息段数：{len(segments)}")
for i, segment in enumerate(segments):
start_str = time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(segment['start_time']))
end_str = time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(segment['end_time']))
last_str = time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(segment['last_msg_time']))
status_lines.append(f" 段{i+1}: {start_str} -> {end_str} (最后消息: {last_str}, 消息数: {segment['message_count']})")
status_lines.append("")
return "\n".join(status_lines)
# ================================
# 主要处理流程
# 统筹各模块协作、对外提供服务接口
# ================================
async def process_info(self, observations: List[Observation] = None, *infos) -> List[InfoBase]:
"""处理信息对象
@@ -140,7 +447,10 @@ class RelationshipProcessor(BaseProcessor):
"""
在回复前进行思考，生成内心想法并收集工具调用结果
"""
# 0. 从观察信息中提取所需数据
# 0. 执行定期清理
self._cleanup_old_segments()
# 1. 从观察信息中提取所需数据
# 需要兼容私聊
chat_observe_info = ""
@@ -149,36 +459,52 @@ class RelationshipProcessor(BaseProcessor):
for observation in observations:
if isinstance(observation, ChattingObservation):
chat_observe_info = observation.get_observe_info()
# 从聊天观察中提取用户信息并更新消息段
# 获取最新的非bot消息来更新消息段
latest_messages = get_raw_msg_by_timestamp_with_chat(
self.subheartflow_id, self.last_processed_message_time, current_time, limit=50 # 获取自上次处理后的消息
)
if latest_messages:
# 处理所有新的非bot消息
for latest_msg in latest_messages:
user_id = latest_msg.get('user_id')
platform = latest_msg.get('user_platform') or latest_msg.get('chat_info_platform')
msg_time = latest_msg.get('time', 0)
if user_id and platform and user_id != global_config.bot.qq_account and msg_time > self.last_processed_message_time:
from src.person_info.person_info import PersonInfoManager
person_id = PersonInfoManager.get_person_id(platform, user_id)
self._update_message_segments(person_id, msg_time)
logger.debug(f"{self.log_prefix} 更新用户 {person_id} 的消息段，消息时间：{time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(msg_time))}")
self.last_processed_message_time = max(self.last_processed_message_time, msg_time)
break
# 1. 处理person_engaged_cache
for record in list(self.person_engaged_cache):
record["rounds"] += 1
time_elapsed = current_time - record["start_time"]
message_count = len(
get_raw_msg_by_timestamp_with_chat(self.subheartflow_id, record["start_time"], current_time)
)
print(record)
# 根据消息数量和时间设置不同的触发条件
should_trigger = (
message_count >= 50 # 50条消息必定满足
or (message_count >= 35 and time_elapsed >= 300) # 35条且10分钟
or (message_count >= 25 and time_elapsed >= 900) # 25条且30分钟
or (message_count >= 10 and time_elapsed >= 2000) # 10条且1小时
)
if should_trigger:
# 1. 检查是否有用户达到关系构建条件（总消息数达到45条）
users_to_build_relationship = []
for person_id, segments in self.person_engaged_cache.items():
total_message_count = self._get_total_message_count(person_id)
if total_message_count >= 45:
users_to_build_relationship.append(person_id)
logger.info(
f"{self.log_prefix} 用户 {record['person_id']} 满足关系构建条件，开始构建关系。消息数：{message_count}，时长：{time_elapsed:.0f}秒"
f"{self.log_prefix} 用户 {person_id} 满足关系构建条件，总消息数：{total_message_count}，消息段数：{len(segments)}"
)
asyncio.create_task(
self.update_impression_on_cache_expiry(
record["person_id"], self.subheartflow_id, record["start_time"], current_time
)
elif total_message_count > 0:
# 记录进度信息
logger.debug(
f"{self.log_prefix} 用户 {person_id} 进度：{total_message_count}/45 条消息，{len(segments)} 个消息段"
)
self.person_engaged_cache.remove(record)
# 2. 为满足条件的用户构建关系
for person_id in users_to_build_relationship:
segments = self.person_engaged_cache[person_id]
# 异步执行关系构建
asyncio.create_task(
self.update_impression_on_segments(person_id, self.subheartflow_id, segments)
)
# 移除已处理的用户缓存
del self.person_engaged_cache[person_id]
self._save_cache()
# 2. 减少info_fetched_cache中所有信息的TTL
for person_id in list(self.info_fetched_cache.keys()):
@@ -260,12 +586,8 @@ class RelationshipProcessor(BaseProcessor):
logger.info(f"{self.log_prefix} 调取用户 {person_name} 的 {info_type} 信息。")
# 检查person_engaged_cache中是否已存在该person_id
person_exists = any(record["person_id"] == person_id for record in self.person_engaged_cache)
if not person_exists:
self.person_engaged_cache.append(
{"person_id": person_id, "start_time": time.time(), "rounds": 0}
)
# 这里不需要检查person_engaged_cache，因为消息段的管理由_update_message_segments处理
# 信息提取和消息段管理是独立的流程
if ENABLE_INSTANT_INFO_EXTRACTION:
# 收集即时提取任务
@@ -336,6 +658,79 @@ class RelationshipProcessor(BaseProcessor):
return persons_infos_str
# ================================
# 关系构建模块
# 负责触发关系构建、整合消息段、更新用户印象
# ================================
async def update_impression_on_segments(self, person_id: str, chat_id: str, segments: List[Dict[str, any]]):
"""
基于消息段更新用户印象
Args:
person_id: 用户ID
chat_id: 聊天ID
segments: 消息段列表
"""
logger.info(f"开始为 {person_id} 基于 {len(segments)} 个消息段更新印象")
try:
processed_messages = []
for i, segment in enumerate(segments):
start_time = segment["start_time"]
end_time = segment["end_time"]
message_count = segment["message_count"]
start_date = time.strftime('%Y-%m-%d %H:%M', time.localtime(start_time))
# 获取该段的消息（包含边界）
segment_messages = get_raw_msg_by_timestamp_with_chat_inclusive(self.subheartflow_id, start_time, end_time)
logger.info(f"消息段 {i+1}: {start_date} - {time.strftime('%Y-%m-%d %H:%M', time.localtime(end_time))}, 消息数: {len(segment_messages)}")
if segment_messages:
# 如果不是第一个消息段，在消息列表前添加间隔标识
if i > 0:
# 创建一个特殊的间隔消息
gap_message = {
"time": start_time - 0.1, # 稍微早于段开始时间
"user_id": "system",
"user_platform": "system",
"user_nickname": "系统",
"user_cardname": "",
"display_message": f"...（中间省略一些消息）{start_date} 之后的消息如下...",
"is_action_record": True,
"chat_info_platform": segment_messages[0].get("chat_info_platform", ""),
"chat_id": chat_id
}
processed_messages.append(gap_message)
# 添加该段的所有消息
processed_messages.extend(segment_messages)
if processed_messages:
# 按时间排序所有消息（包括间隔标识）
processed_messages.sort(key=lambda x: x['time'])
logger.info(f"为 {person_id} 获取到总共 {len(processed_messages)} 条消息（包含间隔标识）用于印象更新")
relationship_manager = get_relationship_manager()
# 调用原有的更新方法
await relationship_manager.update_person_impression(
person_id=person_id,
timestamp=time.time(),
bot_engaged_messages=processed_messages
)
else:
logger.info(f"没有找到 {person_id} 的消息段对应的消息，不更新印象")
except Exception as e:
logger.error(f"为 {person_id} 更新印象时发生错误: {e}")
logger.error(traceback.format_exc())
# ================================
# 信息调取模块
# 负责实时分析对话需求、提取用户信息、管理信息缓存
# ================================
async def _execute_instant_extraction_batch(self, instant_tasks: list):
"""
批量执行即时提取任务
@@ -526,25 +921,7 @@ class RelationshipProcessor(BaseProcessor):
logger.error(f"{self.log_prefix} 执行LLM请求获取用户信息时出错: {e}")
logger.error(traceback.format_exc())
async def update_impression_on_cache_expiry(self, person_id: str, chat_id: str, start_time: float, end_time: float):
"""
在缓存过期时，获取聊天记录并更新用户印象
"""
logger.info(f"缓存过期，开始为 {person_id} 更新印象。时间范围：{start_time} -> {end_time}")
try:
impression_messages = get_raw_msg_by_timestamp_with_chat(chat_id, start_time, end_time)
if impression_messages:
logger.info(f"为 {person_id} 获取到 {len(impression_messages)} 条消息用于印象更新。")
relationship_manager = get_relationship_manager()
await relationship_manager.update_person_impression(
person_id=person_id, timestamp=end_time, bot_engaged_messages=impression_messages
)
else:
logger.info(f"在指定时间范围内没有找到 {person_id} 的消息，不更新印象。")
except Exception as e:
logger.error(f"为 {person_id} 更新印象时发生错误: {e}")
logger.error(traceback.format_exc())
init_prompt()