|
|
|
|
|
|
|
|
import os
|
|
|
import json
|
|
|
import logging
|
|
|
import asyncio
|
|
|
import hashlib
|
|
|
import numpy as np
|
|
|
from datetime import datetime, timedelta
|
|
|
from typing import List, Dict, Any, Optional, Tuple, Set
|
|
|
from collections import defaultdict, deque
|
|
|
import pickle
|
|
|
from dataclasses import dataclass, field
|
|
|
from enum import Enum
|
|
|
import re
|
|
|
import heapq
|
|
|
|
|
|
|
|
|
try:
|
|
|
from sentence_transformers import SentenceTransformer
|
|
|
HAS_SBERT = True
|
|
|
except ImportError:
|
|
|
HAS_SBERT = False
|
|
|
from sklearn.feature_extraction.text import TfidfVectorizer
|
|
|
import warnings
|
|
|
warnings.filterwarnings("ignore")
|
|
|
|
|
|
|
|
|
import data_manager
|
|
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
|
|
|
|
|
|
class MemoryPriority(Enum):
|
|
|
"""اولویت حافظه"""
|
|
|
LOW = 1
|
|
|
MEDIUM = 2
|
|
|
HIGH = 3
|
|
|
CRITICAL = 4
|
|
|
|
|
|
class MessageType(Enum):
|
|
|
"""نوع پیام"""
|
|
|
FACT = "fact"
|
|
|
PREFERENCE = "preference"
|
|
|
EMOTION = "emotion"
|
|
|
QUESTION = "question"
|
|
|
ANSWER = "answer"
|
|
|
DECISION = "decision"
|
|
|
ACTION = "action"
|
|
|
CHITCHAT = "chitchat"
|
|
|
|
|
|
class EmotionType(Enum):
|
|
|
"""نوع احساس"""
|
|
|
POSITIVE = "positive"
|
|
|
NEUTRAL = "neutral"
|
|
|
NEGATIVE = "negative"
|
|
|
EXCITED = "excited"
|
|
|
ANGRY = "angry"
|
|
|
CONFUSED = "confused"
|
|
|
|
|
|
@dataclass
|
|
|
class MessageNode:
|
|
|
"""گره پیام در گراف حافظه"""
|
|
|
id: str
|
|
|
content: str
|
|
|
role: str
|
|
|
timestamp: datetime
|
|
|
message_type: MessageType
|
|
|
importance_score: float = 0.5
|
|
|
emotion_score: Dict[EmotionType, float] = field(default_factory=dict)
|
|
|
tokens: int = 0
|
|
|
embeddings: Optional[np.ndarray] = None
|
|
|
metadata: Dict[str, Any] = field(default_factory=dict)
|
|
|
|
|
|
def __hash__(self):
|
|
|
return hash(self.id)
|
|
|
|
|
|
def __eq__(self, other):
|
|
|
return self.id == other.id
|
|
|
|
|
|
@dataclass
|
|
|
class MemoryConnection:
|
|
|
"""اتصال بین پیامها در گراف حافظه"""
|
|
|
source_id: str
|
|
|
target_id: str
|
|
|
connection_type: str
|
|
|
strength: float = 1.0
|
|
|
metadata: Dict[str, Any] = field(default_factory=dict)
|
|
|
|
|
|
@dataclass
|
|
|
class UserProfile:
|
|
|
"""پروفایل کاربر"""
|
|
|
user_id: int
|
|
|
personality_traits: Dict[str, float] = field(default_factory=dict)
|
|
|
interests: Set[str] = field(default_factory=set)
|
|
|
preferences: Dict[str, Any] = field(default_factory=dict)
|
|
|
conversation_style: str = "balanced"
|
|
|
knowledge_level: Dict[str, float] = field(default_factory=dict)
|
|
|
emotional_patterns: Dict[str, List[float]] = field(default_factory=dict)
|
|
|
learning_style: Optional[str] = None
|
|
|
|
|
|
def update_from_message(self, message: str, analysis: Dict[str, Any]):
|
|
|
"""بهروزرسانی پروفایل بر اساس پیام جدید"""
|
|
|
if 'personality_clues' in analysis:
|
|
|
for trait, score in analysis['personality_clues'].items():
|
|
|
current = self.personality_traits.get(trait, 0.5)
|
|
|
self.personality_traits[trait] = 0.7 * current + 0.3 * score
|
|
|
|
|
|
if 'interests' in analysis:
|
|
|
self.interests.update(analysis['interests'])
|
|
|
|
|
|
if 'preferences' in analysis:
|
|
|
self.preferences.update(analysis['preferences'])
|
|
|
|
|
|
|
|
|
|
|
|
class EmbeddingManager:
|
|
|
"""مدیریت embeddings برای جستجوی معنایی"""
|
|
|
|
|
|
def __init__(self, model_name: str = None):
|
|
|
self.model = None
|
|
|
self.vectorizer = None
|
|
|
self.use_sbert = HAS_SBERT
|
|
|
|
|
|
if self.use_sbert:
|
|
|
try:
|
|
|
self.model = SentenceTransformer('paraphrase-multilingual-MiniLM-L12-v2')
|
|
|
logger.info("Loaded multilingual sentence transformer")
|
|
|
except Exception as e:
|
|
|
logger.warning(f"Failed to load sentence transformer: {e}")
|
|
|
self.use_sbert = False
|
|
|
|
|
|
if not self.use_sbert:
|
|
|
try:
|
|
|
from sklearn.feature_extraction.text import TfidfVectorizer
|
|
|
self.vectorizer = TfidfVectorizer(max_features=1000)
|
|
|
logger.info("Using TF-IDF for embeddings")
|
|
|
except ImportError:
|
|
|
logger.warning("scikit-learn not available, using simple word vectors")
|
|
|
self.vectorizer = None
|
|
|
|
|
|
def get_embedding(self, text: str) -> np.ndarray:
|
|
|
"""دریافت embedding برای متن"""
|
|
|
if self.use_sbert and self.model:
|
|
|
return self.model.encode([text])[0]
|
|
|
elif self.vectorizer:
|
|
|
|
|
|
words = text.lower().split()
|
|
|
unique_words = list(set(words))
|
|
|
embedding = np.zeros(100)
|
|
|
for i, word in enumerate(unique_words[:100]):
|
|
|
embedding[i] = hash(word) % 100 / 100.0
|
|
|
return embedding
|
|
|
else:
|
|
|
|
|
|
text_lower = text.lower()
|
|
|
embedding = np.zeros(50)
|
|
|
|
|
|
embedding[0] = len(text) / 1000.0
|
|
|
embedding[1] = text.count('؟') / 5.0
|
|
|
embedding[2] = text.count('!') / 5.0
|
|
|
|
|
|
embedding[3] = 1.0 if 'چه' in text_lower else 0.0
|
|
|
embedding[4] = 1.0 if 'چرا' in text_lower else 0.0
|
|
|
embedding[5] = 1.0 if 'چگونه' in text_lower else 0.0
|
|
|
return embedding
|
|
|
|
|
|
def cosine_similarity(self, vec1: np.ndarray, vec2: np.ndarray) -> float:
|
|
|
"""محاسبه cosine similarity"""
|
|
|
norm1 = np.linalg.norm(vec1)
|
|
|
norm2 = np.linalg.norm(vec2)
|
|
|
if norm1 == 0 or norm2 == 0:
|
|
|
return 0.0
|
|
|
return np.dot(vec1, vec2) / (norm1 * norm2)
|
|
|
|
|
|
|
|
|
|
|
|
class MemoryGraph:
|
|
|
"""گراف حافظه برای ذخیره و بازیابی ارتباطات"""
|
|
|
|
|
|
def __init__(self):
|
|
|
self.nodes: Dict[str, MessageNode] = {}
|
|
|
self.connections: List[MemoryConnection] = []
|
|
|
self.adjacency: Dict[str, List[MemoryConnection]] = defaultdict(list)
|
|
|
self.topic_clusters: Dict[str, Set[str]] = defaultdict(set)
|
|
|
self.time_index: List[Tuple[datetime, str]] = []
|
|
|
|
|
|
def add_node(self, node: MessageNode):
|
|
|
"""افزودن گره جدید"""
|
|
|
self.nodes[node.id] = node
|
|
|
self.time_index.append((node.timestamp, node.id))
|
|
|
self.time_index.sort(key=lambda x: x[0])
|
|
|
|
|
|
def add_connection(self, connection: MemoryConnection):
|
|
|
"""افزودن اتصال جدید"""
|
|
|
self.connections.append(connection)
|
|
|
self.adjacency[connection.source_id].append(connection)
|
|
|
|
|
|
def find_similar_nodes(self, query_embedding: np.ndarray,
|
|
|
threshold: float = 0.7,
|
|
|
max_results: int = 5) -> List[Tuple[str, float]]:
|
|
|
"""یافتن گرههای مشابه"""
|
|
|
similarities = []
|
|
|
for node_id, node in self.nodes.items():
|
|
|
if node.embeddings is not None:
|
|
|
similarity = self._cosine_similarity(query_embedding, node.embeddings)
|
|
|
if similarity > threshold:
|
|
|
similarities.append((node_id, similarity))
|
|
|
|
|
|
similarities.sort(key=lambda x: x[1], reverse=True)
|
|
|
return similarities[:max_results]
|
|
|
|
|
|
def get_temporal_neighbors(self, node_id: str,
|
|
|
time_window: timedelta = timedelta(hours=24)) -> List[str]:
|
|
|
"""یافتن همسایههای زمانی"""
|
|
|
if node_id not in self.nodes:
|
|
|
return []
|
|
|
|
|
|
node_time = self.nodes[node_id].timestamp
|
|
|
neighbors = []
|
|
|
|
|
|
for timestamp, nid in self.time_index:
|
|
|
if nid != node_id and abs(timestamp - node_time) <= time_window:
|
|
|
neighbors.append(nid)
|
|
|
|
|
|
return neighbors
|
|
|
|
|
|
def get_semantic_cluster(self, node_id: str, min_similarity: float = 0.6) -> Set[str]:
|
|
|
"""دریافت خوشه معنایی یک گره"""
|
|
|
cluster = {node_id}
|
|
|
|
|
|
if node_id not in self.nodes or self.nodes[node_id].embeddings is None:
|
|
|
return cluster
|
|
|
|
|
|
query_embedding = self.nodes[node_id].embeddings
|
|
|
|
|
|
for other_id, other_node in self.nodes.items():
|
|
|
if other_id != node_id and other_node.embeddings is not None:
|
|
|
similarity = self._cosine_similarity(query_embedding, other_node.embeddings)
|
|
|
if similarity > min_similarity:
|
|
|
cluster.add(other_id)
|
|
|
|
|
|
return cluster
|
|
|
|
|
|
def _cosine_similarity(self, vec1: np.ndarray, vec2: np.ndarray) -> float:
|
|
|
"""محاسبه cosine similarity"""
|
|
|
norm1 = np.linalg.norm(vec1)
|
|
|
norm2 = np.linalg.norm(vec2)
|
|
|
if norm1 == 0 or norm2 == 0:
|
|
|
return 0.0
|
|
|
return np.dot(vec1, vec2) / (norm1 * norm2)
|
|
|
|
|
|
|
|
|
|
|
|
class ContextAnalyzer:
|
|
|
"""آنالیزور هوشمند context"""
|
|
|
|
|
|
def __init__(self):
|
|
|
self.keyword_patterns = {
|
|
|
'name': [r'نام\s+من\s+(.*?)(?:است|میباشد)', r'اسم\s+من\s+(.*?)'],
|
|
|
'age': [r'سن\s+من\s+(\d+)', r'(\d+)\s+سالمه'],
|
|
|
'location': [r'اهل\s+(.*?)\s+هستم', r'در\s+(.*?)\s+زندگی میکنم'],
|
|
|
'preference': [r'دوست\s+دارم\s+(.*?)', r'علاقه\s+دارم\s+(.*?)', r'ترجیح\s+میدهم\s+(.*?)'],
|
|
|
'dislike': [r'دوست\s+ندارم\s+(.*?)', r'مخالف\s+(.*?)\s+هستم'],
|
|
|
'goal': [r'میخواهم\s+(.*?)', r'هدف\s+من\s+(.*?)\s+است'],
|
|
|
'question': [r'(چرا|چگونه|چه|کجا|کی|آیا)\s+.*?\?'],
|
|
|
'decision': [r'تصمیم\s+گرفتم\s+(.*?)', r'قصد\s+دارم\s+(.*?)'],
|
|
|
'emotion': [r'احساس\s+(.*?)\s+دارم', r'(خوشحالم|ناراحتم|عصبانیام|خستهام)']
|
|
|
}
|
|
|
|
|
|
self.emotion_keywords = {
|
|
|
EmotionType.POSITIVE: ['خوب', 'عالی', 'خوشحال', 'عالی', 'ممنون', 'مرسی', 'دوست دارم', 'عالیه'],
|
|
|
EmotionType.NEGATIVE: ['بد', 'بدی', 'ناراحت', 'غمگین', 'عصبانی', 'مشکل', 'خطا', 'اشتباه'],
|
|
|
EmotionType.EXCITED: ['هیجانزده', 'هیجان', 'جالب', 'شگفتانگیز', 'عالی', 'وای'],
|
|
|
EmotionType.ANGRY: ['عصبانی', 'خشمگین', 'ناراحت', 'اعصاب', 'دیوانه'],
|
|
|
EmotionType.CONFUSED: ['سردرگم', 'گیج', 'نمیفهمم', 'مشکل دارم', 'کمک']
|
|
|
}
|
|
|
|
|
|
def analyze_message(self, text: str, role: str) -> Dict[str, Any]:
|
|
|
"""آنالیز پیام و استخراج اطلاعات"""
|
|
|
analysis = {
|
|
|
'type': self._detect_message_type(text, role),
|
|
|
'entities': self._extract_entities(text),
|
|
|
'keywords': self._extract_keywords(text),
|
|
|
'emotion': self._analyze_emotion(text),
|
|
|
'importance': self._calculate_importance(text, role),
|
|
|
'topics': self._extract_topics(text),
|
|
|
'intent': self._detect_intent(text),
|
|
|
'complexity': self._calculate_complexity(text),
|
|
|
'has_personal_info': False,
|
|
|
'personal_info': {}
|
|
|
}
|
|
|
|
|
|
|
|
|
personal_info = self._extract_personal_info(text)
|
|
|
if personal_info:
|
|
|
analysis['has_personal_info'] = True
|
|
|
analysis['personal_info'] = personal_info
|
|
|
|
|
|
|
|
|
preferences = self._extract_preferences(text)
|
|
|
if preferences:
|
|
|
analysis['preferences'] = preferences
|
|
|
|
|
|
return analysis
|
|
|
|
|
|
def _detect_message_type(self, text: str, role: str) -> MessageType:
|
|
|
"""تشخیص نوع پیام"""
|
|
|
text_lower = text.lower()
|
|
|
|
|
|
if role == 'user':
|
|
|
if any(q in text_lower for q in ['؟', '?', 'چرا', 'چگونه', 'چه', 'کجا']):
|
|
|
return MessageType.QUESTION
|
|
|
elif any(e in text_lower for e in ['احساس', 'حالم', 'خوشحالم', 'ناراحتم']):
|
|
|
return MessageType.EMOTION
|
|
|
elif any(d in text_lower for d in ['تصمیم', 'قصد', 'میخواهم']):
|
|
|
return MessageType.DECISION
|
|
|
elif any(p in text_lower for p in ['دوست دارم', 'علاقه', 'ترجیح']):
|
|
|
return MessageType.PREFERENCE
|
|
|
else:
|
|
|
return MessageType.FACT
|
|
|
else:
|
|
|
if '؟' in text_lower:
|
|
|
return MessageType.QUESTION
|
|
|
elif any(a in text_lower for a in ['پاسخ', 'جواب', 'راه حل']):
|
|
|
return MessageType.ANSWER
|
|
|
else:
|
|
|
return MessageType.FACT
|
|
|
|
|
|
def _extract_entities(self, text: str) -> List[str]:
|
|
|
"""استخراج موجودیتها"""
|
|
|
entities = []
|
|
|
|
|
|
name_patterns = [
|
|
|
r'نام\s+(?:من|او)\s+(.*?)(?:\s|$|\.|،)',
|
|
|
r'اسم\s+(?:من|او)\s+(.*?)(?:\s|$|\.|،)',
|
|
|
]
|
|
|
|
|
|
for pattern in name_patterns:
|
|
|
matches = re.findall(pattern, text)
|
|
|
entities.extend(matches)
|
|
|
|
|
|
return entities
|
|
|
|
|
|
def _extract_keywords(self, text: str) -> List[str]:
|
|
|
"""استخراج کلمات کلیدی"""
|
|
|
|
|
|
stopwords = {'و', 'در', 'با', 'به', 'از', 'که', 'این', 'آن', 'را', 'برای', 'اما', 'یا'}
|
|
|
words = re.findall(r'\b[\wآ-ی]+\b', text.lower())
|
|
|
keywords = [w for w in words if w not in stopwords and len(w) > 2]
|
|
|
|
|
|
return list(set(keywords))[:10]
|
|
|
|
|
|
def _analyze_emotion(self, text: str) -> Dict[EmotionType, float]:
|
|
|
"""تحلیل احساسات متن"""
|
|
|
emotion_scores = {e: 0.0 for e in EmotionType}
|
|
|
text_lower = text.lower()
|
|
|
|
|
|
for emotion_type, keywords in self.emotion_keywords.items():
|
|
|
score = 0.0
|
|
|
for keyword in keywords:
|
|
|
if keyword in text_lower:
|
|
|
score += 1.0
|
|
|
emotion_scores[emotion_type] = min(score / 3.0, 1.0)
|
|
|
|
|
|
|
|
|
if sum(emotion_scores.values()) == 0:
|
|
|
emotion_scores[EmotionType.NEUTRAL] = 1.0
|
|
|
|
|
|
return emotion_scores
|
|
|
|
|
|
def _calculate_importance(self, text: str, role: str) -> float:
|
|
|
"""محاسبه اهمیت پیام"""
|
|
|
score = 0.0
|
|
|
|
|
|
|
|
|
if role == 'user':
|
|
|
score += 0.3
|
|
|
|
|
|
|
|
|
length = len(text)
|
|
|
if length > 200:
|
|
|
score += 0.3
|
|
|
elif length > 100:
|
|
|
score += 0.2
|
|
|
elif length > 50:
|
|
|
score += 0.1
|
|
|
|
|
|
|
|
|
if '؟' in text or '?' in text:
|
|
|
score += 0.2
|
|
|
|
|
|
|
|
|
important_words = ['مهم', 'لطفا', 'فوری', 'ضروری', 'لازم', 'حتما']
|
|
|
for word in important_words:
|
|
|
if word in text.lower():
|
|
|
score += 0.2
|
|
|
break
|
|
|
|
|
|
|
|
|
if any(pattern in text for pattern in ['نام من', 'سن من', 'اهل']):
|
|
|
score += 0.3
|
|
|
|
|
|
return min(score, 1.0)
|
|
|
|
|
|
def _extract_topics(self, text: str) -> List[str]:
|
|
|
"""استخراج موضوعات"""
|
|
|
topics = []
|
|
|
|
|
|
|
|
|
topic_categories = {
|
|
|
'ورزش': ['فوتبال', 'ورزش', 'تمرین', 'مسابقه', 'تیم'],
|
|
|
'تکنولوژی': ['کامپیوتر', 'برنامه', 'کد', 'پایتون', 'هوش مصنوعی'],
|
|
|
'هنر': ['فیلم', 'موسیقی', 'نقاشی', 'کتاب', 'خواننده'],
|
|
|
'علم': ['تحقیق', 'دانش', 'کشف', 'آزمایش', 'نظریه'],
|
|
|
'غذا': ['غذا', 'رستوران', 'پخت', 'خوراک', 'ناهار'],
|
|
|
'سفر': ['سفر', 'مسافرت', 'کشور', 'شهر', 'هتل'],
|
|
|
'سلامتی': ['سلامت', 'بیماری', 'درمان', 'دکتر', 'بیمارستان'],
|
|
|
'کاری': ['کار', 'شغل', 'شرکت', 'مصاحبه', 'پروژه'],
|
|
|
'تحصیل': ['درس', 'دانشگاه', 'مدرسه', 'آموزش', 'یادگیری'],
|
|
|
}
|
|
|
|
|
|
text_lower = text.lower()
|
|
|
for topic, keywords in topic_categories.items():
|
|
|
if any(keyword in text_lower for keyword in keywords):
|
|
|
topics.append(topic)
|
|
|
|
|
|
return topics[:3]
|
|
|
|
|
|
def _detect_intent(self, text: str) -> str:
|
|
|
"""تشخیص قصد کاربر"""
|
|
|
text_lower = text.lower()
|
|
|
|
|
|
if any(q in text_lower for q in ['چطور', 'چگونه', 'راهنمایی']):
|
|
|
return 'guidance'
|
|
|
elif any(q in text_lower for q in ['چه', 'اطلاعات', 'معرفی']):
|
|
|
return 'information'
|
|
|
elif any(q in text_lower for q in ['چرا', 'دلیل', 'علت']):
|
|
|
return 'explanation'
|
|
|
elif any(q in text_lower for q in ['کمک', 'راه حل', 'مشکل']):
|
|
|
return 'help'
|
|
|
elif any(q in text_lower for q in ['توصیه', 'پیشنهاد', 'نظر']):
|
|
|
return 'advice'
|
|
|
elif any(q in text_lower for q in ['بحث', 'بحث کنیم', 'نظرت']):
|
|
|
return 'discussion'
|
|
|
else:
|
|
|
return 'general'
|
|
|
|
|
|
def _calculate_complexity(self, text: str) -> float:
|
|
|
"""محاسبه پیچیدگی متن"""
|
|
|
|
|
|
words = text.split()
|
|
|
if not words:
|
|
|
return 0.0
|
|
|
|
|
|
avg_word_length = sum(len(w) for w in words) / len(words)
|
|
|
|
|
|
|
|
|
sentences = re.split(r'[.!?]', text)
|
|
|
num_sentences = max(len(sentences), 1)
|
|
|
|
|
|
|
|
|
unique_words = len(set(words))
|
|
|
diversity = unique_words / len(words) if words else 0
|
|
|
|
|
|
complexity = (avg_word_length / 10.0 + diversity + min(len(words) / 50.0, 1.0)) / 3.0
|
|
|
|
|
|
return min(complexity, 1.0)
|
|
|
|
|
|
def _extract_personal_info(self, text: str) -> Dict[str, Any]:
|
|
|
"""استخراج اطلاعات شخصی"""
|
|
|
info = {}
|
|
|
|
|
|
for info_type, patterns in self.keyword_patterns.items():
|
|
|
for pattern in patterns:
|
|
|
matches = re.findall(pattern, text)
|
|
|
if matches:
|
|
|
info[info_type] = matches[0]
|
|
|
break
|
|
|
|
|
|
return info
|
|
|
|
|
|
def _extract_preferences(self, text: str) -> Dict[str, Any]:
|
|
|
"""استخراج ترجیحات"""
|
|
|
preferences = {}
|
|
|
text_lower = text.lower()
|
|
|
|
|
|
|
|
|
if 'دوست دارم' in text_lower:
|
|
|
parts = text_lower.split('دوست دارم')
|
|
|
if len(parts) > 1:
|
|
|
preference = parts[1].split('.')[0].strip()
|
|
|
if preference:
|
|
|
preferences['likes'] = preferences.get('likes', []) + [preference]
|
|
|
|
|
|
if 'دوست ندارم' in text_lower:
|
|
|
parts = text_lower.split('دوست ندارم')
|
|
|
if len(parts) > 1:
|
|
|
preference = parts[1].split('.')[0].strip()
|
|
|
if preference:
|
|
|
preferences['dislikes'] = preferences.get('dislikes', []) + [preference]
|
|
|
|
|
|
return preferences
|
|
|
|
|
|
|
|
|
|
|
|
class IntelligentContextManager:
|
|
|
"""مدیر هوشمند context"""
|
|
|
|
|
|
def __init__(self, user_id: int):
|
|
|
self.user_id = user_id
|
|
|
self.embedding_manager = EmbeddingManager()
|
|
|
self.analyzer = ContextAnalyzer()
|
|
|
self.memory_graph = MemoryGraph()
|
|
|
self.user_profile = UserProfile(user_id=user_id)
|
|
|
|
|
|
|
|
|
self.memory_layers = {
|
|
|
'ephemeral': deque(maxlen=20),
|
|
|
'working': deque(maxlen=50),
|
|
|
'recent': deque(maxlen=100),
|
|
|
'long_term': [],
|
|
|
'core': []
|
|
|
}
|
|
|
|
|
|
|
|
|
self.max_working_tokens = 512
|
|
|
self.max_context_tokens = 2048
|
|
|
self.min_importance_threshold = 0.3
|
|
|
self.semantic_similarity_threshold = 0.7
|
|
|
|
|
|
|
|
|
self.stats = {
|
|
|
'total_messages': 0,
|
|
|
'compressed_messages': 0,
|
|
|
'retrieved_memories': 0,
|
|
|
'profile_updates': 0,
|
|
|
'average_importance': 0.0
|
|
|
}
|
|
|
|
|
|
|
|
|
self._load_saved_data()
|
|
|
|
|
|
def _load_saved_data(self):
|
|
|
"""بارگذاری دادههای ذخیره شده"""
|
|
|
try:
|
|
|
|
|
|
user_data = data_manager.DATA['users'].get(str(self.user_id), {})
|
|
|
|
|
|
if 'smart_context' in user_data:
|
|
|
saved_data = user_data['smart_context']
|
|
|
|
|
|
|
|
|
if 'profile' in saved_data:
|
|
|
self.user_profile = UserProfile(**saved_data['profile'])
|
|
|
|
|
|
|
|
|
if 'stats' in saved_data:
|
|
|
self.stats.update(saved_data['stats'])
|
|
|
|
|
|
logger.info(f"Loaded smart context for user {self.user_id}")
|
|
|
except Exception as e:
|
|
|
logger.error(f"Error loading saved context data: {e}")
|
|
|
|
|
|
def _save_data(self):
|
|
|
"""ذخیره دادهها"""
|
|
|
try:
|
|
|
user_id_str = str(self.user_id)
|
|
|
if user_id_str not in data_manager.DATA['users']:
|
|
|
data_manager.DATA['users'][user_id_str] = {}
|
|
|
|
|
|
|
|
|
data_manager.DATA['users'][user_id_str]['smart_context'] = {
|
|
|
'profile': {
|
|
|
'user_id': self.user_profile.user_id,
|
|
|
'personality_traits': dict(self.user_profile.personality_traits),
|
|
|
'interests': list(self.user_profile.interests),
|
|
|
'preferences': dict(self.user_profile.preferences),
|
|
|
'conversation_style': self.user_profile.conversation_style,
|
|
|
'knowledge_level': dict(self.user_profile.knowledge_level),
|
|
|
'emotional_patterns': dict(self.user_profile.emotional_patterns),
|
|
|
'learning_style': self.user_profile.learning_style
|
|
|
},
|
|
|
'stats': self.stats,
|
|
|
'last_updated': datetime.now().isoformat()
|
|
|
}
|
|
|
|
|
|
data_manager.save_data()
|
|
|
except Exception as e:
|
|
|
logger.error(f"Error saving smart context data: {e}")
|
|
|
|
|
|
async def process_message(self, role: str, content: str) -> Dict[str, Any]:
|
|
|
"""پرداش کامل یک پیام جدید"""
|
|
|
start_time = datetime.now()
|
|
|
|
|
|
|
|
|
analysis = self.analyzer.analyze_message(content, role)
|
|
|
|
|
|
|
|
|
message_id = self._generate_message_id(content)
|
|
|
|
|
|
|
|
|
embedding_task = asyncio.create_task(
|
|
|
self._get_embedding_async(content)
|
|
|
)
|
|
|
|
|
|
node = MessageNode(
|
|
|
id=message_id,
|
|
|
content=content,
|
|
|
role=role,
|
|
|
timestamp=datetime.now(),
|
|
|
message_type=analysis['type'],
|
|
|
importance_score=analysis['importance'],
|
|
|
emotion_score=analysis['emotion'],
|
|
|
tokens=data_manager.count_tokens(content),
|
|
|
embeddings=None,
|
|
|
metadata={
|
|
|
'analysis': analysis,
|
|
|
'topics': analysis['topics'],
|
|
|
'intent': analysis['intent'],
|
|
|
'complexity': analysis['complexity']
|
|
|
}
|
|
|
)
|
|
|
|
|
|
|
|
|
try:
|
|
|
node.embeddings = await asyncio.wait_for(embedding_task, timeout=2.0)
|
|
|
except asyncio.TimeoutError:
|
|
|
logger.warning(f"Embedding generation timeout for message {message_id}")
|
|
|
node.embeddings = self.embedding_manager.get_embedding(content)
|
|
|
|
|
|
|
|
|
await asyncio.to_thread(self._add_to_memory_layers, node, analysis)
|
|
|
await asyncio.to_thread(self.memory_graph.add_node, node)
|
|
|
|
|
|
|
|
|
await asyncio.to_thread(self._create_memory_connections, node)
|
|
|
|
|
|
|
|
|
if role == 'user':
|
|
|
await asyncio.to_thread(self._update_user_profile, content, analysis)
|
|
|
|
|
|
|
|
|
await asyncio.to_thread(self._optimize_memory)
|
|
|
|
|
|
|
|
|
self.stats['total_messages'] += 1
|
|
|
self.stats['average_importance'] = (
|
|
|
self.stats['average_importance'] * (self.stats['total_messages'] - 1) +
|
|
|
analysis['importance']
|
|
|
) / self.stats['total_messages']
|
|
|
|
|
|
|
|
|
await asyncio.to_thread(self._save_data)
|
|
|
|
|
|
processing_time = (datetime.now() - start_time).total_seconds()
|
|
|
logger.info(f"Processed message {message_id} in {processing_time:.2f}s, importance: {analysis['importance']:.2f}")
|
|
|
|
|
|
return {
|
|
|
'node_id': message_id,
|
|
|
'analysis': analysis,
|
|
|
'processing_time': processing_time
|
|
|
}
|
|
|
|
|
|
async def _get_embedding_async(self, text: str) -> np.ndarray:
|
|
|
"""دریافت embedding به صورت async"""
|
|
|
loop = asyncio.get_event_loop()
|
|
|
return await loop.run_in_executor(
|
|
|
None,
|
|
|
self.embedding_manager.get_embedding,
|
|
|
text
|
|
|
)
|
|
|
|
|
|
def _generate_message_id(self, content: str) -> str:
|
|
|
"""تولید شناسه منحصر به فرد برای پیام"""
|
|
|
timestamp = datetime.now().strftime('%Y%m%d%H%M%S%f')
|
|
|
content_hash = hashlib.md5(content.encode()).hexdigest()[:8]
|
|
|
return f"{self.user_id}_{timestamp}_{content_hash}"
|
|
|
|
|
|
def _add_to_memory_layers(self, node: MessageNode, analysis: Dict[str, Any]):
|
|
|
"""افزودن پیام به لایههای حافظه مناسب"""
|
|
|
|
|
|
|
|
|
self.memory_layers['ephemeral'].append(node)
|
|
|
self.memory_layers['working'].append(node)
|
|
|
|
|
|
|
|
|
if analysis['importance'] > 0.2:
|
|
|
self.memory_layers['recent'].append(node)
|
|
|
|
|
|
|
|
|
if analysis['importance'] > self.min_importance_threshold:
|
|
|
self.memory_layers['long_term'].append(node)
|
|
|
|
|
|
|
|
|
if analysis.get('has_personal_info', False) or analysis['importance'] > 0.8:
|
|
|
core_entry = {
|
|
|
'node': node,
|
|
|
'info': analysis.get('personal_info', {}),
|
|
|
'timestamp': datetime.now()
|
|
|
}
|
|
|
self.memory_layers['core'].append(core_entry)
|
|
|
|
|
|
def _create_memory_connections(self, node: MessageNode):
|
|
|
"""ایجاد ارتباطات حافظه برای گره جدید"""
|
|
|
|
|
|
|
|
|
if len(self.memory_graph.nodes) > 1:
|
|
|
|
|
|
last_nodes = list(self.memory_graph.nodes.values())[-5:]
|
|
|
for last_node in last_nodes:
|
|
|
if last_node.id != node.id:
|
|
|
temporal_conn = MemoryConnection(
|
|
|
source_id=last_node.id,
|
|
|
target_id=node.id,
|
|
|
connection_type='temporal',
|
|
|
strength=0.8
|
|
|
)
|
|
|
self.memory_graph.add_connection(temporal_conn)
|
|
|
|
|
|
|
|
|
if node.embeddings is not None:
|
|
|
similar_nodes = self.memory_graph.find_similar_nodes(
|
|
|
node.embeddings,
|
|
|
threshold=self.semantic_similarity_threshold
|
|
|
)
|
|
|
|
|
|
for similar_id, similarity in similar_nodes:
|
|
|
semantic_conn = MemoryConnection(
|
|
|
source_id=node.id,
|
|
|
target_id=similar_id,
|
|
|
connection_type='semantic',
|
|
|
strength=similarity
|
|
|
)
|
|
|
self.memory_graph.add_connection(semantic_conn)
|
|
|
|
|
|
def _update_user_profile(self, content: str, analysis: Dict[str, Any]):
|
|
|
"""بهروزرسانی پروفایل کاربر"""
|
|
|
|
|
|
|
|
|
if 'topics' in analysis:
|
|
|
for topic in analysis['topics']:
|
|
|
self.user_profile.interests.add(topic)
|
|
|
|
|
|
|
|
|
if 'preferences' in analysis:
|
|
|
self.user_profile.preferences.update(analysis['preferences'])
|
|
|
|
|
|
|
|
|
complexity = analysis.get('complexity', 0.5)
|
|
|
if complexity > 0.7:
|
|
|
self.user_profile.conversation_style = "detailed"
|
|
|
elif complexity < 0.3:
|
|
|
self.user_profile.conversation_style = "concise"
|
|
|
|
|
|
|
|
|
emotion = analysis.get('emotion', {})
|
|
|
for emotion_type, score in emotion.items():
|
|
|
if score > 0.3:
|
|
|
emotion_name = emotion_type.value if hasattr(emotion_type, 'value') else str(emotion_type)
|
|
|
if emotion_name not in self.user_profile.emotional_patterns:
|
|
|
self.user_profile.emotional_patterns[emotion_name] = []
|
|
|
self.user_profile.emotional_patterns[emotion_name].append(score)
|
|
|
|
|
|
self.stats['profile_updates'] += 1
|
|
|
|
|
|
def _optimize_memory(self):
|
|
|
"""بهینهسازی و فشردهسازی حافظه"""
|
|
|
|
|
|
|
|
|
working_tokens = sum(node.tokens for node in self.memory_layers['working'])
|
|
|
if working_tokens > self.max_working_tokens:
|
|
|
self._compress_working_memory()
|
|
|
self.stats['compressed_messages'] += 1
|
|
|
|
|
|
|
|
|
if len(self.memory_layers['ephemeral']) > 50:
|
|
|
self.memory_layers['ephemeral'].clear()
|
|
|
|
|
|
|
|
|
self.memory_layers['long_term'].sort(key=lambda x: x.importance_score, reverse=True)
|
|
|
if len(self.memory_layers['long_term']) > 100:
|
|
|
self.memory_layers['long_term'] = self.memory_layers['long_term'][:100]
|
|
|
|
|
|
def _compress_working_memory(self):
|
|
|
"""فشردهسازی هوشمند حافظه فعال"""
|
|
|
if len(self.memory_layers['working']) <= 10:
|
|
|
return
|
|
|
|
|
|
working_memory = list(self.memory_layers['working'])
|
|
|
|
|
|
|
|
|
scored_messages = []
|
|
|
for i, node in enumerate(working_memory):
|
|
|
score = self._calculate_compression_score(node, i, len(working_memory))
|
|
|
scored_messages.append((score, node))
|
|
|
|
|
|
|
|
|
scored_messages.sort(key=lambda x: x[0], reverse=True)
|
|
|
|
|
|
|
|
|
compressed = []
|
|
|
total_tokens = 0
|
|
|
|
|
|
for score, node in scored_messages:
|
|
|
if total_tokens + node.tokens <= self.max_working_tokens:
|
|
|
compressed.append(node)
|
|
|
total_tokens += node.tokens
|
|
|
else:
|
|
|
break
|
|
|
|
|
|
|
|
|
compressed.sort(key=lambda x: x.timestamp)
|
|
|
|
|
|
|
|
|
self.memory_layers['working'] = deque(compressed, maxlen=50)
|
|
|
|
|
|
logger.info(f"Compressed working memory: {len(working_memory)} -> {len(compressed)} messages")
|
|
|
|
|
|
def _calculate_compression_score(self, node: MessageNode, index: int, total: int) -> float:
|
|
|
"""محاسبه امتیاز فشردهسازی برای یک پیام"""
|
|
|
score = 0.0
|
|
|
|
|
|
|
|
|
score += node.importance_score * 0.4
|
|
|
|
|
|
|
|
|
recency = (index / total) * 0.3
|
|
|
score += recency
|
|
|
|
|
|
|
|
|
topics = node.metadata.get('topics', [])
|
|
|
topic_diversity = min(len(topics) / 3.0, 0.2)
|
|
|
score += topic_diversity
|
|
|
|
|
|
|
|
|
type_weights = {
|
|
|
MessageType.QUESTION: 0.2,
|
|
|
MessageType.PREFERENCE: 0.2,
|
|
|
MessageType.DECISION: 0.2,
|
|
|
MessageType.EMOTION: 0.1,
|
|
|
MessageType.FACT: 0.1,
|
|
|
MessageType.ANSWER: 0.1,
|
|
|
MessageType.ACTION: 0.2,
|
|
|
MessageType.CHITCHAT: 0.05
|
|
|
}
|
|
|
score += type_weights.get(node.message_type, 0.1)
|
|
|
|
|
|
|
|
|
if 'has_personal_info' in node.metadata.get('analysis', {}):
|
|
|
if node.metadata['analysis']['has_personal_info']:
|
|
|
score += 0.2
|
|
|
|
|
|
return score
|
|
|
|
|
|
async def retrieve_context(self, query: str, max_tokens: int = None) -> List[Dict[str, Any]]:
|
|
|
"""بازیابی هوشمند context مرتبط با query"""
|
|
|
try:
|
|
|
if max_tokens is None:
|
|
|
max_tokens = self.max_context_tokens
|
|
|
|
|
|
start_time = datetime.now()
|
|
|
|
|
|
|
|
|
try:
|
|
|
embedding_task = asyncio.create_task(
|
|
|
self._get_embedding_async(query)
|
|
|
)
|
|
|
query_embedding = await asyncio.wait_for(embedding_task, timeout=3.0)
|
|
|
except asyncio.TimeoutError:
|
|
|
logger.warning(f"Embedding timeout for query: {query[:50]}")
|
|
|
query_embedding = self.embedding_manager.get_embedding(query)
|
|
|
|
|
|
|
|
|
tasks = []
|
|
|
|
|
|
|
|
|
tasks.append(asyncio.create_task(
|
|
|
asyncio.to_thread(self._retrieve_from_working_memory)
|
|
|
))
|
|
|
|
|
|
|
|
|
tasks.append(self._retrieve_semantic_memories(query_embedding, 'recent'))
|
|
|
tasks.append(self._retrieve_semantic_memories(query_embedding, 'long_term'))
|
|
|
|
|
|
|
|
|
tasks.append(asyncio.create_task(
|
|
|
asyncio.to_thread(self._retrieve_core_memories, query)
|
|
|
))
|
|
|
|
|
|
|
|
|
results = await asyncio.gather(*tasks, return_exceptions=True)
|
|
|
|
|
|
|
|
|
retrieved_memories = []
|
|
|
for result in results:
|
|
|
if isinstance(result, Exception):
|
|
|
logger.error(f"Error retrieving memory: {result}")
|
|
|
continue
|
|
|
retrieved_memories.extend(result)
|
|
|
|
|
|
except Exception as e:
|
|
|
logger.error(f"Error in retrieve_context: {e}")
|
|
|
|
|
|
return self._retrieve_from_working_memory()
|
|
|
|
|
|
|
|
|
query_embedding = self.embedding_manager.get_embedding(query)
|
|
|
|
|
|
|
|
|
retrieved_memories = []
|
|
|
|
|
|
|
|
|
retrieved_memories.extend(self._retrieve_from_working_memory())
|
|
|
|
|
|
|
|
|
recent_memories = await self._retrieve_semantic_memories(query_embedding, 'recent')
|
|
|
retrieved_memories.extend(recent_memories)
|
|
|
|
|
|
|
|
|
long_term_memories = await self._retrieve_semantic_memories(query_embedding, 'long_term')
|
|
|
retrieved_memories.extend(long_term_memories)
|
|
|
|
|
|
|
|
|
core_memories = self._retrieve_core_memories(query)
|
|
|
retrieved_memories.extend(core_memories)
|
|
|
|
|
|
|
|
|
unique_memories = self._deduplicate_memories(retrieved_memories)
|
|
|
prioritized_memories = self._prioritize_memories(unique_memories, query_embedding)
|
|
|
|
|
|
|
|
|
final_context = []
|
|
|
total_tokens = 0
|
|
|
|
|
|
for memory in prioritized_memories:
|
|
|
memory_tokens = memory['node'].tokens if 'node' in memory else 50
|
|
|
|
|
|
if total_tokens + memory_tokens <= max_tokens:
|
|
|
final_context.append(memory)
|
|
|
total_tokens += memory_tokens
|
|
|
else:
|
|
|
break
|
|
|
|
|
|
|
|
|
self.stats['retrieved_memories'] += len(final_context)
|
|
|
|
|
|
retrieval_time = (datetime.now() - start_time).total_seconds()
|
|
|
logger.info(f"Retrieved {len(final_context)} memories in {retrieval_time:.2f}s")
|
|
|
|
|
|
return final_context
|
|
|
|
|
|
|
|
|
def _retrieve_from_working_memory(self) -> List[Dict[str, Any]]:
|
|
|
"""بازیابی از حافظه فعال"""
|
|
|
memories = []
|
|
|
|
|
|
for node in list(self.memory_layers['working'])[-10:]:
|
|
|
memories.append({
|
|
|
'node': node,
|
|
|
'source': 'working',
|
|
|
'relevance': 1.0,
|
|
|
'recency': 1.0
|
|
|
})
|
|
|
|
|
|
return memories
|
|
|
|
|
|
async def _retrieve_semantic_memories(self, query_embedding: np.ndarray,
|
|
|
layer: str) -> List[Dict[str, Any]]:
|
|
|
"""بازیابی حافظههای معنایی"""
|
|
|
memories = []
|
|
|
|
|
|
if layer not in self.memory_layers:
|
|
|
return memories
|
|
|
|
|
|
layer_memories = self.memory_layers[layer]
|
|
|
|
|
|
for item in layer_memories:
|
|
|
node = item if hasattr(item, 'embeddings') else item['node'] if isinstance(item, dict) else None
|
|
|
|
|
|
if node and node.embeddings is not None:
|
|
|
similarity = self.embedding_manager.cosine_similarity(
|
|
|
query_embedding, node.embeddings
|
|
|
)
|
|
|
|
|
|
if similarity > self.semantic_similarity_threshold:
|
|
|
recency_weight = 1.0 if layer == 'working' else 0.7
|
|
|
|
|
|
memories.append({
|
|
|
'node': node,
|
|
|
'source': layer,
|
|
|
'relevance': similarity,
|
|
|
'recency': recency_weight,
|
|
|
'importance': node.importance_score
|
|
|
})
|
|
|
|
|
|
return memories
|
|
|
|
|
|
def _retrieve_core_memories(self, query: str) -> List[Dict[str, Any]]:
|
|
|
"""بازیابی حافظههای هسته"""
|
|
|
memories = []
|
|
|
query_lower = query.lower()
|
|
|
|
|
|
for core_entry in self.memory_layers['core']:
|
|
|
node = core_entry['node']
|
|
|
info = core_entry.get('info', {})
|
|
|
|
|
|
|
|
|
relevance = 0.0
|
|
|
|
|
|
|
|
|
for key, value in info.items():
|
|
|
if isinstance(value, str) and value.lower() in query_lower:
|
|
|
relevance = 0.9
|
|
|
break
|
|
|
|
|
|
|
|
|
if relevance == 0.0 and node.content.lower() in query_lower:
|
|
|
relevance = 0.7
|
|
|
|
|
|
if relevance > 0.5:
|
|
|
memories.append({
|
|
|
'node': node,
|
|
|
'source': 'core',
|
|
|
'relevance': relevance,
|
|
|
'recency': 0.8,
|
|
|
'importance': 1.0,
|
|
|
'personal_info': info
|
|
|
})
|
|
|
|
|
|
return memories
|
|
|
|
|
|
def _deduplicate_memories(self, memories: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
|
|
|
"""حذف حافظههای تکراری"""
|
|
|
seen_ids = set()
|
|
|
unique_memories = []
|
|
|
|
|
|
for memory in memories:
|
|
|
node = memory.get('node')
|
|
|
if node and node.id not in seen_ids:
|
|
|
seen_ids.add(node.id)
|
|
|
unique_memories.append(memory)
|
|
|
|
|
|
return unique_memories
|
|
|
|
|
|
def _prioritize_memories(self, memories: List[Dict[str, Any]],
|
|
|
query_embedding: np.ndarray) -> List[Dict[str, Any]]:
|
|
|
"""اولویتبندی حافظههای بازیابی شده"""
|
|
|
|
|
|
def calculate_priority(memory: Dict[str, Any]) -> float:
|
|
|
"""محاسبه اولویت برای یک حافظه"""
|
|
|
priority = 0.0
|
|
|
|
|
|
|
|
|
priority += memory.get('relevance', 0.0) * 0.4
|
|
|
|
|
|
|
|
|
priority += memory.get('recency', 0.0) * 0.3
|
|
|
|
|
|
|
|
|
priority += memory.get('importance', 0.5) * 0.2
|
|
|
|
|
|
|
|
|
source = memory.get('source', '')
|
|
|
if source == 'core':
|
|
|
priority += 0.2
|
|
|
elif source == 'long_term':
|
|
|
priority += 0.1
|
|
|
|
|
|
return priority
|
|
|
|
|
|
|
|
|
prioritized = []
|
|
|
for memory in memories:
|
|
|
priority = calculate_priority(memory)
|
|
|
prioritized.append((priority, memory))
|
|
|
|
|
|
prioritized.sort(key=lambda x: x[0], reverse=True)
|
|
|
|
|
|
return [memory for _, memory in prioritized]
|
|
|
|
|
|
async def get_context_for_api(self, query: str = None) -> List[Dict[str, Any]]:
|
|
|
"""تهیه context برای ارسال به API"""
|
|
|
|
|
|
|
|
|
if query:
|
|
|
retrieved = await self.retrieve_context(query)
|
|
|
|
|
|
|
|
|
api_messages = []
|
|
|
|
|
|
|
|
|
api_messages.append({
|
|
|
'role': 'system',
|
|
|
'content': f"User profile: {self._format_user_profile()}"
|
|
|
})
|
|
|
|
|
|
|
|
|
for memory in retrieved:
|
|
|
node = memory['node']
|
|
|
api_messages.append({
|
|
|
'role': node.role,
|
|
|
'content': node.content
|
|
|
})
|
|
|
|
|
|
return api_messages
|
|
|
|
|
|
else:
|
|
|
|
|
|
api_messages = []
|
|
|
|
|
|
for node in list(self.memory_layers['working'])[-6:]:
|
|
|
api_messages.append({
|
|
|
'role': node.role,
|
|
|
'content': node.content
|
|
|
})
|
|
|
|
|
|
return api_messages
|
|
|
|
|
|
def _format_user_profile(self) -> str:
|
|
|
"""قالببندی پروفایل کاربر برای سیستم"""
|
|
|
profile_parts = []
|
|
|
|
|
|
if self.user_profile.interests:
|
|
|
interests_str = ', '.join(list(self.user_profile.interests)[:5])
|
|
|
profile_parts.append(f"Interests: {interests_str}")
|
|
|
|
|
|
if self.user_profile.preferences:
|
|
|
prefs = list(self.user_profile.preferences.items())[:3]
|
|
|
prefs_str = ', '.join(f"{k}: {v}" for k, v in prefs)
|
|
|
profile_parts.append(f"Preferences: {prefs_str}")
|
|
|
|
|
|
if self.user_profile.conversation_style:
|
|
|
profile_parts.append(f"Conversation style: {self.user_profile.conversation_style}")
|
|
|
|
|
|
if self.user_profile.learning_style:
|
|
|
profile_parts.append(f"Learning style: {self.user_profile.learning_style}")
|
|
|
|
|
|
if profile_parts:
|
|
|
return ' | '.join(profile_parts)
|
|
|
else:
|
|
|
return "New user, minimal profile information available."
|
|
|
|
|
|
def get_summary(self) -> Dict[str, Any]:
|
|
|
"""دریافت خلاصه وضعیت"""
|
|
|
return {
|
|
|
'user_id': self.user_id,
|
|
|
'total_messages': self.stats['total_messages'],
|
|
|
'working_memory': len(self.memory_layers['working']),
|
|
|
'recent_memory': len(self.memory_layers['recent']),
|
|
|
'long_term_memory': len(self.memory_layers['long_term']),
|
|
|
'core_memory': len(self.memory_layers['core']),
|
|
|
'profile_interests': list(self.user_profile.interests)[:10],
|
|
|
'average_importance': self.stats['average_importance'],
|
|
|
'compression_ratio': self.stats.get('compressed_messages', 0) / max(self.stats['total_messages'], 1),
|
|
|
'retrieval_efficiency': self.stats.get('retrieved_memories', 0) / max(self.stats['total_messages'], 1)
|
|
|
}
|
|
|
|
|
|
def clear_context(self):
|
|
|
"""پاک کردن context کاربر"""
|
|
|
self.memory_layers['working'].clear()
|
|
|
self.memory_layers['ephemeral'].clear()
|
|
|
|
|
|
|
|
|
logger.info(f"Cleared context for user {self.user_id}")
|
|
|
|
|
|
def export_debug_info(self) -> Dict[str, Any]:
|
|
|
"""دریافت اطلاعات دیباگ"""
|
|
|
return {
|
|
|
'memory_graph_size': len(self.memory_graph.nodes),
|
|
|
'memory_graph_connections': len(self.memory_graph.connections),
|
|
|
'user_profile': {
|
|
|
'interests_count': len(self.user_profile.interests),
|
|
|
'preferences_count': len(self.user_profile.preferences),
|
|
|
'personality_traits': dict(self.user_profile.personality_traits)
|
|
|
},
|
|
|
'layer_sizes': {k: len(v) for k, v in self.memory_layers.items()},
|
|
|
'stats': dict(self.stats)
|
|
|
} |