File size: 9,543 Bytes
422e76f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
"""

Advanced caching system with semantic similarity and persistence

"""
import pickle
import hashlib
import time
from typing import Dict, Any, Optional, Tuple
from dataclasses import dataclass
import numpy as np
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import os
import logging
from threading import Lock

logger = logging.getLogger(__name__)

@dataclass
class CacheEntry:
    result: Dict[str, Any]
    timestamp: float
    access_count: int
    semantic_vector: Optional[np.ndarray] = None

class AdvancedCache:
    def __init__(self, max_size: int = 1000, ttl: int = 3600, similarity_threshold: float = 0.995):
        self.max_size = max_size
        self.ttl = ttl
        self.similarity_threshold = similarity_threshold  # Increased from 0.99 to 0.995 for more precise matching
        self.cache: Dict[str, CacheEntry] = {}
        self.access_times: Dict[str, float] = {}
        self.lock = Lock()
        
        # For semantic similarity
        self.vectorizer = TfidfVectorizer(max_features=1000, stop_words='english')
        self.is_vectorizer_fitted = False
        self.cache_file = "cache_persistent.pkl"
        
        # Load persistent cache
        self._load_cache()
        
        # Stats
        self.hits = 0
        self.misses = 0
        self.semantic_hits = 0

    def _load_cache(self):
        """Load cache from disk"""
        try:
            if os.path.exists(self.cache_file):
                with open(self.cache_file, 'rb') as f:
                    data = pickle.load(f)
                    self.cache = data.get('cache', {})
                    self.access_times = data.get('access_times', {})
                    if 'vectorizer' in data and data['vectorizer'] is not None:
                        self.vectorizer = data['vectorizer']
                        self.is_vectorizer_fitted = True
                logger.info(f"Loaded {len(self.cache)} entries from persistent cache")
        except Exception as e:
            logger.warning(f"Failed to load persistent cache: {e}")

    def _save_cache(self):
        """Save cache to disk"""
        try:
            data = {
                'cache': self.cache,
                'access_times': self.access_times,
                'vectorizer': self.vectorizer if self.is_vectorizer_fitted else None
            }
            with open(self.cache_file, 'wb') as f:
                pickle.dump(data, f)
        except Exception as e:
            logger.warning(f"Failed to save persistent cache: {e}")

    def _generate_key(self, prompt: str, response: str, question: str = "") -> str:
        """Generate cache key"""
        combined = f"{prompt}|{response}|{question}".lower().strip()
        return hashlib.sha256(combined.encode()).hexdigest()

    def _create_semantic_vector(self, prompt: str, response: str, question: str = "") -> np.ndarray:
        """Create semantic vector for similarity comparison"""
        combined_text = f"{prompt} {response} {question}"
        
        if not self.is_vectorizer_fitted:
            # Fit vectorizer with current text (bootstrap)
            self.vectorizer.fit([combined_text])
            self.is_vectorizer_fitted = True
        
        try:
            vector = self.vectorizer.transform([combined_text])
            return vector.toarray()[0]
        except Exception:
            # If transform fails, refit with all available texts
            all_texts = [combined_text]
            for entry in self.cache.values():
                if hasattr(entry, 'semantic_vector') and entry.semantic_vector is not None:
                    all_texts.append("dummy")  # Placeholder
            
            self.vectorizer.fit(all_texts)
            vector = self.vectorizer.transform([combined_text])
            return vector.toarray()[0]

    def _find_similar_entry(self, prompt: str, response: str, question: str = "") -> Optional[Tuple[str, CacheEntry]]:
        """Find semantically similar cache entry"""
        if len(self.cache) == 0:
            return None
        
        try:
            query_vector = self._create_semantic_vector(prompt, response, question)
            
            best_similarity = 0
            best_entry = None
            best_key = None
            
            for key, entry in self.cache.items():
                if entry.semantic_vector is None:
                    continue
                
                similarity = cosine_similarity([query_vector], [entry.semantic_vector])[0][0]
                if similarity > best_similarity and similarity >= self.similarity_threshold:
                    best_similarity = similarity
                    best_entry = entry
                    best_key = key
            
            if best_entry:
                logger.debug(f"Found similar entry with {best_similarity:.3f} similarity")
                return best_key, best_entry
            
        except Exception as e:
            logger.warning(f"Semantic similarity search failed: {e}")
        
        return None

    def get(self, prompt: str, response: str, question: str = "") -> Optional[Dict[str, Any]]:
        """Get cached result with semantic similarity fallback"""
        with self.lock:
            key = self._generate_key(prompt, response, question)
            current_time = time.time()
            
            # Direct cache hit
            if key in self.cache:
                entry = self.cache[key]
                if current_time - entry.timestamp <= self.ttl:
                    entry.access_count += 1
                    self.access_times[key] = current_time
                    self.hits += 1
                    logger.debug(f"Cache hit for key: {key[:8]}...")
                    return entry.result
                else:
                    # Expired entry
                    del self.cache[key]
                    if key in self.access_times:
                        del self.access_times[key]
            
            # Semantic similarity search
            similar_result = self._find_similar_entry(prompt, response, question)
            if similar_result:
                similar_key, similar_entry = similar_result
                # Update access info for similar entry
                similar_entry.access_count += 1
                self.access_times[similar_key] = current_time
                self.semantic_hits += 1
                logger.debug(f"Semantic cache hit for key: {similar_key[:8]}...")
                return similar_entry.result
            
            self.misses += 1
            return None

    def set(self, prompt: str, response: str, question: str, result: Dict[str, Any]):
        """Cache result with semantic vector"""
        with self.lock:
            key = self._generate_key(prompt, response, question)
            current_time = time.time()
            
            # Create semantic vector
            semantic_vector = self._create_semantic_vector(prompt, response, question)
            
            # Create cache entry
            entry = CacheEntry(
                result=result,
                timestamp=current_time,
                access_count=1,
                semantic_vector=semantic_vector
            )
            
            # Check if we need to evict entries
            if len(self.cache) >= self.max_size:
                self._evict_entries()
            
            self.cache[key] = entry
            self.access_times[key] = current_time
            
            # Periodically save to disk
            if len(self.cache) % 10 == 0:
                self._save_cache()

    def _evict_entries(self):
        """Evict least recently used entries"""
        if not self.cache:
            return
        
        # Sort by access time and remove oldest 20%
        sorted_keys = sorted(self.access_times.keys(), key=lambda k: self.access_times[k])
        evict_count = max(1, len(sorted_keys) // 5)
        
        for key in sorted_keys[:evict_count]:
            if key in self.cache:
                del self.cache[key]
            if key in self.access_times:
                del self.access_times[key]
        
        logger.info(f"Evicted {evict_count} cache entries")

    def get_stats(self) -> Dict[str, Any]:
        """Get cache statistics"""
        total_requests = self.hits + self.misses
        hit_rate = (self.hits / total_requests * 100) if total_requests > 0 else 0
        semantic_hit_rate = (self.semantic_hits / total_requests * 100) if total_requests > 0 else 0
        
        return {
            "total_entries": len(self.cache),
            "hits": self.hits,
            "misses": self.misses,
            "semantic_hits": self.semantic_hits,
            "hit_rate": hit_rate,
            "semantic_hit_rate": semantic_hit_rate,
            "total_requests": total_requests
        }

    def clear(self):
        """Clear all cache entries"""
        with self.lock:
            self.cache.clear()
            self.access_times.clear()
            if os.path.exists(self.cache_file):
                os.remove(self.cache_file)

    def __del__(self):
        """Save cache when object is destroyed"""
        self._save_cache()

# Global advanced cache instance
advanced_cache = AdvancedCache()