File size: 8,182 Bytes
0c2d7d2 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 |
"""
OmniCoreX Memory Manager
Advanced memory and context management utility to handle short-term and long-term knowledge retention,
supporting adaptive reasoning and continual learning within the OmniCoreX AI system.
Features:
- Dual memory layers: short-term working memory and long-term persistent memory.
- Efficient retrieval with similarity search (vector-based) for context relevance.
- Context window management for input conditioning.
- Supports updating, forgetting, and pruning memory entries.
- Interfaces for integration with external vector stores or databases.
"""
import time
from typing import Any, Dict, List, Optional, Tuple
import numpy as np
from collections import deque
from threading import RLock
class MemoryEntry:
"""
Represents a single memory record with timestamp, content and optional embedding vector.
"""
def __init__(self, content: Any, embedding: Optional[np.ndarray] = None):
self.content = content
self.embedding = embedding
self.timestamp = time.time()
def __repr__(self):
return f"MemoryEntry timestamp={self.timestamp} content={self.content}>"
class ShortTermMemory:
"""
Short-term working memory implemented as a fixed-size deque for recent context.
"""
def __init__(self, max_size: int = 50):
self.max_size = max_size
self.memory = deque(maxlen=max_size)
self.lock = RLock()
def add(self, content: Any, embedding: Optional[np.ndarray] = None):
with self.lock:
entry = MemoryEntry(content, embedding)
self.memory.append(entry)
def get_recent(self, n: Optional[int] = None) -> List[MemoryEntry]:
with self.lock:
if n is None or n > len(self.memory):
return list(self.memory)
else:
return list(self.memory)[-n:]
def clear(self):
with self.lock:
self.memory.clear()
class LongTermMemory:
"""
Long-term memory stores past knowledge persistently.
This simple implementation stores entries with embeddings for similarity search.
"""
def __init__(self):
self.entries: List[MemoryEntry] = []
self.lock = RLock()
def add(self, content: Any, embedding: Optional[np.ndarray] = None):
with self.lock:
entry = MemoryEntry(content, embedding)
self.entries.append(entry)
def retrieve_similar(self, query_embedding: np.ndarray, top_k: int = 5) -> List[MemoryEntry]:
"""
Retrieve top-k memory entries most similar to the query embedding using cosine similarity.
Args:
query_embedding: numpy array vector.
top_k: Number of top similar entries to return.
Returns:
List of MemoryEntry objects.
"""
with self.lock:
if not self.entries or query_embedding is None:
return []
embeddings = np.array([e.embedding for e in self.entries if e.embedding is not None])
if embeddings.size == 0:
return []
# Normalize embeddings
norm_embeddings = embeddings / np.linalg.norm(embeddings, axis=1, keepdims=True)
norm_query = query_embedding / np.linalg.norm(query_embedding)
# Cosine similarity
similarities = np.dot(norm_embeddings, norm_query)
top_indices = similarities.argsort()[-top_k:][::-1]
return [self.entries[i] for i in top_indices]
def prune_older_than(self, max_age_seconds: float):
"""
Remove entries older than max_age_seconds from now.
Args:
max_age_seconds: Maximum age to keep entries.
"""
current_time = time.time()
with self.lock:
self.entries = [e for e in self.entries if current_time - e.timestamp = max_age_seconds]
def clear(self):
with self.lock:
self.entries.clear()
class MemoryManager:
"""
Combined manager providing unified interface over short-term and long-term memory.
Supports addition, retrieval, context building, and pruning.
"""
def __init__(self,
short_term_size: int = 50,
long_term_prune_age: float = 60 * 60 * 24 * 7, # One week default
embedding_function: Optional[Callable[[Any], np.ndarray]] = None):
"""
Initialize MemoryManager.
Args:
short_term_size: Max entries in short-term memory.
long_term_prune_age: Max age for pruning long-term memory in seconds.
embedding_function: Callable converting text/content to vector embeddings.
"""
self.short_term = ShortTermMemory(max_size=short_term_size)
self.long_term = LongTermMemory()
self.prune_age = long_term_prune_age
self.embedding_function = embedding_function
def add_memory(self, content: Any):
"""
Add a memory record to short-term and long-term memory with embedding.
Args:
content: Content string or data representing memory.
"""
embedding = None
if self.embedding_function is not None:
try:
embedding = self.embedding_function(content)
except Exception as e:
embedding = None
self.short_term.add(content, embedding)
self.long_term.add(content, embedding)
def get_recent(self, n: Optional[int] = None) -> List[Any]:
"""
Retrieves recent content entries from short-term memory.
Args:
n: Number of recent entries to retrieve.
Returns:
List of content objects.
"""
entries = self.short_term.get_recent(n)
return [e.content for e in entries]
def retrieve_similar(self, query_content: Any, top_k: int = 5) -> List[Any]:
"""
Retrieve top-k long term similar memories based on query content.
Args:
query_content: Query string or data.
top_k: Number of matches to retrieve.
Returns:
List of content objects.
"""
if self.embedding_function is None:
return []
try:
query_emb = self.embedding_function(query_content)
except Exception:
return []
top_entries = self.long_term.retrieve_similar(query_emb, top_k)
return [e.content for e in top_entries]
def prune_long_term(self):
"""
Prune old entries from long-term memory.
"""
self.long_term.prune_older_than(self.prune_age)
def clear_all(self):
"""
Clear both short-term and long-term memories.
"""
self.short_term.clear()
self.long_term.clear()
if __name__ == "__main__":
# Basic test and demonstration
def dummy_embedding_func(text: str) -> np.ndarray:
# Very simple embedding: vector of char ordinals normalized
vec = np.array([ord(c) for c in text], dtype=np.float32)
if vec.size == 0:
return np.zeros(10)
vec = vec / (np.linalg.norm(vec) + 1e-9)
# Pad or truncate to length 10
if vec.size 10:
vec = np.pad(vec, (0, 10 - vec.size))
else:
vec = vec[:10]
return vec
mem_manager = MemoryManager(short_term_size=5, long_term_prune_age=60*60*24*30, embedding_function=dummy_embedding_func)
print("Adding memories...")
mem_manager.add_memory("Learned about AI models.")
mem_manager.add_memory("OmniCoreX is powerful.")
mem_manager.add_memory("Recall adaptive reasoning.")
mem_manager.add_memory("Real-time decision making enabled.")
mem_manager.add_memory("Deep multi-modal integration.")
print("Recent memories:")
for mem in mem_manager.get_recent():
print("-", mem)
print("Retrieving similar memories for query 'adaptive AI':")
similars = mem_manager.retrieve_similar("adaptive AI", top_k=3)
for s in similars:
print("*", s)
print("Pruning memories older than prune age...")
mem_manager.prune_long_term()
print("Memory manager demo completed.")
|