Memory Systems for LLMs

123456789101112
interface MemorySystem {
  // Working memory - current context window
  workingMemory: {
    systemPrompt: string;
    retrievedContext: Document[];
    conversationHistory: Message[];
    currentQuery: string;
  };
  
  // Short-term memory - session buffer
  shortTermMemory: {
    sessionId: string;

123456789101112
interface Memory {
  id: string;
  type: 'episodic' | 'semantic';
  content: string;
  embedding: number[];
  timestamp: Date;
  importance: number;
  accessCount: number;
  lastAccessed: Date;
}

class MemoryManager {

123456789101112
from dataclasses import dataclass, field
from typing import List, Dict, Optional
from datetime import datetime
import json
import hashlib

@dataclass
class MemoryEntry:
    content: str
    timestamp: datetime
    importance: float
    entry_type: str  # 'user', 'assistant', 'system', 'summary'

123456789101112
from typing import List, Dict, Tuple, Optional
import chromadb
from chromadb.config import Settings
import openai
from datetime import datetime
import numpy as np

class SemanticMemoryStore:
    def __init__(self, collection_name: str = "memories"):
        self.client = chromadb.Client(Settings(anonymized_telemetry=False))
        self.collection = self.client.get_or_create_collection(
            name=collection_name,

123456789101112
from dataclasses import dataclass
from typing import List, Dict, Set
from datetime import datetime, timedelta
import logging

@dataclass
class ConsolidationReport:
    memories_merged: int = 0
    memories_promoted: int = 0
    memories_demoted: int = 0
    memories_deleted: int = 0
    contradictions_resolved: int = 0

123456789101112
from typing import List, Dict, Optional
import tiktoken

class MemoryAwarePromptConstructor:
    def __init__(self, working_memory: 'WorkingMemoryBuffer',
                 semantic_memory: 'SemanticMemoryStore',
                 max_memory_tokens: int = 2000,
                 model: str = "gpt-4"):
        self.working_memory = working_memory
        self.semantic_memory = semantic_memory
        self.max_memory_tokens = max_memory_tokens
        self.tokenizer = tiktoken.encoding_for_model(model)