Overview

AI Agent memory systems are critical for achieving persistent intelligence. Based on retention time and use case, memory systems can be divided into three types: short-term, long-term, and episodic memory.

Comparison of Three Memory Types

Short-term Memory Implementation

class ShortTermMemory:
    def __init__(self, system_prompt: str):
        self.messages = [SystemMessage(content=system_prompt)]
    
    def get_context(self) -> list:
        return self.messages[-20:]

Long-term Memory Implementation

import chromadb

class LongTermMemory:
    def __init__(self):
        self.collection = chromadb.PersistentClient(path="./memory").get_or_create_collection("agent_memory")
    
    def remember(self, key: str, content: str):
        self.collection.add(documents=[content], ids=[key])

References

• LangChain Memory Components
• Chroma Vector Database