Advanced Concepts: Memory in Agentic Systems

Agentic AI systems can maintain different types of memory:

Short-term Memory

• Current conversation context
• Recent decisions
• Temporary state information

Long-term Memory

• User preferences
• Historical interactions
• Learned patterns

Episodic Memory

• Specific past events
• Contextual experiences
• Situation-outcome pairs

Semantic Memory

• General knowledge
• Facts and rules
• Domain expertise

Code:

 
class MemoryState(TypedDict):
    short_term: list  # Current conversation
    long_term: dict   # User preferences
    episodic: list    # Past interactions
 
def memory_manager(state: MemoryState) -> MemoryState:
    """Manage different types of memory"""
    # Update short-term memory
    state["short_term"].append(state["current_input"])
    
    # Check if we should store in long-term
    if is_important(state["current_input"]):
        update_long_term_memory(state)
    
    # Retrieve relevant episodic memories
    relevant_memories = retrieve_episodic(state)
    state["context"] = relevant_memories
    
    return state

For a customer service agent, which type of memory would store "Customer prefers email communication"?

Long-term memoryShort-term memoryEpisodic memorySemantic memory