-
Book Overview & Buying
-
Table Of Contents
AI Agents in Practice
By :
Memory is a foundational component of intelligent AI agents, enabling them to maintain context, personalize interactions, and adapt over time. This chapter explored the spectrum of memory types—short-term and long-term—and their subcategories, including semantic, episodic, and procedural memory.
We examined strategies for managing the limited context window of LLMs, as well as techniques for storing, retrieving, and refreshing memory using both structured and semantic approaches. A hybrid model, combining metadata filtering with vector search, emerged as a powerful method for scalable and relevant memory access.
Finally, we introduced key tools—LangMem, Mem0, and MemGPT—that operationalize memory in different ways, from workflow-aware storage to operating system-inspired context management.
In the next chapter, we will see how to integrate memory with the actual capability of AI agents of “doing things,” by properly defining...