An Illustrated Guide to AI Agents (for Raymond Rhine) by Maarten Grootendorst and Jay Alammar

Author:Maarten Grootendorst and Jay Alammar
Language: eng
Format: epub
Publisher: O'Reilly Media, Inc.
Published: 2025-10-22T00:00:00+00:00

Summary

In this chapter, we explored various methods for giving LLMs and Agents memory. We first covered various techniques for short-term memory, including the conversation history and methods for compressing it and keeping it manageable. Short-term memory is an important, but often underestimated component of enabling memory in intelligent systems.

Then, we explored long-term memory, including methods for RAG (MemoryBank) and agentic RAG (A-MEM and Search-o1). These methods are often inspired by human memory systems and may include techniques for degrading memory or deciding what’s meant to be important.

Finally, we explored context engineering as the next frontier in memory. Where we used to engineer our prompts ourselves, the entire context window now requires careful consideration. We covered why this context is important for agents and various methods for optimizing it.

The context, being much more than just the user’s prompt, contains all previously discussed forms of memory and potentially even more, like tools. In the next chapter, we will explore how tools can further enhance the capabilities of LLMs as an important component of agents. Moreover, we will cover how these tools can be called and the best practices for doing so. In Chapter 6, we cover planning and reflection capabilities of agents, where memory plays an important role.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.