Short-Term Memory

LLMsiGo to glossary for large language model (opens in a new tab)A model trained on text to predict the next token, and so to generate language. are stateless — without memory management, every message is a fresh start. This module covers the conversation memory layer: how to store messages as a linked graph, how to retrieve and search them, and how adding a message automatically seeds the long-term memory layer through entity extraction.

By the end of this module, you will:

• Explain why a linked-list graph structure is more useful for conversation history than a flat table
• Build a conversation memory layer using add_message(), get_conversation(), and search_messages()
• Search message history semantically using vector embeddingsiFull definition for embedding (opens in a new tab)Information represented as a numerical vector, positioned so that similar information sits close together.
• Understand how entity extraction automatically connects short-term and long-term memory
• Store a multi-turn exchange and inspect the FIRST_MESSAGE and NEXT_MESSAGE chain in your own graph