Long-term Memory

Long-term memory is what the agent knows — entities, facts, and preferences, connected in a graph that outlives any conversation.

In this module, you give your agent tools to search and write that knowledge deliberately, choose the extractor that does its reading, and grant it the ability to record its own conclusions. The optional lessons take the layer apart: where entities come from, how near-duplicates get resolved, and the API beneath facts and preferences.

By the end of this module, you will be able to:

• Write and register memory tools the agent chooses to call
• Grant the agent the ability to record facts, with your identity stamped on
• Explain how entities reach the graph, and resolve near-duplicates (optional)
• Choose an extractor and extend the entity vocabulary it looks for