This section maps the kinds of memory an agent can have — working, episodic, semantic, procedural, and more — so you can design storage and retrieval around what each type is for.
What you will learn
- The core memory types used in agent systems and how they differ
- How scope, structure, and time change what a memory means
- Why one memory type is never enough for real products
- How memory types map to different storage and query requirements
Chapters in this section
There are 20 chapters in this part. Open any chapter to read it on its own, or work through them in order.
- 1 Working Memory: The Live Context Window
- 2 Episodic Memory: Recording What Happened
- 3 Semantic Memory: Storing What Is True
- 4 Procedural Memory: Encoding What Works
- 5 Reflective and Meta-Memory: Memory About Memory
- 6 Short-Term vs Long-Term Memory in Agent Systems
- 7 Declarative vs Non-Declarative Memory, Borrowed from Cognitive Science
- 8 Shared Memory vs Private Memory
- 9 Bounded Memory vs Unbounded Memory
- 10 Structured Facts vs Unstructured Narrative Memory
- 11 Scoped Memory: Per-User, Per-Session, Per-Project
- 12 Temporal Memory: Representing Change Over Time
- 13 Profile Memory vs Event Memory
- 14 Preference Memory: A Special Case of Semantic Memory
- 15 Summary Memory vs Atomic Fact Memory
- 16 Tool-Use Memory: Remembering How, Not Just What
- 17 Cross-Session Memory vs Cross-Conversation Memory
- 18 Multi-Modal Memory: Beyond Text
- 19 Why One Memory Type Is Never Enough
- 20 Mapping Memory Types to Storage Requirements