LLM Memory

Memory management for AI agents and multi-agent systems.

Types:

• Short-term (in-context)
• Long-term:
  • Semantic — facts and knowledge
  • Episodic — past interactions
  • Procedural — how-to knowledge

Writing memory:

• In the hot path (blocks the response)
• In the background (async)

Key resources:

• https://www.mongodb.com/resources/basics/artificial-intelligence/agent-memory
• https://www.mongodb.com/company/blog/technical/why-multi-agent-systems-need-memory-engineering
• https://docs.langchain.com/oss/python/concepts/memory
• Checkpoints: https://reference.langchain.com/python/langgraph/checkpoints

Related

• mem0
• Qdrant
• Mempalace
• Agent_Harness
• Context_Window
• Context_Engineering
• AgentStack/Agents/memory-architecture — this taxonomy applied to the vault's own memory setups