Secrets of Agentic Memory Management
summary
Prototype Java Spring memory server implements six memory types (semantic, episodic, procedural, entity and variants), a memory type registry and modules, and supports MCP for connecting external LLMs. Retrieval uses hybrid BM25 + vector search combined with reciprocal rank fusion. System shows concrete tooling: LLM‑driven and code‑driven extraction, TTL and eviction policies, deduplication, promotion between short and long term, emotional tagging, and background consolidation/compaction experiments that actually preserve or remove memories over time.
AI-written summary · about
chapters (10)
- 0:00Agentic memory management
- 1:59Human memory science
- 10:10Long-term vs short-term memory
- 27:46AI memory problem and RAG limits
- 31:56Redis Spring AI memory server
- 38:59Six memory types in agents
- 40:03Live demo: semantic and episodic memories
- 44:43Entity memories and knowledge graphs
- 48:59Cross-memory integration challenges
- 50:04MCP integration and closing
related talks
28mAgentic Memory
54mThe Anatomy of Memory in Humans and AI Agents
18mArchitecting Agent Memory: Principles, Patterns, and Best Practices
32mOpenSearch Agentic Memory: Intelligent Context for AI Agents
47mAgents and Memory Architecture
46mThe Anatomy of Memory in Humans and AI Agents
01h 49mBuilding Memory Augmented Agents with MemoRizz
27mMemory Engineering: Going Beyond Context Engineering