|
Your AI agent forgets everything the moment the session ends. Every user preference, every decision chain, every piece of context -- gone. If you've built session stores or caching layers for web apps, you've already solved a version of this problem. Agent memory architecture applies the same thinking: short-term memory for active context, long-term memory for persistent knowledge, and episodic memory for recalling past interactions.
This workshop covers how to build that memory layer:
- Build short-term memory for active context: session state, user preferences, and decision chains that persist during the current interaction
- Design long-term memory with semantic retrieval: k-NN and hybrid search so agents recall relevant knowledge across sessions, not just the current context window
- Implement episodic memory for interaction recall: agents that remember past conversations and outcomes, then use that history to improve future responses
- Technical demos of memory tools from across the AI landscape such as MongoDB Atlas and Redis Cloud
|
