Definition
Memory RAG is an architectural pattern that treats past conversation logs, session states, and user preferences as the retrieval corpus to provide long-term persistence for AI agents. It balances the trade-off between limited context window sizes and the need for personalized continuity by dynamically injecting relevant historical snippets into the prompt.
Retrieves from a user's specific interaction history rather than a static knowledge base or general document corpus.
"A private secretary's searchable filing cabinet of all your previous meetings."
- Episodic Memory(Component)
- Semantic Memory(Component)
- Vector Database(Prerequisite)
- Context Window(Prerequisite)
Conceptual Overview
Memory RAG is an architectural pattern that treats past conversation logs, session states, and user preferences as the retrieval corpus to provide long-term persistence for AI agents. It balances the trade-off between limited context window sizes and the need for personalized continuity by dynamically injecting relevant historical snippets into the prompt.
Disambiguation
Retrieves from a user's specific interaction history rather than a static knowledge base or general document corpus.
Visual Analog
A private secretary's searchable filing cabinet of all your previous meetings.