memory-systems

To keep agent memory alive across sessions, you can use a memory system that combines a vector store with a temporal knowledge graph. This allows agents to store and retrieve persistent memories, enabling cross-session reasoning and continuity.

A temporal knowledge graph for agent memory stores time-bound facts and relationships. It works by integrating with a vector store to reason with time-aware contexts, allowing agents to track entities and consolidate memories over time.

You can store facts and retrieve memories using APIs designed for persistent memory backends. These APIs allow you to store_fact with timestamps and retrieve_memories while applying optional entity and time filters for precise context retrieval.

Yes, you can integrate a vector store with a knowledge graph to achieve seamless memory consolidation. This combination supports layered memory models including working, short-term, long-term, and entity memory for efficient retrieval.

Yes, numpy is required as a dependency to implement this memory system. It supports the underlying computations needed to combine the vector store with the temporal knowledge graph for cross-session personalization and knowledge retention.

You need a temporal knowledge graph instead of a standard vector store when your agent must reason with time-bound facts and relationships. This approach enables entity tracking and memory consolidation across interactions in time-aware contexts.