memory-systems

Agent memory persistence across sessions is achieved by combining transient context with structured memory layers, specifically a vector store and a temporal knowledge graph. This architecture enables continuity and reliable reasoning over past interactions.

A temporal knowledge graph maintains entity identity and reasons about changes over time within a structured memory layer. It supports time-aware reasoning and entity tracking by connecting vector stores with graph data for contextual retrieval.

Build a layered memory architecture by stacking working memory, long-term memory, and temporal knowledge graphs. This skill provides retrieval, consolidation, and context-injection utilities to connect vector stores with graphs for AI tasks.

Yes, numpy is a required dependency for implementing this agent memory system. You also need Python 3.10 or higher to run the scripts and utilize the vector-based memory and temporal knowledge graph retrieval utilities.

The best way to add context injection is to combine a vector store with a temporal knowledge graph and use dedicated retrieval and consolidation utilities. This connects structured memory layers to supply contextual data for AI tasks.

Avoid using a temporal knowledge graph for agent memory if your application does not require cross-session recall, entity tracking, or time-aware reasoning. For purely stateless or transient interactions, this combined memory stack is unnecessary.