What is Persistent Memory?
A system's ability to retain information across sessions, enabling context-aware responses and cumulative learning.
Detailed Explanation
Persistent memory is the capability of an AI system to store and recall information across multiple sessions, conversations, and tasks. Unlike stateless systems that treat each interaction as isolated, systems with persistent memory build a cumulative knowledge base over time. Hermes Agent implements persistent memory through a three-tier architecture: session memory (current conversation context), persistent memory (long-term storage in SQLite with FTS5 full-text search), and skill memory (encoded reusable patterns). This enables the agent to remember user preferences, reference past projects, maintain continuity across platforms, and develop increasingly personalized responses.
Related Terms
Frequently Asked Questions
How is persistent memory different from a database?
A database stores raw data. Persistent memory includes semantic understanding — the agent knows what information is relevant to the current context and can retrieve it through both keyword and conceptual matching.
Is persistent memory private?
With Hermes, yes. Memory is stored locally on your infrastructure. You control where data lives and can delete it at any time. Nothing is sent to third-party servers unless explicitly configured.
Can persistent memory be shared across devices?
Yes, if the memory store is hosted on a shared database or synchronized across devices. Hermes supports remote databases and cloud storage for multi-device access.