Spectron is a persistent agent memory layer built on SurrealDB. It provides knowledge graphs, automatic entity extraction, GraphRAG support, temporal facts, and hybrid retrieval for accurate context recall. Spectron enables AI agents to maintain long-term memory with high accuracy by combining graph structure, vector similarity, and time-aware facts in a single system.
Why Spectron?
Spectron was built to resolve a problem faced by users of AI agents in which the agents themselves end up being little more than chatbots. Without an awareness of user preferences or structured memory to understand the relationship between entities, agents are unable to accumulate context or coordinate with each other.
Spectron gives your AI agents persistent, queryable memory powered by knowledge graphs, entity extraction, temporal facts, and hybrid retrieval - built directly into SurrealDB rather than bolted on top. Some features of Spectron are:
ACID memory: Atomic writes across entities, facts, and embeddings.
Temporal facts: Bi-temporal history to know what was true and when.
Hybrid retrieval: Graph traversal, vector search, and filters in one call.
Multi-agent: Shared memory for coordinated agent swarms.
How to get started
Spectron will be coming soon, along with the documentation for how to use it. In the meantime, please see the next page to get started as well as this page for a top-level overview.