Loading patterns…
Temporal Knowledge Graph Memory(TKG)
An agent memory substrate that stores facts as edges in a knowledge graph, each edge carrying explicit bi-temporal validity: valid time (when the fact was true in the world) and transaction time (when it was ingested). When new information contradicts an existing fact, the edge is invalidated by writing a t_invalid timestamp rather than deleted, so the graph can answer what was believed and when, and never serves a stale fact as current. This is the Zep and Graphiti design. Distinct from `semantic-memory-networks` and the static `knowledge-graph-construction` pattern in knowledge-representation, which model a graph with no time dimension or fact-supersession.
In 30 seconds
- What
- Stores facts as graph edges with valid-time and transaction-time stamps, invalidating contradictions by marking t_invalid rather than deleting.
- When to use
- Long-running agents tracking entities where facts change over time and you need to query what was believed at any past moment without serving stale data.
- Watch out
- High operational cost: extraction, deduplication, and temporal index maintenance grow with fact volume; stale extraction can create false contradictions and bloat the graph.
Ask the AI expert about this pattern
Opens the assistant with your question prefilled. You review it before sending.
Temporal Knowledge Graph Memory: Overview
An agent memory substrate that stores facts as edges in a knowledge graph, each edge carrying explicit bi-temporal validity: valid time (when the fact was true in the world) and transaction time (when it was ingested). When new information contradicts an existing fact, the edge is invalidated by writing a t_invalid timestamp rather than deleted, so the graph can answer what was believed and when, and never serves a stale fact as current. This is the Zep and Graphiti design. Distinct from `semantic-memory-networks` and the static `knowledge-graph-construction` pattern in knowledge-representation, which model a graph with no time dimension or fact-supersession.
- Facts stored as graph edges, not opaque vectors
- Bi-temporal stamps: valid time plus transaction time
- Contradiction invalidates an edge (writes t_invalid) instead of deleting it
- Point-in-time queries reconstruct past belief states
- Current-view queries filter to edges with no t_invalid
- Incremental entity and relationship extraction from new events
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
References
The papers, specifications, and repositories this pattern is based on.
From the engineer behind this catalog
Get your agent architecture reviewed
This page documents one pattern. Your system runs dozens, and most failures live in how they fit together. Have the whole design reviewed against the 288 patterns in this catalog: architecture, reliability, evaluation and cost, every finding mapped to the pattern that fixes it.
€750 instead of €1,500, one week, written report and walkthrough call, until 30 September