Beyond retrieval
Most AI applications utilize RAG (Retrieval-Augmented Generation). In RAG, the agent searches for relevant documents, reads them, and then generates an answer. This approach is similar to an agent having a bookshelf it can consult, but it lacks a structured, persistent internal model of the information. Worlds Platform provides a stateful memory layer. A “World” is a stateful memory environment. Instead of storing flat text documents, Worlds organizes information as a Graph.Functional comparison
| Feature | Traditional RAG | Worlds Platform |
|---|---|---|
| Data structure | Flat text / Embeddings | Structured Graph (RDF) |
| Persistence | Session-based or static | Truly persistent & stateful |
| Reasoning | Vector similarity | Symbolic logic, via SPARQL |
| Interaction | Read-only retrieval | Dynamic updates & tool-calling |
Why “Worlds”?
The term “Worlds” represents the environment where your AI agent operates. Just as humans understand their surroundings through relationships (identity, location, ownership), an AI agent uses a World to maintain a consistent, structured understanding of its context. For a deeper dive into the underlying structures of these environments, refer to our Knowledge Graphs guide.Memory types
Worlds mirrors human memory systems to provide a complete “memory stack” for autonomous agents:| Memory type | Agent perspective | Worlds implementation |
|---|---|---|
| Semantic | What it knows | RDF Store: Structured facts and reasoning. |
| Episodic | What it did | Append-only Log: Temporal history of events. |
| Working | What it is processing | Context Window: Live distillation of focus. |
| Procedural | What it can do | Tools: Skills for graph and agent operations. |
| Sensory | What it perceives | Ingestion: Raw data streams and vector indexing. |