Worlds Platform™

Worlds Platform™ is a REST API designed to manage, query, update, and reason over SPARQL-compatible knowledge bases at the edge. It places a malleable context graph within arm's reach of your AI agent.

Design

Bring your own brain (BYOB). Worlds Platform™ is agnostic to the agent using it.

Powered by N3. Worlds Platform™ leverages N3 for high-performance store operations.

Usage

You can use the Worlds SDK to interact with your knowledge bases programmatically.

import { WorldsSdk } from "@wazoo/sdk";

// Initialize the client.
const sdk = new WorldsSdk({
  baseUrl: "http://localhost:8000",
  apiKey: "your-api-key",
});

const worldId = "my-knowledge-base";

// Add some knowledge (triples) to your world.
await sdk.worlds.sparql(
  worldId,
  `
  INSERT DATA {
    <http://example.com/ethan> <http://schema.org/relatedTo> <http://example.com/gregory> .
    <http://example.com/gregory> <http://schema.org/givenName> "Gregory" .
  }
`,
);

// Search your world to find the named node for Gregory.
const searchResult = await sdk.worlds.search(worldId, "Gregory");

console.log(searchResult);
// [
//   {
//     subject: "http://example.com/gregory",
//     predicate: "http://schema.org/givenName",
//     object: "Gregory",
//     vecRank: 0.1,
//     ftsRank: 0.1,
//     score: 0.9
//   }
// ]

// Reason over your world using SPARQL.
const result = await sdk.worlds.sparql(
  worldId,
  `
  SELECT ?name WHERE {
    <http://example.com/ethan> <http://schema.org/relatedTo> ?person .
    ?person <http://schema.org/givenName> ?name .
  }
`,
);

console.log(result);
// {
//   head: { vars: [ "name" ] },
//   results: {
//     bindings: [
//       {
//         name: { type: "literal", value: "Gregory" }
//       }
//     ]
//   }
// }

Development

Contributions are welcome! Please open an issue or submit a pull request.

Generate code:

deno task generate

Start the development server:

deno task start

Format, lint, and test before committing:

deno task precommit

Etymology

We named the Worlds API™ after "World Models as a Service" as a nod to the Many-worlds interpretation.

Research

This work is inspired by the intersection of neuro-symbolic AI and knowledge graphs:

• Thinking with Knowledge Graphs (Arxiv)
• World Models (Ha & Schmidhuber)
• MemGPT: Towards LLMs as Operating Systems (Arxiv)

For further information, please refer to our whitepaper.

Glossary

• Neuro-symbolic AI: An AI system that combines the strengths of neural networks and structured data.
• RDF: A W3C standard for representing information.
• SPARQL: A W3C standard for querying and updating RDF data.
• Ontology: A formal description of a domain of knowledge.

---

Developed with 🧪 @wazootech