Skip to main content
Before reasoning about facts, agents must find the correct starting point.

Hosted private-beta status

The hosted Worlds Data API currently uses keyword full-text search. It does not have an embedding provider configured, so vector ranking and the hybrid flow described below are not active in the hosted beta.

Hybrid retrieval design

When vector search is configured, Worlds can combine keyword matching, vector similarity, and graph filters to resolve a natural-language query to a semantic entity (IRI) before executing symbolic logic.

Reciprocal rank fusion

When multiple signals are available, Worlds merges their rankings using Reciprocal Rank Fusion (RRF): score=∑d∈D160+rank(d)score = \sum_{d \in D} \frac{1}{60 + rank(d)} This lifts results ranked highly by multiple signals to the top and suppresses noise from any single signal. Use the data-plane search endpoint for keyword search in the hosted private beta. Data-plane requests authenticate with a wzw_ world token:
Watch the shell quoting: a JSON string inside single quotes breaks the moment the text contains an apostrophe (“Ethan’s manager” terminates the quote early). If a query genuinely needs an apostrophe, end the single-quoted string, escape the apostrophe, and reopen it — or put the payload in a file and use -d @payload.json. Results return the matching subject and predicate IRIs plus the literal content, so an agent can bind the exact entity before running SPARQL.