Skip to main content
Before reasoning about facts, agents must find the correct starting point. Hybrid search disambiguates natural language queries, resolving them to the correct item (IRI). Standard vector search retrieves semantically similar text, but agents require factual precision. If an agent searches for “Ethan’s manager”, a pure vector query might return “Gregory’s manager” because their text embeddings are nearly identical in vector space.

Why not just RDF and SPARQL?

While SPARQL supports structured logic and keyword search, it cannot perform semantic vector retrieval. Worlds provides open-source infrastructure for indexed RDF stores built specifically for autonomous agents.

Hybrid retrieval

Rather than relying on a single index, Worlds fuses three complementary signals to guarantee retrieval accuracy. By layering keyword precision (BM25) and structural graph filters on top of vector embeddings, hybrid search ensures the agent disambiguates the exact item it needs before executing strict symbolic logic.

Signals

SignalTechniqueWhat it captures
SemanticVector embeddings, 1536-dimConceptual meaning
KeywordFTS5 / BM25Exact term matches
Graph contextRDF relationship filtersStructural relationships

Reciprocal rank fusion

Results from each signal are merged using Reciprocal Rank Fusion (RRF), an algorithm that produces a single unified ranking: score=dD160+rank(d)score = \sum_{d \in D} \frac{1}{60 + rank(d)} This ensures that a result ranked highly by multiple signals surfaces to the top, while noise from any single signal is suppressed.