Struct EmbeddingEntityResolver 

pub struct EmbeddingEntityResolver<E, C> { /* private fields */ }

Resolves entities by cosine similarity over candidate embeddings.

The production impl: embeds the candidate string and compares it against the embeddings of existing nodes in the same scope, merging into the closest node above MIN_ENTITY_SIMILARITY. This collapses surface-form variants onto one node where the exact-string resolver would fragment them.

Candidate embeddings are stored as a node property and scanned exactly here; a FalkorDB-native vector index can later replace the catalog’s linear scan behind the EntityCatalog seam without changing this resolver.

Generic over the embedder and catalog so tests inject stubs.

Implementations

impl<E: EmbeddingModel, C: EntityCatalog> EmbeddingEntityResolver<E, C>

pub fn new(embedder: E, catalog: C) -> Self

Builds a resolver over embedder and catalog with the default floor.

pub fn with_min_similarity(self, min_similarity: f32) -> Self

Overrides the minimum cosine similarity for a match.

Trait Implementations

impl<E: EmbeddingModel, C: EntityCatalog> EntityResolver for EmbeddingEntityResolver<E, C>

async fn resolve( &self, scope: &Scope, entity: &str, ) -> Result<Resolution, ResolveError>

Resolves entity within scope to an existing or new node.