Struct SemanticKnowledgeGraph 

pub struct SemanticKnowledgeGraph {
    pub entities: HashMap<u64, GraphEntity>,
    pub edges: Vec<GraphEdge>,
    pub next_entity_id: u64,
    pub next_edge_id: u64,
}

A queryable semantic knowledge graph that links entities via typed, weighted edges.

Supports:

• Adding/removing entities and edges.
• Neighbour lookup with optional relation filtering (sorted by weight desc).
• BFS traversal with relation and entity-kind filters.
• Cosine-similarity–based entity retrieval for entities that carry embeddings.

Fields

entities: HashMap<u64, GraphEntity>

All entities, keyed by their entity_id.

edges: Vec<GraphEdge>

All directed edges in insertion order.

next_entity_id: u64

Monotonically increasing counter for entity ids.

next_edge_id: u64

Monotonically increasing counter for edge ids.

Implementations

impl SemanticKnowledgeGraph

pub fn new() -> Self

Create a new, empty knowledge graph.

pub fn add_entity( &mut self, name: &str, kind: EntityKind, embedding: Option<Vec<f32>>, ) -> u64

Add a new entity and return its freshly assigned entity_id.

pub fn add_edge( &mut self, from_id: u64, to_id: u64, relation: &str, weight: f32, ) -> u64

Add a directed edge between two entities and return its edge_id.

The entities referenced by from_id and to_id need not exist yet; no validation is performed here so that callers can build the graph in any order.

pub fn get_entity(&self, id: u64) -> Option<&GraphEntity>

Look up an entity by id.

pub fn remove_entity(&mut self, entity_id: u64) -> bool

Remove an entity and all edges that touch it.

Returns true if the entity existed, false otherwise.

pub fn neighbors( &self, entity_id: u64, relation: Option<&str>, ) -> Vec<&GraphEntity>

Return all entities reachable via a single outgoing edge from entity_id.

• When relation is Some(r), only edges whose relation == r are considered.
• The result list is sorted by edge weight descending.

pub fn traverse(&self, query: &GraphQuery) -> Vec<&GraphEntity>

BFS traversal from query.start_entity_id up to query.max_hops hops.

At every hop the relation filter and entity-kind filter (from the GraphQuery) are applied to the destination entity. The start entity itself is excluded from the result. Results are sorted by entity_id ascending.

pub fn similar_entities( &self, entity_id: u64, threshold: f32, ) -> Vec<(&GraphEntity, f32)>

Find all entities (excluding entity_id itself) whose embedding has a cosine similarity ≥ threshold with the embedding of entity_id.

Returns an empty vec if the query entity does not exist or has no embedding. Results are sorted by similarity descending.

pub fn stats(&self) -> KnowledgeGraphStats

Return aggregate statistics for this graph.

Trait Implementations

impl Default for SemanticKnowledgeGraph

fn default() -> Self

Returns the “default value” for a type.