Struct VectorSearchEngine 

pub struct VectorSearchEngine { /* private fields */ }

VectorSearchEngine with HNSW indexing and fastembed embeddings

Features:

• Pure Rust embeddings via fastembed (no Python dependency)
• HNSW index for O(log n) approximate nearest neighbor search
• Background indexing with async channel-based processing
• Multi-tenant support with tenant isolation
• Configurable similarity thresholds

Implementations

impl VectorSearchEngine

pub fn new() -> Result<Self>

Create a VectorSearchEngine without the embedding model (for testing)

pub fn with_config(config: VectorSearchEngineConfig) -> Result<Self>

pub fn config(&self) -> &VectorSearchEngineConfig

Get the engine configuration

pub fn embed_text(&self, _text: &str) -> Result<Vec<f32>>

pub fn embed_event(&self, payload: &Value) -> Result<Vec<f32>>

Generate embedding from an event payload

Extracts text content from the event payload and generates an embedding. Handles various payload structures by extracting relevant text fields.

pub async fn index_event( &self, event_id: Uuid, tenant_id: &str, embedding: Vec<f32>, source_text: Option<String>, ) -> Result<()>

Index an event with its embedding

pub fn search_similar( &self, query: &SimilarityQuery, ) -> Result<Vec<SimilarityResult>>

Fallback search using brute force (when HNSW not available or for testing)

pub fn start_background_indexer(&mut self) -> Sender<IndexRequest>

Start background indexing worker

Returns a sender that can be used to submit index requests. The worker runs asynchronously and processes requests in batches.

pub fn count(&self, tenant_id: Option<&str>) -> usize

Get the number of indexed vectors

pub fn stats(&self) -> (u64, u64, u64)

Get engine statistics

pub fn delete(&self, event_id: Uuid) -> Result<bool>

Delete a vector by event ID

pub fn delete_by_tenant(&self, tenant_id: &str) -> Result<usize>

Delete all vectors for a tenant

pub fn health_check(&self) -> Result<()>

Health check