Struct SqliteVectorStore 

pub struct SqliteVectorStore { /* private fields */ }

SQLite-backed implementation of the VectorStore trait.

Stores document embeddings as BLOBs (little-endian f32 sequences) and computes cosine similarity in Rust. An FTS5 virtual table provides full-text search for hybrid_search.

Implementations

impl SqliteVectorStore

pub fn new(config: SqliteVectorStoreConfig) -> Result<Self, SynapticError>

Create a new SqliteVectorStore from the given configuration.

Opens (or creates) the SQLite database and initializes the vectors and FTS5 tables if they do not already exist.

pub async fn hybrid_search( &self, query: &str, k: usize, embeddings: &dyn Embeddings, alpha: f32, ) -> Result<Vec<(Document, f32)>, SynapticError>

Hybrid search combining cosine similarity and BM25 full-text scoring.

alpha controls the balance:

• 1.0 = pure vector similarity
• 0.0 = pure BM25 text relevance
• 0.5 = balanced (typical default)

The final score is alpha * cosine + (1 - alpha) * normalized_bm25.

Trait Implementations

impl VectorStore for SqliteVectorStore

fn add_documents<'life0, 'life1, 'async_trait>( &'life0 self, docs: Vec<Document>, embeddings: &'life1 dyn Embeddings, ) -> Pin<Box<dyn Future<Output = Result<Vec<String>, SynapticError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Add documents to the store, computing their embeddings.

fn similarity_search<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, query: &'life1 str, k: usize, embeddings: &'life2 dyn Embeddings, ) -> Pin<Box<dyn Future<Output = Result<Vec<Document>, SynapticError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Search for similar documents by query string.

fn similarity_search_with_score<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, query: &'life1 str, k: usize, embeddings: &'life2 dyn Embeddings, ) -> Pin<Box<dyn Future<Output = Result<Vec<(Document, f32)>, SynapticError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Search with similarity scores (higher = more similar).

fn similarity_search_by_vector<'life0, 'life1, 'async_trait>( &'life0 self, embedding: &'life1 [f32], k: usize, ) -> Pin<Box<dyn Future<Output = Result<Vec<Document>, SynapticError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Search by pre-computed embedding vector instead of text query.

fn delete<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, ids: &'life1 [&'life2 str], ) -> Pin<Box<dyn Future<Output = Result<(), SynapticError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Delete documents by ID.