Struct VectorStore 

pub struct VectorStore { /* private fields */ }

A vector store for in-memory vector search.

Performance

Uses contiguous memory layout for optimal cache locality and fast serialization. Vector data is stored in a single buffer rather than individual Vec allocations.

Implementations

impl VectorStore

pub fn new(dimension: usize, metric: &str) -> Result<VectorStore, JsValue>

Creates a new vector store with the specified dimension and distance metric.

Arguments

• dimension - Vector dimension (e.g., 768 for BERT, 1536 for GPT)
• metric - Distance metric: “cosine”, “euclidean”, or “dot”

Errors

Returns an error if the metric is not recognized.

pub fn len(&self) -> usize

Returns the number of vectors in the store.

pub fn is_empty(&self) -> bool

Returns true if the store is empty.

pub fn dimension(&self) -> usize

Returns the vector dimension.

pub fn insert(&mut self, id: u64, vector: &[f32]) -> Result<(), JsValue>

Inserts a vector with the given ID.

Arguments

• id - Unique identifier for the vector
• vector - Float32Array of the vector data

Errors

Returns an error if vector dimension doesn’t match store dimension.

pub fn search(&self, query: &[f32], k: usize) -> Result<JsValue, JsValue>

Searches for the k nearest neighbors to the query vector.

Arguments

• query - Query vector as Float32Array
• k - Number of results to return

Returns

Array of [id, score] pairs sorted by relevance.

Errors

Returns an error if query dimension doesn’t match store dimension.

pub fn remove(&mut self, id: u64) -> bool

Removes a vector by ID.

pub fn clear(&mut self)

Clears all vectors from the store.

pub fn memory_usage(&self) -> usize

Returns memory usage estimate in bytes.

pub fn with_capacity( dimension: usize, metric: &str, capacity: usize, ) -> Result<VectorStore, JsValue>

Creates a new vector store with pre-allocated capacity.

This is more efficient when you know the approximate number of vectors you’ll be inserting, as it avoids repeated memory allocations.

Arguments

• dimension - Vector dimension
• metric - Distance metric: “cosine”, “euclidean”, or “dot”
• capacity - Number of vectors to pre-allocate space for

Errors

Returns an error if the metric is not recognized.

pub fn reserve(&mut self, additional: usize)

Pre-allocates memory for the specified number of additional vectors.

Call this before bulk insertions to avoid repeated allocations.

Arguments

• additional - Number of additional vectors to reserve space for

pub fn insert_batch(&mut self, batch: JsValue) -> Result<(), JsValue>

Inserts multiple vectors in a single batch operation.

This is significantly faster than calling insert() multiple times because it pre-allocates memory and reduces per-call overhead.

Arguments

• batch - JavaScript array of [id, Float32Array] pairs

Errors

Returns an error if any vector dimension doesn’t match store dimension.

pub fn export_to_bytes(&self) -> Result<Vec<u8>, JsValue>

Exports the vector store to a binary format.

The binary format contains:

• Header: dimension (u32), metric (u8), count (u64)
• For each vector: id (u64), data (f32 array)

Use this to persist data to IndexedDB or localStorage.

Errors

This function currently does not return errors but uses Result for future extensibility.

Performance

Perf: Pre-allocates exact buffer size to avoid reallocations. Throughput: ~1600 MB/s on 10k vectors (768D)

pub async fn save(&self, db_name: &str) -> Result<(), JsValue>

Saves the vector store to IndexedDB.

This method persists all vectors to the browser’s IndexedDB, enabling offline-first applications.

Arguments

• db_name - Name of the IndexedDB database

Errors

Returns an error if IndexedDB is not available or the save fails.

Example

const store = new VectorStore(768, "cosine");
store.insert(1n, vector1);
await store.save("my-vectors");

pub async fn load(db_name: &str) -> Result<VectorStore, JsValue>

Loads a vector store from IndexedDB.

This method restores all vectors from the browser’s IndexedDB.

Arguments

• db_name - Name of the IndexedDB database

Errors

Returns an error if the database doesn’t exist or is corrupted.

Example

const store = await VectorStore.load("my-vectors");
console.log(store.len); // Number of restored vectors

pub async fn delete_database(db_name: &str) -> Result<(), JsValue>

Deletes the IndexedDB database.

Use this to clear all persisted data.

Arguments

• db_name - Name of the IndexedDB database to delete

Errors

Returns an error if the deletion fails.

pub fn import_from_bytes(bytes: &[u8]) -> Result<VectorStore, JsValue>

Imports a vector store from binary format.

Use this to restore data from IndexedDB or localStorage.

Errors

Returns an error if:

• The data is too short or corrupted
• The magic number is invalid
• The version is unsupported
• The metric byte is invalid

Trait Implementations

impl From<VectorStore> for JsValue

fn from(value: VectorStore) -> Self

Converts to this type from the input type.

impl FromWasmAbi for VectorStore

type Abi = u32

The Wasm ABI type that this converts from when coming back out from the ABI boundary.

unsafe fn from_abi(js: u32) -> Self

Recover a Self from Self::Abi.

impl IntoWasmAbi for VectorStore

type Abi = u32

The Wasm ABI type that this converts into when crossing the ABI boundary.

fn into_abi(self) -> u32

Convert self into Self::Abi so that it can be sent across the wasm ABI boundary.

impl LongRefFromWasmAbi for VectorStore

type Abi = u32

Same as RefFromWasmAbi::Abi

type Anchor = RcRef<VectorStore>

Same as RefFromWasmAbi::Anchor

unsafe fn long_ref_from_abi(js: Self::Abi) -> Self::Anchor

Same as RefFromWasmAbi::ref_from_abi

impl OptionFromWasmAbi for VectorStore

fn is_none(abi: &Self::Abi) -> bool

Tests whether the argument is a “none” instance. If so it will be deserialized as None, and otherwise it will be passed to FromWasmAbi.

impl OptionIntoWasmAbi for VectorStore

fn none() -> Self::Abi

Returns an ABI instance indicating “none”, which JS will interpret as the None branch of this option.

impl RefFromWasmAbi for VectorStore

type Abi = u32

The Wasm ABI type references to Self are recovered from.

type Anchor = RcRef<VectorStore>

The type that holds the reference to Self for the duration of the invocation of the function that has an &Self parameter. This is required to ensure that the lifetimes don’t persist beyond one function call, and so that they remain anonymous.

unsafe fn ref_from_abi(js: Self::Abi) -> Self::Anchor

Recover a Self::Anchor from Self::Abi.

impl RefMutFromWasmAbi for VectorStore

type Abi = u32

Same as RefFromWasmAbi::Abi

type Anchor = RcRefMut<VectorStore>

Same as RefFromWasmAbi::Anchor

unsafe fn ref_mut_from_abi(js: Self::Abi) -> Self::Anchor

Same as RefFromWasmAbi::ref_from_abi

impl TryFromJsValue for VectorStore

fn try_from_js_value(value: JsValue) -> Result<Self, JsValue>

Performs the conversion.

fn try_from_js_value_ref(value: &JsValue) -> Option<Self>

Performs the conversion.

impl VectorFromWasmAbi for VectorStore

type Abi = <Box<[JsValue]> as FromWasmAbi>::Abi

unsafe fn vector_from_abi(js: Self::Abi) -> Box<[VectorStore]>

impl VectorIntoWasmAbi for VectorStore

type Abi = <Box<[JsValue]> as IntoWasmAbi>::Abi

fn vector_into_abi(vector: Box<[VectorStore]>) -> Self::Abi

impl WasmDescribe for VectorStore

fn describe()

impl WasmDescribeVector for VectorStore

fn describe_vector()

impl SupportsConstructor for VectorStore

impl SupportsInstanceProperty for VectorStore

impl SupportsStaticProperty for VectorStore