API Reference

new VectorDB(config: VectorDBConfig)

Creates a database instance.

• dbName – name of the IndexedDB database.
• dimension – length of the stored vectors.
• storeName – optional object store name (defaults to "vectors").
• distanceType – default distance metric (cosine, l2, l1, dot, hamming, minkowski).
• minkowskiP – power parameter for Minkowski distance (default 3).

add(vector, metadata?) => Promise<string>

Add a vector with optional metadata. Returns the generated id.

get(id) => Promise<VectorEntry | undefined>

Retrieve a stored entry.

update(id, update) => Promise<void>

Partially update an entry.

delete(id) => Promise<void>

Remove an entry from the database.

buildIndex(numHashes?) => Promise<void>

Build an LSH index from all entries. numHashes controls the number of hyperplanes (default 10).

search(query, k?, distanceType?) => Promise<SearchResult[]>

Exact similarity search. distanceType can be 'cosine', 'l2', 'l1', 'dot', 'hamming', or 'minkowski'.

annSearch(query, k?, radius?, distanceType?) => Promise<SearchResult[]>

Approximate nearest neighbour search using the LSH index. The index is built lazily when first needed.

close() => Promise<void>

Close the underlying IndexedDB connection.

Types

• VectorDBConfig
• VectorEntry
• SearchResult
• DistanceType