ReadonlynameThe name of the store (e.g. 'mongodb', 'redis', 'dynamodb')
ReadonlylibraryThe client library being used
Returns the underlying weaviate-client WeaviateClient for anything not wrapped here.
Build a FilterValue (see nearVector()/fetchObjects()/deleteMany()'s
filters option) against a collection's schema, e.g.
store.filter('Article').byProperty('title').equal('Hello').
Create a new collection ("class"). properties declares the object
schema (name/dataType/indexing per field); vectorizer configures how
vectors are produced - pass weaviate.configure.vectorizer.none() for
bring-your-own vectors, or a module config like
weaviate.configure.vectorizer.text2VecOpenAI() to have Weaviate call
out to an embedding provider (which requires the corresponding API key
in headers, e.g. X-OpenAI-Api-Key, at connect time).
Optionalvectorizer: WeaviateVectorizerConfigDelete a collection and all of its objects.
List every collection's configuration.
Whether a collection with this name exists.
Detailed schema/config for one collection (properties, vectorizer, replication, etc).
Insert a single object. object.vector (renamed vectors on the wire)
may be omitted entirely for collections with a server-side vectorizer
module, or supplied explicitly (a plain array, or a { name: vector }
map for named-vector collections) for none/self-provided vectorizers.
Returns the object's UUID (generated if object.id wasn't given).
Insert multiple objects in a single request (no client-side batching -
all objects are sent together). Per-object failures are reported in
the returned errors/hasErrors rather than rejecting the whole call;
check hasErrors to detect partial failures.
Find the objects whose vector is closest to vector, optionally constrained by options.filters.
Find the objects whose vector is closest to the vectorized form of
query (a single string, or multiple strings averaged together).
Requires the collection to have a text vectorizer module configured
(this is not a keyword/BM25 search - use getClient() for that).
Fetch objects without a similarity search - the vector-store analogue
of SELECT ... WHERE .... Supports limit/offset/cursor-based
pagination (after), filters, and returnProperties to select which
fields come back.
Fetch a single object by its UUID. Returns null if no object with that ID exists.
Partially update an object (PATCH semantics) - only the fields in
fields.properties/fields.vectors are changed; everything else on
the object is left as-is.
Fully replace an object (PUT semantics) - any property not present in
fields.properties is cleared, unlike updateObject().
Whether an object with this UUID exists in the collection.
Delete a single object by its UUID. Returns true if an object was
actually deleted.
The real weaviate-client data.deleteById() issues a REST DELETE
and resolves true unconditionally (Weaviate's DELETE /objects/{id} returns success whether or not an object with that id
existed) - it does not itself report whether anything was deleted.
Check existence first so the documented "returns whether something
was actually deleted" contract holds against the real driver, not
just the test mock (whose Map.delete() happens to return that for
free).
Delete every object matching where (build one with filter()). Pass
{ verbose: true } to get back the per-object outcome in
result.objects; otherwise only the aggregate counts are populated.
WeaviateStorewraps aweaviate-client(v3)WeaviateClientand exposes Weaviate's collection management, object insert/query/update/delete, and near-vector/near-text similarity search as a typed, promise-based API.