v6: Implement users' feedback

[bevzzz]

Reference property static factory

// Before
.properties(Property.reference("hasAward", "Oscars", "Grammys"))
// After
.properties(ReferenceProperty.to("hasAward", "Oscars", "Grammys"))

Improved String representation of Vectors

var vector = Vectors.of(new float[][] { { 1, 2, 3 }, { 1, 2, 3 } });
vector.toString();
// Outputs:
//     Vectors(default=[[1.0, 2.0, 3.0], [1.0, 2.0, 3.0]])

Renamed connection helpers + liveness check

// Before
WeaviateClient.local();
WeaviateClient.wcd("cluster-url", "api-key");
WeaviateClient.custom(conn -> conn);
// After
WeaviateClient.connectToLocal();
WeaviateClient.connectToWeaviateCloud("cluster-url", "api-key");
WeaviateClient.connectToCustom(conn -> conn);

If the instance does not respond to a liveness check (GET /live) the constructor (and the helper factories) will throw an exception and dispose of any resources obtained thus far (HTTP client).

// Before (Weaviate is down)
try (var client = WeaviateClient.local()) {
  client.collections.create("FirstCollection");  // <- throws IOException 
} catch (IOException e) {
  // Weaviate is down, but had to wait until the first query to find out.
}

// After (Weaviate is down)
try (var client = WeaviateClient.local()) { // <- throws WeaviateConnectException
    client.collections.create("FirstCollection");
} catch (WeaviateConnectException e) {
    // Weaviate is down!
} catch (IOException e) {
   // Something else is wrong.
}

The examples show WeaviateClient, but the same is of course true for WeaviateAsyncClient.

Local configuration parameters

// Before
WeaviateClient.connectToLocal(conn -> conn.httpPort(8081));
// After
WeaviateClient.connectToLocal(conn -> conn.port(8081));

VectorConfig

// Before
.vectors(Vectorizer.text2vecWeaviate(t2v -> t2v.sourceProperties("title")))
// After
.vectorConfig(VectorConfig.text2vecWeaviate(t2v -> t2v.sourceProperties("title")))

Quantization

.vectorConfig(
  VectorConfig.text2vecWeaviate(
    "title_vec",
    t2v -> t2v.quantization(Quantization.bq(bq -> bq.rescoreLimit(6)))
  ),
  VectorConfig.text2vecCohere(
    "body_vec",
    t2v -> t2v.quantization(Quantization.sq(sq -> sq.cache(true)))
  )
)

Max gRPC message size

The client now fetches server's metadata at startup and sets an inbound/outbound grpc_max_message_size to match the value configured on the server.

[orca-security-eu]

Orca Security Scan Summary

Status	Check	Issues by priority
Passed	Secrets	0   0   0   0	View in Orca

[sebawita]

Hey @bevzzz this is all great 😃