EmbeddingVector
@RequiresFeature(enforcement = "androidx.appsearch.app.Features#isFeatureSupported", name = Features.SCHEMA_EMBEDDING_PROPERTY_CONFIG)
public final class EmbeddingVector
Embeddings are vector representations of data, such as text, images, and audio, which can be generated by machine learning models and used for semantic search. This class represents an embedding vector, which wraps a float array for the values of the embedding vector and a model signature that can be any string to distinguish between embedding vectors generated by different models.
For more details on how embedding search works, check search and setRankingStrategy.
Summary
Nested types |
|---|
@ExperimentalAppSearchApiHolds the pre-quantized embedding values along with the quantization parameters. |
Public constructors |
|---|
@ExperimentalAppSearchApiCreates a new |
EmbeddingVector(@NonNull float[] values, @NonNull String modelSignature)Creates a new |
Public methods |
|
|---|---|
boolean |
|
@NonNull String |
Returns the model signature of this embedding vector, which is an arbitrary string to distinguish between embedding vectors generated by different models. |
@Nullable EmbeddingVector.QuantizedData |
Returns the quantized data of this embedding vector, or null if it only contains standard float values. |
@NonNull float[] |
Returns the values of this embedding vector. |
int |
hashCode() |
Public constructors
EmbeddingVector
@ExperimentalAppSearchApi
@RequiresFeature(enforcement = "androidx.appsearch.app.Features#isFeatureSupported", name = Features.SCHEMA_EMBEDDING_PRE_QUANTIZED_DATA)
public EmbeddingVector(
@NonNull EmbeddingVector.QuantizedData quantizedData,
@NonNull String modelSignature
)
Creates a new EmbeddingVector from pre-quantized embedding data.
Storing 8-bit quantized values directly helps reduce memory and disk space usage by approximately 75% compared to storing full 32-bit floats.
| Parameters | |
|---|---|
@NonNull EmbeddingVector.QuantizedData quantizedData |
The pre-quantized embedding data. |
@NonNull String modelSignature |
An arbitrary string used to distinguish between embedding vectors generated by different models. |
| Throws | |
|---|---|
java.lang.IllegalArgumentException |
if |
EmbeddingVector
public EmbeddingVector(@NonNull float[] values, @NonNull String modelSignature)
Creates a new EmbeddingVector containing standard 32-bit float values.
| Parameters | |
|---|---|
@NonNull float[] values |
The float values of the embedding vector. |
@NonNull String modelSignature |
An arbitrary string used to distinguish between embedding vectors generated by different models. |
| Throws | |
|---|---|
java.lang.IllegalArgumentException |
if |
Public methods
getModelSignature
public @NonNull String getModelSignature()
Returns the model signature of this embedding vector, which is an arbitrary string to distinguish between embedding vectors generated by different models.
getQuantizedData
@ExperimentalAppSearchApi
public @Nullable EmbeddingVector.QuantizedData getQuantizedData()
Returns the quantized data of this embedding vector, or null if it only contains standard float values.