Skip to content

EmbeddingIndex#

[source]

EmbeddingIndex#

hsfs.embedding.EmbeddingIndex(index_name=None, features=None, col_prefix=None)

Represents an index for managing embedding features.

Arguments

  • index_name str | None: The name of the embedding index. The name of the project index is used if not provided.
  • features List[hsfs.embedding.EmbeddingFeature] | None: A list of EmbeddingFeature objects for the features that contain embeddings that should be indexed for similarity search.
  • col_prefix str | None: The prefix to be added to column names when using project index. It is managed by Hopsworks and should not be provided.

Example

embedding_index = EmbeddingIndex()
embedding_index.add_embedding(name="user_vector", dimension=256)
embeddings = embedding_index.get_embeddings()

Properties#

[source]

col_prefix#

str: The prefix to be added to column names.

[source]

feature_group#

FeatureGroup: The feature group object that contains the embedding feature.

[source]

index_name#

str: The name of the embedding index.

Methods#

[source]

add_embedding#

EmbeddingIndex.add_embedding(name, dimension, similarity_function_type="l2_norm", model=None)

Adds a new embedding feature to the index.

Example: 

embedding_index = EmbeddingIndex()
embedding_index.add_embedding(name="user_vector", dimension=256)

# Attach a hsml model to the embedding feature
embedding_index = EmbeddingIndex()
embedding_index.add_embedding(name="user_vector", dimension=256, model=hsml_model)

Arguments

  • name str: The name of the embedding feature.
  • dimension int: The dimensionality of the embedding feature.
  • similarity_function_type hsfs.embedding.SimilarityFunctionType | None: The type of similarity function to be used.
  • model (hsml.model.Model, optional): The hsml model used to generate the embedding. Defaults to None.

[source]

count#

EmbeddingIndex.count(options=None)

Count the number of records in the feature group.

Arguments

  • options map | None: The options used for the request to the vector database. The keys are attribute values of the hsfs.core.opensearch.OpensearchRequestOption class.

Returns

int: The number of records in the feature group.

Raises:

ValueError: If the feature group is not initialized. FeaturestoreException: If an error occurs during the count operation.

[source]

get_embedding#

EmbeddingIndex.get_embedding(name)

Returns the hsfs.embedding.EmbeddingFeature object associated with the feature name.

Arguments

  • name (str): The name of the embedding feature.

Returns

hsfs.embedding.EmbeddingFeature object

[source]

get_embeddings#

EmbeddingIndex.get_embeddings()

Returns the list of hsfs.embedding.EmbeddingFeature objects associated with the index.

Returns

A list of hsfs.embedding.EmbeddingFeature objects

[source]

json#

EmbeddingIndex.json()

Serialize the EmbeddingIndex object to a JSON string.

[source]

to_dict#

EmbeddingIndex.to_dict()

Convert the EmbeddingIndex object to a dictionary.

Returns: dict: A dictionary representation of the EmbeddingIndex object.