Skip to content

Embedding Retrievers¤

CouchbaseSearchEmbeddingRetriever¤

Retrieves documents from the CouchbaseSearchDocumentStore by embedding similarity.

Uses Search Vector Index (FTS-based) for hybrid searches combining vector, full-text, and geospatial queries. See CouchbaseSearchDocumentStore for more information.

Usage example:

import numpy as np
from couchbase_haystack import CouchbaseSearchDocumentStore, CouchbaseSearchEmbeddingRetriever, CouchbasePasswordAuthenticator
from haystack.utils import Secret

store = CouchbaseSearchDocumentStore(
    cluster_connection_string=Secret.from_env_var("CB_CONNECTION_STRING"),
    authenticator=CouchbasePasswordAuthenticator(
        username=Secret.from_env_var("CB_USERNAME"),
        password=Secret.from_env_var("CB_PASSWORD")
    ),
    bucket="haystack_test_bucket",
    scope="scope_name",
    collection="collection_name",
    vector_search_index="vector_index"
)
retriever = CouchbaseSearchEmbeddingRetriever(document_store=store)

results = retriever.run(query_embedding=np.random.random(768).tolist())
print(results["documents"])

The example above retrieves the 10 most similar documents to a random query embedding from the CouchbaseSearchDocumentStore. Note that dimensions of the query_embedding must match the dimensions of the embeddings stored in the CouchbaseSearchDocumentStore.

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
@component
class CouchbaseSearchEmbeddingRetriever:
    """Retrieves documents from the CouchbaseSearchDocumentStore by embedding similarity.

    Uses Search Vector Index (FTS-based) for hybrid searches combining vector, full-text, and geospatial queries.
    See CouchbaseSearchDocumentStore for more information.

    Usage example:

    ```python
    import numpy as np
    from couchbase_haystack import CouchbaseSearchDocumentStore, CouchbaseSearchEmbeddingRetriever, CouchbasePasswordAuthenticator
    from haystack.utils import Secret

    store = CouchbaseSearchDocumentStore(
        cluster_connection_string=Secret.from_env_var("CB_CONNECTION_STRING"),
        authenticator=CouchbasePasswordAuthenticator(
            username=Secret.from_env_var("CB_USERNAME"),
            password=Secret.from_env_var("CB_PASSWORD")
        ),
        bucket="haystack_test_bucket",
        scope="scope_name",
        collection="collection_name",
        vector_search_index="vector_index"
    )
    retriever = CouchbaseSearchEmbeddingRetriever(document_store=store)

    results = retriever.run(query_embedding=np.random.random(768).tolist())
    print(results["documents"])
    ```

    The example above retrieves the 10 most similar documents to a random query embedding from the
    CouchbaseSearchDocumentStore. Note that dimensions of the query_embedding must match the dimensions of the embeddings
    stored in the CouchbaseSearchDocumentStore.
    """

    def __init__(
        self,
        *,
        document_store: CouchbaseSearchDocumentStore,
        top_k: int = 10,
    ):
        """Create the CouchbaseSearchDocumentStore component.

        Note: Currently, the filter option is not supported with embedding queries.
        Instead, you can provide a couchbase search query while running the embedding query.
        The embedding query and search query are combined using an OR operation.

        Args:
            document_store: An instance of CouchbaseSearchDocumentStore.
            top_k: Maximum number of Documents to return.

        Raises:
            ValueError: If document_store is not an instance of CouchbaseSearchDocumentStore.
        """
        if not isinstance(document_store, CouchbaseSearchDocumentStore):
            msg = "document_store must be an instance of CouchbaseSearchDocumentStore"
            raise ValueError(msg)

        self.document_store = document_store
        self.top_k = top_k

    def to_dict(self) -> Dict[str, Any]:
        """Serializes the component to a dictionary.

        Returns:
            Dictionary with serialized data.
        """
        return default_to_dict(
            self,
            top_k=self.top_k,
            document_store=self.document_store.to_dict(),
        )

    @classmethod
    def from_dict(cls, data: Dict[str, Any]) -> "CouchbaseSearchEmbeddingRetriever":
        """Deserializes the component from a dictionary.

        Args:
            data: Dictionary to deserialize from.

        Returns:
            Deserialized component.
        """
        data["init_parameters"]["document_store"] = CouchbaseSearchDocumentStore.from_dict(
            data["init_parameters"]["document_store"]
        )
        return default_from_dict(cls, data)

    @component.output_types(documents=List[Document])
    def run(
        self,
        query_embedding: List[float],
        top_k: Optional[int] = None,
        filters: Optional[Dict[str, Any]] = None,
        search_query: Optional[SearchQuery] = None,
        limit: Optional[int] = None,
    ) -> Dict[str, List[Document]]:
        """Retrieve documents from the CouchbaseSearchDocumentStore, based on the provided embedding similarity.

        Args:
            query_embedding: Embedding of the query.
            top_k: Maximum number of Documents to be returned from vector query.
                  Overrides the value specified at initialization.
            filters: Optional dictionary of filters to apply before the vector search.
                     Refer to Haystack documentation for filter structure (https://docs.haystack.deepset.ai/v2.0/docs/metadata-filtering).
            search_query: Search filters param which is parsed to the Couchbase search query.
                        The vector query and search query are ORed operation.
            limit: Maximum number of Documents to be return by the couchbase fts search request.
                  Default value is top_k.

        Returns:
            A dictionary with the following keys:
            - documents: List of Documents most similar to the given query_embedding
        """
        top_k = top_k or self.top_k

        docs = self.document_store._embedding_retrieval(
            query_embedding=query_embedding, top_k=top_k, search_query=search_query, filters=filters, limit=limit
        )
        return {"documents": docs}

__init__ ¤

__init__(*, document_store: CouchbaseSearchDocumentStore, top_k: int = 10)

Note: Currently, the filter option is not supported with embedding queries. Instead, you can provide a couchbase search query while running the embedding query. The embedding query and search query are combined using an OR operation.

Parameters:

  • document_store (CouchbaseSearchDocumentStore) –

    An instance of CouchbaseSearchDocumentStore.

  • top_k (int, default: 10 ) –

    Maximum number of Documents to return.

Raises:

  • ValueError –

    If document_store is not an instance of CouchbaseSearchDocumentStore.

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
def __init__(
    self,
    *,
    document_store: CouchbaseSearchDocumentStore,
    top_k: int = 10,
):
    """Create the CouchbaseSearchDocumentStore component.

    Note: Currently, the filter option is not supported with embedding queries.
    Instead, you can provide a couchbase search query while running the embedding query.
    The embedding query and search query are combined using an OR operation.

    Args:
        document_store: An instance of CouchbaseSearchDocumentStore.
        top_k: Maximum number of Documents to return.

    Raises:
        ValueError: If document_store is not an instance of CouchbaseSearchDocumentStore.
    """
    if not isinstance(document_store, CouchbaseSearchDocumentStore):
        msg = "document_store must be an instance of CouchbaseSearchDocumentStore"
        raise ValueError(msg)

    self.document_store = document_store
    self.top_k = top_k

to_dict ¤

to_dict() -> Dict[str, Any]

Serializes the component to a dictionary.

Returns:

  • Dict[str, Any] –

    Dictionary with serialized data.

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
78
79
80
81
82
83
84
85
86
87
88
def to_dict(self) -> Dict[str, Any]:
    """Serializes the component to a dictionary.

    Returns:
        Dictionary with serialized data.
    """
    return default_to_dict(
        self,
        top_k=self.top_k,
        document_store=self.document_store.to_dict(),
    )

from_dict classmethod ¤

from_dict(data: Dict[str, Any]) -> CouchbaseSearchEmbeddingRetriever

Deserializes the component from a dictionary.

Parameters:

  • data (Dict[str, Any]) –

    Dictionary to deserialize from.

Returns:

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
@classmethod
def from_dict(cls, data: Dict[str, Any]) -> "CouchbaseSearchEmbeddingRetriever":
    """Deserializes the component from a dictionary.

    Args:
        data: Dictionary to deserialize from.

    Returns:
        Deserialized component.
    """
    data["init_parameters"]["document_store"] = CouchbaseSearchDocumentStore.from_dict(
        data["init_parameters"]["document_store"]
    )
    return default_from_dict(cls, data)

run ¤

run(
    query_embedding: List[float],
    top_k: Optional[int] = None,
    filters: Optional[Dict[str, Any]] = None,
    search_query: Optional[SearchQuery] = None,
    limit: Optional[int] = None,
) -> Dict[str, List[Document]]

Retrieve documents from the CouchbaseSearchDocumentStore, based on the provided embedding similarity.

Parameters:

  • query_embedding (List[float]) –

    Embedding of the query.

  • top_k (Optional[int], default: None ) –

    Maximum number of Documents to be returned from vector query. Overrides the value specified at initialization.

  • filters (Optional[Dict[str, Any]], default: None ) –

    Optional dictionary of filters to apply before the vector search. Refer to Haystack documentation for filter structure (https://docs.haystack.deepset.ai/v2.0/docs/metadata-filtering).

  • search_query (Optional[SearchQuery], default: None ) –

    Search filters param which is parsed to the Couchbase search query. The vector query and search query are ORed operation.

  • limit (Optional[int], default: None ) –

    Maximum number of Documents to be return by the couchbase fts search request. Default value is top_k.

Returns:

  • Dict[str, List[Document]] –

    A dictionary with the following keys:

  • Dict[str, List[Document]] –
    • documents: List of Documents most similar to the given query_embedding
Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
@component.output_types(documents=List[Document])
def run(
    self,
    query_embedding: List[float],
    top_k: Optional[int] = None,
    filters: Optional[Dict[str, Any]] = None,
    search_query: Optional[SearchQuery] = None,
    limit: Optional[int] = None,
) -> Dict[str, List[Document]]:
    """Retrieve documents from the CouchbaseSearchDocumentStore, based on the provided embedding similarity.

    Args:
        query_embedding: Embedding of the query.
        top_k: Maximum number of Documents to be returned from vector query.
              Overrides the value specified at initialization.
        filters: Optional dictionary of filters to apply before the vector search.
                 Refer to Haystack documentation for filter structure (https://docs.haystack.deepset.ai/v2.0/docs/metadata-filtering).
        search_query: Search filters param which is parsed to the Couchbase search query.
                    The vector query and search query are ORed operation.
        limit: Maximum number of Documents to be return by the couchbase fts search request.
              Default value is top_k.

    Returns:
        A dictionary with the following keys:
        - documents: List of Documents most similar to the given query_embedding
    """
    top_k = top_k or self.top_k

    docs = self.document_store._embedding_retrieval(
        query_embedding=query_embedding, top_k=top_k, search_query=search_query, filters=filters, limit=limit
    )
    return {"documents": docs}

CouchbaseQueryEmbeddingRetriever¤

Retrieves documents from the CouchbaseQueryDocumentStore using vector similarity search.

Works with both Hyperscale Vector Index and Composite Vector Index. Supports ANN (approximate) and KNN (exact) search with various similarity metrics. See CouchbaseQueryDocumentStore for more details.

Usage example:

import numpy as np
from couchbase_haystack import (
    CouchbaseQueryDocumentStore,
    CouchbaseQueryEmbeddingRetriever,
    CouchbasePasswordAuthenticator,
    QueryVectorSearchType,
    CouchbaseQueryOptions,
    QueryVectorSearchSimilarity
)
from haystack.utils import Secret

# Assume a Couchbase GSI index named "vector_gsi_index" exists on the "embedding" field
# with dimension 768 and using cosine similarity.
store = CouchbaseQueryDocumentStore(
    cluster_connection_string=Secret.from_env_var("CB_CONNECTION_STRING"),
    authenticator=CouchbasePasswordAuthenticator(
        username=Secret.from_env_var("CB_USERNAME"),
        password=Secret.from_env_var("CB_PASSWORD")
    ),
    bucket="haystack_test_bucket",
    scope="scope_name",
    collection="collection_name",
    search_type=QueryVectorSearchType.ANN, # Or KNN depending on index
    similarity=QueryVectorSearchSimilarity.COSINE, # Or DOT, L2, EUCLIDEAN, L2_SQUARED or EUCLIDEAN_SQUARED
    nprobes=10, # optional Number of probes for the ANN search
    query_options=CouchbaseQueryOptions() # Optional query options
)
retriever = CouchbaseQueryEmbeddingRetriever(document_store=store, top_k=5)

# Generate a random query embedding matching the dimension
random_embedding = np.random.rand(768).tolist()

# Example without filters
results_no_filter = retriever.run(query_embedding=random_embedding)
print("Documents found without filters:", results_no_filter["documents"])

# Example with filters
filters = {"field": "meta.genre", "operator": "==", "value": "fiction"}
results_with_filter = retriever.run(query_embedding=random_embedding, filters=filters)
print("Documents found with filters:", results_with_filter["documents"])

The example above retrieves the 5 most similar documents to a random query embedding from the CouchbaseQueryDocumentStore. Note that the dimensions of the query_embedding must match the dimensions configured in the query_vector_search_params of the CouchbaseQueryDocumentStore. Filters are applied before the vector search.

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
@component
class CouchbaseQueryEmbeddingRetriever:
    """Retrieves documents from the CouchbaseQueryDocumentStore using vector similarity search.

    Works with both Hyperscale Vector Index and Composite Vector Index.
    Supports ANN (approximate) and KNN (exact) search with various similarity metrics.
    See CouchbaseQueryDocumentStore for more details.

    Usage example:

    ```python
    import numpy as np
    from couchbase_haystack import (
        CouchbaseQueryDocumentStore,
        CouchbaseQueryEmbeddingRetriever,
        CouchbasePasswordAuthenticator,
        QueryVectorSearchType,
        CouchbaseQueryOptions,
        QueryVectorSearchSimilarity
    )
    from haystack.utils import Secret

    # Assume a Couchbase GSI index named "vector_gsi_index" exists on the "embedding" field
    # with dimension 768 and using cosine similarity.
    store = CouchbaseQueryDocumentStore(
        cluster_connection_string=Secret.from_env_var("CB_CONNECTION_STRING"),
        authenticator=CouchbasePasswordAuthenticator(
            username=Secret.from_env_var("CB_USERNAME"),
            password=Secret.from_env_var("CB_PASSWORD")
        ),
        bucket="haystack_test_bucket",
        scope="scope_name",
        collection="collection_name",
        search_type=QueryVectorSearchType.ANN, # Or KNN depending on index
        similarity=QueryVectorSearchSimilarity.COSINE, # Or DOT, L2, EUCLIDEAN, L2_SQUARED or EUCLIDEAN_SQUARED
        nprobes=10, # optional Number of probes for the ANN search
        query_options=CouchbaseQueryOptions() # Optional query options
    )
    retriever = CouchbaseQueryEmbeddingRetriever(document_store=store, top_k=5)

    # Generate a random query embedding matching the dimension
    random_embedding = np.random.rand(768).tolist()

    # Example without filters
    results_no_filter = retriever.run(query_embedding=random_embedding)
    print("Documents found without filters:", results_no_filter["documents"])

    # Example with filters
    filters = {"field": "meta.genre", "operator": "==", "value": "fiction"}
    results_with_filter = retriever.run(query_embedding=random_embedding, filters=filters)
    print("Documents found with filters:", results_with_filter["documents"])
    ```

    The example above retrieves the 5 most similar documents to a random query embedding from the
    CouchbaseQueryDocumentStore. Note that the dimensions of the `query_embedding` must match the dimensions
    configured in the `query_vector_search_params` of the `CouchbaseQueryDocumentStore`.
    Filters are applied before the vector search.
    """

    def __init__(
        self,
        *,
        document_store: CouchbaseQueryDocumentStore,
        top_k: int = 10,
    ):
        """Create the CouchbaseQueryEmbeddingRetriever component.

        Args:
            document_store: An instance of CouchbaseQueryDocumentStore.
            top_k: Maximum number of Documents to return based on vector similarity.

        Raises:
            ValueError: If document_store is not an instance of CouchbaseQueryDocumentStore.
        """
        if not isinstance(document_store, CouchbaseQueryDocumentStore):
            msg = "document_store must be an instance of CouchbaseQueryDocumentStore"
            raise ValueError(msg)

        self.document_store = document_store
        self.top_k = top_k

    def to_dict(self) -> Dict[str, Any]:
        """Serializes the component to a dictionary.

        Returns:
            Dictionary with serialized data.
        """
        return default_to_dict(
            self,
            top_k=self.top_k,
            document_store=self.document_store.to_dict(),
        )

    @classmethod
    def from_dict(cls, data: Dict[str, Any]) -> "CouchbaseQueryEmbeddingRetriever":
        """Deserializes the component from a dictionary.

        Args:
            data: Dictionary to deserialize from.

        Returns:
            Deserialized component.
        """
        data["init_parameters"]["document_store"] = CouchbaseQueryDocumentStore.from_dict(
            data["init_parameters"]["document_store"]
        )
        return default_from_dict(cls, data)

    @component.output_types(documents=List[Document])
    def run(
        self,
        query_embedding: List[float],
        top_k: Optional[int] = None,
        filters: Optional[Dict[str, Any]] = None,
        nprobes: Optional[int] = None,
    ) -> Dict[str, List[Document]]:
        """Retrieve documents from the CouchbaseQueryDocumentStore based on embedding similarity using GSI.

        Args:
            query_embedding: Embedding of the query.
            top_k: Maximum number of Documents to be returned based on similarity score.
                   Overrides the value specified at initialization.
            filters: Optional dictionary of filters to apply before the vector search.
                     Refer to Haystack documentation for filter structure (https://docs.haystack.deepset.ai/v2.0/docs/metadata-filtering).
            nprobes: Number of probes for the ANN search. If None, uses the value set at index creation time
            or the value set at the document store level.
        Returns:
            A dictionary with the following keys:
            - documents: List of Documents most similar to the given `query_embedding`, potentially filtered.
        """
        top_k = top_k or self.top_k

        docs = self.document_store._embedding_retrieval(
            query_embedding=query_embedding, top_k=top_k, filters=filters, nprobes=nprobes
        )
        return {"documents": docs}

__init__ ¤

__init__(*, document_store: CouchbaseQueryDocumentStore, top_k: int = 10)

Parameters:

  • document_store (CouchbaseQueryDocumentStore) –

    An instance of CouchbaseQueryDocumentStore.

  • top_k (int, default: 10 ) –

    Maximum number of Documents to return based on vector similarity.

Raises:

  • ValueError –

    If document_store is not an instance of CouchbaseQueryDocumentStore.

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
def __init__(
    self,
    *,
    document_store: CouchbaseQueryDocumentStore,
    top_k: int = 10,
):
    """Create the CouchbaseQueryEmbeddingRetriever component.

    Args:
        document_store: An instance of CouchbaseQueryDocumentStore.
        top_k: Maximum number of Documents to return based on vector similarity.

    Raises:
        ValueError: If document_store is not an instance of CouchbaseQueryDocumentStore.
    """
    if not isinstance(document_store, CouchbaseQueryDocumentStore):
        msg = "document_store must be an instance of CouchbaseQueryDocumentStore"
        raise ValueError(msg)

    self.document_store = document_store
    self.top_k = top_k

to_dict ¤

to_dict() -> Dict[str, Any]

Serializes the component to a dictionary.

Returns:

  • Dict[str, Any] –

    Dictionary with serialized data.

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
220
221
222
223
224
225
226
227
228
229
230
def to_dict(self) -> Dict[str, Any]:
    """Serializes the component to a dictionary.

    Returns:
        Dictionary with serialized data.
    """
    return default_to_dict(
        self,
        top_k=self.top_k,
        document_store=self.document_store.to_dict(),
    )

from_dict classmethod ¤

from_dict(data: Dict[str, Any]) -> CouchbaseQueryEmbeddingRetriever

Deserializes the component from a dictionary.

Parameters:

  • data (Dict[str, Any]) –

    Dictionary to deserialize from.

Returns:

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
232
233
234
235
236
237
238
239
240
241
242
243
244
245
@classmethod
def from_dict(cls, data: Dict[str, Any]) -> "CouchbaseQueryEmbeddingRetriever":
    """Deserializes the component from a dictionary.

    Args:
        data: Dictionary to deserialize from.

    Returns:
        Deserialized component.
    """
    data["init_parameters"]["document_store"] = CouchbaseQueryDocumentStore.from_dict(
        data["init_parameters"]["document_store"]
    )
    return default_from_dict(cls, data)

run ¤

run(
    query_embedding: List[float],
    top_k: Optional[int] = None,
    filters: Optional[Dict[str, Any]] = None,
    nprobes: Optional[int] = None,
) -> Dict[str, List[Document]]

Retrieve documents from the CouchbaseQueryDocumentStore based on embedding similarity using GSI.

Parameters:

  • query_embedding (List[float]) –

    Embedding of the query.

  • top_k (Optional[int], default: None ) –

    Maximum number of Documents to be returned based on similarity score. Overrides the value specified at initialization.

  • filters (Optional[Dict[str, Any]], default: None ) –

    Optional dictionary of filters to apply before the vector search. Refer to Haystack documentation for filter structure (https://docs.haystack.deepset.ai/v2.0/docs/metadata-filtering).

  • nprobes (Optional[int], default: None ) –

    Number of probes for the ANN search. If None, uses the value set at index creation time

Returns: A dictionary with the following keys: - documents: List of Documents most similar to the given query_embedding, potentially filtered.

Source code in src/couchbase_haystack/components/retrievers/embedding_retriever.py 
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
@component.output_types(documents=List[Document])
def run(
    self,
    query_embedding: List[float],
    top_k: Optional[int] = None,
    filters: Optional[Dict[str, Any]] = None,
    nprobes: Optional[int] = None,
) -> Dict[str, List[Document]]:
    """Retrieve documents from the CouchbaseQueryDocumentStore based on embedding similarity using GSI.

    Args:
        query_embedding: Embedding of the query.
        top_k: Maximum number of Documents to be returned based on similarity score.
               Overrides the value specified at initialization.
        filters: Optional dictionary of filters to apply before the vector search.
                 Refer to Haystack documentation for filter structure (https://docs.haystack.deepset.ai/v2.0/docs/metadata-filtering).
        nprobes: Number of probes for the ANN search. If None, uses the value set at index creation time
        or the value set at the document store level.
    Returns:
        A dictionary with the following keys:
        - documents: List of Documents most similar to the given `query_embedding`, potentially filtered.
    """
    top_k = top_k or self.top_k

    docs = self.document_store._embedding_retrieval(
        query_embedding=query_embedding, top_k=top_k, filters=filters, nprobes=nprobes
    )
    return {"documents": docs}