from __future__ import annotations
import logging
from typing import Any, Callable, Dict, Iterable, List, Optional, Tuple, Union
import numpy as np
try:
import deeplake
from deeplake import VectorStore as DeepLakeVectorStore
from deeplake.core.fast_forwarding import version_compare
from deeplake.util.exceptions import SampleExtendError
_DEEPLAKE_INSTALLED = True
except ImportError:
_DEEPLAKE_INSTALLED = False
from langchain_core.documents import Document
from langchain_core.embeddings import Embeddings
from langchain_core.vectorstores import VectorStore
from langchain_community.vectorstores.utils import maximal_marginal_relevance
logger = logging.getLogger(__name__)
[docs]
class DeepLake(VectorStore):
"""`Activeloop Deep Lake` vector store.
We integrated deeplake's similarity search and filtering for fast prototyping.
Now, it supports Tensor Query Language (TQL) for production use cases
over billion rows.
Why Deep Lake?
- Not only stores embeddings, but also the original data with version control.
- Serverless, doesn't require another service and can be used with major
cloud providers (S3, GCS, etc.)
- More than just a multi-modal vector store. You can use the dataset
to fine-tune your own LLM models.
To use, you should have the ``deeplake`` python package installed.
Example:
.. code-block:: python
from langchain_community.vectorstores import DeepLake
from langchain_community.embeddings.openai import OpenAIEmbeddings
embeddings = OpenAIEmbeddings()
vectorstore = DeepLake("langchain_store", embeddings.embed_query)
"""
_LANGCHAIN_DEFAULT_DEEPLAKE_PATH: str = "./deeplake/"
_valid_search_kwargs = ["lambda_mult"]
[docs]
def __init__(
self,
dataset_path: str = _LANGCHAIN_DEFAULT_DEEPLAKE_PATH,
token: Optional[str] = None,
embedding: Optional[Embeddings] = None,
embedding_function: Optional[Embeddings] = None,
read_only: bool = False,
ingestion_batch_size: int = 1024,
num_workers: int = 0,
verbose: bool = True,
exec_option: Optional[str] = None,
runtime: Optional[Dict] = None,
index_params: Optional[Dict[str, Union[int, str]]] = None,
**kwargs: Any,
) -> None:
"""Creates an empty DeepLakeVectorStore or loads an existing one.
The DeepLakeVectorStore is located at the specified ``path``.
Examples:
>>> # Create a vector store with default tensors
>>> deeplake_vectorstore = DeepLake(
... path = <path_for_storing_Data>,
... )
>>>
>>> # Create a vector store in the Deep Lake Managed Tensor Database
>>> data = DeepLake(
... path = "hub://org_id/dataset_name",
... runtime = {"tensor_db": True},
... )
Args:
dataset_path (str): The full path for storing to the Deep Lake
Vector Store. It can be:
- a Deep Lake cloud path of the form ``hub://org_id/dataset_name``.
Requires registration with Deep Lake.
- an s3 path of the form ``s3://bucketname/path/to/dataset``.
Credentials are required in either the environment or passed to
the creds argument.
- a local file system path of the form ``./path/to/dataset``
or ``~/path/to/dataset`` or ``path/to/dataset``.
- a memory path of the form ``mem://path/to/dataset`` which doesn't
save the dataset but keeps it in memory instead.
Should be used only for testing as it does not persist.
Defaults to _LANGCHAIN_DEFAULT_DEEPLAKE_PATH.
token (str, optional): Activeloop token, for fetching credentials
to the dataset at path if it is a Deep Lake dataset.
Tokens are normally autogenerated. Optional.
embedding (Embeddings, optional): Function to convert
either documents or query. Optional.
embedding_function (Embeddings, optional): Function to convert
either documents or query. Optional. Deprecated: keeping this
parameter for backwards compatibility.
read_only (bool): Open dataset in read-only mode. Default is False.
ingestion_batch_size (int): During data ingestion, data is divided
into batches. Batch size is the size of each batch.
Default is 1024.
num_workers (int): Number of workers to use during data ingestion.
Default is 0.
verbose (bool): Print dataset summary after each operation.
Default is True.
exec_option (str, optional): Default method for search execution.
It could be either ``"auto"``, ``"python"``, ``"compute_engine"``
or ``"tensor_db"``. Defaults to ``"auto"``.
If None, it's set to "auto".
- ``auto``- Selects the best execution method based on the storage
location of the Vector Store. It is the default option.
- ``python`` - Pure-python implementation that runs on the client and
can be used for data stored anywhere. WARNING: using this option
with big datasets is discouraged because it can lead to
memory issues.
- ``compute_engine`` - Performant C++ implementation of the Deep Lake
Compute Engine that runs on the client and can be used for any data
stored in or connected to Deep Lake. It cannot be used with
in-memory or local datasets.
- ``tensor_db`` - Performant and fully-hosted Managed Tensor Database
that is responsible for storage and query execution. Only available
for data stored in the Deep Lake Managed Database. Store datasets
in this database by specifying runtime = {"tensor_db": True}
during dataset creation.
runtime (Dict, optional): Parameters for creating the Vector Store in
Deep Lake's Managed Tensor Database. Not applicable when loading an
existing Vector Store. To create a Vector Store in the Managed Tensor
Database, set `runtime = {"tensor_db": True}`.
index_params (Optional[Dict[str, Union[int, str]]], optional): Dictionary
containing information about vector index that will be created. Defaults
to None, which will utilize ``DEFAULT_VECTORSTORE_INDEX_PARAMS`` from
``deeplake.constants``. The specified key-values override the default
ones.
- threshold: The threshold for the dataset size above which an index
will be created for the embedding tensor. When the threshold value
is set to -1, index creation is turned off. Defaults to -1, which
turns off the index.
- distance_metric: This key specifies the method of calculating the
distance between vectors when creating the vector database (VDB)
index. It can either be a string that corresponds to a member of
the DistanceType enumeration, or the string value itself.
- If no value is provided, it defaults to "L2".
- "L2" corresponds to DistanceType.L2_NORM.
- "COS" corresponds to DistanceType.COSINE_SIMILARITY.
- additional_params: Additional parameters for fine-tuning the index.
**kwargs: Other optional keyword arguments.
Raises:
ValueError: If some condition is not met.
"""
self.ingestion_batch_size = ingestion_batch_size
self.num_workers = num_workers
self.verbose = verbose
if _DEEPLAKE_INSTALLED is False:
raise ImportError(
"Could not import deeplake python package. "
"Please install it with `pip install deeplake[enterprise]`."
)
if (
runtime == {"tensor_db": True}
and version_compare(deeplake.__version__, "3.6.7") == -1
):
raise ImportError(
"To use tensor_db option you need to update deeplake to `3.6.7` or "
"higher. "
f"Currently installed deeplake version is {deeplake.__version__}. "
)
self.dataset_path = dataset_path
if embedding_function:
logger.warning(
"Using embedding function is deprecated and will be removed "
"in the future. Please use embedding instead."
)
self.vectorstore = DeepLakeVectorStore(
path=self.dataset_path,
embedding_function=embedding_function or embedding,
read_only=read_only,
token=token,
exec_option=exec_option,
verbose=verbose,
runtime=runtime,
index_params=index_params,
**kwargs,
)
self._embedding_function = embedding_function or embedding
self._id_tensor_name = "ids" if "ids" in self.vectorstore.tensors() else "id"
@property
def embeddings(self) -> Optional[Embeddings]:
return self._embedding_function
[docs]
def add_texts(
self,
texts: Iterable[str],
metadatas: Optional[List[dict]] = None,
ids: Optional[List[str]] = None,
**kwargs: Any,
) -> List[str]:
"""Run more texts through the embeddings and add to the vectorstore.
Examples:
>>> ids = deeplake_vectorstore.add_texts(
... texts = <list_of_texts>,
... metadatas = <list_of_metadata_jsons>,
... ids = <list_of_ids>,
... )
Args:
texts (Iterable[str]): Texts to add to the vectorstore.
metadatas (Optional[List[dict]], optional): Optional list of metadatas.
ids (Optional[List[str]], optional): Optional list of IDs.
embedding_function (Optional[Embeddings], optional): Embedding function
to use to convert the text into embeddings.
**kwargs (Any): Any additional keyword arguments passed is not supported
by this method.
Returns:
List[str]: List of IDs of the added texts.
"""
self._validate_kwargs(kwargs, "add_texts")
kwargs = {}
if ids:
if self._id_tensor_name == "ids": # for backwards compatibility
kwargs["ids"] = ids
else:
kwargs["id"] = ids
if metadatas is None:
metadatas = [{}] * len(list(texts))
if not isinstance(texts, list):
texts = list(texts)
if texts is None:
raise ValueError("`texts` parameter shouldn't be None.")
elif len(texts) == 0:
raise ValueError("`texts` parameter shouldn't be empty.")
try:
return self.vectorstore.add(
text=texts,
metadata=metadatas,
embedding_data=texts,
embedding_tensor="embedding",
embedding_function=self._embedding_function.embed_documents, # type: ignore
return_ids=True,
**kwargs,
)
except SampleExtendError as e:
if "Failed to append a sample to the tensor 'metadata'" in str(e):
msg = (
"**Hint: You might be using invalid type of argument in "
"document loader (e.g. 'pathlib.PosixPath' instead of 'str')"
)
raise ValueError(e.args[0] + "\n\n" + msg)
else:
raise e
def _search_tql(
self,
tql: Optional[str],
exec_option: Optional[str] = None,
**kwargs: Any,
) -> List[Document]:
"""Function for performing tql_search.
Args:
tql (str): TQL Query string for direct evaluation.
Available only for `compute_engine` and `tensor_db`.
exec_option (str, optional): Supports 3 ways to search.
Could be "python", "compute_engine" or "tensor_db". Default is "python".
- ``python`` - Pure-python implementation for the client.
WARNING: not recommended for big datasets due to potential memory
issues.
- ``compute_engine`` - C++ implementation of Deep Lake Compute
Engine for the client. Not for in-memory or local datasets.
- ``tensor_db`` - Hosted Managed Tensor Database for storage
and query execution. Only for data in Deep Lake Managed Database.
Use runtime = {"db_engine": True} during dataset creation.
return_score (bool): Return score with document. Default is False.
Returns:
Tuple[List[Document], List[Tuple[Document, float]]] - A tuple of two lists.
The first list contains Documents, and the second list contains
tuples of Document and float score.
Raises:
ValueError: If return_score is True but some condition is not met.
"""
result = self.vectorstore.search(
query=tql,
exec_option=exec_option,
)
metadatas = result["metadata"]
texts = result["text"]
docs = [
Document(
page_content=text,
metadata=metadata,
)
for text, metadata in zip(texts, metadatas)
]
if kwargs:
unsupported_argument = next(iter(kwargs))
if kwargs[unsupported_argument] is not False:
raise ValueError(
f"specifying {unsupported_argument} is "
"not supported with tql search."
)
return docs
def _search(
self,
query: Optional[str] = None,
embedding: Optional[Union[List[float], np.ndarray]] = None,
embedding_function: Optional[Callable] = None,
k: int = 4,
distance_metric: Optional[str] = None,
use_maximal_marginal_relevance: bool = False,
fetch_k: Optional[int] = 20,
filter: Optional[Union[Dict, Callable]] = None,
return_score: bool = False,
exec_option: Optional[str] = None,
deep_memory: bool = False,
**kwargs: Any,
) -> Any[List[Document], List[Tuple[Document, float]]]:
"""
Return docs similar to query.
Args:
query (str, optional): Text to look up similar docs.
embedding (Union[List[float], np.ndarray], optional): Query's embedding.
embedding_function (Callable, optional): Function to convert `query`
into embedding.
k (int): Number of Documents to return.
distance_metric (Optional[str], optional): `L2` for Euclidean, `L1` for
Nuclear, `max` for L-infinity distance, `cos` for cosine similarity,
'dot' for dot product.
filter (Union[Dict, Callable], optional): Additional filter prior
to the embedding search.
- ``Dict`` - Key-value search on tensors of htype json, on an
AND basis (a sample must satisfy all key-value filters to be True)
Dict = {"tensor_name_1": {"key": value},
"tensor_name_2": {"key": value}}
- ``Function`` - Any function compatible with `deeplake.filter`.
use_maximal_marginal_relevance (bool): Use maximal marginal relevance.
fetch_k (int): Number of Documents for MMR algorithm.
return_score (bool): Return the score.
exec_option (str, optional): Supports 3 ways to perform searching.
Could be "python", "compute_engine" or "tensor_db".
- ``python`` - Pure-python implementation for the client.
WARNING: not recommended for big datasets.
- ``compute_engine`` - C++ implementation of Deep Lake Compute
Engine for the client. Not for in-memory or local datasets.
- ``tensor_db`` - Hosted Managed Tensor Database for storage
and query execution. Only for data in Deep Lake Managed Database.
Use runtime = {"db_engine": True} during dataset creation.
deep_memory (bool): Whether to use the Deep Memory model for improving
search results. Defaults to False if deep_memory is not specified in
the Vector Store initialization. If True, the distance metric is set
to "deepmemory_distance", which represents the metric with which the
model was trained. The search is performed using the Deep Memory model.
If False, the distance metric is set to "COS" or whatever distance
metric user specifies.
kwargs: Additional keyword arguments.
Returns:
List of Documents by the specified distance metric,
if return_score True, return a tuple of (Document, score)
Raises:
ValueError: if both `embedding` and `embedding_function` are not specified.
"""
if kwargs.get("tql_query"):
logger.warning("`tql_query` is deprecated. Please use `tql` instead.")
kwargs["tql"] = kwargs.pop("tql_query")
if kwargs.get("tql"):
return self._search_tql(
tql=kwargs["tql"],
exec_option=exec_option,
return_score=return_score,
embedding=embedding,
embedding_function=embedding_function,
distance_metric=distance_metric,
use_maximal_marginal_relevance=use_maximal_marginal_relevance,
filter=filter,
)
self._validate_kwargs(kwargs, "search")
if embedding_function:
if isinstance(embedding_function, Embeddings):
_embedding_function = embedding_function.embed_query
else:
_embedding_function = embedding_function
elif self._embedding_function:
_embedding_function = self._embedding_function.embed_query
else:
_embedding_function = None
if embedding is None:
if _embedding_function is None:
raise ValueError(
"Either `embedding` or `embedding_function` needs to be"
" specified."
)
embedding = _embedding_function(query) if query else None
if isinstance(embedding, list):
embedding = np.array(embedding, dtype=np.float32)
if len(embedding.shape) > 1:
embedding = embedding[0]
result = self.vectorstore.search(
embedding=embedding,
k=fetch_k if use_maximal_marginal_relevance else k,
distance_metric=distance_metric,
filter=filter,
exec_option=exec_option,
return_tensors=["embedding", "metadata", "text", self._id_tensor_name],
deep_memory=deep_memory,
)
scores = result["score"]
embeddings = result["embedding"]
metadatas = result["metadata"]
texts = result["text"]
if use_maximal_marginal_relevance:
lambda_mult = kwargs.get("lambda_mult", 0.5)
indices = maximal_marginal_relevance( # type: ignore
embedding, # type: ignore
embeddings,
k=min(k, len(texts)),
lambda_mult=lambda_mult,
)
scores = [scores[i] for i in indices]
texts = [texts[i] for i in indices]
metadatas = [metadatas[i] for i in indices]
docs = [
Document(
page_content=text,
metadata=metadata,
)
for text, metadata in zip(texts, metadatas)
]
if return_score:
if not isinstance(scores, list):
scores = [scores]
return [(doc, score) for doc, score in zip(docs, scores)]
return docs
[docs]
def similarity_search(
self,
query: str,
k: int = 4,
**kwargs: Any,
) -> List[Document]:
"""
Return docs most similar to query.
Examples:
>>> # Search using an embedding
>>> data = vector_store.similarity_search(
... query=<your_query>,
... k=<num_items>,
... exec_option=<preferred_exec_option>,
... )
>>> # Run tql search:
>>> data = vector_store.similarity_search(
... query=None,
... tql="SELECT * WHERE id == <id>",
... exec_option="compute_engine",
... )
Args:
k (int): Number of Documents to return. Defaults to 4.
query (str): Text to look up similar documents.
kwargs: Additional keyword arguments include:
embedding (Callable): Embedding function to use. Defaults to None.
distance_metric (str): 'L2' for Euclidean, 'L1' for Nuclear, 'max'
for L-infinity, 'cos' for cosine, 'dot' for dot product.
Defaults to 'L2'.
filter (Union[Dict, Callable], optional): Additional filter
before embedding search.
- Dict: Key-value search on tensors of htype json,
(sample must satisfy all key-value filters)
Dict = {"tensor_1": {"key": value}, "tensor_2": {"key": value}}
- Function: Compatible with `deeplake.filter`.
Defaults to None.
exec_option (str): Supports 3 ways to perform searching.
'python', 'compute_engine', or 'tensor_db'. Defaults to 'python'.
- 'python': Pure-python implementation for the client.
WARNING: not recommended for big datasets.
- 'compute_engine': C++ implementation of the Compute Engine for
the client. Not for in-memory or local datasets.
- 'tensor_db': Managed Tensor Database for storage and query.
Only for data in Deep Lake Managed Database.
Use `runtime = {"db_engine": True}` during dataset creation.
deep_memory (bool): Whether to use the Deep Memory model for improving
search results. Defaults to False if deep_memory is not specified
in the Vector Store initialization. If True, the distance metric
is set to "deepmemory_distance", which represents the metric with
which the model was trained. The search is performed using the Deep
Memory model. If False, the distance metric is set to "COS" or
whatever distance metric user specifies.
Returns:
List[Document]: List of Documents most similar to the query vector.
"""
return self._search(
query=query,
k=k,
use_maximal_marginal_relevance=False,
return_score=False,
**kwargs,
)
[docs]
def similarity_search_by_vector(
self,
embedding: Union[List[float], np.ndarray],
k: int = 4,
**kwargs: Any,
) -> List[Document]:
"""
Return docs most similar to embedding vector.
Examples:
>>> # Search using an embedding
>>> data = vector_store.similarity_search_by_vector(
... embedding=<your_embedding>,
... k=<num_items_to_return>,
... exec_option=<preferred_exec_option>,
... )
Args:
embedding (Union[List[float], np.ndarray]):
Embedding to find similar docs.
k (int): Number of Documents to return. Defaults to 4.
kwargs: Additional keyword arguments including:
filter (Union[Dict, Callable], optional):
Additional filter before embedding search.
- ``Dict`` - Key-value search on tensors of htype json. True
if all key-value filters are satisfied.
Dict = {"tensor_name_1": {"key": value},
"tensor_name_2": {"key": value}}
- ``Function`` - Any function compatible with
`deeplake.filter`.
Defaults to None.
exec_option (str): Options for search execution include
"python", "compute_engine", or "tensor_db". Defaults to
"python".
- "python" - Pure-python implementation running on the client.
Can be used for data stored anywhere. WARNING: using this
option with big datasets is discouraged due to potential
memory issues.
- "compute_engine" - Performant C++ implementation of the Deep
Lake Compute Engine. Runs on the client and can be used for
any data stored in or connected to Deep Lake. It cannot be
used with in-memory or local datasets.
- "tensor_db" - Performant, fully-hosted Managed Tensor Database.
Responsible for storage and query execution. Only available
for data stored in the Deep Lake Managed Database.
To store datasets in this database, specify
`runtime = {"db_engine": True}` during dataset creation.
distance_metric (str): `L2` for Euclidean, `L1` for Nuclear,
`max` for L-infinity distance, `cos` for cosine similarity,
'dot' for dot product. Defaults to `L2`.
deep_memory (bool): Whether to use the Deep Memory model for improving
search results. Defaults to False if deep_memory is not specified
in the Vector Store initialization. If True, the distance metric
is set to "deepmemory_distance", which represents the metric with
which the model was trained. The search is performed using the Deep
Memory model. If False, the distance metric is set to "COS" or
whatever distance metric user specifies.
Returns:
List[Document]: List of Documents most similar to the query vector.
"""
return self._search(
embedding=embedding,
k=k,
use_maximal_marginal_relevance=False,
return_score=False,
**kwargs,
)
[docs]
def similarity_search_with_score(
self,
query: str,
k: int = 4,
**kwargs: Any,
) -> List[Tuple[Document, float]]:
"""
Run similarity search with Deep Lake with distance returned.
Examples:
>>> data = vector_store.similarity_search_with_score(
... query=<your_query>,
... embedding=<your_embedding_function>
... k=<number_of_items_to_return>,
... exec_option=<preferred_exec_option>,
... )
Args:
query (str): Query text to search for.
k (int): Number of results to return. Defaults to 4.
kwargs: Additional keyword arguments. Some of these arguments are:
distance_metric: `L2` for Euclidean, `L1` for Nuclear, `max` L-infinity
distance, `cos` for cosine similarity, 'dot' for dot product.
Defaults to `L2`.
filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None.
embedding_function (Callable): Embedding function to use. Defaults
to None.
exec_option (str): DeepLakeVectorStore supports 3 ways to perform
searching. It could be either "python", "compute_engine" or
"tensor_db". Defaults to "python".
- "python" - Pure-python implementation running on the client.
Can be used for data stored anywhere. WARNING: using this
option with big datasets is discouraged due to potential
memory issues.
- "compute_engine" - Performant C++ implementation of the Deep
Lake Compute Engine. Runs on the client and can be used for
any data stored in or connected to Deep Lake. It cannot be used
with in-memory or local datasets.
- "tensor_db" - Performant, fully-hosted Managed Tensor Database.
Responsible for storage and query execution. Only available for
data stored in the Deep Lake Managed Database. To store datasets
in this database, specify `runtime = {"db_engine": True}`
during dataset creation.
deep_memory (bool): Whether to use the Deep Memory model for improving
search results. Defaults to False if deep_memory is not specified
in the Vector Store initialization. If True, the distance metric
is set to "deepmemory_distance", which represents the metric with
which the model was trained. The search is performed using the Deep
Memory model. If False, the distance metric is set to "COS" or
whatever distance metric user specifies.
Returns:
List[Tuple[Document, float]]: List of documents most similar to the query
text with distance in float."""
return self._search(
query=query,
k=k,
return_score=True,
**kwargs,
)
[docs]
def max_marginal_relevance_search_by_vector(
self,
embedding: List[float],
k: int = 4,
fetch_k: int = 20,
lambda_mult: float = 0.5,
exec_option: Optional[str] = None,
**kwargs: Any,
) -> List[Document]:
"""
Return docs selected using the maximal marginal relevance. Maximal marginal
relevance optimizes for similarity to query AND diversity among selected docs.
Examples:
>>> data = vector_store.max_marginal_relevance_search_by_vector(
... embedding=<your_embedding>,
... fetch_k=<elements_to_fetch_before_mmr_search>,
... k=<number_of_items_to_return>,
... exec_option=<preferred_exec_option>,
... )
Args:
embedding: Embedding to look up documents similar to.
k: Number of Documents to return. Defaults to 4.
fetch_k: Number of Documents to fetch for MMR algorithm.
lambda_mult: Number between 0 and 1 determining the degree of diversity.
0 corresponds to max diversity and 1 to min diversity. Defaults to 0.5.
exec_option (str): DeepLakeVectorStore supports 3 ways for searching.
Could be "python", "compute_engine" or "tensor_db". Defaults to
"python".
- "python" - Pure-python implementation running on the client.
Can be used for data stored anywhere. WARNING: using this
option with big datasets is discouraged due to potential
memory issues.
- "compute_engine" - Performant C++ implementation of the Deep
Lake Compute Engine. Runs on the client and can be used for
any data stored in or connected to Deep Lake. It cannot be used
with in-memory or local datasets.
- "tensor_db" - Performant, fully-hosted Managed Tensor Database.
Responsible for storage and query execution. Only available for
data stored in the Deep Lake Managed Database. To store datasets
in this database, specify `runtime = {"db_engine": True}`
during dataset creation.
deep_memory (bool): Whether to use the Deep Memory model for improving
search results. Defaults to False if deep_memory is not specified
in the Vector Store initialization. If True, the distance metric
is set to "deepmemory_distance", which represents the metric with
which the model was trained. The search is performed using the Deep
Memory model. If False, the distance metric is set to "COS" or
whatever distance metric user specifies.
kwargs: Additional keyword arguments.
Returns:
List[Documents] - A list of documents.
"""
return self._search(
embedding=embedding,
k=k,
fetch_k=fetch_k,
use_maximal_marginal_relevance=True,
lambda_mult=lambda_mult,
exec_option=exec_option,
**kwargs,
)
[docs]
def max_marginal_relevance_search(
self,
query: str,
k: int = 4,
fetch_k: int = 20,
lambda_mult: float = 0.5,
exec_option: Optional[str] = None,
**kwargs: Any,
) -> List[Document]:
"""Return docs selected using maximal marginal relevance.
Maximal marginal relevance optimizes for similarity to query AND diversity
among selected documents.
Examples:
>>> # Search using an embedding
>>> data = vector_store.max_marginal_relevance_search(
... query = <query_to_search>,
... embedding_function = <embedding_function_for_query>,
... k = <number_of_items_to_return>,
... exec_option = <preferred_exec_option>,
... )
Args:
query: Text to look up documents similar to.
k: Number of Documents to return. Defaults to 4.
fetch_k: Number of Documents for MMR algorithm.
lambda_mult: Value between 0 and 1. 0 corresponds
to maximum diversity and 1 to minimum.
Defaults to 0.5.
exec_option (str): Supports 3 ways to perform searching.
- "python" - Pure-python implementation running on the client.
Can be used for data stored anywhere. WARNING: using this
option with big datasets is discouraged due to potential
memory issues.
- "compute_engine" - Performant C++ implementation of the Deep
Lake Compute Engine. Runs on the client and can be used for
any data stored in or connected to Deep Lake. It cannot be
used with in-memory or local datasets.
- "tensor_db" - Performant, fully-hosted Managed Tensor Database.
Responsible for storage and query execution. Only available
for data stored in the Deep Lake Managed Database. To store
datasets in this database, specify
`runtime = {"db_engine": True}` during dataset creation.
deep_memory (bool): Whether to use the Deep Memory model for improving
search results. Defaults to False if deep_memory is not specified
in the Vector Store initialization. If True, the distance metric
is set to "deepmemory_distance", which represents the metric with
which the model was trained. The search is performed using the Deep
Memory model. If False, the distance metric is set to "COS" or
whatever distance metric user specifies.
kwargs: Additional keyword arguments
Returns:
List of Documents selected by maximal marginal relevance.
Raises:
ValueError: when MRR search is on but embedding function is
not specified.
"""
embedding_function = kwargs.get("embedding") or self._embedding_function
if embedding_function is None:
raise ValueError(
"For MMR search, you must specify an embedding function on"
" `creation` or during add call."
)
return self._search(
query=query,
k=k,
fetch_k=fetch_k,
use_maximal_marginal_relevance=True,
lambda_mult=lambda_mult,
exec_option=exec_option,
embedding_function=embedding_function, # type: ignore
**kwargs,
)
[docs]
@classmethod
def from_texts(
cls,
texts: List[str],
embedding: Optional[Embeddings] = None,
metadatas: Optional[List[dict]] = None,
ids: Optional[List[str]] = None,
dataset_path: str = _LANGCHAIN_DEFAULT_DEEPLAKE_PATH,
**kwargs: Any,
) -> DeepLake:
"""Create a Deep Lake dataset from a raw documents.
If a dataset_path is specified, the dataset will be persisted in that location,
otherwise by default at `./deeplake`
Examples:
>>> # Search using an embedding
>>> vector_store = DeepLake.from_texts(
... texts = <the_texts_that_you_want_to_embed>,
... embedding_function = <embedding_function_for_query>,
... k = <number_of_items_to_return>,
... exec_option = <preferred_exec_option>,
... )
Args:
dataset_path (str): - The full path to the dataset. Can be:
- Deep Lake cloud path of the form ``hub://username/dataset_name``.
To write to Deep Lake cloud datasets,
ensure that you are logged in to Deep Lake
(use 'activeloop login' from command line)
- AWS S3 path of the form ``s3://bucketname/path/to/dataset``.
Credentials are required in either the environment
- Google Cloud Storage path of the form
``gcs://bucketname/path/to/dataset`` Credentials are required
in either the environment
- Local file system path of the form ``./path/to/dataset`` or
``~/path/to/dataset`` or ``path/to/dataset``.
- In-memory path of the form ``mem://path/to/dataset`` which doesn't
save the dataset, but keeps it in memory instead.
Should be used only for testing as it does not persist.
texts (List[Document]): List of documents to add.
embedding (Optional[Embeddings]): Embedding function. Defaults to None.
Note, in other places, it is called embedding_function.
metadatas (Optional[List[dict]]): List of metadatas. Defaults to None.
ids (Optional[List[str]]): List of document IDs. Defaults to None.
kwargs: Additional keyword arguments.
Returns:
DeepLake: Deep Lake dataset.
"""
deeplake_dataset = cls(dataset_path=dataset_path, embedding=embedding, **kwargs)
deeplake_dataset.add_texts(
texts=texts,
metadatas=metadatas,
ids=ids,
)
return deeplake_dataset
[docs]
def delete(self, ids: Optional[List[str]] = None, **kwargs: Any) -> bool:
"""Delete the entities in the dataset.
Args:
ids (Optional[List[str]], optional): The document_ids to delete.
Defaults to None.
**kwargs: Other keyword arguments that subclasses might use.
- filter (Optional[Dict[str, str]], optional): The filter to delete by.
- delete_all (Optional[bool], optional): Whether to drop the dataset.
Returns:
bool: Whether the delete operation was successful.
"""
filter = kwargs.get("filter")
delete_all = kwargs.get("delete_all")
self.vectorstore.delete(ids=ids, filter=filter, delete_all=delete_all)
return True
[docs]
@classmethod
def force_delete_by_path(cls, path: str) -> None:
"""Force delete dataset by path.
Args:
path (str): path of the dataset to delete.
Raises:
ValueError: if deeplake is not installed.
"""
try:
import deeplake
except ImportError:
raise ImportError(
"Could not import deeplake python package. "
"Please install it with `pip install deeplake`."
)
deeplake.delete(path, large_ok=True, force=True)
[docs]
def delete_dataset(self) -> None:
"""Delete the collection."""
self.delete(delete_all=True)
[docs]
def ds(self) -> Any:
logger.warning(
"this method is deprecated and will be removed, "
"better to use `db.vectorstore.dataset` instead."
)
return self.vectorstore.dataset
@classmethod
def _validate_kwargs(cls, kwargs, method_name): # type: ignore[no-untyped-def]
if kwargs:
valid_items = cls._get_valid_args(method_name)
unsupported_items = cls._get_unsupported_items(kwargs, valid_items)
if unsupported_items:
raise TypeError(
f"`{unsupported_items}` are not a valid "
f"argument to {method_name} method"
)
@classmethod
def _get_valid_args(cls, method_name): # type: ignore[no-untyped-def]
if method_name == "search":
return cls._valid_search_kwargs
else:
return []
@staticmethod
def _get_unsupported_items(kwargs, valid_items): # type: ignore[no-untyped-def]
kwargs = {k: v for k, v in kwargs.items() if k not in valid_items}
unsupported_items = None
if kwargs:
unsupported_items = "`, `".join(set(kwargs.keys()))
return unsupported_items
