mirror of
https://github.com/hwchase17/langchain
synced 2024-11-18 09:25:54 +00:00
493afe4d8d
Implemented the ability to enable full-text search within the SingleStore vector store, offering users a versatile range of search strategies. This enhancement allows users to seamlessly combine full-text search with vector search, enabling the following search strategies: * Search solely by vector similarity. * Conduct searches exclusively based on text similarity, utilizing Lucene internally. * Filter search results by text similarity score, with the option to specify a threshold, followed by a search based on vector similarity. * Filter results by vector similarity score before conducting a search based on text similarity. * Perform searches using a weighted sum of vector and text similarity scores. Additionally, integration tests have been added to comprehensively cover all scenarios. Updated notebook with examples. CC: @baskaryan, @hwchase17 --------- Co-authored-by: Volodymyr Tkachuk <vtkachuk-ua@singlestore.com> Co-authored-by: Bagatur <baskaryan@gmail.com>
1001 lines
44 KiB
Python
1001 lines
44 KiB
Python
from __future__ import annotations
|
|
|
|
import json
|
|
import re
|
|
from enum import Enum
|
|
from typing import (
|
|
Any,
|
|
Callable,
|
|
Iterable,
|
|
List,
|
|
Optional,
|
|
Tuple,
|
|
Type,
|
|
)
|
|
|
|
from langchain_core.documents import Document
|
|
from langchain_core.embeddings import Embeddings
|
|
from langchain_core.vectorstores import VectorStore, VectorStoreRetriever
|
|
from sqlalchemy.pool import QueuePool
|
|
|
|
from langchain_community.vectorstores.utils import DistanceStrategy
|
|
|
|
DEFAULT_DISTANCE_STRATEGY = DistanceStrategy.DOT_PRODUCT
|
|
|
|
ORDERING_DIRECTIVE: dict = {
|
|
DistanceStrategy.EUCLIDEAN_DISTANCE: "",
|
|
DistanceStrategy.DOT_PRODUCT: "DESC",
|
|
}
|
|
|
|
|
|
class SingleStoreDB(VectorStore):
|
|
"""`SingleStore DB` vector store.
|
|
|
|
The prerequisite for using this class is the installation of the ``singlestoredb``
|
|
Python package.
|
|
|
|
The SingleStoreDB vectorstore can be created by providing an embedding function and
|
|
the relevant parameters for the database connection, connection pool, and
|
|
optionally, the names of the table and the fields to use.
|
|
"""
|
|
|
|
class SearchStrategy(str, Enum):
|
|
"""Enumerator of the Search strategies for searching in the vectorstore."""
|
|
|
|
VECTOR_ONLY = "VECTOR_ONLY"
|
|
TEXT_ONLY = "TEXT_ONLY"
|
|
FILTER_BY_TEXT = "FILTER_BY_TEXT"
|
|
FILTER_BY_VECTOR = "FILTER_BY_VECTOR"
|
|
WEIGHTED_SUM = "WEIGHTED_SUM"
|
|
|
|
def _get_connection(self: SingleStoreDB) -> Any:
|
|
try:
|
|
import singlestoredb as s2
|
|
except ImportError:
|
|
raise ImportError(
|
|
"Could not import singlestoredb python package. "
|
|
"Please install it with `pip install singlestoredb`."
|
|
)
|
|
return s2.connect(**self.connection_kwargs)
|
|
|
|
def __init__(
|
|
self,
|
|
embedding: Embeddings,
|
|
*,
|
|
distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY,
|
|
table_name: str = "embeddings",
|
|
content_field: str = "content",
|
|
metadata_field: str = "metadata",
|
|
vector_field: str = "vector",
|
|
id_field: str = "id",
|
|
use_vector_index: bool = False,
|
|
vector_index_name: str = "",
|
|
vector_index_options: Optional[dict] = None,
|
|
vector_size: int = 1536,
|
|
use_full_text_search: bool = False,
|
|
pool_size: int = 5,
|
|
max_overflow: int = 10,
|
|
timeout: float = 30,
|
|
**kwargs: Any,
|
|
):
|
|
"""Initialize with necessary components.
|
|
|
|
Args:
|
|
embedding (Embeddings): A text embedding model.
|
|
|
|
distance_strategy (DistanceStrategy, optional):
|
|
Determines the strategy employed for calculating
|
|
the distance between vectors in the embedding space.
|
|
Defaults to DOT_PRODUCT.
|
|
Available options are:
|
|
- DOT_PRODUCT: Computes the scalar product of two vectors.
|
|
This is the default behavior
|
|
- EUCLIDEAN_DISTANCE: Computes the Euclidean distance between
|
|
two vectors. This metric considers the geometric distance in
|
|
the vector space, and might be more suitable for embeddings
|
|
that rely on spatial relationships. This metric is not
|
|
compatible with the WEIGHTED_SUM search strategy.
|
|
|
|
table_name (str, optional): Specifies the name of the table in use.
|
|
Defaults to "embeddings".
|
|
content_field (str, optional): Specifies the field to store the content.
|
|
Defaults to "content".
|
|
metadata_field (str, optional): Specifies the field to store metadata.
|
|
Defaults to "metadata".
|
|
vector_field (str, optional): Specifies the field to store the vector.
|
|
Defaults to "vector".
|
|
id_field (str, optional): Specifies the field to store the id.
|
|
Defaults to "id".
|
|
|
|
use_vector_index (bool, optional): Toggles the use of a vector index.
|
|
Works only with SingleStoreDB 8.5 or later. Defaults to False.
|
|
If set to True, vector_size parameter is required to be set to
|
|
a proper value.
|
|
|
|
vector_index_name (str, optional): Specifies the name of the vector index.
|
|
Defaults to empty. Will be ignored if use_vector_index is set to False.
|
|
|
|
vector_index_options (dict, optional): Specifies the options for
|
|
the vector index. Defaults to {}.
|
|
Will be ignored if use_vector_index is set to False. The options are:
|
|
index_type (str, optional): Specifies the type of the index.
|
|
Defaults to IVF_PQFS.
|
|
For more options, please refer to the SingleStoreDB documentation:
|
|
https://docs.singlestore.com/cloud/reference/sql-reference/vector-functions/vector-indexing/
|
|
|
|
vector_size (int, optional): Specifies the size of the vector.
|
|
Defaults to 1536. Required if use_vector_index is set to True.
|
|
Should be set to the same value as the size of the vectors
|
|
stored in the vector_field.
|
|
|
|
use_full_text_search (bool, optional): Toggles the use a full-text index
|
|
on the document content. Defaults to False. If set to True, the table
|
|
will be created with a full-text index on the content field,
|
|
and the simularity_search method will all using TEXT_ONLY,
|
|
FILTER_BY_TEXT, FILTER_BY_VECTOR, and WIGHTED_SUM search strategies.
|
|
If set to False, the simularity_search method will only allow
|
|
VECTOR_ONLY search strategy.
|
|
|
|
Following arguments pertain to the connection pool:
|
|
|
|
pool_size (int, optional): Determines the number of active connections in
|
|
the pool. Defaults to 5.
|
|
max_overflow (int, optional): Determines the maximum number of connections
|
|
allowed beyond the pool_size. Defaults to 10.
|
|
timeout (float, optional): Specifies the maximum wait time in seconds for
|
|
establishing a connection. Defaults to 30.
|
|
|
|
Following arguments pertain to the database connection:
|
|
|
|
host (str, optional): Specifies the hostname, IP address, or URL for the
|
|
database connection. The default scheme is "mysql".
|
|
user (str, optional): Database username.
|
|
password (str, optional): Database password.
|
|
port (int, optional): Database port. Defaults to 3306 for non-HTTP
|
|
connections, 80 for HTTP connections, and 443 for HTTPS connections.
|
|
database (str, optional): Database name.
|
|
|
|
Additional optional arguments provide further customization over the
|
|
database connection:
|
|
|
|
pure_python (bool, optional): Toggles the connector mode. If True,
|
|
operates in pure Python mode.
|
|
local_infile (bool, optional): Allows local file uploads.
|
|
charset (str, optional): Specifies the character set for string values.
|
|
ssl_key (str, optional): Specifies the path of the file containing the SSL
|
|
key.
|
|
ssl_cert (str, optional): Specifies the path of the file containing the SSL
|
|
certificate.
|
|
ssl_ca (str, optional): Specifies the path of the file containing the SSL
|
|
certificate authority.
|
|
ssl_cipher (str, optional): Sets the SSL cipher list.
|
|
ssl_disabled (bool, optional): Disables SSL usage.
|
|
ssl_verify_cert (bool, optional): Verifies the server's certificate.
|
|
Automatically enabled if ``ssl_ca`` is specified.
|
|
ssl_verify_identity (bool, optional): Verifies the server's identity.
|
|
conv (dict[int, Callable], optional): A dictionary of data conversion
|
|
functions.
|
|
credential_type (str, optional): Specifies the type of authentication to
|
|
use: auth.PASSWORD, auth.JWT, or auth.BROWSER_SSO.
|
|
autocommit (bool, optional): Enables autocommits.
|
|
results_type (str, optional): Determines the structure of the query results:
|
|
tuples, namedtuples, dicts.
|
|
results_format (str, optional): Deprecated. This option has been renamed to
|
|
results_type.
|
|
|
|
Examples:
|
|
Basic Usage:
|
|
|
|
.. code-block:: python
|
|
|
|
from langchain_openai import OpenAIEmbeddings
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
|
|
vectorstore = SingleStoreDB(
|
|
OpenAIEmbeddings(),
|
|
host="https://user:password@127.0.0.1:3306/database"
|
|
)
|
|
|
|
Advanced Usage:
|
|
|
|
.. code-block:: python
|
|
|
|
from langchain_openai import OpenAIEmbeddings
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
|
|
vectorstore = SingleStoreDB(
|
|
OpenAIEmbeddings(),
|
|
distance_strategy=DistanceStrategy.EUCLIDEAN_DISTANCE,
|
|
host="127.0.0.1",
|
|
port=3306,
|
|
user="user",
|
|
password="password",
|
|
database="db",
|
|
table_name="my_custom_table",
|
|
pool_size=10,
|
|
timeout=60,
|
|
)
|
|
|
|
Using environment variables:
|
|
|
|
.. code-block:: python
|
|
|
|
from langchain_openai import OpenAIEmbeddings
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
|
|
os.environ['SINGLESTOREDB_URL'] = 'me:p455w0rd@s2-host.com/my_db'
|
|
vectorstore = SingleStoreDB(OpenAIEmbeddings())
|
|
|
|
Using vector index:
|
|
|
|
.. code-block:: python
|
|
|
|
from langchain_openai import OpenAIEmbeddings
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
|
|
os.environ['SINGLESTOREDB_URL'] = 'me:p455w0rd@s2-host.com/my_db'
|
|
vectorstore = SingleStoreDB(
|
|
OpenAIEmbeddings(),
|
|
use_vector_index=True,
|
|
)
|
|
|
|
Using full-text index:
|
|
|
|
.. code-block:: python
|
|
from langchain_openai import OpenAIEmbeddings
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
|
|
os.environ['SINGLESTOREDB_URL'] = 'me:p455w0rd@s2-host.com/my_db'
|
|
vectorstore = SingleStoreDB(
|
|
OpenAIEmbeddings(),
|
|
use_full_text_search=True,
|
|
)
|
|
"""
|
|
|
|
self.embedding = embedding
|
|
self.distance_strategy = distance_strategy
|
|
self.table_name = self._sanitize_input(table_name)
|
|
self.content_field = self._sanitize_input(content_field)
|
|
self.metadata_field = self._sanitize_input(metadata_field)
|
|
self.vector_field = self._sanitize_input(vector_field)
|
|
self.id_field = self._sanitize_input(id_field)
|
|
|
|
self.use_vector_index = bool(use_vector_index)
|
|
self.vector_index_name = self._sanitize_input(vector_index_name)
|
|
self.vector_index_options = dict(vector_index_options or {})
|
|
self.vector_index_options["metric_type"] = self.distance_strategy
|
|
self.vector_size = int(vector_size)
|
|
|
|
self.use_full_text_search = bool(use_full_text_search)
|
|
|
|
# Pass the rest of the kwargs to the connection.
|
|
self.connection_kwargs = kwargs
|
|
|
|
# Add program name and version to connection attributes.
|
|
if "conn_attrs" not in self.connection_kwargs:
|
|
self.connection_kwargs["conn_attrs"] = dict()
|
|
|
|
self.connection_kwargs["conn_attrs"]["_connector_name"] = "langchain python sdk"
|
|
self.connection_kwargs["conn_attrs"]["_connector_version"] = "2.0.0"
|
|
|
|
# Create connection pool.
|
|
self.connection_pool = QueuePool(
|
|
self._get_connection,
|
|
max_overflow=max_overflow,
|
|
pool_size=pool_size,
|
|
timeout=timeout,
|
|
)
|
|
self._create_table()
|
|
|
|
@property
|
|
def embeddings(self) -> Embeddings:
|
|
return self.embedding
|
|
|
|
def _sanitize_input(self, input_str: str) -> str:
|
|
# Remove characters that are not alphanumeric or underscores
|
|
return re.sub(r"[^a-zA-Z0-9_]", "", input_str)
|
|
|
|
def _select_relevance_score_fn(self) -> Callable[[float], float]:
|
|
return self._max_inner_product_relevance_score_fn
|
|
|
|
def _create_table(self: SingleStoreDB) -> None:
|
|
"""Create table if it doesn't exist."""
|
|
conn = self.connection_pool.connect()
|
|
try:
|
|
cur = conn.cursor()
|
|
try:
|
|
full_text_index = ""
|
|
if self.use_full_text_search:
|
|
full_text_index = ", FULLTEXT({})".format(self.content_field)
|
|
if self.use_vector_index:
|
|
index_options = ""
|
|
if self.vector_index_options and len(self.vector_index_options) > 0:
|
|
index_options = "INDEX_OPTIONS '{}'".format(
|
|
json.dumps(self.vector_index_options)
|
|
)
|
|
cur.execute(
|
|
"""CREATE TABLE IF NOT EXISTS {}
|
|
({} BIGINT AUTO_INCREMENT PRIMARY KEY, {} LONGTEXT CHARACTER
|
|
SET utf8mb4 COLLATE utf8mb4_general_ci, {} VECTOR({}, F32)
|
|
NOT NULL, {} JSON, VECTOR INDEX {} ({}) {}{});""".format(
|
|
self.table_name,
|
|
self.id_field,
|
|
self.content_field,
|
|
self.vector_field,
|
|
self.vector_size,
|
|
self.metadata_field,
|
|
self.vector_index_name,
|
|
self.vector_field,
|
|
index_options,
|
|
full_text_index,
|
|
),
|
|
)
|
|
else:
|
|
cur.execute(
|
|
"""CREATE TABLE IF NOT EXISTS {}
|
|
({} BIGINT AUTO_INCREMENT PRIMARY KEY, {} LONGTEXT CHARACTER
|
|
SET utf8mb4 COLLATE utf8mb4_general_ci, {} BLOB, {} JSON{});
|
|
""".format(
|
|
self.table_name,
|
|
self.id_field,
|
|
self.content_field,
|
|
self.vector_field,
|
|
self.metadata_field,
|
|
full_text_index,
|
|
),
|
|
)
|
|
finally:
|
|
cur.close()
|
|
finally:
|
|
conn.close()
|
|
|
|
def add_images(
|
|
self,
|
|
uris: List[str],
|
|
metadatas: Optional[List[dict]] = None,
|
|
embeddings: Optional[List[List[float]]] = None,
|
|
**kwargs: Any,
|
|
) -> List[str]:
|
|
"""Run images through the embeddings and add to the vectorstore.
|
|
|
|
Args:
|
|
uris List[str]: File path to images.
|
|
Each URI will be added to the vectorstore as document content.
|
|
metadatas (Optional[List[dict]], optional): Optional list of metadatas.
|
|
Defaults to None.
|
|
embeddings (Optional[List[List[float]]], optional): Optional pre-generated
|
|
embeddings. Defaults to None.
|
|
|
|
Returns:
|
|
List[str]: empty list
|
|
"""
|
|
# Set embeddings
|
|
if (
|
|
embeddings is None
|
|
and self.embedding is not None
|
|
and hasattr(self.embedding, "embed_image")
|
|
):
|
|
embeddings = self.embedding.embed_image(uris=uris)
|
|
return self.add_texts(uris, metadatas, embeddings, **kwargs)
|
|
|
|
def add_texts(
|
|
self,
|
|
texts: Iterable[str],
|
|
metadatas: Optional[List[dict]] = None,
|
|
embeddings: Optional[List[List[float]]] = None,
|
|
**kwargs: Any,
|
|
) -> List[str]:
|
|
"""Add more texts to the vectorstore.
|
|
|
|
Args:
|
|
texts (Iterable[str]): Iterable of strings/text to add to the vectorstore.
|
|
metadatas (Optional[List[dict]], optional): Optional list of metadatas.
|
|
Defaults to None.
|
|
embeddings (Optional[List[List[float]]], optional): Optional pre-generated
|
|
embeddings. Defaults to None.
|
|
|
|
Returns:
|
|
List[str]: empty list
|
|
"""
|
|
conn = self.connection_pool.connect()
|
|
try:
|
|
cur = conn.cursor()
|
|
try:
|
|
# Write data to singlestore db
|
|
for i, text in enumerate(texts):
|
|
# Use provided values by default or fallback
|
|
metadata = metadatas[i] if metadatas else {}
|
|
embedding = (
|
|
embeddings[i]
|
|
if embeddings
|
|
else self.embedding.embed_documents([text])[0]
|
|
)
|
|
cur.execute(
|
|
"""INSERT INTO {}({}, {}, {})
|
|
VALUES (%s, JSON_ARRAY_PACK(%s), %s)""".format(
|
|
self.table_name,
|
|
self.content_field,
|
|
self.vector_field,
|
|
self.metadata_field,
|
|
),
|
|
(
|
|
text,
|
|
"[{}]".format(",".join(map(str, embedding))),
|
|
json.dumps(metadata),
|
|
),
|
|
)
|
|
if self.use_vector_index or self.use_full_text_search:
|
|
cur.execute("OPTIMIZE TABLE {} FLUSH;".format(self.table_name))
|
|
finally:
|
|
cur.close()
|
|
finally:
|
|
conn.close()
|
|
return []
|
|
|
|
def similarity_search(
|
|
self,
|
|
query: str,
|
|
k: int = 4,
|
|
filter: Optional[dict] = None,
|
|
search_strategy: SearchStrategy = SearchStrategy.VECTOR_ONLY,
|
|
filter_threshold: float = 0,
|
|
text_weight: float = 0.5,
|
|
vector_weight: float = 0.5,
|
|
vector_select_count_multiplier: int = 10,
|
|
**kwargs: Any,
|
|
) -> List[Document]:
|
|
"""Returns the most similar indexed documents to the query text.
|
|
|
|
Uses cosine similarity.
|
|
|
|
Args:
|
|
query (str): The query text for which to find similar documents.
|
|
k (int): The number of documents to return. Default is 4.
|
|
filter (dict): A dictionary of metadata fields and values to filter by.
|
|
Default is None.
|
|
search_strategy (SearchStrategy): The search strategy to use.
|
|
Default is SearchStrategy.VECTOR_ONLY.
|
|
Available options are:
|
|
- SearchStrategy.VECTOR_ONLY: Searches only by vector similarity.
|
|
- SearchStrategy.TEXT_ONLY: Searches only by text similarity. This
|
|
option is only available if use_full_text_search is True.
|
|
- SearchStrategy.FILTER_BY_TEXT: Filters by text similarity and
|
|
searches by vector similarity. This option is only available if
|
|
use_full_text_search is True.
|
|
- SearchStrategy.FILTER_BY_VECTOR: Filters by vector similarity and
|
|
searches by text similarity. This option is only available if
|
|
use_full_text_search is True.
|
|
- SearchStrategy.WEIGHTED_SUM: Searches by a weighted sum of text and
|
|
vector similarity. This option is only available if
|
|
use_full_text_search is True and distance_strategy is DOT_PRODUCT.
|
|
filter_threshold (float): The threshold for filtering by text or vector
|
|
similarity. Default is 0. This option has effect only if search_strategy
|
|
is SearchStrategy.FILTER_BY_TEXT or SearchStrategy.FILTER_BY_VECTOR.
|
|
text_weight (float): The weight of text similarity in the weighted sum
|
|
search strategy. Default is 0.5. This option has effect only if
|
|
search_strategy is SearchStrategy.WEIGHTED_SUM.
|
|
vector_weight (float): The weight of vector similarity in the weighted sum
|
|
search strategy. Default is 0.5. This option has effect only if
|
|
search_strategy is SearchStrategy.WEIGHTED_SUM.
|
|
vector_select_count_multiplier (int): The multiplier for the number of
|
|
vectors to select when using the vector index. Default is 10.
|
|
This parameter has effect only if use_vector_index is True and
|
|
search_strategy is SearchStrategy.WEIGHTED_SUM or
|
|
SearchStrategy.FILTER_BY_TEXT.
|
|
The number of vectors selected will
|
|
be k * vector_select_count_multiplier.
|
|
This is needed due to the limitations of the vector index.
|
|
|
|
|
|
Returns:
|
|
List[Document]: A list of documents that are most similar to the query text.
|
|
|
|
Examples:
|
|
|
|
Basic Usage:
|
|
.. code-block:: python
|
|
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
from langchain_openai import OpenAIEmbeddings
|
|
|
|
s2 = SingleStoreDB.from_documents(
|
|
docs,
|
|
OpenAIEmbeddings(),
|
|
host="username:password@localhost:3306/database"
|
|
)
|
|
results = s2.similarity_search("query text", 1,
|
|
{"metadata_field": "metadata_value"})
|
|
|
|
Different Search Strategies:
|
|
.. code-block:: python
|
|
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
from langchain_openai import OpenAIEmbeddings
|
|
|
|
s2 = SingleStoreDB.from_documents(
|
|
docs,
|
|
OpenAIEmbeddings(),
|
|
host="username:password@localhost:3306/database",
|
|
use_full_text_search=True,
|
|
use_vector_index=True,
|
|
)
|
|
results = s2.similarity_search("query text", 1,
|
|
search_strategy=SingleStoreDB.SearchStrategy.FILTER_BY_TEXT,
|
|
filter_threshold=0.5)
|
|
|
|
Weighted Sum Search Strategy:
|
|
.. code-block:: python
|
|
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
from langchain_openai import OpenAIEmbeddings
|
|
|
|
s2 = SingleStoreDB.from_documents(
|
|
docs,
|
|
OpenAIEmbeddings(),
|
|
host="username:password@localhost:3306/database",
|
|
use_full_text_search=True,
|
|
use_vector_index=True,
|
|
)
|
|
results = s2.similarity_search("query text", 1,
|
|
search_strategy=SingleStoreDB.SearchStrategy.WEIGHTED_SUM,
|
|
text_weight=0.3,
|
|
vector_weight=0.7)
|
|
"""
|
|
docs_and_scores = self.similarity_search_with_score(
|
|
query=query,
|
|
k=k,
|
|
filter=filter,
|
|
search_strategy=search_strategy,
|
|
filter_threshold=filter_threshold,
|
|
text_weight=text_weight,
|
|
vector_weight=vector_weight,
|
|
vector_select_count_multiplier=vector_select_count_multiplier,
|
|
**kwargs,
|
|
)
|
|
return [doc for doc, _ in docs_and_scores]
|
|
|
|
def similarity_search_with_score(
|
|
self,
|
|
query: str,
|
|
k: int = 4,
|
|
filter: Optional[dict] = None,
|
|
search_strategy: SearchStrategy = SearchStrategy.VECTOR_ONLY,
|
|
filter_threshold: float = 1,
|
|
text_weight: float = 0.5,
|
|
vector_weight: float = 0.5,
|
|
vector_select_count_multiplier: int = 10,
|
|
**kwargs: Any,
|
|
) -> List[Tuple[Document, float]]:
|
|
"""Return docs most similar to query. Uses cosine similarity.
|
|
|
|
Args:
|
|
query: Text to look up documents similar to.
|
|
k: Number of Documents to return. Defaults to 4.
|
|
filter: A dictionary of metadata fields and values to filter by.
|
|
Defaults to None.
|
|
search_strategy (SearchStrategy): The search strategy to use.
|
|
Default is SearchStrategy.VECTOR_ONLY.
|
|
Available options are:
|
|
- SearchStrategy.VECTOR_ONLY: Searches only by vector similarity.
|
|
- SearchStrategy.TEXT_ONLY: Searches only by text similarity. This
|
|
option is only available if use_full_text_search is True.
|
|
- SearchStrategy.FILTER_BY_TEXT: Filters by text similarity and
|
|
searches by vector similarity. This option is only available if
|
|
use_full_text_search is True.
|
|
- SearchStrategy.FILTER_BY_VECTOR: Filters by vector similarity and
|
|
searches by text similarity. This option is only available if
|
|
use_full_text_search is True.
|
|
- SearchStrategy.WEIGHTED_SUM: Searches by a weighted sum of text and
|
|
vector similarity. This option is only available if
|
|
use_full_text_search is True and distance_strategy is DOT_PRODUCT.
|
|
filter_threshold (float): The threshold for filtering by text or vector
|
|
similarity. Default is 0. This option has effect only if search_strategy
|
|
is SearchStrategy.FILTER_BY_TEXT or SearchStrategy.FILTER_BY_VECTOR.
|
|
text_weight (float): The weight of text similarity in the weighted sum
|
|
search strategy. Default is 0.5. This option has effect only if
|
|
search_strategy is SearchStrategy.WEIGHTED_SUM.
|
|
vector_weight (float): The weight of vector similarity in the weighted sum
|
|
search strategy. Default is 0.5. This option has effect only if
|
|
search_strategy is SearchStrategy.WEIGHTED_SUM.
|
|
vector_select_count_multiplier (int): The multiplier for the number of
|
|
vectors to select when using the vector index. Default is 10.
|
|
This parameter has effect only if use_vector_index is True and
|
|
search_strategy is SearchStrategy.WEIGHTED_SUM or
|
|
SearchStrategy.FILTER_BY_TEXT.
|
|
The number of vectors selected will
|
|
be k * vector_select_count_multiplier.
|
|
This is needed due to the limitations of the vector index.
|
|
Returns:
|
|
List of Documents most similar to the query and score for each
|
|
document.
|
|
|
|
Raises:
|
|
ValueError: If the search strategy is not supported with the
|
|
distance strategy.
|
|
|
|
Examples:
|
|
Basic Usage:
|
|
.. code-block:: python
|
|
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
from langchain_openai import OpenAIEmbeddings
|
|
|
|
s2 = SingleStoreDB.from_documents(
|
|
docs,
|
|
OpenAIEmbeddings(),
|
|
host="username:password@localhost:3306/database"
|
|
)
|
|
results = s2.similarity_search_with_score("query text", 1,
|
|
{"metadata_field": "metadata_value"})
|
|
|
|
Different Search Strategies:
|
|
|
|
.. code-block:: python
|
|
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
from langchain_openai import OpenAIEmbeddings
|
|
|
|
s2 = SingleStoreDB.from_documents(
|
|
docs,
|
|
OpenAIEmbeddings(),
|
|
host="username:password@localhost:3306/database",
|
|
use_full_text_search=True,
|
|
use_vector_index=True,
|
|
)
|
|
results = s2.similarity_search_with_score("query text", 1,
|
|
search_strategy=SingleStoreDB.SearchStrategy.FILTER_BY_VECTOR,
|
|
filter_threshold=0.5)
|
|
|
|
Weighted Sum Search Strategy:
|
|
.. code-block:: python
|
|
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
from langchain_openai import OpenAIEmbeddings
|
|
|
|
s2 = SingleStoreDB.from_documents(
|
|
docs,
|
|
OpenAIEmbeddings(),
|
|
host="username:password@localhost:3306/database",
|
|
use_full_text_search=True,
|
|
use_vector_index=True,
|
|
)
|
|
results = s2.similarity_search_with_score("query text", 1,
|
|
search_strategy=SingleStoreDB.SearchStrategy.WEIGHTED_SUM,
|
|
text_weight=0.3,
|
|
vector_weight=0.7)
|
|
"""
|
|
|
|
if (
|
|
search_strategy != SingleStoreDB.SearchStrategy.VECTOR_ONLY
|
|
and not self.use_full_text_search
|
|
):
|
|
raise ValueError(
|
|
"""Search strategy {} is not supported
|
|
when use_full_text_search is False""".format(search_strategy)
|
|
)
|
|
|
|
if (
|
|
search_strategy == SingleStoreDB.SearchStrategy.WEIGHTED_SUM
|
|
and self.distance_strategy != DistanceStrategy.DOT_PRODUCT
|
|
):
|
|
raise ValueError(
|
|
"Search strategy {} is not supported with distance strategy {}".format(
|
|
search_strategy, self.distance_strategy
|
|
)
|
|
)
|
|
|
|
# Creates embedding vector from user query
|
|
embedding = []
|
|
if search_strategy != SingleStoreDB.SearchStrategy.TEXT_ONLY:
|
|
embedding = self.embedding.embed_query(query)
|
|
|
|
self.embedding.embed_query(query)
|
|
conn = self.connection_pool.connect()
|
|
result = []
|
|
where_clause: str = ""
|
|
where_clause_values: List[Any] = []
|
|
if filter or search_strategy in [
|
|
SingleStoreDB.SearchStrategy.FILTER_BY_TEXT,
|
|
SingleStoreDB.SearchStrategy.FILTER_BY_VECTOR,
|
|
]:
|
|
where_clause = "WHERE "
|
|
arguments = []
|
|
|
|
if search_strategy == SingleStoreDB.SearchStrategy.FILTER_BY_TEXT:
|
|
arguments.append(
|
|
"MATCH ({}) AGAINST (%s) > %s".format(self.content_field)
|
|
)
|
|
where_clause_values.append(query)
|
|
where_clause_values.append(float(filter_threshold))
|
|
|
|
if search_strategy == SingleStoreDB.SearchStrategy.FILTER_BY_VECTOR:
|
|
condition = "{}({}, JSON_ARRAY_PACK(%s)) ".format(
|
|
self.distance_strategy.name
|
|
if isinstance(self.distance_strategy, DistanceStrategy)
|
|
else self.distance_strategy,
|
|
self.vector_field,
|
|
)
|
|
if self.distance_strategy == DistanceStrategy.EUCLIDEAN_DISTANCE:
|
|
condition += "< %s"
|
|
else:
|
|
condition += "> %s"
|
|
arguments.append(condition)
|
|
where_clause_values.append("[{}]".format(",".join(map(str, embedding))))
|
|
where_clause_values.append(float(filter_threshold))
|
|
|
|
def build_where_clause(
|
|
where_clause_values: List[Any],
|
|
sub_filter: dict,
|
|
prefix_args: Optional[List[str]] = None,
|
|
) -> None:
|
|
prefix_args = prefix_args or []
|
|
for key in sub_filter.keys():
|
|
if isinstance(sub_filter[key], dict):
|
|
build_where_clause(
|
|
where_clause_values, sub_filter[key], prefix_args + [key]
|
|
)
|
|
else:
|
|
arguments.append(
|
|
"JSON_EXTRACT_JSON({}, {}) = %s".format(
|
|
self.metadata_field,
|
|
", ".join(["%s"] * (len(prefix_args) + 1)),
|
|
)
|
|
)
|
|
where_clause_values += prefix_args + [key]
|
|
where_clause_values.append(json.dumps(sub_filter[key]))
|
|
|
|
if filter:
|
|
build_where_clause(where_clause_values, filter)
|
|
where_clause += " AND ".join(arguments)
|
|
|
|
try:
|
|
cur = conn.cursor()
|
|
try:
|
|
if (
|
|
search_strategy == SingleStoreDB.SearchStrategy.VECTOR_ONLY
|
|
or search_strategy == SingleStoreDB.SearchStrategy.FILTER_BY_TEXT
|
|
):
|
|
search_options = ""
|
|
if (
|
|
self.use_vector_index
|
|
and search_strategy
|
|
== SingleStoreDB.SearchStrategy.FILTER_BY_TEXT
|
|
):
|
|
search_options = "SEARCH_OPTIONS '{\"k\":%d}'" % (
|
|
k * vector_select_count_multiplier
|
|
)
|
|
cur.execute(
|
|
"""SELECT {}, {}, {}({}, JSON_ARRAY_PACK(%s)) as __score
|
|
FROM {} {} ORDER BY __score {}{} LIMIT %s""".format(
|
|
self.content_field,
|
|
self.metadata_field,
|
|
self.distance_strategy.name
|
|
if isinstance(self.distance_strategy, DistanceStrategy)
|
|
else self.distance_strategy,
|
|
self.vector_field,
|
|
self.table_name,
|
|
where_clause,
|
|
search_options,
|
|
ORDERING_DIRECTIVE[self.distance_strategy],
|
|
),
|
|
("[{}]".format(",".join(map(str, embedding))),)
|
|
+ tuple(where_clause_values)
|
|
+ (k,),
|
|
)
|
|
elif (
|
|
search_strategy == SingleStoreDB.SearchStrategy.FILTER_BY_VECTOR
|
|
or search_strategy == SingleStoreDB.SearchStrategy.TEXT_ONLY
|
|
):
|
|
cur.execute(
|
|
"""SELECT {}, {}, MATCH ({}) AGAINST (%s) as __score
|
|
FROM {} {} ORDER BY __score DESC LIMIT %s""".format(
|
|
self.content_field,
|
|
self.metadata_field,
|
|
self.content_field,
|
|
self.table_name,
|
|
where_clause,
|
|
),
|
|
(query,) + tuple(where_clause_values) + (k,),
|
|
)
|
|
elif search_strategy == SingleStoreDB.SearchStrategy.WEIGHTED_SUM:
|
|
cur.execute(
|
|
"""SELECT {}, {}, __score1 * %s + __score2 * %s as __score
|
|
FROM (
|
|
SELECT {}, {}, {}, MATCH ({}) AGAINST (%s) as __score1
|
|
FROM {} {}) r1 FULL OUTER JOIN (
|
|
SELECT {}, {}({}, JSON_ARRAY_PACK(%s)) as __score2
|
|
FROM {} {} ORDER BY __score2 {} LIMIT %s
|
|
) r2 ON r1.{} = r2.{} ORDER BY __score {} LIMIT %s""".format(
|
|
self.content_field,
|
|
self.metadata_field,
|
|
self.id_field,
|
|
self.content_field,
|
|
self.metadata_field,
|
|
self.content_field,
|
|
self.table_name,
|
|
where_clause,
|
|
self.id_field,
|
|
self.distance_strategy.name
|
|
if isinstance(self.distance_strategy, DistanceStrategy)
|
|
else self.distance_strategy,
|
|
self.vector_field,
|
|
self.table_name,
|
|
where_clause,
|
|
ORDERING_DIRECTIVE[self.distance_strategy],
|
|
self.id_field,
|
|
self.id_field,
|
|
ORDERING_DIRECTIVE[self.distance_strategy],
|
|
),
|
|
(text_weight, vector_weight, query)
|
|
+ tuple(where_clause_values)
|
|
+ ("[{}]".format(",".join(map(str, embedding))),)
|
|
+ tuple(where_clause_values)
|
|
+ (k * vector_select_count_multiplier, k),
|
|
)
|
|
else:
|
|
raise ValueError(
|
|
"Invalid search strategy: {}".format(search_strategy)
|
|
)
|
|
|
|
for row in cur.fetchall():
|
|
doc = Document(page_content=row[0], metadata=row[1])
|
|
result.append((doc, float(row[2])))
|
|
finally:
|
|
cur.close()
|
|
finally:
|
|
conn.close()
|
|
return result
|
|
|
|
@classmethod
|
|
def from_texts(
|
|
cls: Type[SingleStoreDB],
|
|
texts: List[str],
|
|
embedding: Embeddings,
|
|
metadatas: Optional[List[dict]] = None,
|
|
distance_strategy: DistanceStrategy = DEFAULT_DISTANCE_STRATEGY,
|
|
table_name: str = "embeddings",
|
|
content_field: str = "content",
|
|
metadata_field: str = "metadata",
|
|
vector_field: str = "vector",
|
|
id_field: str = "id",
|
|
use_vector_index: bool = False,
|
|
vector_index_name: str = "",
|
|
vector_index_options: Optional[dict] = None,
|
|
vector_size: int = 1536,
|
|
use_full_text_search: bool = False,
|
|
pool_size: int = 5,
|
|
max_overflow: int = 10,
|
|
timeout: float = 30,
|
|
**kwargs: Any,
|
|
) -> SingleStoreDB:
|
|
"""Create a SingleStoreDB vectorstore from raw documents.
|
|
This is a user-friendly interface that:
|
|
1. Embeds documents.
|
|
2. Creates a new table for the embeddings in SingleStoreDB.
|
|
3. Adds the documents to the newly created table.
|
|
This is intended to be a quick way to get started.
|
|
Args:
|
|
texts (List[str]): List of texts to add to the vectorstore.
|
|
embedding (Embeddings): A text embedding model.
|
|
metadatas (Optional[List[dict]], optional): Optional list of metadatas.
|
|
Defaults to None.
|
|
distance_strategy (DistanceStrategy, optional):
|
|
Determines the strategy employed for calculating
|
|
the distance between vectors in the embedding space.
|
|
Defaults to DOT_PRODUCT.
|
|
Available options are:
|
|
- DOT_PRODUCT: Computes the scalar product of two vectors.
|
|
This is the default behavior
|
|
- EUCLIDEAN_DISTANCE: Computes the Euclidean distance between
|
|
two vectors. This metric considers the geometric distance in
|
|
the vector space, and might be more suitable for embeddings
|
|
that rely on spatial relationships. This metric is not
|
|
compatible with the WEIGHTED_SUM search strategy.
|
|
table_name (str, optional): Specifies the name of the table in use.
|
|
Defaults to "embeddings".
|
|
content_field (str, optional): Specifies the field to store the content.
|
|
Defaults to "content".
|
|
metadata_field (str, optional): Specifies the field to store metadata.
|
|
Defaults to "metadata".
|
|
vector_field (str, optional): Specifies the field to store the vector.
|
|
Defaults to "vector".
|
|
id_field (str, optional): Specifies the field to store the id.
|
|
Defaults to "id".
|
|
use_vector_index (bool, optional): Toggles the use of a vector index.
|
|
Works only with SingleStoreDB 8.5 or later. Defaults to False.
|
|
If set to True, vector_size parameter is required to be set to
|
|
a proper value.
|
|
vector_index_name (str, optional): Specifies the name of the vector index.
|
|
Defaults to empty. Will be ignored if use_vector_index is set to False.
|
|
vector_index_options (dict, optional): Specifies the options for
|
|
the vector index. Defaults to {}.
|
|
Will be ignored if use_vector_index is set to False. The options are:
|
|
index_type (str, optional): Specifies the type of the index.
|
|
Defaults to IVF_PQFS.
|
|
For more options, please refer to the SingleStoreDB documentation:
|
|
https://docs.singlestore.com/cloud/reference/sql-reference/vector-functions/vector-indexing/
|
|
vector_size (int, optional): Specifies the size of the vector.
|
|
Defaults to 1536. Required if use_vector_index is set to True.
|
|
Should be set to the same value as the size of the vectors
|
|
stored in the vector_field.
|
|
use_full_text_search (bool, optional): Toggles the use a full-text index
|
|
on the document content. Defaults to False. If set to True, the table
|
|
will be created with a full-text index on the content field,
|
|
and the simularity_search method will all using TEXT_ONLY,
|
|
FILTER_BY_TEXT, FILTER_BY_VECTOR, and WIGHTED_SUM search strategies.
|
|
If set to False, the simularity_search method will only allow
|
|
VECTOR_ONLY search strategy.
|
|
|
|
pool_size (int, optional): Determines the number of active connections in
|
|
the pool. Defaults to 5.
|
|
max_overflow (int, optional): Determines the maximum number of connections
|
|
allowed beyond the pool_size. Defaults to 10.
|
|
timeout (float, optional): Specifies the maximum wait time in seconds for
|
|
establishing a connection. Defaults to 30.
|
|
|
|
Additional optional arguments provide further customization over the
|
|
database connection:
|
|
|
|
pure_python (bool, optional): Toggles the connector mode. If True,
|
|
operates in pure Python mode.
|
|
local_infile (bool, optional): Allows local file uploads.
|
|
charset (str, optional): Specifies the character set for string values.
|
|
ssl_key (str, optional): Specifies the path of the file containing the SSL
|
|
key.
|
|
ssl_cert (str, optional): Specifies the path of the file containing the SSL
|
|
certificate.
|
|
ssl_ca (str, optional): Specifies the path of the file containing the SSL
|
|
certificate authority.
|
|
ssl_cipher (str, optional): Sets the SSL cipher list.
|
|
ssl_disabled (bool, optional): Disables SSL usage.
|
|
ssl_verify_cert (bool, optional): Verifies the server's certificate.
|
|
Automatically enabled if ``ssl_ca`` is specified.
|
|
ssl_verify_identity (bool, optional): Verifies the server's identity.
|
|
conv (dict[int, Callable], optional): A dictionary of data conversion
|
|
functions.
|
|
credential_type (str, optional): Specifies the type of authentication to
|
|
use: auth.PASSWORD, auth.JWT, or auth.BROWSER_SSO.
|
|
autocommit (bool, optional): Enables autocommits.
|
|
results_type (str, optional): Determines the structure of the query results:
|
|
tuples, namedtuples, dicts.
|
|
results_format (str, optional): Deprecated. This option has been renamed to
|
|
results_type.
|
|
|
|
Example:
|
|
.. code-block:: python
|
|
|
|
from langchain_community.vectorstores import SingleStoreDB
|
|
from langchain_openai import OpenAIEmbeddings
|
|
|
|
s2 = SingleStoreDB.from_texts(
|
|
texts,
|
|
OpenAIEmbeddings(),
|
|
host="username:password@localhost:3306/database"
|
|
)
|
|
"""
|
|
|
|
instance = cls(
|
|
embedding,
|
|
distance_strategy=distance_strategy,
|
|
table_name=table_name,
|
|
content_field=content_field,
|
|
metadata_field=metadata_field,
|
|
vector_field=vector_field,
|
|
id_field=id_field,
|
|
pool_size=pool_size,
|
|
max_overflow=max_overflow,
|
|
timeout=timeout,
|
|
use_vector_index=use_vector_index,
|
|
vector_index_name=vector_index_name,
|
|
vector_index_options=vector_index_options,
|
|
vector_size=vector_size,
|
|
use_full_text_search=use_full_text_search,
|
|
**kwargs,
|
|
)
|
|
instance.add_texts(texts, metadatas, embedding.embed_documents(texts), **kwargs)
|
|
return instance
|
|
|
|
|
|
# SingleStoreDBRetriever is not needed, but we keep it for backwards compatibility
|
|
SingleStoreDBRetriever = VectorStoreRetriever
|