from __future__ import annotations
import uuid
import warnings
from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Tuple
import numpy as np
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
if TYPE_CHECKING:
from azure.cosmos.cosmos_client import CosmosClient
[docs]
class AzureCosmosDBNoSqlVectorSearch(VectorStore):
"""`Azure Cosmos DB for NoSQL` vector store.
To use, you should have both:
- the ``azure-cosmos`` python package installed
You can read more about vector search using AzureCosmosDBNoSQL here:
https://learn.microsoft.com/en-us/azure/cosmos-db/nosql/vector-search
"""
[docs]
def __init__(
self,
*,
cosmos_client: CosmosClient,
embedding: Embeddings,
vector_embedding_policy: Dict[str, Any],
indexing_policy: Dict[str, Any],
cosmos_container_properties: Dict[str, Any],
cosmos_database_properties: Dict[str, Any],
database_name: str = "vectorSearchDB",
container_name: str = "vectorSearchContainer",
create_container: bool = True,
):
"""
Constructor for AzureCosmosDBNoSqlVectorSearch
Args:
cosmos_client: Client used to connect to azure cosmosdb no sql account.
database_name: Name of the database to be created.
container_name: Name of the container to be created.
embedding: Text embedding model to use.
vector_embedding_policy: Vector Embedding Policy for the container.
indexing_policy: Indexing Policy for the container.
cosmos_container_properties: Container Properties for the container.
cosmos_database_properties: Database Properties for the container.
"""
self._cosmos_client = cosmos_client
self._database_name = database_name
self._container_name = container_name
self._embedding = embedding
self._vector_embedding_policy = vector_embedding_policy
self._indexing_policy = indexing_policy
self._cosmos_container_properties = cosmos_container_properties
self._cosmos_database_properties = cosmos_database_properties
self._create_container = create_container
if self._create_container:
if (
indexing_policy["vectorIndexes"] is None
or len(indexing_policy["vectorIndexes"]) == 0
):
raise ValueError(
"vectorIndexes cannot be null or empty in the indexing_policy."
)
if (
vector_embedding_policy is None
or len(vector_embedding_policy["vectorEmbeddings"]) == 0
):
raise ValueError(
"vectorEmbeddings cannot be null "
"or empty in the vector_embedding_policy."
)
if self._cosmos_container_properties["partition_key"] is None:
raise ValueError(
"partition_key cannot be null or empty for a container."
)
# Create the database if it already doesn't exist
self._database = self._cosmos_client.create_database_if_not_exists(
id=self._database_name,
offer_throughput=self._cosmos_database_properties.get("offer_throughput"),
session_token=self._cosmos_database_properties.get("session_token"),
initial_headers=self._cosmos_database_properties.get("initial_headers"),
etag=self._cosmos_database_properties.get("etag"),
match_condition=self._cosmos_database_properties.get("match_condition"),
)
# Create the collection if it already doesn't exist
self._container = self._database.create_container_if_not_exists(
id=self._container_name,
partition_key=self._cosmos_container_properties["partition_key"],
indexing_policy=self._indexing_policy,
default_ttl=self._cosmos_container_properties.get("default_ttl"),
offer_throughput=self._cosmos_container_properties.get("offer_throughput"),
unique_key_policy=self._cosmos_container_properties.get(
"unique_key_policy"
),
conflict_resolution_policy=self._cosmos_container_properties.get(
"conflict_resolution_policy"
),
analytical_storage_ttl=self._cosmos_container_properties.get(
"analytical_storage_ttl"
),
computed_properties=self._cosmos_container_properties.get(
"computed_properties"
),
etag=self._cosmos_container_properties.get("etag"),
match_condition=self._cosmos_container_properties.get("match_condition"),
session_token=self._cosmos_container_properties.get("session_token"),
initial_headers=self._cosmos_container_properties.get("initial_headers"),
vector_embedding_policy=self._vector_embedding_policy,
)
self._embedding_key = self._vector_embedding_policy["vectorEmbeddings"][0][
"path"
][1:]
[docs]
def add_texts(
self,
texts: Iterable[str],
metadatas: Optional[List[dict]] = None,
**kwargs: Any,
) -> List[str]:
"""Run more texts through the embeddings and add to the vectorstore.
Args:
texts: Iterable of strings to add to the vectorstore.
metadatas: Optional list of metadatas associated with the texts.
Returns:
List of ids from adding the texts into the vectorstore.
"""
_metadatas = list(metadatas if metadatas is not None else ({} for _ in texts))
return self._insert_texts(list(texts), _metadatas)
def _insert_texts(
self, texts: List[str], metadatas: List[Dict[str, Any]]
) -> List[str]:
"""Used to Load Documents into the collection
Args:
texts: The list of documents strings to load
metadatas: The list of metadata objects associated with each document
Returns:
List of ids from adding the texts into the vectorstore.
"""
# If the texts is empty, throw an error
if not texts:
raise Exception("Texts can not be null or empty")
# Embed and create the documents
embeddings = self._embedding.embed_documents(texts)
text_key = "text"
to_insert = [
{
"id": str(uuid.uuid4()),
text_key: t,
self._embedding_key: embedding,
"metadata": m,
}
for t, m, embedding in zip(texts, metadatas, embeddings)
]
# insert the documents in CosmosDB No Sql
doc_ids: List[str] = []
for item in to_insert:
created_doc = self._container.create_item(item)
doc_ids.append(created_doc["id"])
return doc_ids
@classmethod
def _from_kwargs(
cls,
embedding: Embeddings,
*,
cosmos_client: CosmosClient,
vector_embedding_policy: Dict[str, Any],
indexing_policy: Dict[str, Any],
cosmos_container_properties: Dict[str, Any],
cosmos_database_properties: Dict[str, Any],
database_name: str = "vectorSearchDB",
container_name: str = "vectorSearchContainer",
create_container: bool = True,
**kwargs: Any,
) -> AzureCosmosDBNoSqlVectorSearch:
if kwargs:
warnings.warn(
"Method 'from_texts' of AzureCosmosDBNoSql vector "
"store invoked with "
f"unsupported arguments "
f"({', '.join(sorted(kwargs))}), "
"which will be ignored."
)
return cls(
embedding=embedding,
cosmos_client=cosmos_client,
vector_embedding_policy=vector_embedding_policy,
indexing_policy=indexing_policy,
cosmos_container_properties=cosmos_container_properties,
cosmos_database_properties=cosmos_database_properties,
database_name=database_name,
container_name=container_name,
create_container=create_container,
)
[docs]
@classmethod
def from_texts(
cls,
texts: List[str],
embedding: Embeddings,
metadatas: Optional[List[dict]] = None,
**kwargs: Any,
) -> AzureCosmosDBNoSqlVectorSearch:
"""Create an AzureCosmosDBNoSqlVectorSearch vectorstore from raw texts.
Args:
texts: the texts to insert.
embedding: the embedding function to use in the store.
metadatas: metadata dicts for the texts.
**kwargs: you can pass any argument that you would
to :meth:`~add_texts` and/or to the 'AstraDB' constructor
(see these methods for details). These arguments will be
routed to the respective methods as they are.
Returns:
an `AzureCosmosDBNoSqlVectorSearch` vectorstore.
"""
vectorstore = AzureCosmosDBNoSqlVectorSearch._from_kwargs(embedding, **kwargs)
vectorstore.add_texts(
texts=texts,
metadatas=metadatas,
)
return vectorstore
[docs]
def delete(self, ids: Optional[List[str]] = None, **kwargs: Any) -> Optional[bool]:
if ids is None:
raise ValueError("No document ids provided to delete.")
for document_id in ids:
self._container.delete_item(document_id)
return True
[docs]
def delete_document_by_id(self, document_id: Optional[str] = None) -> None:
"""Removes a Specific Document by id
Args:
document_id: The document identifier
"""
if document_id is None:
raise ValueError("No document ids provided to delete.")
self._container.delete_item(document_id, partition_key=document_id)
def _similarity_search_with_score(
self,
embeddings: List[float],
k: int = 4,
pre_filter: Optional[Dict] = None,
with_embedding: bool = False,
) -> List[Tuple[Document, float]]:
query = "SELECT "
# If limit_offset_clause is not specified, add TOP clause
if pre_filter is None or pre_filter.get("limit_offset_clause") is None:
query += "TOP @limit "
query += (
"c.id, c[@embeddingKey], c.text, c.metadata, "
"VectorDistance(c[@embeddingKey], @embeddings) AS SimilarityScore FROM c"
)
# Add where_clause if specified
if pre_filter is not None and pre_filter.get("where_clause") is not None:
query += " {}".format(pre_filter["where_clause"])
query += " ORDER BY VectorDistance(c[@embeddingKey], @embeddings)"
# Add limit_offset_clause if specified
if pre_filter is not None and pre_filter.get("limit_offset_clause") is not None:
query += " {}".format(pre_filter["limit_offset_clause"])
parameters = [
{"name": "@limit", "value": k},
{"name": "@embeddingKey", "value": self._embedding_key},
{"name": "@embeddings", "value": embeddings},
]
docs_and_scores = []
items = list(
self._container.query_items(
query=query, parameters=parameters, enable_cross_partition_query=True
)
)
for item in items:
text = item["text"]
metadata = item["metadata"]
score = item["SimilarityScore"]
if with_embedding:
metadata[self._embedding_key] = item[self._embedding_key]
docs_and_scores.append(
(Document(page_content=text, metadata=metadata), score)
)
return docs_and_scores
[docs]
def similarity_search_with_score(
self,
query: str,
k: int = 4,
pre_filter: Optional[Dict] = None,
with_embedding: bool = False,
) -> List[Tuple[Document, float]]:
embeddings = self._embedding.embed_query(query)
docs_and_scores = self._similarity_search_with_score(
embeddings=embeddings,
k=k,
pre_filter=pre_filter,
with_embedding=with_embedding,
)
return docs_and_scores
[docs]
def similarity_search(
self,
query: str,
k: int = 4,
pre_filter: Optional[Dict] = None,
with_embedding: bool = False,
**kwargs: Any,
) -> List[Document]:
docs_and_scores = self.similarity_search_with_score(
query,
k=k,
pre_filter=pre_filter,
with_embedding=with_embedding,
)
return [doc for doc, _ in docs_and_scores]
[docs]
def max_marginal_relevance_search_by_vector(
self,
embedding: List[float],
k: int = 4,
fetch_k: int = 20,
lambda_mult: float = 0.5,
**kwargs: Any,
) -> List[Document]:
# Retrieves the docs with similarity scores
pre_filter = {}
with_embedding = False
if kwargs["pre_filter"]:
pre_filter = kwargs["pre_filter"]
if kwargs["with_embedding"]:
with_embedding = kwargs["with_embedding"]
docs = self._similarity_search_with_score(
embeddings=embedding,
k=fetch_k,
pre_filter=pre_filter,
with_embedding=with_embedding,
)
# Re-ranks the docs using MMR
mmr_doc_indexes = maximal_marginal_relevance(
np.array(embedding),
[doc.metadata[self._embedding_key] for doc, _ in docs],
k=k,
lambda_mult=lambda_mult,
)
mmr_docs = [docs[i][0] for i in mmr_doc_indexes]
return mmr_docs
[docs]
def max_marginal_relevance_search(
self,
query: str,
k: int = 4,
fetch_k: int = 20,
lambda_mult: float = 0.5,
**kwargs: Any,
) -> List[Document]:
# compute the embeddings vector from the query string
pre_filter = {}
with_embedding = False
if kwargs["pre_filter"]:
pre_filter = kwargs["pre_filter"]
if kwargs["with_embedding"]:
with_embedding = kwargs["with_embedding"]
embeddings = self._embedding.embed_query(query)
docs = self.max_marginal_relevance_search_by_vector(
embeddings,
k=k,
fetch_k=fetch_k,
lambda_mult=lambda_mult,
pre_filter=pre_filter,
with_embedding=with_embedding,
)
return docs
