Skip to content

Azure mongodb vector store driver

AzureMongoDbVectorStoreDriver

Bases: MongoDbAtlasVectorStoreDriver

A Vector Store Driver for CosmosDB with MongoDB vCore API.

Source code in griptape/drivers/vector/azure_mongodb_vector_store_driver.py 
12
13
14
15
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
@define
class AzureMongoDbVectorStoreDriver(MongoDbAtlasVectorStoreDriver):
    """A Vector Store Driver for CosmosDB with MongoDB vCore API."""

    def query(
        self,
        query: str,
        count: Optional[int] = None,
        namespace: Optional[str] = None,
        include_vectors: bool = False,
        offset: Optional[int] = None,
        **kwargs,
    ) -> list[BaseVectorStoreDriver.QueryResult]:
        """Queries the MongoDB collection for documents that match the provided query string.

        Results can be customized based on parameters like count, namespace, inclusion of vectors, offset, and index.
        """
        collection = self.get_collection()

        # Using the embedding driver to convert the query string into a vector
        vector = self.embedding_driver.embed_string(query)

        count = count if count else BaseVectorStoreDriver.DEFAULT_QUERY_COUNT
        offset = offset if offset else 0

        pipeline = []

        pipeline.append(
            {
                "$search": {
                    "cosmosSearch": {
                        "vector": vector,
                        "path": self.vector_path,
                        "k": min(count * self.num_candidates_multiplier, self.MAX_NUM_CANDIDATES),
                    },
                    "returnStoredSource": True,
                }
            }
        )

        if namespace:
            pipeline.append({"$match": {"namespace": namespace}})

        pipeline.append({"$project": {"similarityScore": {"$meta": "searchScore"}, "document": "$$ROOT"}})

        return [
            BaseVectorStoreDriver.QueryResult(
                id=str(doc["_id"]),
                vector=doc[self.vector_path] if include_vectors else [],
                score=doc["similarityScore"],
                meta=doc["document"]["meta"],
                namespace=namespace,
            )
            for doc in collection.aggregate(pipeline)
        ]

query(query, count=None, namespace=None, include_vectors=False, offset=None, **kwargs)

Queries the MongoDB collection for documents that match the provided query string.

Results can be customized based on parameters like count, namespace, inclusion of vectors, offset, and index.

Source code in griptape/drivers/vector/azure_mongodb_vector_store_driver.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
def query(
    self,
    query: str,
    count: Optional[int] = None,
    namespace: Optional[str] = None,
    include_vectors: bool = False,
    offset: Optional[int] = None,
    **kwargs,
) -> list[BaseVectorStoreDriver.QueryResult]:
    """Queries the MongoDB collection for documents that match the provided query string.

    Results can be customized based on parameters like count, namespace, inclusion of vectors, offset, and index.
    """
    collection = self.get_collection()

    # Using the embedding driver to convert the query string into a vector
    vector = self.embedding_driver.embed_string(query)

    count = count if count else BaseVectorStoreDriver.DEFAULT_QUERY_COUNT
    offset = offset if offset else 0

    pipeline = []

    pipeline.append(
        {
            "$search": {
                "cosmosSearch": {
                    "vector": vector,
                    "path": self.vector_path,
                    "k": min(count * self.num_candidates_multiplier, self.MAX_NUM_CANDIDATES),
                },
                "returnStoredSource": True,
            }
        }
    )

    if namespace:
        pipeline.append({"$match": {"namespace": namespace}})

    pipeline.append({"$project": {"similarityScore": {"$meta": "searchScore"}, "document": "$$ROOT"}})

    return [
        BaseVectorStoreDriver.QueryResult(
            id=str(doc["_id"]),
            vector=doc[self.vector_path] if include_vectors else [],
            score=doc["similarityScore"],
            meta=doc["document"]["meta"],
            namespace=namespace,
        )
        for doc in collection.aggregate(pipeline)
    ]