Azure Cosmos DB MongoDB vCore Vector Database

Use Azure Cosmos DB MongoDB vCore as a vector database for your Knowledge Base.

Setup

Follow the instructions in the Azure Cosmos DB Setup Guide to get the connection string.

Install MongoDB packages:

1uv pip install "pymongo[srv]"

Example

1import urllib.parse
2from kern.agent import Agent
3from kern.knowledge.knowledge import Knowledge
4from kern.vectordb.mongodb import MongoVectorDb
5
6# Azure Cosmos DB MongoDB connection string
7"""
8Example connection strings:
9"mongodb+srv://<username>:<encoded_password>@cluster0.mongocluster.cosmos.azure.com/?tls=true&authMechanism=SCRAM-SHA-256&retrywrites=false&maxIdleTimeMS=120000"
10"""
11mdb_connection_string = f"mongodb+srv://<username>:<encoded_password>@cluster0.mongocluster.cosmos.azure.com/?tls=true&authMechanism=SCRAM-SHA-256&retrywrites=false&maxIdleTimeMS=120000"
12
13knowledge_base = Knowledge(
14    vector_db=MongoVectorDb(
15        collection_name="recipes",
16        db_url=mdb_connection_string,
17        search_index_name="recipes",
18        cosmos_compatibility=True,
19    ),
20)
21
22knowledge.insert(
23    url="https://kern-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
24)
25
26# Create and use the agent
27agent = Agent(knowledge=knowledge_base)
28agent.print_response("How to make Thai curry?", markdown=True)

Azure Cosmos DB MongoDB vCore Params

Parameter	Type	Description	Default
`collection_name`	`str`	Name of the MongoDB collection	Required
`name`	`Optional[str]`	Name of the vector database	`None`
`description`	`Optional[str]`	Description of the vector database	`None`
`id`	`Optional[str]`	Unique identifier for the vector database	Auto-generated
`db_url`	`Optional[str]`	MongoDB connection string	`"mongodb://localhost:27017/"`
`database`	`str`	Database name	`"kern"`
`embedder`	`Optional[Embedder]`	Embedder instance for generating embeddings	`OpenAIEmbedder()`
`distance_metric`	`str`	Distance metric for similarity	`Distance.cosine`
`overwrite`	`bool`	Overwrite existing collection and index if True	`False`
`wait_until_index_ready_in_seconds`	`Optional[float]`	Time in seconds to wait until the index is ready	`3`
`wait_after_insert_in_seconds`	`Optional[float]`	Time in seconds to wait after inserting documents	`3`
`max_pool_size`	`int`	Maximum number of connections in the connection pool	`100`
`retry_writes`	`bool`	Whether to retry write operations	`True`
`client`	`Optional[MongoClient]`	An existing MongoClient instance	`None`
`search_index_name`	`Optional[str]`	Name of the search index	`"vector_index_1"`
`cosmos_compatibility`	`Optional[bool]`	Whether to use Azure Cosmos DB MongoDB vCore compatibility mode	`False`
`search_type`	`SearchType`	The search type to use when searching for documents	`SearchType.vector`
`hybrid_vector_weight`	`float`	Default weight for vector search results in hybrid search	`0.5`
`hybrid_keyword_weight`	`float`	Default weight for keyword search results in hybrid search	`0.5`
`hybrid_rank_constant`	`int`	Default rank constant (k) for Reciprocal Rank Fusion in hybrid search	`60`