Python V4 - Set api version for Azure OpenAI vectorizer?

Support
, ,
1

Description

Is there a way to set the api version for v4 Azure OpenAI vectorizer? I tried
passing the OPENAI_API_VERSION but Weaviate seems to default to:

/embeddings?api-version=2024-02-01

…but our instance of Azure OpenAI uses 2023-05-15

Server Setup Information

  • Weaviate Server Version: 1.26.1
  • Deployment Method: Docker
  • Multi Node? Number of Running Nodes: Single
  • Client Language and Version: Python weaviate_client-4.7.1
  • Multitenancy?:

Any additional Information


  weaviate:
    image: cr.weaviate.io/semitechnologies/weaviate:latest
    ports:
      - 8080:8080
      - 50051:50051
    environment:
      QUERY_DEFAULTS_LIMIT: 25
      AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED: 'true'
      PERSISTENCE_DATA_PATH: '/var/lib/weaviate'
      AZURE_APIKEY: ${AZURE_OPENAI_API_KEY}
      OPENAI_API_TYPE: 'azure'
      ENABLE_MODULES: 'text2vec-azure-openai, text2vec-openai'
      DEFAULT_VECTORIZER_MODULE: 'none'
      OPENAI_API_VERSION: ${AZURE_OPENAI_API_VERSION}
      AZURE_ENDPOINT: ${AZURE_OPENAI_ENDPOINT}
      AZURE_OPENAI_BASE_URL: ${AZURE_OPENAI_BASE_URL}
      AZURE_OPENAI_RESOURCE_NAME: ${AZURE_OPENAI_RESOURCE_NAME}
      AZURE_OPENAI_API_VERSION: ${AZURE_OPENAI_API_VERSION}
      AZURE_OPENAI_EMBEDDING_DEPLOYMENT: ${AZURE_OPENAI_EMBEDDING_DEPLOYMENT}
      CLUSTER_HOSTNAME: 'node1'
    volumes:
      - weaviate_data:/var/lib/weaviate
      


class WeaviateStore:
    def __init__(self, settings: Settings):
        self.settings = settings
        self.client: weaviate.WeaviateAsyncClient = None
        self.class_name = settings.WEAVIATE_CLASS_NAME

    async def initialize(self):
        if self.client is None:
            connection_params = ConnectionParams.from_params(
                http_host=self.settings.WEAVIATE_HOST,
                http_port=self.settings.WEAVIATE_PORT,
                http_secure=False,
                grpc_host=self.settings.WEAVIATE_HOST,
                grpc_port=self.settings.WEAVIATE_GRPC_PORT,
                grpc_secure=False,
            )

            self.client = weaviate.WeaviateAsyncClient(
                connection_params=connection_params,
                additional_headers={
                    "X-Azure-Api-Key": self.settings.AZURE_OPENAI_API_KEY,
                },
            )
            
            await self.client.connect()
            await self._create_or_update_schema()

    async def _create_or_update_schema(self):
        try:
            collection = self.client.collections.get(self.class_name)
            logger.info(f"Collection '{self.class_name}' already exists. Updating configuration.")
            await self.client.collections.delete(self.class_name)
        except UnexpectedStatusCodeError as e:
            if e.status_code != 404:
                logger.error(f"Unexpected error when checking for existing collection: {str(e)}")
                raise
        try:
            properties = [
                Property(name="content", data_type=DataType.TEXT),
                Property(name="metadata", data_type=DataType.OBJECT, nested_properties=[
                    Property(name="source", data_type=DataType.TEXT),
                    Property(name="document_type", data_type=DataType.TEXT),
                    Property(name="npi", data_type=DataType.TEXT),
                ]),
                Property(name="npi", data_type=DataType.TEXT),
            ]

            vectorizer_config = Configure.Vectorizer.text2vec_azure_openai(
                resource_name=self.settings.AZURE_OPENAI_RESOURCE_NAME,
                deployment_id=self.settings.AZURE_OPENAI_EMBEDDING_DEPLOYMENT,
                base_url=self.settings.AZURE_OPENAI_ENDPOINT
            )

            logger.info(f"Creating new collection '{self.class_name}'")
            await self.client.collections.create(
                name=self.class_name,
                properties=properties,
                vectorizer_config=vectorizer_config
            )
            logger.info(f"Successfully created collection '{self.class_name}'")

        except UnexpectedStatusCodeError as e:
            logger.error(f"Failed to create collection: {str(e)}")
            raise
        except Exception as e:
            logger.error(f"Unexpected error during schema creation/update: {str(e)}", exc_info=True)
            raise
2

hi @Taylor_Hickman !!

I believe we have found an undocumented feature :slight_smile:

apiVersion

I am not sure you can pass apiVersion at query time. :thinking: I will have to ask internally as I couldn’t figure out from the PR:

You can also change the baseUrl at query time, as documented here:

I have also found some related forum threads:

Let me know if this helps!

3

Hi!

You cannot pass the apiVersion on query time :frowning:

Our team is looking into this.

but right now, you will need to set the apiversion while creating a collection.