Data source - Azure AI Search

Стаття
09/20/2024

The configurable options of Azure AI Search when using Azure OpenAI On Your Data. This data source is supported in API version 2024-02-01.

Name	Type	Required	Description
`parameters`	Parameters	True	The parameters to use when configuring Azure Search.
`type`	string	True	Must be `azure_search`.

Parameters

Name	Type	Required	Description
`endpoint`	string	True	The absolute endpoint path for the Azure Search resource to use.
`index_name`	string	True	The name of the index to use in the referenced Azure Search resource.
`authentication`	One of ApiKeyAuthenticationOptions, SystemAssignedManagedIdentityAuthenticationOptions, UserAssignedManagedIdentityAuthenticationOptions, onYourDataAccessTokenAuthenticationOptions	True	The authentication method to use when accessing the defined data source.
`embedding_dependency`	One of DeploymentNameVectorizationSource, EndpointVectorizationSource	False	The embedding dependency for vector search. Required when `query_type` is `vector`, `vector_simple_hybrid`, or `vector_semantic_hybrid`.
`fields_mapping`	FieldsMappingOptions	False	Customized field mapping behavior to use when interacting with the search index.
`filter`	string	False	Search filter.
`in_scope`	boolean	False	Whether queries should be restricted to use of indexed data. Default is `True`.
`query_type`	QueryType	False	The query type to use with Azure Search. Default is `simple`
`role_information`	string	False	Give the model instructions about how it should behave and any context it should reference when generating a response. You can describe the assistant's personality and tell it how to format responses.
`semantic_configuration`	string	False	The semantic configuration for the query. Required when `query_type` is `semantic` or `vector_semantic_hybrid`.
`strictness`	integer	False	The configured strictness of the search relevance filtering. The higher of strictness, the higher of the precision but lower recall of the answer. Default is `3`.
`top_n_documents`	integer	False	The configured top number of documents to feature for the configured query. Default is `5`.
`max_search_queries`	integer	False	The max number of rewritten queries should be send to search provider for one user message. If not specified, the system will decide the number of queries to send.
`allow_partial_result`	integer	False	If specified as true, the system will allow partial search results to be used and the request fails if all the queries fail. If not specified, or specified as false, the request will fail if any search query fails.
`include_contexts`	array	False	The included properties of the output context. If not specified, the default value is `citations` and `intent`. Values can be `citations`,`intent`, `all_retrieved_documents`.

API key authentication options

The authentication options for Azure OpenAI On Your Data when using an API key.

Name	Type	Required	Description
`key`	string	True	The API key to use for authentication.
`type`	string	True	Must be `api_key`.

System assigned managed identity authentication options

The authentication options for Azure OpenAI On Your Data when using a system-assigned managed identity.

Name	Type	Required	Description
`type`	string	True	Must be `system_assigned_managed_identity`.

User assigned managed identity authentication options

The authentication options for Azure OpenAI On Your Data when using a user-assigned managed identity.

Name	Type	Required	Description
`managed_identity_resource_id`	string	True	The resource ID of the user-assigned managed identity to use for authentication.
`type`	string	True	Must be `user_assigned_managed_identity`.

Access token authentication options

The authentication options for Azure OpenAI On Your Data when using access token.

Name	Type	Required	Description
`access_token`	string	True	The access token to use for authentication.
`type`	string	True	Must be `access_token`.

Deployment name vectorization source

The details of the vectorization source, used by Azure OpenAI On Your Data when applying vector search. This vectorization source is based on an internal embeddings model deployment name in the same Azure OpenAI resource. This vectorization source enables you to use vector search without Azure OpenAI api-key and without Azure OpenAI public network access.

Name	Type	Required	Description
`deployment_name`	string	True	The embedding model deployment name within the same Azure OpenAI resource.
`type`	string	True	Must be `deployment_name`.
`dimensions`	integer	False	The number of dimensions the embeddings should have. Only supported in `text-embedding-3` and later models.

Endpoint vectorization source

The details of the vectorization source, used by Azure OpenAI On Your Data when applying vector search. This vectorization source is based on the Azure OpenAI embedding API endpoint.

Name	Type	Required	Description
`endpoint`	string	True	Specifies the resource endpoint URL from which embeddings should be retrieved. It should be in the format of `https://{YOUR_RESOURCE_NAME}.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT_NAME/embeddings`. The api-version query parameter isn't allowed.
`authentication`	ApiKeyAuthenticationOptions	True	Specifies the authentication options to use when retrieving embeddings from the specified endpoint.
`type`	string	True	Must be `endpoint`.
`dimensions`	integer	False	The number of dimensions the embeddings should have. Only supported in `text-embedding-3` and later models.

Fields mapping options

Optional settings to control how fields are processed when using a configured Azure Search resource.

Name	Type	Required	Description
`content_fields`	string[]	False	The names of index fields that should be treated as content.
`vector_fields`	string[]	False	The names of fields that represent vector data.
`content_fields_separator`	string	False	The separator pattern that content fields should use. Default is `\n`.
`filepath_field`	string	False	The name of the index field to use as a filepath.
`title_field`	string	False	The name of the index field to use as a title.
`url_field`	string	False	The name of the index field to use as a URL.

Query type

The type of Azure Search retrieval query that should be executed when using it as an Azure OpenAI On Your Data.

Enum Value	Description
`simple`	Represents the default, simple query parser.
`semantic`	Represents the semantic query parser for advanced semantic modeling.
`vector`	Represents vector search over computed data.
`vector_simple_hybrid`	Represents a combination of the simple query strategy with vector data.
`vector_semantic_hybrid`	Represents a combination of semantic search and vector data querying.

Examples

Prerequisites:

Configure the role assignments from Azure OpenAI system assigned managed identity to Azure search service. Required roles: Search Index Data Reader, Search Service Contributor.
Configure the role assignments from the user to the Azure OpenAI resource. Required role: Cognitive Services OpenAI User.
Install Az CLI, and run az login.
Define the following environment variables: AzureOpenAIEndpoint, ChatCompletionsDeploymentName,SearchEndpoint, SearchIndex.

export AzureOpenAIEndpoint=https://example.openai.azure.com/
export ChatCompletionsDeploymentName=turbo
export SearchEndpoint=https://example.search.windows.net
export SearchIndex=example-index

Python 1.x
REST

Install the latest pip packages openai, azure-identity.

import os
from openai import AzureOpenAI
from azure.identity import DefaultAzureCredential, get_bearer_token_provider

endpoint = os.environ.get("AzureOpenAIEndpoint")
deployment = os.environ.get("ChatCompletionsDeploymentName")
search_endpoint = os.environ.get("SearchEndpoint")
search_index = os.environ.get("SearchIndex")

token_provider = get_bearer_token_provider(DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default")

client = AzureOpenAI(
    azure_endpoint=endpoint,
    azure_ad_token_provider=token_provider,
    api_version="2024-02-01",
)

completion = client.chat.completions.create(
    model=deployment,
    messages=[
        {
            "role": "user",
            "content": "Who is DRI?",
        },
    ],
    extra_body={
        "data_sources": [
            {
                "type": "azure_search",
                "parameters": {
                    "endpoint": search_endpoint,
                    "index_name": search_index,
                    "authentication": {
                        "type": "system_assigned_managed_identity"
                    }
                }
            }
        ]
    }
)

print(completion.model_dump_json(indent=2))

az rest --method POST \
 --uri $AzureOpenAIEndpoint/openai/deployments/$ChatCompletionsDeploymentName/chat/completions?api-version=2024-02-01 \
 --resource https://cognitiveservices.azure.com/ \
 --body \
'
{
    "data_sources": [
        {
            "type": "azure_search",
            "parameters": {
                "endpoint": "'$SearchEndpoint'",
                "index_name": "'$SearchIndex'",
                "authentication": {
                    "type": "system_assigned_managed_identity"
                }
            }
        }
    ],
    "messages": [
        {
            "role": "user",
            "content": "Who is DRI?",
        }
    ]
}
'

Поділитися через